S/390: Fix conditional returns on z196+
[official-gcc.git] / gcc / ChangeLog
blob7b4c98b6d884ada9adfec5a71dba2a8c17b73bd8
1 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3         PR target/80080
4         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5         RETURN+USE when returning via %r14.
7 2018-09-24  Martin Liska  <mliska@suse.cz>
9         * gcov.c (output_lines): Print colorization legend
10         for both flag_use_colors and flag_use_hotness_colors.
11         Reword the help.
13 2018-09-24  Martin Liska  <mliska@suse.cz>
15         * coverage.c (get_coverage_counts): Use warning_at
16         with current_function_decl location. Use %qD in warning
17         message.
19 2018-09-24  Martin Liska  <mliska@suse.cz>
21         * memory-block.h (memory_block_pool::release): Annotate with
22         valgrind that the memory is not accessible.
24 2018-09-24  Martin Liska  <mliska@suse.cz>
26         PR sanitizer/85774
27         * asan.c: Make asan_handled_variables extern.
28         * asan.h: Likewise.
29         * cfgexpand.c (expand_stack_vars): Make sure
30         a representative is unpoison if another
31         variable in the partition is handled by
32         use-after-scope sanitization.
34 2018-09-24  Richard Biener  <rguenther@suse.de>
36         PR tree-optimization/63155
37         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
38         the worklist when the edge of the respective argument isn't
39         executable.
41 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
43         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
44         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
45         (MASK_CLASS_P): Update for rename.
46         (MAYBE_MASK_CLASS_P): Ditto.
47         (REG_CLASS_NAMES): Update.
48         (REG_CLASS_CONTENT): Update.
49         * config/i386/i386.c (regclass_map): Update for MASK_REG
50         and ALL_MASK_REGS rename.
51         * config/i386/constraints.md (Yk): Update for rename.
52         (k): Ditto.
54 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
56         * config/i386/i386.h (enum reg_class): Remove
57         EVEX_SSE_REGS and MOD4_SSE_REGS.
58         (REG_CLASS_NAMES): Update.
59         (REG_CLASS_CONTENT): Update.
60         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
61         registers as ALL_SSE_REGS.
62         (ix86_additional_allocno_class_p): Remove.
63         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
64         (ix86_register_priority): Lower priority of EVEX SSE registers.
65         Use IN_RANGE macro where appropriate.
66         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
67         AVX-5124VNNIW checks.
68         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
69         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
70         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
71         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
72         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
73         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
74         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
75         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
76         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
77         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
78         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
79         * config/i386/constraints.md (Yh): Remove.
81 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
83         * config/i386/i386.c (regclass_map): Declare integer REX registers
84         as GENERAL_REGS.
86 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
88         * doc/service.texi (Service): Switch the fsf.org link to https.
90 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
92         PR target/86798
93         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
94         Define to speculation_safe_value_not_needed.
96 2018-09-21  Florian Weimer  <fweimer@redhat.com>
98         PR middle-end/81035
99         * doc/extend.texi (Common Function Attributes): Mention that
100         noreturn suppresses tail call optimization.
102 2018-09-21  Jeff Law  <law@redhat.com>
104         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
105         vr_values::cleanup_edges_and_switches.
106         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
107         vr_values class.
108         (identify_jump_threads): Remove EDGE_IGNORE handling.
109         (execute_vrp): Move handling of to_remove_edges and
110         to_update_switch_stmts into vr_values class member functions.
111         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
112         (to_update_switch_stmts): Likewise.
113         * vr-values.c: Include cfghooks.h.
114         (vr_values::vr_values): Initialize to_remove_edges and
115         to_update_switch_stmts.
116         (vr_values::~vr_values): Verify to_remove_edges and
117         to_update_switch_stmts are empty.
118         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
119         (vr_values::cleanup_edges_and_switches): New member function.
120         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
121         function.  Add new data members.
123 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
125         PR tree-optimization/87309
126         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
127         calls with pflags and alt_flags.
128         (selftest::test_capture_of_dump_calls): Add test of interaction of
129         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
131 2018-09-21  Olivier Hainque  <hainque@adacore.com>
133         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
135 2018-09-21  Olivier Hainque  <hainque@adacore.com>
137         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
139 2018-09-21  Olivier Hainque  <hainque@adacore.com>
141         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
142         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
143         (VXWORKS_LIBS_RTP): Minor reordering.
145 2018-09-21  Olivier Hainque  <hainque@adacore.com>
147         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
148         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
150 2018-09-21  Olivier Hainque  <hainque@adacore.com>
152         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
153         (PTRDIFF_TYPE): Likewise.
155 2018-09-21  Olivier Hainque  <hainque@adacore.com>
157         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
158         triplet, similar to support for VxWorks7.
159         * config/vxworks-dummy.h: Provide a default definition
160         of TARGET_VXWORKS64 to 0.
162 2018-09-21  Olivier Hainque  <hainque@adacore.com>
164         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
165         * config/vxworks-dummy.h: here.
167 2018-09-21  Olivier Hainque  <hainque@adacore.com>
169         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
171 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
172             Bo Zhou  <zbo.zhou@hisilicon.com>
174         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
175         * config/aarch64/aarch64-tune.md: Regenerated.
176         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
177         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
178         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
180 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
181             Julian Brown  <julian@codesourcery.com>
183         * builtins.c (get_builtin_sync_mem): Handle address spaces.
185 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
187         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
188         if the call takes a static chain.
190 2018-09-21  Martin Liska  <mliska@suse.cz>
192         * auto-profile.c (autofdo_source_profile::read): Do not
193         set sum_all.
194         (read_profile): Do not add working sets.
195         (read_autofdo_file): Remove sum_all.
196         (afdo_callsite_hot_enough_for_early_inline): Remove const
197         qualifier.
198         * coverage.c (struct counts_entry): Remove gcov_summary.
199         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
200         do not support GCOV_TAG_PROGRAM_SUMMARY.
201         (get_coverage_counts): Remove summary and expected
202         arguments.
203         * coverage.h (get_coverage_counts): Likewise.
204         * doc/gcov-dump.texi: Remove -w option.
205         * gcov-dump.c (dump_working_sets): Remove.
206         (main): Do not support '-w' option.
207         (print_usage): Likewise.
208         (tag_summary): Likewise.
209         * gcov-io.c (gcov_write_summary): Do not dump
210         histogram.
211         (gcov_read_summary): Likewise.
212         (gcov_histo_index): Remove.
213         (gcov_histogram_merge): Likewise.
214         (compute_working_sets): Likewise.
215         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
216         it not obsolete.
217         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
218         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
219         (GCOV_HISTOGRAM_SIZE): Remove.
220         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
221         (struct gcov_summary): Simplify rapidly just
222         to runs and sum_max fields.
223         (gcov_histo_index): Remove.
224         (NUM_GCOV_WORKING_SETS): Likewise.
225         (compute_working_sets): Likewise.
226         * gcov-tool.c (print_overlap_usage_message): Remove
227         trailing empty line.
228         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
229         (output_lines): Remove program related line.
230         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
231         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
232         histogram.
233         (input_profile_summary): Do not read it.
234         (merge_profile_summaries): And do not merge it.
235         (input_symtab): Do not call removed function.
236         * modulo-sched.c (sms_schedule): Do not print sum_max.
237         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
238         removed when histogram method was invented.
239         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
240         mode.
241         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
242         GCOV coding style.
243         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
244         and dump selected value.
245         * profile.c (add_working_set): Remove.
246         (get_working_sets): Likewise.
247         (find_working_set): Likewise.
248         (get_exec_counts): Do not work with working sets.
249         (read_profile_edge_counts): Do not inform as sum_max is removed.
250         (compute_branch_probabilities): Likewise.
251         (compute_value_histograms): Remove argument for call of
252         get_coverage_counts.
253         * profile.h: Do not make gcov_summary const.
255 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
257         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
259 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
261         PR tree-optimization/86990
262         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
263         Check that the entire merged store group is made of constants only for
264         overlapping stores.
266 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
268         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
269         (VTABLE_VERIFICATION_SPEC): Likewise.
270         (SANITIZER_EARLY_SPEC): Likewise.
271         (SANITIZER_SPEC): Likewise.
272         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
273         * doc/invoke.texi (Link Options): Document -r.
275 2018-09-20  Richard Biener <rguenther@suse.de>
277         PR middle-end/87054
278         * gimplify.c (gimplify_expr): Retain alignment of
279         addressable lvalue in dereference.
281 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
283         PR bootstrap/87013
284         * configure.ac: Check for .loc is_stmt support.
285         * configure, config.in: Rebuilt.
286         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
287         if not supported.
289 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
291         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
292         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
293         -misel=no.
295 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
297         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
298         VECTOR_OTHER.
299         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
300         case VECTOR_OTHER.
302 2018-09-20  Marek Polacek  <polacek@redhat.com>
304         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
306 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
308         PR tree-optimization/87288
309         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
310         into account when determining PEELING_FOR_NITERS.
312 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
314         PR tree-optimization/86877
315         * tree-vect-loop.c (vect_analyze_loop_2): Call
316         vect_verify_datarefs_alignment.
318 2018-09-19  Marek Polacek  <polacek@redhat.com>
320         * doc/invoke.texi: Document -Wclass-conversion.
322 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
324         * config/pa/pa.c (pa_adjust_priority): Delete.
325         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
327         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
328         (atomic_storehi): Likewise.
329         (atomic_storesi): Likewise.
330         (atomic_loaddi): Restore compare and swap exchange loop code.
332 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
334         PR rtl-optimization/86902
335         * combine.c (try_combine): When changing the CC mode used, don't change
336         an unrelated mode in other_insn to that new CC mode.
338 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
340         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
341         with %T in place of calls to dump_generic_expr.
342         (prune_runtime_alias_test_list): Likewise.
343         (create_runtime_alias_checks): Likewise.
344         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
345         (vect_analyze_data_ref_dependence): Likewise.
346         (vect_slp_analyze_data_ref_dependence): Likewise.
347         (vect_record_base_alignment): Likewise.  Use %G in place of call
348         to dump_gimple_stmt.
349         (vect_compute_data_ref_alignment): Likewise.
350         (verify_data_ref_alignment): Likewise.
351         (vect_find_same_alignment_drs): Likewise.
352         (vect_analyze_group_access_1): Likewise.
353         (vect_analyze_data_ref_accesses): Likewise.
354         (dependence_distance_ge_vf): Likewise.
355         (dump_lower_bound): Likewise.
356         (vect_prune_runtime_alias_test_list): Likewise.
357         (vect_find_stmt_data_reference): Likewise.
358         (vect_analyze_data_refs): Likewise.
359         (vect_create_addr_base_for_vector_ref): Likewise.
360         (vect_create_data_ref_ptr): Likewise.
361         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
362         (vect_can_advance_ivs_p): Likewise.
363         (vect_update_ivs_after_vectorizer): Likewise.
364         (vect_gen_prolog_loop_niters): Likewise.
365         (vect_prepare_for_masked_peels): Likewise.
366         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
367         (vect_determine_vectorization_factor): Likewise.
368         (vect_is_simple_iv_evolution): Likewise.
369         (vect_analyze_scalar_cycles_1): Likewise.
370         (vect_analyze_loop_operations): Likewise.
371         (report_vect_op): Likewise.
372         (vect_is_slp_reduction): Likewise.
373         (check_reduction_path): Likewise.
374         (vect_is_simple_reduction): Likewise.
375         (vect_create_epilog_for_reduction): Likewise.
376         (vect_finalize_reduction:): Likewise.
377         (vectorizable_induction): Likewise.
378         (vect_transform_loop_stmt): Likewise.
379         (vect_transform_loop): Likewise.
380         (optimize_mask_stores): Likewise.
381         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
382         (vect_split_statement): Likewise.
383         (vect_recog_over_widening_pattern): Likewise.
384         (vect_recog_average_pattern): Likewise.
385         (vect_determine_min_output_precision_1): Likewise.
386         (vect_determine_precisions_from_range): Likewise.
387         (vect_determine_precisions_from_users): Likewise.
388         (vect_mark_pattern_stmts): Likewise.
389         (vect_pattern_recog_1): Likewise.
390         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
391         (vect_record_max_nunits): Likewise.
392         (vect_build_slp_tree_1): Likewise.
393         (vect_build_slp_tree_2): Likewise.
394         (vect_print_slp_tree): Likewise.
395         (vect_analyze_slp_instance): Likewise.
396         (vect_detect_hybrid_slp_stmts): Likewise.
397         (vect_detect_hybrid_slp_1): Likewise.
398         (vect_slp_analyze_operations): Likewise.
399         (vect_slp_analyze_bb_1): Likewise.
400         (vect_transform_slp_perm_load): Likewise.
401         (vect_schedule_slp_instance): Likewise.
402         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
403         (vect_mark_stmts_to_be_vectorized): Likewise.
404         (vect_init_vector_1): Likewise.
405         (vect_get_vec_def_for_operand): Likewise.
406         (vect_finish_stmt_generation_1): Likewise.
407         (vect_check_load_store_mask): Likewise.
408         (vectorizable_call): Likewise.
409         (vectorizable_conversion): Likewise.
410         (vectorizable_operation): Likewise.
411         (vectorizable_load): Likewise.
412         (vect_analyze_stmt): Likewise.
413         (vect_is_simple_use): Likewise.
414         (vect_get_vector_types_for_stmt): Likewise.
415         (vect_get_mask_type_for_stmt): Likewise.
416         * tree-vectorizer.c (increase_alignment): Likewise.
418 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
420         * doc/rtl.texi: Adjust vec_select description.
421         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
422         non-constant selectors.
424 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
426         * config/aarch64/aarch64-protos.h
427         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
428         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
429         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
430         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
431         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
432         (AARCH64_FL_PROFILE): Move index so flags are ordered.
433         (AARCH64_ISA_RCPC8_4): New flag.
434         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
435         to aarch64_offset_9bit_signed_unscaled_p.
436         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
437         and use stlur.
438         * config/aarch64/constraints.md (Ust): New constraint.
439         * config/aarch64/predicates.md.
440         (aarch64_9bit_offset_memory_operand): New predicate.
441         (aarch64_rcpc_memory_operand): New predicate.
443 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
445         PR rtl-optimization/87361
446         * rtlanal.c (nonzero_bits1): Revert accidental change.
448 2018-09-19  Richard Biener  <rguenther@suse.de>
450         PR tree-optimization/87349
451         PR tree-optimization/87342
452         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
454 2018-09-18  Marek Polacek  <polacek@redhat.com>
456         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
457         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
459 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
461         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
463 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
465         PR rtl-optimization/86882
466         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
468 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
470         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
471         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
473 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
475         PR other/87353
476         * doc/invoke.texi (Link Options): Fix formatting and grammar.
478 2018-09-18  Richard Biener  <rguenther@suse.de>
480         PR middle-end/63155
481         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
482         (compute_samebase_partition_bases): Likewise.
483         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
484         (gimple_can_coalesce_p): Simplify.
486 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
488         Handle a library implementation of ffs calling __builtin_ffs.
489         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
490         (mmix_init_libfuncs): New function: make __builtin_ffs expand
491         to __ffsdi2.
493 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
495         * diagnostic-show-locus.c (class layout_range): Add field
496         "m_original_idx".
497         (layout_range::layout_range): Add "original_idx" param and use it
498         to initialize new field.
499         (make_range): Use 0 for original_idx.
500         (layout::layout): Pass in index to calls to
501         maybe_add_location_range.
502         (layout::maybe_add_location_range): Add param "original_idx" and
503         pass it on to layout_range.
504         (layout::print_any_labels): Pass on range->m_original_idx to
505         get_text call.
506         (gcc_rich_location::add_location_if_nearby): Use 0 for
507         original_idx.
508         * gcc-rich-location.h (text_range_label::get_text): Update for new
509         param.
510         (range_label_for_type_mismatch::get_text): Likewise.
512 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
514         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
516 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
518         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
519         format_string_diagnostic_t.
520         (fmtwarn_n): Likewise.
521         * substring-locations.c
522         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
523         (format_warning_n_va): Convert to...
524         (format_string_diagnostic_t::emit_warning_n_va): ...this.
525         (format_warning_va): Convert to...
526         (format_string_diagnostic_t::emit_warning_va): ...this.
527         (format_warning_at_substring): Convert to...
528         (format_string_diagnostic_t::emit_warning): ...this.
529         (format_warning_at_substring_n): Convert to...
530         (format_string_diagnostic_t::emit_warning_n): ...this.
531         * substring-locations.h (class format_string_diagnostic_t): New
532         class.
533         (format_warning_va): Convert to
534         format_string_diagnostic_t::emit_warning_va.
535         (format_warning_n_va): Convert to
536         format_string_diagnostic_t::emit_warning_n_va.
537         (format_warning_at_substring): Convert to
538         format_string_diagnostic_t::emit_warning.
539         (format_warning_at_substring_n): Convert to
540         format_string_diagnostic_t::emit_warning_n.
542 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
543             Bernd Schmidt <bernds_cb1@t-online.de>
545         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
546         SImode args.
548 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
550         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
551         operand 0 predicate to nonimmediate operand.
552         (rint<mode>2_frndint): Remove insn pattern.
553         (rint<mode>2): Change operand 1 predicate to general_operand.
554         Extend operand 1 to XFmode and generate rintxf2 insn.
555         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
556         Do not use X87MODEF mode macro.
557         (frndintxf2_<rounding>_i387): Rename from
558         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
559         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
560         to XFmode and generate significandxf3 insn.
562 2018-09-17  Richard Biener  <rguenther@suse.de>
564         PR tree-optimization/87328
565         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
566         visiting unexecutable backedges when not iterating.
567         (do_rpo_vn): Mark all edges not executable even when not
568         iterating.
570 2018-09-17  Martin Jambor  <mjambor@suse.cz>
572         PR c/63886
573         * doc/invoke.texi (Warning Options): Likewise.
575 2018-09-17  Richard Biener  <rguenther@suse.de>
577         PR tree-optimization/87301
578         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
579         clean EH info from leftover copy assignments.
581 2018-09-17  Martin Liska  <mliska@suse.cz>
583         PR gcov-profile/85871
584         * gcov.c (output_intermediate_file): Fix out of bounds
585         access.
587 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
589         * config/arc/arc.c: Object attributes for core4 not reflected
590         correctly.
591         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
592         core3).
594 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
596         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
598 2018-09-17  Martin Liska  <mliska@suse.cz>
600         * doc/gcov.texi: Document new option --use-hotness-colors.
601         * gcov.c (struct source_info): Declare new field.
602         (source_info::source_info): Set default for maximum_count.
603         (print_usage): Add new -q option.
604         (process_args): Process it.
605         (accumulate_line_info): Save src->maximum_count.
606         (output_line_beginning): Make color line number if
607         flag_use_hotness_colors is set.
608         (output_line_details): Pass default argument value.
609         (output_lines): Pass src->maximum_count.
611 2018-09-17  Martin Liska  <mliska@suse.cz>
613         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
614         Use processor_names table.
615         * config/i386/i386.c (ix86_default_align): Use
616         processor_cost_table for alignment values.
617         (ix86_option_override_internal): Use processor_names.
618         (ix86_function_specific_print): Likewise.
619         * config/i386/i386.h (struct processor_costs):
620         Add alignment values.
621         (struct ptt): Remove and replace with const char *.
622         * config/i386/x86-tune-costs.h (struct processor_costs):
623         Declare default alignments for all costs.
625 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
627         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
628         symbolics or VR_VARYING ranges for ABS_EXPR.
629         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
630         when range will wrap.
632 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
634         PR middle-end/86864
635         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
636         before and after a JUMP_TABLE_DATA.
638 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
640         PR middle-end/87188
641         * dojump.c (do_compare_and_jump): Canonicalize function pointers
642         when one operand is a function pointer.  Use POINTER_TYPE_P and
643         FUNC_OR_METHOD_TYPE_P.
644         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
645         * fold-const.c (build_range_check): Likewise.
646         * match.pd (simple_comparison): Likewise.
648 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
650         PR c/82967
651         * spellcheck.c (get_edit_distance_cutoff): New function.
652         (selftest::test_edit_distance_unit_test_oneway): Rename to...
653         (selftest::test_get_edit_distance_one_way): ...this.
654         (selftest::test_get_edit_distance_unit): Rename to...
655         (selftest::test_get_edit_distance_both_ways): ...this.
656         (selftest::test_edit_distances): Move tests to this new function,
657         and test some more pairs of strings.  Update for above renaming.
658         (selftest::get_old_cutoff): New function.
659         (selftest::test_get_edit_distance_cutoff): New function.
660         (selftest::assert_suggested_for): New function.
661         (ASSERT_SUGGESTED_FOR): New macro.
662         (selftest::assert_not_suggested_for): New function.
663         (ASSERT_NOT_SUGGESTED_FOR): New macro.
664         (selftest::test_suggestions): New function.
665         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
666         tests to selftest::test_edit_distances and call it.  Add calls to
667         selftest::test_get_edit_distance_cutoff and
668         selftest::test_suggestions.
669         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
670         (best_match::consider): Replace hard-coded cutoff calculation with
671         a call to...
672         (best_match::get_cutoff): New declaration.
673         (best_match::get_best_meaningful_candidate): Likewise.
675 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
677         * builtins.c (fold_builtin_strlen): Remove TODO comment.
679 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
681         revert:
682         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
684         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
685         terminated string literal.
687 2018-09-14  Martin Sebor  <msebor@redhat.com>
689         * builtins.c (unterminated_array): Handle ARRAY_REF.
690         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
691         * builtins.h (unterminated_array): Declare extern.
692         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
693         arrays.
694         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
695         calls.
697 2018-09-14  Martin Sebor  <msebor@redhat.com>
698             Jeff Law  <law@redhat.com>
700         * builtins.c (unterminated_array): New.
701         (expand_builtin_strcpy): Adjust.
702         (expand_builtin_strcpy_args): Detect unterminated arrays.
703         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
704         unterminated arrays.
705         * gimple-fold.h (get_maxval_strlen): Add argument.
706         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
708         * gimple-fold.c (get_range_strlen): Add argument.
709         (get_maxval_strlen): Adjust.
710         * gimple-fold.h (get_range_strlen): Add argument.
712 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
714         * config/i386/movdirintrin.h: Fix copyright year.
716 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
718         * reg-stack.c: Include regs.h.
719         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
720         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
721         FIRST_STACK_REG, not DFmode.
722         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
723         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
724         (change stack): Default register mode to the reg_raw_mode of
725         FIRST_STACK_REG, not DFmode.
726         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
727         (*swapxf): Rename from swapxf.
729 2018-09-14  Carl Love  <cel@us.ibm.com>
731         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
732         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
734 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
736         PR target/87224
737         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
738         alternatives.
740 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
742         PR target/85628
743         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
745 2018-09-14  Jason Merrill  <jason@redhat.com>
747         Fix --enable-gather-detailed-mem-stats.
748         * hash-table.c (hash_table_usage): Change from variable to function.
749         * hash-table.h: Adjust.
750         * Makefile.in: Add missing dependencies on hash-table.h.
752 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
754         PR tree-optimization/87259
755         PR lto/87283
756         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
757         execute_cse_reciprocals_1 has tried transforming.
759 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
761         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
762         VR_VARYING for PLUS/MINUS_EXPR.
764 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
766         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
767         formatting.
769 2018-09-14  Richard Biener  <rguenther@suse.de>
771         PR middle-end/63155
772         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
773         bits for the merged partition.
775 2018-09-13  Martin Sebor  <msebor@redhat.com>
776             Bernd Edlinger  <bernd.edlinger@hotmail.de>
778         * builtins.h (c_srlen): Add argument.
779         * builtins.c (warn_string_no_nul): New function.
780         (c_strlen): Add argument and use it.  Update recursive calls.
781         Pass DECL argument to string_constant to get info on non
782         terminated strings.  Update *NONSTR as needed.
783         (fold_builtin_strlen): Add argument to calls to c_strlen.
784         Warn for unterminated arrays.
785         (warn_string_no_null): Add prototype.
786         * expr.c (string_constant): Update arguments.  Update recursive
787         calls appropriately.  Detect missing NUL terminator and outermost
788         declaration its missing in.
789         Improve checks for arrays with nonzero lower bound or elements
790         that are not a single byte.  Simplify offset computation.
791         Simplify checks for non-NUL terminated strings.
792         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
793         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
795 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
797         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
798         correctly.
799         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
800         argument.  Fix range checks.
801         * fold-const.h (c_getstr): Adjust protoype.
802         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
803         string is constant but contains no NUL byte.
805         * expr.c (string_constant): Adjust function comment.
806         Remove bogus check for zero termination.
808         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
810         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
811         (get_constant_size): Don't make STRING_CSTs larger than they are.
812         (check_string_literal): New check function for STRING_CSTs.
813         (output_constant): Use it.
815 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
817         PR target/86812
818         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
820 2018-09-13  Richard Biener  <rguenther@suse.de>
822         PR tree-optimization/87263
823         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
824         (struct unwind_state): Add max_rpo field.
825         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
826         Compute max_rpo, the max RPO number a block can be backwards reached
827         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
828         separating it from the iterating mode.
830 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
832         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
833         (rfs_decision): New scheduling decision.
835 2018-09-13  Richard Biener  <rguenther@suse.de>
837         PR bootstrap/87134
838         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
839         (vn_nary_op_insert_pieces_predicated): Do not write useless
840         valid_dominated_by_p entry outside of the allocated storage.
842 2018-09-13  Omar Sandoval  <osandov@osandov.com>
843             Tom de Vries  <tdevries@suse.de>
845         PR debug/86985
846         * dwarf2out.c (is_c): New function.
847         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
849 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
851         PR target/85628
852         * config/aarch64/aarch64.md (*aarch64_bfxil):
853         Define.
854         * config/aarch64/constraints.md (Ulc): Define.
855         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
856         Define.
857         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
858         New function.
860 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
862         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
863         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
864         aarch64_layout_frame call.
865         (aarch64_expand_epilogue): Likewise.
866         (aarch64_initial_elimination_offset): Likewise.
867         (aarch64_get_separate_components): Likewise.
868         (aarch64_use_return_insn_p): Likewise.
869         (aarch64_layout_frame): Remove unneeded check.
871 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
873         * configure.ac: Only append
874         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
875         gcc_config_arguments if it was never reconfigured or last reconfigure
876         was with different arguments.
877         * configure: Regenerated.
879 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
880             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
882         PR middle-end/87290
883         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
884         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
886 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
888         PR tree-optimization/87287
889         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
890         X % C == 0 to X % (unsigned) C == 0 optimization to ...
891         * match.pd (X % C == 0): ... here.  New optimization.
893 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
895         PR middle-end/82853
896         * expr.h (maybe_optimize_mod_cmp): Declare.
897         * expr.c (mod_inv): New function.
898         (maybe_optimize_mod_cmp): New function.
899         (do_store_flag): Use it.
900         * cfgexpand.c (expand_gimple_cond): Likewise.
902 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
903             Julian Brown  <julian@codesourcery.com>
905         PR middle-end/86336
906         * gimplify.c (gimplify_scan_omp_clauses): Set
907         target_firstprivatize_array_bases in OpenACC parallel and kernels
908         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
909         OpenACC data regions.
911 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
913         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
914         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
916 2018-09-12  Richard Biener  <rguenther@suse.de>
918         PR tree-optimization/87280
919         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
920         edge but unreachable target.
921         (do_rpo_vn): For conservatively handling a PHI only mark
922         the backedge executable but not the block reachable.
924 2018-09-12  Richard Biener  <rguenther@suse.de>
926         PR tree-optimization/87266
927         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
928         visited blocks.
930 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
932         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
933         constants.
934         ("trunc<BFP:mode><DFP_ALL:mode>2")
935         ("trunc<DFP_ALL:mode><BFP:mode>2")
936         ("extend<BFP:mode><DFP_ALL:mode>2")
937         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
938         according to the target operand type.
940 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
941             Andreas Krebbel  <krebbel@linux.ibm.com>
943         PR tree-optimization/86844
944         * gimple-ssa-store-merging.c
945         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
946         there are any overlapping stores in between them, make sure they are
947         also coalesced or we give up completely.
949 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
951         PR middle-end/87248
952         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
953         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
955 2018-09-12  Tom de Vries  <tdevries@suse.de>
957         * common.opt (gdescribe-dies): Add option.
958         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
959         attribute for artifical and nameless decls.
960         (dwarf2out_register_external_die): Add description attribute to
961         external reference die.
962         (add_desc_attribute): New functions.
963         (gen_subprogram_die): Add description attribute to
964         DW_TAG_call_site_parameter.
965         * tree-pretty-print.c (print_generic_expr_to_str): New function.
966         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
967         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
968         -gno-describe-dies.
969         (@item -gdescribe-dies): Add.
971 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
973         * tree-vrp.c (vrp_shift_undefined_p): Remove.
974         (extract_range_from_binary_expr_1: Call
975         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
976         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
977         depend on sign.
979 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
981         * gimple-ssa-warn-alloca.c
982         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
983         field for ALLOCA_BOUND_*_LARGE.
985 2018-09-11  Nathan Sidwell  <nathan@acm.org>
987         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
989 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
991         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
992         for clobbers.  Remove obsolete comment.
994 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
996         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
997         mpxchk, mpxld and mpxst types.
998         (define_attr length_immediate): Remove all processing of mpx types.
999         (define_attr prefix_0f): Ditto.
1000         (define_attr memory): Ditto.
1002 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
1004         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
1005         (log<mode>2): Change operand 1 predicate to general_operand.
1006         Extend operand 1 to XFmode and generate logxf3 insn.
1007         (log10<mode>2): Change operand 1 predicate to general_operand.
1008         Extend operand 1 to XFmode and generate log10xf3 insn.
1009         (log2<mode>2): Change operand 1 predicate to general_operand.
1010         Extend operand 1 to XFmode and generate log2xf3 insn.
1011         (fyl2xp1_extend<mode>xf3_i387): Remove.
1012         (log1p<mode>2): Change operand 1 predicate to general_operand.
1013         Extend operand 1 to XFmode and generate log1pxf3 insn.
1014         (fxtract_extend<mode>xf3_i387): Remove.
1015         (logb<mode>2): Change operand 1 predicate to general_operand.
1016         Extend operand 1 to XFmode and generate logbxf3 insn.
1017         (ilogb<mode>2): Change operand 1 predicate to general_operand.
1018         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
1019         (significand<mode>2): Change operand 1 predicate to general_operand.
1020         Extend operand 1 to XFmode and generate significandxf3 insn.
1022 2018-09-11  Nathan Sidwell  <nathan@acm.org>
1024         * gcc.c (perror_with_name, pfatal_with_name): Delete.
1025         (load_specs): Use fatal_error.
1026         (DELETE_IF_ORDINARY, process_command): Use error.
1027         (execute, run_attempt): Use fatal_error.
1029 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
1031         * diagnostic-core.h (sorry_at): New prototype.
1032         * diagnostic.c (sorry_at): New function.
1034 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
1036         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
1037         by zero as VR_UNDEFINED.
1039 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
1041         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
1042         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
1043         (<sincos>mode2): New expander.
1044         (sincos_extend<mode>xf3_i387): Remove insn pattern.
1045         (sincos -> sin, cos splitters): Remove splitter patterns.
1046         (sincos<mode>3): Change operand 2 predicate to general_operand.
1047         Extend operand 2 to XFmode and generate sincosxf3 insn.
1048         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
1049         Change operand 3 predicate to const1_operand.
1050         (fptan_extend<mode>xf4_i387): Remove insn pattern.
1051         (tanxf2): Update operands in the call to fptanxf4_i387.
1052         (tan<mode>2): Change operand 1 predicate to general_operand.
1053         Extend operand 1 to XFmode and generate tanxf3 insn.
1054         (atan2xf3): Rename from *fpatanxf3_i387.
1055         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
1056         (atan2xf3): Remove expander.
1057         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
1058         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
1059         (atan<mode>2): Change operand 1 predicate to general_operand.
1060         Extend operand 1 to XFmode and generate atanxf3 insn.
1062 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
1064         * config/i386/i386.md (x87/SSE constant load splitter): Use
1065         memory_operand instead of nonimmediate_operand for input operand
1066         predicate.
1068 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
1070         * config/i386/i386.md (float partial SSE register stall splitter): Move
1071         splitter near its instruction pattern.
1072         (float_extend partial SSE register stall splitter): Ditto.
1073         (float_truncate partial SSE register stall splitter): Ditto.
1075 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
1077         PR target/86794
1078         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1079         to speculation_safe_value_not_needed.
1081         PR target/85666
1082         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
1083         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
1084         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
1085         leaf_function_p, instead use has_hard_reg_initial_val.
1087 2018-09-09  Nathan Sidwell  <nathan@acm.org>
1089         * gcc.h (pfatal_with_name): Don't declare here.
1090         * gcc.c (pfatal_with_name): Make static.
1092 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1094         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
1095         earlyclobber.
1097 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
1099         PR rtl-optimization/85458
1100         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
1101         priority hook to reduce the priority of EXPR.
1103 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
1105         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
1106         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
1107         with X87_ENABLE_FLOAT.  Remove preparation code.
1108         (*float<SWI48:mode><MODEF:mode>2): Rename from
1109         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
1110         math using "enabled" attribute.
1111         (*floatdi<MODEF:mode>2_i387): Rename from
1112         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
1113         enable for 32bit targets only.
1114         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
1115         splitter.
1116         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
1117         as operand 1 predicate.  Rewrite as define_insn_and_split.
1118         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
1120 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
1122         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
1123         to fallthru to FLOAT case.
1125 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
1127         PR target/86731
1128         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
1129         around folding of vec_sl to handle out of range shift values.
1131 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
1133         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
1134         Update callers to gen_fix_trunc<mode>_i387_fisttp
1135         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
1136         nonimmediate_operand.
1137         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
1138         and corresponding splitters.
1139         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
1140         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
1141         (fix_truncdi_i387_with_temp): Remove insn pattern
1142         and corresponding splitters.
1143         (fix_trunc<mode>_i387): Change operand 0 predicate to
1144         nonimmediate_operand.
1145         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
1146         and corresponding splitters.
1147         (*fistdi2_1): Remove.
1148         (fistdi2): Ditto.
1149         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
1150         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
1151         (*fist<mode>2_1): Remove.
1152         (fist<mode>2): Ditto.
1153         (fist<mode>2_with_temp): Remove insn pattern and corresponding
1154         splitters.
1155         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
1156         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
1157         (fistdi2_<rounding>): Change operand 0 predicate to
1158         nonimmediate_operand.
1159         (fistdi2_<rounding>_with_temp): Remove insn pattern
1160         and corresponding splitters.
1161         (fist<mode>2_<rounding>): Change operand 0 predicate to
1162         nonimmediate_operand.
1163         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
1164         and corresponding splitters.
1166         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
1168 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1170         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1171         the init value.
1173 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
1175         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1176         early gimple folding of vec_splat().
1177         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
1178         * gimple-fold.h: Add an extern define for tree_vec_extract().
1180 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
1182         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1183         wrappers around TREE_TYPE comparisons.
1185 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1187         PR target/80080
1188         * config/s390/predicates.md: Add nonsym_memory_operand.
1189         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1190         contains a SYMBOL_REF, load it into an intermediate pseudo.
1191         (s390_emit_compare_and_swap): Legitimize operand.
1192         * config/s390/s390.md: Use the new nonsym_memory_operand
1193         with UNSPECV_CAS patterns.
1195 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1197         PR target/80080
1198         * config/s390/s390-passes.def: New file.
1199         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1200         declaration.
1201         (make_pass_s390_early_mach): Add declaration.
1202         * config/s390/s390.c (make_pass_s390_early_mach):
1203         (s390_option_override): Remove dynamic registration.
1204         * config/s390/t-s390: Add s390-passes.def.
1206 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1208         * config/s390/s390.c (s390_decompose_constant_pool_ref):
1209         Remove UNSPEC_LTREL_BASE check.
1210         (annotate_constant_pool_refs): Likewise.
1211         (find_constant_pool_ref): Likewise.
1212         (find_ltrel_base): Removed.
1213         (replace_ltrel_base): Removed.
1214         (s390_mainpool_finish): Remove replace_ltrel_base call.
1215         (s390_chunkify_start): Remove pending LTREL_BASE logic.
1216         (s390_chunkify_finish): Remove replace_ltrel_base call.
1217         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1219 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
1221         PR target/86779
1222         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1223         to speculation_safe_value_not_needed.
1225 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
1226             Bernd Schmidt  <bernds_cb1@t-online.de>
1228         * config/nvptx/nvptx-opts.h: New file.
1229         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1230         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1231         (ASM_SPEC): Define.
1232         (TARGET_SM35): New macro.
1233         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1234         correct predicate.
1235         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1236         values.
1237         (misa=): New option.
1238         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1240 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
1242         * config/i386/i386.md (truncdfsf2): Remove expander.
1243         (truncdfsf2_with_temp): Ditto.
1244         (truncxf<mode>2): Ditto.
1245         (*truncdfsf_fast_mixed): Remove insn pattern.
1246         (*truncdfsf_fast_i387): Ditto.
1247         (*truncdfsf_mixed): Ditto.
1248         (*truncdfsf_i387): Ditto.
1249         (*truncdfsf2_i387_1): Ditto.
1250         (*truncxfsf2_mixed): Ditto.
1251         (*truncxfdf2_mixed): Ditto.
1252         (*truncxf<mode>2_i387_noop): Ditto. Update callers
1253         to call gen_truncxf<mode>2 instead.
1254         (*truncxf<mode>2_i387): Remove.
1255         (reg->reg splitters): Remove splitter pattern.
1256         (reg->mem splitters): Ditto.
1258         (truncdfsf2): New insn pattern.
1259         (truncxf<mode>2): Ditto.
1261 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1263         * tree-ssa-math-opts.c (is_mult_by): New function.
1264         (is_square_of): Use the above.
1265         (optimize_recip_sqrt): New function.
1266         (pass_cse_reciprocals::execute): Use the above.
1268 2018-09-05  Richard Biener  <rguenther@suse.de>
1270         PR bootstrap/87134
1271         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1272         to zero-init the emplaced vec.
1274 2018-09-05  Martin Liska  <mliska@suse.cz>
1276         PR tree-optimization/87205
1277         * tree-switch-conversion.c (pass_lower_switch::execute):
1278         Group cases for switch statements.
1280 2018-09-05  Richard Biener  <rguenther@suse.de>
1282         PR tree-optimization/87217
1283         * tree-ssa-sccvn.c (vuse_valueize): New.
1284         (vn_reference_lookup_pieces): Use it.
1285         (vn_reference_lookup): Likewise.
1287 2018-09-05  Nathan Sidwell  <nathan@acm.org>
1289         PR c++/87137
1290         * stor-layout.c (place_field): Scan forwards to check last
1291         bitfield when ms_bitfield_placement is in effect.
1293 2018-09-05  Richard Biener  <rguenther@suse.de>
1295         PR bootstrap/87225
1296         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1297         return.
1299 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
1300             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
1302         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1303         * config.gcc (extra_objs): Build it.
1304         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1305         Likewise.
1306         * config/aarch64/aarch64-passes.def
1307         (pass_tag_collision_avoidance): New pass.
1308         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1309         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1310         (aarch64_classify_address): Remove static qualifier.
1311         (aarch64_address_info, aarch64_address_type): Move to...
1312         * config/aarch64/aarch64-protos.h: ... here.
1313         (make_pass_tag_collision_avoidance): New function.
1314         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1315         New tuning flag.
1317 2018-09-05  Martin Liska  <mliska@suse.cz>
1319         * doc/gcov.texi: Update documentation of humar
1320         readable mode.
1321         * gcov.c (format_count): Print one decimal place, it provides
1322         more fine number of situations like '1G' vs. '1.4G'.
1324 2018-09-05  Martin Liska  <mliska@suse.cz>
1326         PR target/87164
1327         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1328         * optc-gen.awk: Allow 'Var' for Deprecated options in order
1329         to generate a MASK value.
1331 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
1333         PR debug/86593
1334         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1335         if frame pointer isn't used.
1336         (compute_frame_pointer_to_fb_displacement): Likewise.
1338 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
1340         PR target/87198
1341         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1342         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1343         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1344         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1345         and OPTION_MASK_ISA_XSAVEC_UNSET.
1347 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
1349         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1350         XOR operations in NAND case.
1352 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
1354         * wide-int-range.cc (wide_int_range_convert): New.
1355         * wide-int-range.h (wide_int_range_convert): New.
1356         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1357         code into wide_int_range_convert.
1358         (extract_range_into_wide_ints): Do not munge anti range constants
1359         into the entire domain.  Just return the range back.
1361 2018-09-04  Martin Liska  <mliska@suse.cz>
1363         * genmatch.c (output_line_directive): Add new argument
1364         fnargs.
1365         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1367 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
1369         * doc/invoke.texi (Option Summary): Add whitespace.
1371         * doc/invoke.texi (Option Summary): Add -Waligned-new.
1373 2018-09-04  Richard Biener  <rguenther@suse.de>
1375         PR tree-optimization/87211
1376         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1377         backedge value we're supposed to treat as VARYING also number
1378         the PHI to VARYING in case it got a different value-number already.
1380 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
1382         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
1383         (extract_range_from_binary_expr_1): Do not call
1384         vrp_can_optimize_bit_op.
1385         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
1386         static.
1387         (wide_int_range_get_mask_and_bounds): New.
1388         (wide_int_range_optimize_bit_op): New.
1389         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
1390         (wide_int_range_bit_and): Same.
1391         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
1392         (wide_int_range_optimize_bit_op): New.
1393         (wide_int_range_get_mask_and_bounds): New.
1395 2018-09-04  Richard Biener  <rguenther@suse.de>
1397         PR tree-optimization/87176
1398         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
1399         variable.  When value-numbering a virtual PHI node make sure
1400         to not value-number to the backedge value.
1402 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
1404         * doc/extend.texi (Long Long, Hex Floats): Document support for
1405         long long and hex floats in more recent versions of ISO C++.
1407 2018-09-03  Richard Biener  <rguenther@suse.de>
1409         PR tree-optimization/87177
1410         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
1411         cleanup.
1413 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1415         * bb-reorder.c (edge_order): Convert to C-qsort-style
1416         tri-state comparator.
1417         (reorder_basic_blocks_simple): Change std::stable_sort to
1418         gcc_stablesort.
1420 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1422         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
1423         tri-state comparator.
1424         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
1426 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1428         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
1429         (mergesort): ... here as maximum count for using netsort.
1430         (gcc_qsort): Set nlim to 3 if stable sort is requested.
1431         (gcc_stablesort): New.
1432         * system.h (gcc_stablesort): Declare.
1434 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1436         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
1437         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
1438         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
1440 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
1442         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
1443         lxsdx and stxsdx alternatives.
1444         (*mov<mode>_hardfloat64): Ditto.
1445         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
1447 2018-09-03  Richard Biener  <rguenther@suse.de>
1449         PR tree-optimization/87200
1450         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1451         simplify result.
1453 2018-09-03  Martin Liska  <mliska@suse.cz>
1455         PR tree-optimization/87201
1456         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1457         Fix parenthesis in an expression.
1459 2018-09-03  Richard Biener  <rguenther@suse.de>
1461         PR tree-optimization/87197
1462         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1463         visited.  CSE the VN_INFO hashtable lookup.
1465         PR tree-optimization/87169
1466         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1467         iterating make sure there's no extra backedges from irreducible
1468         regions feeding the header.  Mark the destination block
1469         executable.
1471 2018-09-03  Martin Liska  <mliska@suse.cz>
1473         PR driver/83193
1474         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1475         * common/common-targhooks.c (default_get_valid_option_values):
1476         New function.
1477         * common/common-targhooks.h (default_get_valid_option_values):
1478         Likewise.
1479         * common/config/i386/i386-common.c: Move processor_target_table
1480         from i386.c.
1481         (ix86_get_valid_option_values): New function.
1482         (TARGET_GET_VALID_OPTION_VALUES): New macro.
1483         * config/i386/i386.c (struct ptt): Move to i386-common.c.
1484         (PTA_*): Move all defined masks into i386-common.c.
1485         (ix86_function_specific_restore): Use new processor_cost_table.
1486         * config/i386/i386.h (struct ptt): Moved from i386.c.
1487         (struct pta): Likewise.
1488         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1489         * doc/tm.texi.in: Likewise.
1490         * opt-suggestions.c (option_proposer::suggest_option):
1491         Pass prefix to build_option_suggestions.
1492         (option_proposer::get_completions): Likewise.
1493         (option_proposer::build_option_suggestions): Use the new target
1494         hook.
1495         * opts.c (struct option_help_tuple): New struct.
1496         (print_filtered_help): Use the new target hook.
1498 2018-09-03  Martin Liska  <mliska@suse.cz>
1500         PR middle-end/59521
1501         * predict.c (set_even_probabilities): Add likely_edges
1502         argument and handle cases where we have precisely one
1503         likely edge.
1504         (combine_predictions_for_bb): Catch also likely_edges.
1505         (tree_predict_by_opcode): Handle gswitch statements.
1506         * tree-cfg.h (find_case_label_for_value): New declaration.
1507         (find_taken_edge_switch_expr): Likewise.
1508         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1509         Find pivot in decision tree based on probabily, not by number of
1510         nodes.
1512 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
1514         * doc/standards.texi (Standards): Update Objective-C reference.
1516 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1518         * doc/install.texi (Prerequisites): Update link for MPC.
1520 2018-09-01  Michael Matz  <matz@suse.de>
1522         PR tree-optimization/87074
1523         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1524         PHIs for outer-loop uses.
1526 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1528         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1529         * doc/invoke.texi (C Dialect Options): Ditto.
1531 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1533         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1535 2018-08-31  Richard Biener  <rguenther@suse.de>
1537         PR tree-optimization/87168
1538         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1539         (rpo_elim::eliminate_avail): When OP was not visited it must
1540         be available.
1542 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
1544         * tree-vrp.c (copy_value_range): Convert param "from" from
1545         "value_range *" to "const value_range *".
1546         (range_is_null): Likewise for param "vr".
1547         (range_int_cst_p): Likewise.
1548         (range_int_cst_singleton_p): Likewise.
1549         (symbolic_range_p): Likewise.
1550         (value_ranges_intersect_p): Likewise for both params.
1551         (value_range_nonnegative_p): Likewise for param "vr".
1552         (value_range_constant_singleton): Likewise.
1553         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1554         (extract_range_into_wide_ints): Likewise for param "vr".
1555         (extract_range_from_multiplicative_op): Likewise for params "vr0"
1556         and "vr1".
1557         (vrp_can_optimize_bit_op): Likewise.
1558         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1559         "vr1_".
1560         (extract_range_from_unary_expr): Likewise.
1561         (debug_value_range): Likewise for param "vr".
1562         (value_range::dump): Add "const" qualifier.
1563         (vrp_prop::check_array_ref): Convert local "vr" from
1564         "value_range *" to "const value_range *".
1565         (vrp_prop::check_mem_ref): Likewise.
1566         (vrp_prop::visit_stmt): Likewise for local "old_vr".
1567         (vrp_intersect_ranges_1): Likewise for param "vr_1".
1568         (vrp_intersect_ranges): Likewise.
1569         (simplify_stmt_for_jump_threading): Likewise for local "vr".
1570         (vrp_prop::vrp_finalize): Likewise.
1571         * tree-vrp.h (value_range::dump): Add "const" qualifier.
1572         (vrp_intersect_ranges): Add "const" qualifier to params as above.
1573         (extract_range_from_unary_expr): Likewise.
1574         (value_range_constant_singleton): Likewise.
1575         (symbolic_range_p): Likewise.
1576         (copy_value_range): Likewise.
1577         (extract_range_from_binary_expr_1): Likewise.
1578         (range_int_cst_p): Likewise.
1579         (vrp_set_zero_nonzero_bits): Likewise.
1580         (range_int_cst_singleton_p): Likewise.
1582 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
1584         * config/aarch64/arm_neon.h (vabsd_s64): New.
1585         (vnegd_s64): Likewise.
1587 2018-08-31  Martin Jambor  <mjambor@suse.cz>
1589         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1591 2018-08-31  Martin Liska  <mliska@suse.cz>
1593         * ipa-icf.c (sem_item::add_type): Use
1594         sem_item::m_type_hash_cache.
1595         * ipa-icf.h: Move the cache from sem_item_optimizer
1596         to sem_item.
1598 2018-08-31  Nathan Sidwell  <nathan@acm.org>
1600         * doc/extend.texi (Backwards Compatibility): Remove implicit
1601         extern C leeway of () being (...).
1603 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1605         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1607 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
1609         PR target/86684
1610         PR target/87149
1611         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1613 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
1615         PR middle-end/87138
1616         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1617         gen_int_mode.  Formatting fixes.
1619 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
1621         * target.def (custom_function_descriptors): Improve documentation.
1622         * doc/tm.texi.in (Trampolines): Expand discussion of function
1623         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1624         beginning of the section.
1625         * doc/tm.texi: Regenerated.
1627 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
1629         * cfg.h (class auto_edge_flag): Spell out the template-id of the
1630         base class in the initializer list.  This is a workaround for
1631         building with older GCC.
1632         (class auto_bb_flag): Likewise.
1634 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
1636         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1637         (altivec_vcmpequ<VI_char>_p): Remove star.
1638         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1639         vector load modes.
1640         (expand_strncmp_vec_sequence): New function.
1641         (emit_final_str_compare_vec): New function.
1642         (expand_strn_compare): Add support for vector strncmp.
1643         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1644         length specification to bytes.
1645         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1646         (vcmpnezb_p): New pattern.
1647         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1648         for option -mstring-compare-inline-limit.
1650 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
1652         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1653         (PTA_SKYLAKE): Add PTA_AES.
1654         (PTA_GOLDMONT): Likewise.
1656 2018-08-29  Jan Hubicka  <jh@suse.cz>
1658         PR lto/86517
1659         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1660         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1662 2018-08-29  Jan Hubicka  <jh@suse.cz>
1664         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1665         TYPE_STUB_DECL.
1666         (hash_tree): Do not visit TYPE_STUB_DECL.
1667         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1668         stream TYPE_STUB_DECL.
1669         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1670         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1671         after free_lang_data.
1672         (type_in_anonymous_namespace_p): Likewise.
1674 2018-08-29  Jan Hubicka  <jh@suse.cz>
1676         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1677         comment that it has to be even number.
1678         (class sreal): Change m_sig type to int32_t.
1679         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1680         int64_t for temporary calculations.
1681         (sreal_verify_basics): Drop one bit from minimum and maximum.
1683 2018-08-30  Richard Biener  <rguenther@suse.de>
1685         PR tree-optimization/87147
1686         * tree-ssa-sccvn.c (SSA_VISITED): New function.
1687         (visit_phi): When the degenerate result is from the backedge and
1688         we didn't visit its definition yet drop to VARYING.
1689         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1691 2018-08-29  Jan Hubicka  <jh@suse.cz>
1693         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1694         DECL_VINDEX.
1695         (hash_tree): Likewise.
1697 2018-08-29  Jan Hubicka  <jh@suse.cz>
1699         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1700         and TYPE_NEXT_REF_TO.
1702 2018-08-29  Jan Hubicka  <jh@suse.cz>
1704         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1705         comment that it has to be even number.
1706         (class sreal): Change m_sig type to int32_t.
1707         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1708         int64_t for temporary calculations.
1709         (sreal_verify_basics): Drop one bit from minimum and maximum.
1711 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
1713         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1715 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
1717         PR middle-end/86995
1718         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1719         if to_add is negative.
1721 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1723         PR middle-end/87053
1724         * builtins.c (c_strlen): Improve range checks.
1726 2018-08-29  Martin Sebor  <msebor@redhat.com>
1727             Jeff Law  <law@redhat.com>
1729         PR tree-optimization/86714
1730         PR tree-optimization/86711
1731         * builtins.c (c_strlen): Add arguments to call to string_constant.
1732         * expr.c (string_constant): Add argument.  Detect missing nul
1733         terminator and outermost declaration it's missing in.
1734         * expr.h (string_constant): Add argument.
1735         * fold-const.c (read_from_constant_string): Add arguments to call to
1736         string_constant.
1737         (c_getstr): Likewise.
1738         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1739         to string_constant.
1740         * tree-ssa-strlen.c (get_stridx): Likewise.
1742 2018-08-29  Jan Hubicka  <jh@suse.cz>
1744         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1745         Do not stream DECL_VINDEX.
1746         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1747         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1748         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1750 2018-08-29  Richard Biener  <rguenther@suse.de>
1752         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1753         virtual operands that are not default defs to honor region
1754         boundaries.
1755         (rpo_vn_valueize): Remove ineffective code here.
1757 2018-08-29  Richard Biener  <rguenther@suse.de>
1759         PR tree-optimization/87132
1760         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1761         when skipping defs reachable over backedges.
1763 2018-08-29  Richard Biener  <rguenther@suse.de>
1765         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1766         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1767         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1768         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1769         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1770         (vn_reference_lookup_3): Remove use of const_parms.
1771         (free_rpo_vn): Do not free const_parms.
1772         (do_rpo_vn): Do not call init_const_parms.
1773         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1774         SSA_NAME_POINTS_TO_READONLY_MEMORY.
1775         (call_may_clobber_ref_p_1): Likewise.
1777 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
1779         PR other/86726
1780         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1781         (-O): Ditto.
1782         (-ftree-scev-cprop): Document.
1784 2018-08-29  Jan Hubicka  <jh@suse.cz>
1786         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1787         parameters.
1788         (sreal constructor): Update.
1789         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1790         sreal:operator/): Update.
1792 2018-08-29  Martin Liska  <mliska@suse.cz>
1794         * tree-switch-conversion.c (switch_conversion::expand):
1795         Strenghten assumption about gswitch statements.
1797 2018-08-29  Richard Biener  <rguenther@suse.de>
1799         PR tree-optimization/87117
1800         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1801         re-value-number released SSA VDEFs.
1803 2018-08-29  Richard Biener  <rguenther@suse.de>
1805         PR tree-optimization/87126
1806         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1808 2018-08-28  Jim Wilson  <jimw@sifive.com>
1810         * config/riscv/pic.md: Rewrite.
1811         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1812         invalid address.
1813         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1814         (SOFTF, default_load, softload, softstore): New.
1816 2018-08-28  Jeff Law  <law@redhat.com>
1818         * fold-const.c (fold_binary_loc): Remove recently added assert.
1820 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
1822         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1823         to OP parmeter of generated function.
1825 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
1827         PR tree-optimization/87009
1828         * match.pd: Add boolean optimizations.
1830 2018-08-28  Martin Sebor  <msebor@redhat.com>
1832         PR middle-end/86631
1833         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1834         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1835         (pass_walloca::gate): Use it.
1836         (alloca_call_type): Same.
1837         (pass_walloca::execute): Same.
1838         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1840 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
1842         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1843         GCC_VERSION for usage of "__gcc_dump_printf__" format from
1844         >= 3005 to >= 9000.
1846 2018-08-28  Richard Biener  <rguenther@suse.de>
1848         PR tree-optimization/87124
1849         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1850         constants before looking up avail.
1852 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
1854         PR middle-end/87099
1855         * calls.c (maybe_warn_nonstring_arg): Punt early if
1856         warn_stringop_overflow is zero.  Don't call get_range_strlen
1857         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1858         Swap comparison operands to have constants on rhs.  Only use
1859         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
1860         increment lenrng[0].
1862 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
1864         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1865         use of tree_to_shwi.  Remove duplicated test for the size being
1866         a whole number of bytes.
1868 2018-08-28  Richard Biener  <rguenther@suse.de>
1870         PR tree-optimization/87117
1871         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1872         Handle removed stmt without LHS (GIMPLE_NOP).
1874 2018-08-28  Richard Biener  <rguenther@suse.de>
1876         PR tree-optimization/87117
1877         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1878         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1880 2018-08-28  Richard Biener  <rguenther@suse.de>
1882         PR tree-optimization/87117
1883         * tree-ssa-pre.c (compute_avail): Do not make expressions
1884         with predicated values available.
1885         (get_expr_value_id): Assert we do not run into predicated value
1886         expressions.
1888 2018-08-28  Richard Biener  <rguenther@suse.de>
1890         PR tree-optimization/87117
1891         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1892         get virtual operands.
1893         (get_expr_operands): Handle STRING_CST like other decls.
1895 2018-08-28  Martin Liska  <mliska@suse.cz>
1897         * tree.h: Update documentation of fndecl_built_in_p
1898         functions.
1901 2018-08-27  Jeff Law  <law@redhat.com>
1902         PR tree-optimization/87110
1903         * tree-ssa-dse.c (compute_trims): Handle non-constant
1904         TYPE_SIZE_UNIT.
1906 2018-08-27  Martin Sebor  <msebor@redhat.com>
1908         PR tree-optimization/86914
1909         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1911 2018-08-27  Martin Sebor  <msebor@redhat.com>
1913         PR tree-optimization/87112
1914         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1915         the type of the bound argument.
1917 2018-08-27  Jeff Law  <law@redhat.com>
1919         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1920         type does not have a TYPE_SIZE_UNIT.
1922 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
1924         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1925         with include of backend.h.
1927 2018-08-27  Richard Biener  <rguenther@suse.de>
1929         PR tree-optimization/86927
1930         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1931         use const cond reduction code.
1933 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
1935         PR tree-optimization/85758
1936         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1938 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
1940         PR c++/87091
1941         * diagnostic-show-locus.c (class layout_range): Update for
1942         conversion of show_caret_p to a tri-state.
1943         (layout_range::layout_range): Likewise.
1944         (make_range): Likewise.
1945         (layout::maybe_add_location_range): Likewise.
1946         (layout::should_print_annotation_line_p): Don't show annotation
1947         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1948         (layout::get_state_at_point): Update for conversion of
1949         show_caret_p to a tri-state.  Bail out early for
1950         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1951         underlining or source colorization.
1952         (gcc_rich_location::add_location_if_nearby): Update for conversion
1953         of show_caret_p to a tri-state.
1954         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1955         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1956         Likewise.
1957         (selftest::test_one_liner_labels): Likewise.
1958         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1959         conversion of show_caret_p to a tri-state.
1960         * pretty-print.c (text_info::set_location): Likewise.
1961         * pretty-print.h (text_info::set_location): Likewise.
1962         * substring-locations.c (format_warning_n_va): Likewise.
1963         * tree-diagnostic.c (default_tree_printer): Likewise.
1964         * tree-pretty-print.c (newline_and_indent): Likewise.
1966 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
1968         PR c++/87091
1969         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1970         line above for line-insertion fix-it hints.
1971         (selftest::test_fixit_insert_containing_newline): Update the
1972         expected results, and add a test with line-numbering enabled.
1974 2018-08-27  Martin Liska  <mliska@suse.cz>
1976         PR sanitizer/86962
1977         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1978         params with DECL_HAS_VALUE_EXPR_P.
1980 2018-08-27  Martin Liska  <mliska@suse.cz>
1982         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1983         selected expansion strategy.
1985 2018-08-27  Martin Liska  <mliska@suse.cz>
1987         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1988         * builtins.c (is_builtin_fn): Likewise.
1989         * attribs.c (diag_attr_exclusions): Use new function
1990         fndecl_built_in_p and remove check for FUNCTION_DECL if
1991         possible.
1992         (builtin_mathfn_code): Likewise.
1993         (fold_builtin_expect): Likewise.
1994         (fold_call_expr): Likewise.
1995         (fold_builtin_call_array): Likewise.
1996         (fold_call_stmt): Likewise.
1997         (set_builtin_user_assembler_name): Likewise.
1998         (is_simple_builtin): Likewise.
1999         * calls.c (gimple_alloca_call_p): Likewise.
2000         (maybe_warn_nonstring_arg): Likewise.
2001         * cfgexpand.c (expand_call_stmt): Likewise.
2002         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
2003         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
2004         (cgraph_node::verify_node): Likewise.
2005         * cgraphclones.c (build_function_decl_skip_args): Likewise.
2006         (cgraph_node::create_clone): Likewise.
2007         * config/arm/arm.c (arm_insert_attributes): Likewise.
2008         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
2009         * dse.c (scan_insn): Likewise.
2010         * expr.c (expand_expr_real_1): Likewise.
2011         * fold-const.c (operand_equal_p): Likewise.
2012         (fold_binary_loc): Likewise.
2013         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
2014         * gimple-low.c (lower_stmt): Likewise.
2015         * gimple-pretty-print.c (dump_gimple_call): Likewise.
2016         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
2017         Likewise.
2018         * gimple.c (gimple_build_call_from_tree): Likewise.
2019         (gimple_call_builtin_p): Likewise.
2020         (gimple_call_combined_fn): Likewise.
2021         * gimplify.c (gimplify_call_expr): Likewise.
2022         (gimple_boolify): Likewise.
2023         (gimplify_modify_expr): Likewise.
2024         (gimplify_addr_expr): Likewise.
2025         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
2026         * ipa-cp.c (determine_versionability): Likewise.
2027         * ipa-fnsummary.c (compute_fn_summary): Likewise.
2028         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
2029         * ipa-split.c (visit_bb): Likewise.
2030         (split_function): Likewise.
2031         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2032         * lto-cgraph.c (input_node): Likewise.
2033         * lto-streamer-out.c (write_symbol): Likewise.
2034         * omp-low.c (setjmp_or_longjmp_p): Likewise.
2035         (lower_omp_1): Likewise.
2036         * predict.c (strip_predict_hints): Likewise.
2037         * print-tree.c (print_node): Likewise.
2038         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
2039         * trans-mem.c (is_tm_irrevocable): Likewise.
2040         (is_tm_load): Likewise.
2041         (is_tm_simple_load): Likewise.
2042         (is_tm_store): Likewise.
2043         (is_tm_simple_store): Likewise.
2044         (is_tm_abort): Likewise.
2045         (tm_region_init_1): Likewise.
2046         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
2047         * tree-cfg.c (verify_gimple_call): Likewise.
2048         (move_stmt_r): Likewise.
2049         (stmt_can_terminate_bb_p): Likewise.
2050         * tree-eh.c (lower_eh_constructs_2): Likewise.
2051         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2052         * tree-inline.c (remap_gimple_stmt): Likewise.
2053         (copy_bb): Likewise.
2054         (estimate_num_insns): Likewise.
2055         (fold_marked_statements): Likewise.
2056         * tree-sra.c (scan_function): Likewise.
2057         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
2058         (optimize_stack_restore): Likewise.
2059         (pass_fold_builtins::execute): Likewise.
2060         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
2061         (mark_all_reaching_defs_necessary_1): Likewise.
2062         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
2063         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2064         (pass_forwprop::execute): Likewise.
2065         * tree-ssa-loop-im.c (stmt_cost): Likewise.
2066         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2067         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
2068         * tree-ssa-strlen.c (get_string_length): Likewise.
2069         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
2070         (find_func_aliases_for_call): Likewise.
2071         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
2072         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
2073         * tree-tailcall.c (find_tail_calls): Likewise.
2074         * tree.c (need_assembler_name_p): Likewise.
2075         (free_lang_data_in_decl): Likewise.
2076         (get_call_combined_fn): Likewise.
2077         * ubsan.c (is_ubsan_builtin_p): Likewise.
2078         * varasm.c (incorporeal_function_p): Likewise.
2079         * tree.h (DECL_BUILT_IN): Remove and replace with
2080         fndecl_built_in_p.
2081         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
2082         (fndecl_built_in_p): New.
2084 2018-08-27  Martin Liska  <mliska@suse.cz>
2086         PR tree-optimization/86847
2087         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
2088         Dump also subtree probability.
2089         (switch_decision_tree::do_jump_if_equal): New function.
2090         (switch_decision_tree::emit_case_nodes): Handle special
2091         situations in balanced tree that can be emitted much simpler.
2092         Fix calculation of probabilities that happen in tree expansion.
2093         * tree-switch-conversion.h (struct cluster): Add
2094         is_single_value_p.
2095         (struct simple_cluster): Likewise.
2096         (struct case_tree_node): Add new function has_child.
2097         (do_jump_if_equal): New.
2099 2018-08-27  Martin Liska  <mliska@suse.cz>
2101         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
2102         Add new argument to bit_test_cluster constructor.
2103         (bit_test_cluster::emit): Set bits really number of values
2104         handlel by a test.
2105         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
2106         probability argument.
2107         * tree-switch-conversion.h (struct bit_test_cluster):
2108         Add m_handles_entire_switch.
2110 2018-08-27  Martin Liska  <mliska@suse.cz>
2112         PR tree-optimization/86702
2113         * tree-switch-conversion.c (jump_table_cluster::emit):
2114         Make probabilities even for values in jump table
2115         according to number of cases handled.
2116         (switch_decision_tree::compute_cases_per_edge): Pass
2117         argument to reset_out_edges_aux function.
2118         (switch_decision_tree::analyze_switch_statement): Likewise.
2119         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
2120         Make it static.
2122 2018-08-27  Martin Liska  <mliska@suse.cz>
2124         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
2125         cfun argument explicitly.
2126         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
2127         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
2128         function gimple_switch_default_bb.
2129         (convert_switch_statements):
2130         (expand_builtins):
2131         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
2132         * stmt.c (label_to_block_fn): Use label_to_block and pass
2133         cfun argument explicitly and use gimple_switch_label_bb.
2134         (expand_case): Likewise.
2135         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
2136         cfun argument explicitly. Likewise.
2137         (make_edges_bb): Likewise.
2138         (make_cond_expr_edges): Likewise.
2139         (get_cases_for_edge): Likewise.
2140         (make_gimple_switch_edges): Likewise.
2141         (label_to_block_fn): Likewise.
2142         (label_to_block): Likewise.
2143         (make_goto_expr_edges): Likewise.
2144         (make_gimple_asm_edges): Likewise.
2145         (main_block_label): Likewise.
2146         (group_case_labels_stmt): Likewise.
2147         (find_taken_edge_computed_goto): Likewise.
2148         (find_taken_edge_switch_expr): Likewise.
2149         (gimple_verify_flow_info): Likewise.
2150         (gimple_redirect_edge_and_branch): Likewise.
2151         (gimple_switch_label_bb): New function.
2152         (gimple_switch_default_bb): Likewise.
2153         (gimple_switch_edge): Likewise.
2154         (gimple_switch_default_edge): Likewise.
2155         * tree-cfg.h (label_to_block_fn): Remove and replace ...
2156         (label_to_block): ... with this.
2157         (gimple_switch_label_bb): New.
2158         (gimple_switch_default_bb): Likewise.
2159         (gimple_switch_edge): Likewise.
2160         (gimple_switch_default_edge): Likewise.
2161         * tree-cfgcleanup.c (convert_single_case_switch): Use
2162         new gimple functions and pass new argument to label_to_block.
2163         (cleanup_control_flow_bb):
2164         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
2165         cfun argument explicitly.
2166         (make_eh_edges): Likewise.
2167         (redirect_eh_dispatch_edge): Likewise.
2168         (lower_resx): Likewise.
2169         (lower_eh_dispatch): Likewise.
2170         (maybe_remove_unreachable_handlers): Likewise.
2171         (unsplit_eh): Likewise.
2172         (cleanup_empty_eh): Likewise.
2173         (verify_eh_edges): Likewise.
2174         (verify_eh_dispatch_edge): Likewise.
2175         * tree-ssa-dom.c (record_edge_info): Likewise.
2176         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2177         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
2178         (thread_through_normal_block): Likewise.
2179         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
2180         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2181         * tree-switch-conversion.c (switch_conversion::collect): Use new
2182         gimple functions.
2183         (switch_conversion::check_final_bb): Likewise.
2184         (switch_conversion::gather_default_values): Pass new argument
2185         to label_to_block.
2186         (switch_conversion::build_constructors): Likewise.
2187         (switch_decision_tree::compute_cases_per_edge): Use new
2188         gimple_switch_edge function.
2189         (switch_decision_tree::analyze_switch_statement): Pass new argument
2190         to label_to_block.
2191         (switch_decision_tree::try_switch_expansion): Use
2192         gimple_switch_default_edge.
2193         * tree-vrp.c (find_switch_asserts): Pass new argument
2194         to label_to_block.
2195         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2196         (vr_values::simplify_switch_using_ranges): Likewise.
2198 2018-08-27  Richard Biener  <rguenther@suse.de>
2200         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2201         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2203         * tree-ssa-sccvn.h (struct vn_pval): New structure.
2204         (struct vn_nary_op_s): Add unwind_to member.  Add
2205         predicated_values flag and put result into a union together
2206         with a linked list of vn_pval.
2207         (struct vn_ssa_aux): Add name member to make maintaining
2208         a map of SSA name to vn_ssa_aux possible.  Remove no longer
2209         needed info, dfsnum, low, visited, on_sccstack, use_processed
2210         and range_info_anti_range_p members.
2211         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2212         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2213         New functions.
2214         (vn_valueize): New global.
2215         (vn_context_bb): Likewise.
2216         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2217         VN_INFO_PTR_INFO): Remove.
2218         * tree-ssa-sccvn.c: ... (rewrite)
2219         (pass_fre::execute): For -O2+ initialize loops and run
2220         RPO VN in optimistic mode (iterating).  For -O1 and -Og
2221         run RPO VN in non-optimistic mode.
2222         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2223         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2224         * doc/invoke.texi (sccvn-max-scc-size): Remove.
2225         (rpo-vn-max-loop-depth): Document.
2226         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2227         when valuezing the VUSE signals we walked out of the region.
2228         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2229         (phi_translate): Set VN context block to use for availability
2230         lookup.
2231         (compute_avail): Likewise.
2232         (pre_valueize): New function.
2233         (pass_pre::execute): Adjust to the RPO VN API.
2235         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2236         (propagate_constants_for_unrolling): Remove.
2237         (tree_unroll_loops_completely): Perform value-numbering
2238         on the unrolled bodies loop parent.
2240 2018-08-27  Richard Biener  <rguenther@suse.de>
2242         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2243         for partial antic compute.
2245 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
2247         PR rtl-optimization/87065
2248         * combine.c (simplify_if_then_else): Formatting fix.
2249         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2250         check.
2251         (known_cond): Don't return const_true_rtx for vector modes.  Use
2252         CONST0_RTX instead of const0_rtx.  Formatting fixes.
2254 2018-08-27  Martin Liska  <mliska@suse.cz>
2256         PR gcov-profile/87069
2257         * gcov.c (process_file): Record files already processed
2258         and warn about a file being processed multiple times.
2260 2018-08-27  Martin Liska  <mliska@suse.cz>
2262         PR driver/83193
2263         * config/aarch64/aarch64.c (aarch64_override_options_internal):
2264         Set default values for x_aarch64_*_string strings.
2265         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2266         prefix.  For -mabi do not print '=ABI' in help and use
2267         <option_value> format for -msve-vector-bits and -moverride
2268         options.
2270 2018-08-26  Jeff Law  <law@redhat.com>
2272         * config/mips/frame-header-opt.c: Include "backend.h" rather than
2273         "cfg.h"
2275 2018-08-26  Marek Polacek  <polacek@redhat.com>
2277         PR c++/87029, Implement -Wredundant-move.
2278         * doc/invoke.texi: Document -Wredundant-move.
2280 2018-08-25  Martin Sebor  <msebor@redhat.com>
2282         PR tree-optimization/87059
2283         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2284         to the same type as the other.
2285         * fold-const.c (fold_binary_loc): Assert expectation.
2287 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
2289         * config/darwin.c (machopic_legitimize_pic_address): Clean up
2290         extraneous parentheses, dead code section and formatting.
2292 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
2294         PR c++/87091
2295         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2296         wide enough for jumps in the line-numbering to be visible.
2297         (layout::print_gap_in_line_numbering): New member function.
2298         (layout::calculate_line_spans): When using line numbering, merge
2299         line spans that are only 1 line apart.
2300         (diagnostic_show_locus): When printing line numbers, show gaps in
2301         line numbering directly, rather than printing headers.
2302         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2303         line-numbering with multiple line spans.
2304         (selftest::test_fixit_insert_containing_newline_2): Add test of
2305         line-numbering, in which the spans are close enough to be merged.
2307 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
2309         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2310         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
2311         * tree-vrp.c (range_is_nonnull): Remove.
2312         (range_includes_zero_p): Accept value_range instead of min/max.
2313         (extract_range_from_binary_expr_1): Do not early bail on
2314         POINTER_PLUS_EXPR.
2315         Use range_includes_zero_p instead of range_is_nonnull.
2316         (extract_range_from_unary_expr): Use range_includes_zero_p instead
2317         of range_is_nonnull.
2318         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
2319         special case VR_ANTI_RANGE.
2320         (vrp_finalize): Same.
2321         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2322         instead of min/max.
2323         (range_is_nonnull): Remove.
2324         * vr-values.c (vrp_stmt_computes_nonzero): Use
2325         range_includes_zero_p instead of range_is_nonnull.
2326         (extract_range_basic): Pass value_range to range_includes_zero_p
2327         instead of range_is_nonnull.
2329 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
2331         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2332         * emit-rtl.h (rtl_data): Remove return_bnd.
2333         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2334         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2335         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2336         (POINTER_BOUNDS_MODE): Remove definition.
2337         (make_pointer_bounds_mode): Remove.
2338         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2339         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2340         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2341         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2342         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2343         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2344         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2346         * config/i386/i386-modes.def (BND32, BND64): Remove.
2347         * config/i386/i386.c (dbx_register_map): Remove bound registers.
2348         (dbx64_register_map): Ditto.
2349         (svr4_dbx_register_map): Ditto.
2350         (indirect_thunk_bnd_needed): Remove.
2351         (indirect_thunks_bnd_used): Ditto.
2352         (indirect_return_bnd_needed): Ditto.
2353         (indirect_return_via_cx_bnd): Ditto.
2354         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2355         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2356         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
2357         (output_indirect_thunk_function): Remove handling of
2358         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2359         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2360         (ix86_save_reg): Remove handling of crtl->return_bnd.
2361         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2362         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2363         and UNSPEC_BNDLX_ADDR.
2364         (ix86_output_indirect_branch_via_reg): Remove handling of
2365         indirect_thunk_prefix_bnd.
2366         (ix86_output_indirect_branch_via_push): Ditto.
2367         (ix86_output_function_return): Ditto.
2368         (ix86_output_indirect_function_return): Ditto.
2369         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2370         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2371         (CALL_USED_REGISTERS): Ditto.
2372         (REG_ALLOC_ORDER): Update for removal of bound registers.
2373         (HI_REGISTER_NAMES): Ditto.
2374         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2375         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2376         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2377         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2378         (FIRST_PSEUDO_REG): Update.
2379         (BND): Remove mode iterator.
2380         * config/i386/predicates.md (bnd_mem_operator): Remove.
2382 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
2384         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
2385         vectors.
2387 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
2389         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
2390         the case in which the permute needs only a single element and
2391         repeats for every vector of the result.  Extend that case to
2392         handle variable-length vectors.
2393         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2395 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
2397         PR debug/79342
2398         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
2399         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
2401 2018-08-24  Richard Biener  <rguenther@suse.de>
2403         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
2404         bb_flags_allocated members.
2405         (auto_flag): New RAII class for allocating flags.
2406         (auto_edge_flag): New RAII class for allocating edge flags.
2407         (auto_bb_flag): New RAII class for allocating bb flags.
2408         * cfgloop.c (verify_loop_structure): Allocate temporary edge
2409         flag dynamically.
2410         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
2411         in favor of temporarily allocated BB flag.
2412         * hsa-brig.c: Re-order includes.
2413         * hsa-dump.c: Likewise.
2414         * hsa-regalloc.c: Likewise.
2415         * print-rtl.c: Likewise.
2416         * profile-count.c: Likewise.
2418 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
2420         PR target/86989
2421         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2422         the TOC register.
2424 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
2426         PR 87073/bootstrap
2427         * wide-int-range.cc (wide_int_range_div): Do not ignore result
2428         from wide_int_range_multiplicative_op.
2430 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2432         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
2433         "permutaion".
2435 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
2437         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
2438         to 'expanded'.
2440 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
2442         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
2443         full GENERIC expression used for replacement.
2445 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
2447         * tree-vrp.c (abs_extent_range): Remove.
2448         (extract_range_into_wide_ints): Pass wide ints by reference.
2449         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2450         Pass wide ints by reference in all calls to
2451         extract_range_into_wide_ints.
2452         * wide-int-range.cc (wide_int_range_div): New.
2453         * wide-int-range.h (wide_int_range_div): New.
2454         (wide_int_range_includes_zero_p): New.
2455         (wide_int_range_zero_p): New.
2457 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
2459         * config/aarch64/aarch64.md (arches): New enum.
2460         (arch): New enum attr.
2461         (arch_enabled): New attr.
2462         (enabled): Now uses arch_enabled only.
2463         (simd, sve, fp16): Removed attribute.
2464         (fp): Attr now defined in terms of 'arch'.
2465         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2466         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2467         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2468         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2469         attributes into 'arch'.
2470         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2471         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2472         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2473         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2474         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2475         'simd' attribute into 'arch'.
2476         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2477         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2478         Convert use of 'fp' attribute to 'arch'.
2479         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2480         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2481         into 'arch'.
2482         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2483         (different modes) Merge 'fp' and 'simd' into 'arch'.
2484         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2485         'simd' into 'arch'.
2487 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
2489         PR rtl-optimization/87026
2490         * expmed.c (canonicalize_comparison): If we can no longer create
2491         pseudoregisters, don't.
2493 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
2495         PR target/86951
2496         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2497         prototype.
2498         * config/arm/arm.c (speculation_barrier_libfunc): New static
2499         variable.
2500         (arm_init_libfuncs): Initialize it.
2501         (arm_emit_speculation_barrier): New function.
2502         * config/arm/arm.md (speculation_barrier): Call
2503         arm_emit_speculation_barrier for architectures that do not have
2504         DSB or ISB.
2505         (speculation_barrier_insn): Only match on Armv7 or later.
2507 2018-08-23  Richard Biener  <rguenther@suse.de>
2509         PR middle-end/87024
2510         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2511         calls.
2513 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
2515         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2516         of single-vector TBLs.
2517         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2518         one input is given.
2520 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
2522         PR target/85910
2523         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2524         aarch64_evpc_tbl guard.
2526 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2528         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2529         behaviour.
2531 2018-08-22  Martin Sebor  <msebor@redhat.com>
2533         PR middle-end/87052
2534         * tree-pretty-print.c (pretty_print_string): Add argument.
2535         (dump_generic_node): Call to pretty_print_string with string size.
2537 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
2539         PR rtl-optimization/86771
2540         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2541         of two SETs into those two SETs, one to be placed at i2, if that SETs
2542         destination is modified between i2 and i3.
2544 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2546         PR tree-optimization/86725
2547         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2548         function.
2549         (vect_analyze_scalar_cycles_1): Check it.
2551 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2553         PR tree-optimization/86725
2554         * tree-vect-loop.c (vect_is_simple_reduction): When treating
2555         an outer loop phi as a double reduction, make sure that the
2556         single user of the phi result is an inner loop phi.
2558 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2560         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2561         grouped stores with gaps to a strided group.
2563 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2565         * tree-vect-stmts.c (get_group_load_store_type)
2566         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2567         first statement in a group.
2569 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2571         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2572         the sequence used in gcc/gcc.c.
2574 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2576         PR other/704
2577         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2578         building it.
2580 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2582         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2583         Darwin10-specific unwinder-shim.
2584         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2585         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2586         New to cater for Darwin10 Rosetta.
2588 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2590         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2591         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2593 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2595         PR bootstrap/81033
2596         PR target/81733
2597         PR target/52795
2598         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2599         (dwarf2out_switch_text_section): Generate a local label for the
2600         second function sub-section and apply it as the second FDE start
2601         label.
2602         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2603         second sub-section start.
2605 2018-08-22  Richard Biener  <rguenther@suse.de>
2607         PR tree-optimization/86988
2608         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2610 2018-08-22  Richard Biener  <rguenther@suse.de>
2612         PR tree-optimization/86945
2613         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2615 2018-08-22  Alexandre Oliva <oliva@adacore.com>
2617         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2618         a comment about how uses of r2 for .sdata2 come about.
2620 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
2622         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2624 2018-08-21  Marek Polacek  <polacek@redhat.com>
2626         PR c++/86981, Implement -Wpessimizing-move.
2627         * doc/invoke.texi: Document -Wpessimizing-move.
2629 2018-08-21  Jan Hubicka  <jh@suse.cz>
2631         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2632         * tree.h (is_redundant_typedef): Remove.
2633         * dwarf2out.c (is_redundant_typedef): Turn into static function.
2635 2018-08-21  Jan Hubicka  <jh@suse.cz>
2637         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2638         when possible.
2640 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
2642         * expmed.c (extract_low_bits): Reject invalid subregs early.
2644 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2646         PR middle-end/86121
2647         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2648         behaviour.
2650 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2652         * config/vxworks.h: Guard vxworks_asm_out_constructor and
2653         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2654         * config/vxworks.c: Likewise.
2656 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2658         * config/vxworks.c: Set targetm.have_ctors_dtors
2659         if HAVE_INITFINI_ARRAY_SUPPORT.
2660         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2661         if HAVE_INITFINI_ARRAY_SUPPORT.
2663 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2665         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2666         default search path for VxWorks < 7.
2668 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2670         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2671         (get_format_string): Refer to c_getstr.
2673 2018-08-21  Tom de Vries  <tdevries@suse.de>
2675         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2676         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2677         (debuginfo_early_stop): Declare.
2678         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2679         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2680         (debuginfo_early_stop): New function.
2681         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2682         and debuginfo_early_stop.
2683         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2684         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2685         (general_init): Call debuginfo_early_init.
2686         (finalize): Call debuginfo_fini.
2687         (do_compile): Call debuginfo_init.
2688         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2689         -fdump-early-debug.
2690         (@item -fdump-debug, @item -fdump-earlydebug): Add.
2692 2018-08-21  Tom de Vries  <tdevries@suse.de>
2694         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2695         flag_dump_noaddr and flag_dump_unnumbered.
2697 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
2699         * wide-int-range.cc (wide_int_range_abs): New.
2700         (wide_int_range_order_set): Rename from wide_int_range_min_max.
2701         * wide-int-range.h (wide_int_range_abs): New.
2702         (wide_int_range_min_max): New.
2703         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2704         case to call wide_int_range_abs.
2705         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2706         (extract_range_from_abs_expr): Delete.
2708 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
2710         PR target/87033
2711         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2712         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2713         for indexed loads.
2715 2018-08-20  Nathan Sidwell  <nathan@acm.org>
2716             Jeff Law <law@redhat.com>
2718         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2719         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2721 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
2723         PR other/84889
2724         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2725         (decl_attributes): Likewise.
2726         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2727         instance.
2728         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2729         * diagnostic-core.h (class auto_diagnostic_group): New class.
2730         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2731         (diagnostic_report_diagnostic): Handle the first diagnostics within
2732         a group.
2733         (emit_diagnostic): Add auto_diagnostic_group instance.
2734         (inform): Likewise.
2735         (inform_n): Likewise.
2736         (warning): Likewise.
2737         (warning_at): Likewise.
2738         (warning_n): Likewise.
2739         (pedwarn): Likewise.
2740         (permerror): Likewise.
2741         (error): Likewise.
2742         (error_n): Likewise.
2743         (error_at): Likewise.
2744         (sorry): Likewise.
2745         (fatal_error): Likewise.
2746         (internal_error): Likewise.
2747         (internal_error_no_backtrace): Likewise.
2748         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2749         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2750         * diagnostic.h (struct diagnostic_context): Add fields
2751         "diagnostic_group_nesting_depth",
2752         "diagnostic_group_emission_count", "begin_group_cb",
2753         "end_group_cb".
2754         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2755         Add auto_diagnostic_group instance(s).
2756         (find_explicit_erroneous_behavior): Likewise.
2757         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2758         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2759         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2760         (gimplify_va_arg_expr): Likewise.
2761         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2762         (HSA_SORRY_AT): Likewise.
2763         * ipa-devirt.c (compare_virtual_tables): Likewise.
2764         (warn_odr): Likewise.
2765         * multiple_target.c (expand_target_clones): Likewise.
2766         * opts-common.c (cmdline_handle_error): Likewise.
2767         * reginfo.c (globalize_reg): Likewise.
2768         * substring-locations.c (format_warning_n_va): Likewise.
2769         * tree-inline.c (expand_call_inline): Likewise.
2770         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2771         * tree-ssa-loop-niter.c
2772         (do_warn_aggressive_loop_optimizations): Likewise.
2773         * tree-ssa-uninit.c (warn_uninit): Likewise.
2774         * tree.c (warn_deprecated_use): Likewise.
2776 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
2778         PR target/87014
2779         * config/i386/i386.md (eh_return): Always update EH return
2780         address in word_mode.
2782 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
2784         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2785         TARGET_SPLIT_COMPLEX_ARG is defined.
2787 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2789         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2791 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2793         PR target/86984
2794         * expr.c (expand_assignment): Assert that bitpos is positive.
2795         (store_field): Likewise
2796         (expand_expr_real_1): Make sure that bitpos is positive.
2797         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2798         integer overflow.
2800 2018-08-20  Nathan Sidwell  <nathan@acm.org>
2802         * Makefile.in (CPP_ID_DATA_H): Delete.
2803         (CPP_INTERNAL_H): Don't add it.
2804         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2805         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2807 2018-08-20  Richard Biener  <rguenther@suse.de>
2809         PR tree-optimization/78655
2810         * tree-vrp.c (extract_range_from_binary_expr_1): Make
2811         pointer + offset nonnull if either operand is nonnull work.
2813 2018-08-20  Tom de Vries  <tdevries@suse.de>
2815         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2816         unless the referenced die describes the added property using
2817         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
2818         Otherwise, add a DW_AT_location to the referenced die.
2820 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
2822         PR target/86994
2823         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2824         register_operand when calling ix86_set_reg_reg_cost.
2825         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2826         Set *total to 0 for operands that satisfy x86_64_immediate_operand
2827         predicate and to 1 otherwise.
2829 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
2831         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2832         emit a diagnostic that it is not supported and reset the option.
2833         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2834         supported and consume the option.  (ASM_FINAL_SPEC): New.
2836 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
2838         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2839         a sentence.
2841 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
2843         C-SKY port: Documentation
2845         * doc/extend.texi (C-SKY Function Attributes): New section.
2846         * doc/invoke.texi (Option Summary): Add C-SKY options.
2847         (C-SKY Options): New section.
2848         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2850 2018-08-17  Jojo  <jijie_rong@c-sky.com>
2851             Huibin Wang  <huibin_wang@c-sky.com>
2852             Sandra Loosemore  <sandra@codesourcery.com>
2853             Chung-Lin Tang  <cltang@codesourcery.com>
2855         C-SKY port: Backend implementation
2857         * config/csky/*: New.
2858         * common/config/csky/*: New.
2860 2018-08-17  Jojo  <jijie_rong@c-sky.com>
2861             Huibin Wang  <huibin_wang@c-sky.com>
2862             Sandra Loosemore  <sandra@codesourcery.com>
2863             Chung-Lin Tang  <cltang@codesourcery.com>
2864             Andrew Jenner  <andrew@codesourcery.com>
2866         C-SKY port: Configury
2868         * config.gcc (csky-*-*): New.
2869         * configure.ac: Add csky to targets for dwarf2 debug_line support.
2870         * configure: Regenerated.
2872 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
2874         * dump-context.h: Include "dumpfile.h".
2875         (dump_context::dump_printf_va): Convert final param from va_list
2876         to va_list *.  Convert from ATTRIBUTE_PRINTF to
2877         ATTRIBUTE_GCC_DUMP_PRINTF.
2878         (dump_context::dump_printf_loc_va): Likewise.
2879         * dumpfile.c: Include "stringpool.h".
2880         (make_item_for_dump_printf_va): Delete.
2881         (make_item_for_dump_printf): Delete.
2882         (class dump_pretty_printer): New class.
2883         (dump_pretty_printer::dump_pretty_printer): New ctor.
2884         (dump_pretty_printer::emit_items): New member function.
2885         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2886         function.
2887         (dump_pretty_printer::emit_item): New member function.
2888         (dump_pretty_printer::stash_item): New member function.
2889         (dump_pretty_printer::format_decoder_cb): New member function.
2890         (dump_pretty_printer::decode_format): New member function.
2891         (dump_context::dump_printf_va): Reimplement in terms of
2892         dump_pretty_printer.
2893         (dump_context::dump_printf_loc_va): Convert final param from va_list
2894         to va_list *.
2895         (dump_context::begin_scope): Reimplement call to
2896         make_item_for_dump_printf.
2897         (dump_printf): Update for change to dump_printf_va.
2898         (dump_printf_loc): Likewise.
2899         (selftest::test_capture_of_dump_calls): Convert "stmt" from
2900         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
2901         with %T, %E, and %G.
2902         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2903         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2904         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2905         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2906         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2907         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2908         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2909         within a dump_printf_loc call to "%wu".
2910         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2911         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
2912         missing space after "=".
2913         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2914         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2915         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2916         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2917         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
2918         duplicate "vectorized" from message.
2920 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
2922         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2923         polyNxK_t element's TYPE_STRING_FLAG.
2925 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
2927         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2928         (they were unnamed before).  Fix comments.
2930 2018-08-17  Nathan Sidwell  <nathan@acm.org>
2932         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2934 2018-08-17  Richard Biener  <rguenther@suse.de>
2936         PR tree-optimization/86841
2937         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2939 2018-08-17  Martin Liska  <mliska@suse.cz>
2941         * common.opt: Remove Warn, Init and Report for options with
2942         Ignore/Deprecated flag. Warning is done automatically for
2943         Deprecated flags.
2944         * config/i386/i386.opt: Likewise.
2945         * config/ia64/ia64.opt: Likewise.
2946         * config/rs6000/rs6000.opt: Likewise.
2947         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2948         Remove usage of flag_check_pointer_bounds.
2949         * lto-wrapper.c (merge_and_complain): Do not handle
2950         OPT_fcheck_pointer_bounds.
2951         (append_compiler_options): Likewise.
2952         * opt-functions.awk: Do not handle Deprecated.
2953         * optc-gen.awk: Check that Var, Report and Init are not
2954         used for an option with Ignore/Deprecated flag.
2955         * opts-common.c (decode_cmdline_option): Do not report
2956         CL_ERR_DEPRECATED.
2957         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2958         options.
2959         * opts.h (struct cl_option): Remove cl_deprecated flag.
2960         (CL_ERR_DEPRECATED): Remove error enum value.
2962 2018-08-17  Richard Biener  <rguenther@suse.de>
2964         PR middle-end/86505
2965         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2966         across a va-arg-pack using call adjust its return value accordingly.
2968 2018-08-16  Martin Sebor  <msebor@redhat.com>
2970         PR tree-optimization/86853
2971         * gimple-ssa-sprintf.c (struct format_result): Rename member.
2972         (struct fmtresult): Add member and initialize it in ctors.
2973         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
2974         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
2975         (format_directive): Set POSUNDER4K when MAYFAIL is set.
2976         (parse_directive): Handle %C same as %c.
2977         (sprintf_dom_walker::compute_format_length): Adjust.
2978         (is_call_safe): Adjust.
2980 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2982         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
2983         for determining how to count the elements.
2984         * builtins.h (c_strlen): Adjust prototype.
2985         * expr.c (string_constant): Add new parameter mem_size.
2986         Set *mem_size appropriately.
2987         * expr.h (string_constant): Adjust protoype.
2988         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2989         * gimple-fold.h (get_range_strlen): Adjust prototype.
2990         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2991         (format_string): Call get_string_length with eltsize.
2993 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
2995         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2996         to emit the span, rather than setting it as the prefix.
2998 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
3000         * diagnostic-show-locus.c (layout::start_annotation_line): Add
3001         "margin_char" parameter, defaulting to space.  Use it in place
3002         of pp_space for the initial part of the margin.
3003         (layout::print_leading_fixits): Use '+' when filling the margin
3004         of line-insertion fix-it hints.
3006 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3008         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
3009         Delete.
3011 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3013         * config/rs6000/altivec.md: Don't set length attribute to the default
3014         value.
3015         * config/rs6000/darwin.md: Ditto.
3016         * config/rs6000/dfp.md: Ditto.
3017         * config/rs6000/htm.md: Ditto.
3018         * config/rs6000/rs6000.md: Ditto.
3019         * config/rs6000/sync.md: Ditto.
3020         * config/rs6000/vsx.md: Ditto.
3022 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3024         * config/rs6000/altivec.md: Don't set length attribute to the default
3025         value, for branch instructions.
3026         * config/rs6000/darwin.md: Ditto.
3027         * config/rs6000/rs6000.md: Ditto.
3029 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3031         * config/rs6000/rs6000.md (length): Always define as const_int 4.
3032         (unnamed conditional branch define_insn): Set length to 4 or 8
3033         depending on offset.
3034         (<bd>_<mode>): Similar, for alternative 0.
3035         (<bd>tf_<mode>): Ditto.
3037 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
3039         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
3041 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
3043         * doc/rtl.texi: Replace old RTX class names with new names.
3046 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
3048         * expmed.h (canonicalize_comparison): New declaration.
3049         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
3050         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
3051         * optabs.c (prepare_cmp_insn): Likewise.
3052         * rtl.h (unsigned_condition_p): New function which checks if a
3053         comparison operator is unsigned.
3055 2018-08-16  Nathan Sidwell  <nathan@acm.org>
3057         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
3058         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
3060 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
3062         PR target/84711
3063         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
3064         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
3065         (mov<mov>): ..this and enable unconditionally.
3067 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
3069         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
3071 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
3073         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
3074         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
3075         "Common" with "Target".
3077 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
3079         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
3080         * doc/invoke.texi (mmitigate-rop): Remove.
3081         * config/i386/i386.c: Do not include "regrename.h".
3082         (ix86_rop_should_change_byte_p, reg_encoded_number)
3083         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
3084         Remove.
3085         (ix86_reorg): Remove call to ix86_mitigate_rop.
3086         * config/i386/i386.md (attr "modrm_class"): Remove.
3087         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
3088         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
3089         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
3091 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
3093         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
3094         allow folding of mergeh() and mergel() for the float and double types.
3095         (fold_mergehl_helper): Rework to handle building a permute tree
3096         for float vectors.
3098 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
3100         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
3101         for TARGET_SSE.
3103 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
3105         * common.opt (fdiagnostics-show-labels): New option.
3106         * diagnostic-show-locus.c (class layout_range): Add field
3107         "m_label".
3108         (class layout): Add field "m_show_labels_p".
3109         (layout_range::layout_range): Add param "label" and use it to
3110         initialize m_label.
3111         (make_range): Pass in NULL for new "label" param of layout_range's
3112         ctor.
3113         (layout::layout): Initialize m_show_labels_p.
3114         (layout::maybe_add_location_range): Pass in loc_range->m_label
3115         when constructing layout_range instances.
3116         (struct line_label): New struct.
3117         (layout::print_any_labels): New member function.
3118         (layout::print_line): Call it if label-printing is enabled.
3119         (selftest::test_one_liner_labels): New test.
3120         (selftest::test_diagnostic_show_locus_one_liner): Call it.
3121         * diagnostic.c (diagnostic_initialize): Initialize
3122         context->show_labels_p.
3123         * diagnostic.h (struct diagnostic_context): Add field
3124         "show_labels_p".
3125         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3126         -fno-diagnostics-show-labels.
3127         * dwarf2out.c (gen_producer_string): Add
3128         OPT_fdiagnostics_show_labels to the ignored options.
3129         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
3130         param.
3131         (gcc_rich_location::maybe_add_expr): Likewise.
3132         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
3133         label" param, defaulting to NULL.
3134         (gcc_rich_location::add_expr): Add "label" param.
3135         (gcc_rich_location::maybe_add_expr): Likewise.
3136         (class text_range_label): New class.
3137         (class range_label_for_type_mismatch): New class.
3138         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
3139         of format_warning_va.
3140         (fmtwarn_n): Likewise for new params of format_warning_n_va.
3141         * lto-wrapper.c (merge_and_complain): Add
3142         OPT_fdiagnostics_show_labels to the "pick one setting" options.
3143         (append_compiler_options): Likewise to the dropped options.
3144         (append_diag_options): Likewise to the passed-on options.
3145         * opts.c (common_handle_option): Handle the new option.
3146         * selftest-diagnostic.c
3147         (test_diagnostic_context::test_diagnostic_context): Enable
3148         show_labels_p.
3149         * substring-locations.c: Include "gcc-rich-location.h".
3150         (format_warning_n_va): Add "fmt_label" and "param_label" params
3151         and use them as appropriate.
3152         (format_warning_va): Add "fmt_label" and "param_label" params,
3153         passing them on to format_warning_n_va.
3154         (format_warning_at_substring): Likewise.
3155         (format_warning_at_substring_n): Likewise.
3156         * substring-locations.h (format_warning_va): Add "fmt_label" and
3157         "param_label" params.
3158         (format_warning_n_va): Likewise.
3159         (format_warning_at_substring): Likewise.
3160         (format_warning_at_substring_n): Likewise.
3161         * toplev.c (general_init): Initialize global_dc->show_labels_p.
3163 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
3165         PR testsuite/86519
3166         * builtins.c (expand_builtin_memcmp): Do not expand the call
3167         when overflow is detected.
3169 2018-08-15  Martin Sebor  <msebor@redhat.com>
3171         PR tree-optimization/71625
3172         * config/aarch64/aarch64-builtins.c
3173         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3175 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
3177         * config/s390/s390.c (s390_reorg): Remove loop.
3179 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
3181         * config/darwin.c
3182          (darwin_function_switched_text_sections): Delete.
3183         * gcc/config/darwin.h
3184          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3186 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
3188         PR target/81685
3189         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3190         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
3191         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3193 2018-08-15  Martin Liska  <mliska@suse.cz>
3195         PR tree-optimization/86925
3196         * predict.c (expr_expected_value_1): When taking
3197         later predictor, assign also probability.
3198         Use fold_build2_initializer_loc in order to fold
3199         the expression in -frounding-math.
3201 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
3203         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3204         patterns.
3205         (expand_vec_perm_1): Try the new method.
3207 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
3209         PR target/86547
3210         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3211         Check whether lra_live_max_point is 0 before dividing.
3213 2018-08-14  Martin Sebor  <msebor@redhat.com>
3215         PR tree-optimization/86650
3216         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3217         (vrp_prop::check_mem_ref): Same.
3219 2018-08-13  Liu Hao <lh_mouse@126.com>
3221         * pretty-print.c (eat_esc_sequence): Swap the foreground and
3222         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3223         and clear it thereafter, as it only works for DBCS.
3225 2018-08-13  Liu Hao <lh_mouse@126.com>
3227         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3228         handle returned by _get_osf_handle().
3230 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
3232         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3233         for folding vec_perm.
3235 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
3237         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3238         Add support for gimple-folding of vec_pack() and vec_unpack()
3239         intrinsics.
3241 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
3243         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3244         vec_xst variants to the list.
3245         (rs6000_gimple_fold_builtin): Add support for folding unaligned
3246         vector loads and stores.
3248 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
3250         * config.gcc (rs6000-ibm-aix4.x): Delete.
3251         (rs6000-ibm-aix5.1): Delete.
3252         (rs6000-ibm-aix5.2): Delete.
3253         (rs6000-ibm-aix5.3): Delete.
3254         * config/rs6000/aix43.h: Delete.
3255         * config/rs6000/aix51.h: Delete.
3256         * config/rs6000/aix52.h: Delete.
3257         * config/rs6000/t-aix43: Delete.
3259 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
3261         * config/s390/s390.c (s390_decompose_constant_pool_ref):
3262         New function.
3263         (s390_decompose_address): Factor out constant pool ref
3264         decomposition.
3266 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3268         * config/nds32/nds32-predicates.c
3269         (nds32_can_use_bclr_p): Change return type as bool.
3270         (nds32_can_use_bset_p): Ditto.
3271         (nds32_can_use_btgl_p): Ditto.
3272         (nds32_can_use_bitci_p): Ditto.
3273         * config/nds32/nds32-protos.h
3274         (nds32_can_use_bclr_p): Change declaration.
3275         (nds32_can_use_bset_p): Ditto.
3276         (nds32_can_use_btgl_p): Ditto.
3277         (nds32_can_use_bitci_p): Ditto.
3279 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3281         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3282         Support -msched-prolog-epilog option.
3283         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3285 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3287         * common/config/nds32/nds32-common.c
3288         (nds32_option_optimization_table): Enalbe -malways-align.
3290 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3292         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3293         extra_headers.
3294         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3295         OPT_misr_secure_ case.
3296         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3297         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3298         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3299         secure attribute.
3300         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3301         (nds32_isr_info): New field security_level.
3302         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3303         * config/nds32/nds32.md (return_internal): Consider critical attribute.
3304         * config/nds32/nds32.opt (misr-secure): New option.
3305         * config/nds32/nds32_init.inc: New file.
3306         * config/nds32/nds32_isr.h: New file.
3308 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
3310         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3311         Update comment for atomic instructions.
3312         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3313         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3314         Remove.
3315         (atomic_loaddi): Revise fence expansion to only emit fence prior to
3316         load for __ATOMIC_SEQ_CST model.
3317         (atomic_loaddi_1): Remove float register target.
3318         (atomic_storedi): Handle CONST_INT values.
3319         (atomic_storedi_1): Remove float register source.  Add special case
3320         for zero value.
3321         (memory_barrier): New expander and insn.
3323 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
3325         PR tree-optimization/86835
3326         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3327         new_stmt after def_gsi, make sure to insert new_square_stmt after
3328         that stmt, not 2 stmts before it.
3330 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
3332         PR target/82418
3333         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3334         instead of SWI48.
3336 2018-08-10  Martin Liska  <mliska@suse.cz>
3338         PR target/83610
3339         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3340         function type.
3341         * builtins.c (expand_builtin_expect_with_probability):
3342         New function.
3343         (expand_builtin_expect_with_probability): New function.
3344         (build_builtin_expect_predicate): Add new argumnet probability
3345         for BUILT_IN_EXPECT_WITH_PROBABILITY.
3346         (fold_builtin_expect):
3347         (fold_builtin_2):
3348         (fold_builtin_3):
3349         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3350         * builtins.h (fold_builtin_expect): Set new argument.
3351         * doc/extend.texi: Document __builtin_expect_with_probability.
3352         * doc/invoke.texi: Likewise.
3353         * gimple-fold.c (gimple_fold_call): Pass new argument.
3354         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3355         also BUILT_IN_EXPECT_WITH_PROBABILITY.
3356         * predict.c (get_predictor_value): New function.
3357         (expr_expected_value): Add new argument probability. Assume
3358         that predictor and probability are always non-null.
3359         (expr_expected_value_1): Likewise.  For __builtin_expect and
3360         __builtin_expect_with_probability set probability.  Handle
3361         combination in binary expressions.
3362         (tree_predict_by_opcode): Simplify code by simply calling
3363         get_predictor_value.
3364         (pass_strip_predict_hints::execute): Add handling of
3365         BUILT_IN_EXPECT_WITH_PROBABILITY.
3366         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3367         new predictor.
3368         * tree.h (DECL_BUILT_IN_P): New function.
3370 2018-08-10  Martin Liska  <mliska@suse.cz>
3372         PR tree-optimization/85799
3373         * passes.def: Add argument for pass_strip_predict_hints.
3374         * predict.c (class pass_strip_predict_hints): Add new argument
3375         early_p.
3376         (strip_predictor_early): New function.
3377         (pass_strip_predict_hints::execute): Call the function to
3378         strip predictors.
3379         (strip_predict_hints): New function.
3380         * predict.def: Fix comment.
3382 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
3384         * Makefile.in: Clarify which tm.texi to copy over to assert the
3385         right to grant a GFDL license for all.
3387 2018-08-09  Jeff Law  <law@redhat.com>
3389         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
3390         unused variable.
3392 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
3394         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
3395         prototype.
3397 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3399         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
3400         reductions for variable-length vectors.
3402 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
3404         PR other/84889
3405         * common.opt (fdiagnostics-show-line-numbers): New option.
3406         * diagnostic-show-locus.c (class layout): Add fields
3407         "m_show_line_numbers_p" and "m_linenum_width";
3408         (num_digits): New function.
3409         (test_num_digits): New function.
3410         (layout::layout): Initialize new fields.  Update m_x_offset
3411         logic to handle any left margin.
3412         (layout::print_source_line): Print line number when requested.
3413         (layout::start_annotation_line): New member function.
3414         (layout::print_annotation_line): Call it.
3415         (layout::print_leading_fixits): Likewise.
3416         (layout::print_trailing_fixits): Likewise.  Update calls to
3417         move_to_column for new parameter.
3418         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
3419         it to potentially call start_annotation_line.
3420         (layout::show_ruler): Call start_annotation_line.
3421         (selftest::test_line_numbers_multiline_range): New selftest.
3422         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
3423         and selftest::test_line_numbers_multiline_range.
3424         * diagnostic.c (diagnostic_initialize): Initialize
3425         show_line_numbers_p.
3426         * diagnostic.h (struct diagnostic_context): Add field
3427         "show_line_numbers_p".
3428         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3429         -fno-diagnostics-show-line-numbers.
3430         * dwarf2out.c (gen_producer_string): Add
3431         OPT_fdiagnostics_show_line_numbers to the ignored options.
3432         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
3433         one setting" options.
3434         (append_compiler_options): Likewise to the dropped options.
3435         (append_diag_options): Likewise to the passed-on options.
3436         * opts.c (common_handle_option): Handle the new option.
3437         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
3439 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3441         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
3442         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
3443         third argument of type "const signed char" to descriptions of
3444         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
3445         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
3446         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
3447         __builtin_bcdsub_ov functions.
3449 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3451         PR tree-optimization/86858
3452         * tree-vect-loop.c (vect_is_simple_reduction): Restore
3453         flow_bb_inside_loop_p calls.
3455 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3457         PR tree-optimization/86871
3458         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3459         instead of gimple_assign_lhs.
3461 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
3463         PR target/86887
3464         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3465         register constraint to operand 0.
3466         (add<mode>3_carryinC): Likewise.
3467         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3469 2018-08-09  Martin Liska  <mliska@suse.cz>
3471         PR c/86895
3472         * common.opt: Remove extra line.
3474 2018-08-09  Martin Liska  <mliska@suse.cz>
3476         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3477         at the end of a line, make first letter capital and end up
3478         a sentence with a dot.
3479         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3480         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3481         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3482         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3483         (PARAM_MAX_ISL_OPERATIONS): Likewise.
3484         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3485         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3486         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3487         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3488         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3489         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3490         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3491         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3492         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3493         (PARAM_TREE_REASSOC_WIDTH): Likewise.
3494         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3495         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3496         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3498 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
3500         PR target/84332
3501         * config/s390/s390.c (s390_option_override_internal): Reduce the
3502         stack-clash-protection-probe-interval param if it would be too big
3503         for z900.
3505 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
3507         PR target/46179
3508         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3509         * config/m68k/m68k.c (handle_move_double): Don't call
3510         m68k_final_prescan_insn.
3511         (m68k_adjust_decorated_operand): Renamed from
3512         m68k_final_prescan_insn, remove first and third operand and
3513         simplify.
3514         (print_operand): Call it.
3515         (print_operand_address): Call it.
3517 2018-08-08  Nathan Sidwell  <nathan@acm.org>
3519         * diagnostic.c (diagnostic_report_current_module): Use
3520         linemap_included_from & linemap_included_from_linemap.
3522 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
3524         * config/aarch64/aarch64-cores.def: Add phecda core.
3525         * config/aarch64/aarch64-tune.md: Regenerate.
3526         * doc/invoke.texi: Add phecda core.
3528 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
3530         PR target/85295
3531         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3532         definitions.
3533         * config/s390/s390.md ("movti"): Add more alternatives for
3534         constant to GPR copies.
3536 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
3538         * config/s390/s390.c: Fix whitespace damage throughout the file.
3539         * config/s390/s390.h: Likewise.
3540         * config/s390/tpf.h: Likewise.
3542 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
3544         * config/s390/s390.c (s390_loadrelative_operand_p):
3545         Remove TARGET_CPU_ZARCH usages.
3546         (s390_rtx_costs): Likewise.
3547         (s390_legitimate_constant_p): Likewise.
3548         (s390_cannot_force_const_mem): Likewise.
3549         (legitimate_reload_constant_p): Likewise.
3550         (s390_preferred_reload_class): Likewise.
3551         (legitimize_pic_address): Likewise.
3552         (legitimize_tls_address): Likewise.
3553         (s390_split_branches): Removed.
3554         (s390_add_execute): Removed.
3555         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3556         (s390_mainpool_start): Likewise.
3557         (s390_mainpool_finish): Likewise.
3558         (s390_mainpool_cancel): Removed.
3559         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3560         (s390_chunkify_cancel): Likewise.
3561         (s390_return_addr_rtx): Likewise.
3562         (s390_register_info): Remove split_branches_pending_p uages.
3563         (s390_optimize_register_info): Likewise.
3564         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3565         split_branches_pending_p usages.
3566         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3567         (s390_load_got): Likewise.
3568         (s390_expand_split_stack_prologue): Likewise.
3569         (output_asm_nops): Likewise.
3570         (s390_function_profiler): Likewise.
3571         (s390_emit_call): Likewise.
3572         (s390_conditional_register_usage): Likewise.
3573         (s390_optimize_prologue): Likewise.
3574         (s390_reorg): Remove TARGET_CPU_ZARCH and
3575         split_branches_pending_p usages.
3576         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3577         usages.
3578         (s390_output_indirect_thunk_function): Likewise.
3579         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3580         (TARGET_CPU_ZARCH_P): Removed.
3581         (struct machine_function): Remove split_branches_pending_p.
3582         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3584 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
3586         * common/config/s390/s390-common.c (processor_flags_table):
3587         Remove flags.
3588         * config.gcc: Remove with_arch/with_tune support.
3589         * config/s390/2064.md: Remove cpu attribute comparisons.
3590         * config/s390/driver-native.c (s390_host_detect_local_cpu):
3591         Remove MTN.
3592         * config/s390/linux.h (ASM_SPEC):
3593         Remove -march support.
3594         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3595         Use a table to get an arch level.
3596         * config/s390/s390-opts.h (enum processor_type):
3597         Remove enum values.
3598         * config/s390/s390.c
3599         (processor_table): Remove entries, add arch_level values.
3600         (s390_issue_rate): Remove cases.
3601         (s390_option_override): Adjust
3602         s390_option_override_internal() call.
3603         (s390_option_override_internal): Remove deprecation warning.
3604         (s390_valid_target_attribute_tree): Adjust
3605         s390_option_override_internal() call.
3606         * config/s390/s390.h (struct s390_processor):
3607         Share with s390-c.c, add arch_level field.
3608         * config/s390/s390.md:
3609         Remove occurrences in cpu attribute.
3610         * config/s390/s390.opt: Remove -march/-mtune support.
3611         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3612         * doc/invoke.texi: Remove deprecation warning.
3614 2018-08-08  Luis Machado  <luis.machado@linaro.org>
3616         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3617         global.
3618         (qdf24xx_tunings): Set vector cost structure to
3619         qdf24xx_vector_cost.
3621         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3622         <register_sextend>: Set to 3.
3624 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
3626         PR target/86838
3627         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3628         * config/aarch64/aarch64-simd.md
3629         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3630         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3631         (aarch64_frecpx<mode>): ...this new pattern.
3632         * config/aarch64/aarch64-simd-builtins.def: Remove comment
3633         about aarch64_frecp<FRECP:frecp_suffix><mode>.
3635 2018-08-07  Martin Liska  <mliska@suse.cz>
3637         PR middle-end/83023
3638         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3639         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3640         * predict.def (PRED_MALLOC_NONNULL): New predictor.
3641         * doc/extend.texi: Document that malloc attribute adds
3642         hit to compiler.
3644 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
3646         PR target/86785
3647         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3648         Define to speculation_safe_value_not_needed.
3650 2018-08-06  Jeff Law  <law@redhat.com>
3652         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3653         the vr_values instance to cprop_into_stmt.
3654         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3655         (cprop_operand): Also query EVRP to determine if OP is a constant.
3657 2018-08-06  Nathan Sidwell  <nathan@acm.org>
3659         * diagnostic.c (diagnostic_report_current_module): Reroll
3660         included-at loop.  Translate text.
3662 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
3664         * function-tests.c (selftest::test_expansion_to_rtl): Call
3665         free_after_compilation.
3667 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3669         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3671 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
3673         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3674         loops with memory block operations from getting unrolled.
3676 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
3678         PR target/86807
3679         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3680         Define to speculation_safe_value_not_needed.
3682 2018-08-06  Jeff Law  <law@redhat.com>
3684         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3685         assert.
3687 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3689         PR target/86662
3690         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3691         with all enabled __intN types.
3693         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3695 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3697         * alias.c (record_set): Check for clobber high.
3698         * cfgexpand.c (expand_gimple_stmt): Likewise.
3699         * combine-stack-adj.c (single_set_for_csa): Likewise.
3700         * combine.c (find_single_use_1): Likewise.
3701         (set_nonzero_bits_and_sign_copies): Likewise.
3702         (get_combine_src_dest): Likewise.
3703         (is_parallel_of_n_reg_sets): Likewise.
3704         (try_combine): Likewise.
3705         (record_dead_and_set_regs_1): Likewise.
3706         (reg_dead_at_p_1): Likewise.
3707         (reg_dead_at_p): Likewise.
3708         * dce.c (deletable_insn_p): Likewise.
3709         (mark_nonreg_stores_1): Likewise.
3710         (mark_nonreg_stores_2): Likewise.
3711         * df-scan.c (df_find_hard_reg_defs): Likewise.
3712         (df_uses_record): Likewise.
3713         (df_get_call_refs): Likewise.
3714         * dwarf2out.c (mem_loc_descriptor): Likewise.
3715         * haifa-sched.c (haifa_classify_rtx): Likewise.
3716         * ira-build.c (create_insn_allocnos): Likewise.
3717         * ira-costs.c (scan_one_insn): Likewise.
3718         * ira.c (equiv_init_movable_p): Likewise.
3719         (rtx_moveable_p): Likewise.
3720         (interesting_dest_for_shprep): Likewise.
3721         * jump.c (mark_jump_label_1): Likewise.
3722         * postreload-gcse.c (record_opr_changes): Likewise.
3723         * postreload.c (reload_cse_simplify): Likewise.
3724         (struct reg_use): Add source expr.
3725         (reload_combine): Check for clobber high.
3726         (reload_combine_note_use): Likewise.
3727         (reload_cse_move2add): Likewise.
3728         (move2add_note_store): Likewise.
3729         * print-rtl.c (print_pattern): Likewise.
3730         * recog.c (decode_asm_operands): Likewise.
3731         (store_data_bypass_p): Likewise.
3732         (if_test_bypass_p): Likewise.
3733         * regcprop.c (kill_clobbered_value): Likewise.
3734         (kill_set_value): Likewise.
3735         * reginfo.c (reg_scan_mark_refs): Likewise.
3736         * reload1.c (maybe_fix_stack_asms): Likewise.
3737         (eliminate_regs_1): Likewise.
3738         (elimination_effects): Likewise.
3739         (mark_not_eliminable): Likewise.
3740         (scan_paradoxical_subregs): Likewise.
3741         (forget_old_reloads_1): Likewise.
3742         * reorg.c (find_end_label): Likewise.
3743         (try_merge_delay_insns): Likewise.
3744         (redundant_insn): Likewise.
3745         (own_thread_p): Likewise.
3746         (fill_simple_delay_slots): Likewise.
3747         (fill_slots_from_thread): Likewise.
3748         (dbr_schedule): Likewise.
3749         * resource.c (update_live_status): Likewise.
3750         (mark_referenced_resources): Likewise.
3751         (mark_set_resources): Likewise.
3752         * rtl.c (copy_rtx): Likewise.
3753         * rtlanal.c (reg_referenced_p): Likewise.
3754         (single_set_2): Likewise.
3755         (noop_move_p): Likewise.
3756         (note_stores): Likewise.
3757         * sched-deps.c (sched_analyze_reg): Likewise.
3758         (sched_analyze_insn): Likewise.
3760 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3762         * cse.c (invalidate_reg): New function extracted from...
3763         (invalidate): ...here.
3764         (canonicalize_insn): Check for clobber high.
3765         (invalidate_from_clobbers): invalidate clobber highs.
3766         (invalidate_from_sets_and_clobbers): Likewise.
3767         (count_reg_usage): Check for clobber high.
3768         (insn_live_p): Likewise.
3769         * cselib.c (cselib_expand_value_rtx_1):Likewise.
3770         (cselib_invalidate_regno): Check for clobber in setter.
3771         (cselib_invalidate_rtx): Pass through setter.
3772         (cselib_invalidate_rtx_note_stores):
3773         (cselib_process_insn): Check for clobber high.
3774         * cselib.h (cselib_invalidate_rtx): Add operand.
3776 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3778         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3779         (mark_not_eliminable): Likewise.
3780         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3781         * lra-lives.c (process_bb_lives): Check for clobber high.
3782         * lra.c (new_insn_reg): Remember clobber highs.
3783         (collect_non_operand_hard_regs): Check for clobber high.
3784         (lra_set_insn_recog_data): Likewise.
3785         (add_regs_to_insn_regno_info): Likewise.
3786         (lra_update_insn_regno_info): Likewise.
3788 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3790         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3791         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3793 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3795         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3796         (copy_insn_1): Likewise.
3797         (gen_hard_reg_clobber_high): New gen function.
3798         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3799         * genemit.c (gen_exp): Likewise.
3800         (gen_emit_seq): Pass through info.
3801         (gen_insn): Check for CLOBBER_HIGH.
3802         (gen_expand): Pass through info.
3803         (gen_split): Likewise.
3804         (output_add_clobbers): Likewise.
3805         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3806         (remove_clobbers): Likewise.
3807         * rtl.h (gen_hard_reg_clobber_high): New declaration.
3809 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3811         * doc/rtl.texi (clobber_high): Add.
3812         (parallel): Add in clobber high
3813         * rtl.c (rtl_check_failed_code3): Add function.
3814         * rtl.def (CLOBBER_HIGH): Add expression.
3815         * rtl.h (RTL_CHECKC3): Add macro.
3816         (rtl_check_failed_code3): Add declaration.
3817         (XC3EXP): Add macro.
3819 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
3821         PR target/86386
3822         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3823         cfun->machine->max_used_stack_alignment if needed.
3825 2018-08-04  Martin Sebor  <msebor@redhat.com>
3827         PR tree-optimization/86571
3828         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3829         NaN output to 4.
3831 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
3833         PR target/86799
3834         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3835         Define.
3837 2018-08-03  Jeff Law  <law@redhat.com>
3839         PR target/86795
3840         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3841         Define to speculation_safe_value_not_needed.
3843 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
3845         * doc/gcov.texi (-x): Remove duplicate "to".
3846         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3847         (-Wif-not-aligned): Remove duplicate "is".
3848         (-flto): Remove duplicate "the".
3849         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3850         duplicate "v5.00.b".
3851         (MSP430 Options): Remove duplicate "and" from the description
3852         of "-mgprel-sec=regexp".
3853         (x86 Options): Remove duplicate copies of "vmldLog102" and
3854         vmlsLog104 from description of "-mveclibabi=type".
3856 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
3858         * internal-fn.h (first_commutative_argument): Declare.
3859         * internal-fn.c (first_commutative_argument): New function.
3860         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3861         restrictions for pattern statements.  Use first_commutative_argument
3862         to look for commutative operands in calls to internal functions.
3864 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
3866         * Makefile.in (wide-int-range.o): New.
3867         * tree-vrp.c: Move all the wide_int_* functions to...
3868         * wide-int-range.cc: ...here.
3869         * tree-vrp.h: Move all the wide_int_* prototypes to...
3870         * wide-int-range.h: ...here.
3872 2018-08-03  Tom de Vries  <tdevries@suse.de>
3874         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3875         UI_NONE.
3876         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3877         * except.c (output_function_exception_table): Do early exit if
3878         targetm_common.except_unwind_info (&global_options) == UI_NONE.
3880 2018-08-03  Martin Liska  <mliska@suse.cz>
3882         * predict.c (dump_prediction): Change to 2 digits
3883         in fraction part.
3885 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
3887         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3888         neon_dup_q to...
3889         (falkor_am_1_gtov_gtov): ... a new insn reservation.
3891 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3893         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3894         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3895         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3896         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3897         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3899 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
3901         * diagnostic-show-locus.c (diagnostic_show_locus): Use
3902         pp_take_prefix when saving the existing prefix.
3903         * diagnostic.c (diagnostic_append_note): Likewise.
3904         * langhooks.c (lhd_print_error_function): Likewise.
3905         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3906         param's type.  Free the existing prefix.
3907         (pp_take_prefix): New function.
3908         (pretty_printer::pretty_printer): Drop the prefix parameter.
3909         Rename the length parameter to match the comment.
3910         (pretty_printer::~pretty_printer): Free the prefix.
3911         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3912         parameter.
3913         (struct pretty_printer): Drop the "const" from "prefix" field's
3914         type and clarify memory management.
3915         (pp_set_prefix): Drop the "const" from the 2nd param.
3916         (pp_take_prefix): New decl.
3918 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3920         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3921         for word_mode_ok here instead of passing as argument.
3922         (expand_block_compare): Change select_block_compare_mode() call.
3923         (expand_strncmp_gpr_sequence): New function.
3924         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3926 2018-08-02  Jeff Law  <law@redhat.com>
3928         PR target/86790
3929         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3930         Define to speculation_safe_value_not_needed.
3932         PR target/86784
3933         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3934         Define to speculation_safe_value_not_needed.
3936 2018-08-02  Tom de Vries  <tdevries@suse.de>
3938         PR target/86660
3939         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3940         function.  Return UI_TARGET unconditionally.
3941         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3942         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3944 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
3946         * genemit.c (print_overload_test): Fix typo.
3948 2018-08-02  Richard Biener  <rguenther@suse.de>
3950         PR tree-optimization/86816
3951         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3952         which checks for value availability before querying it.
3953         (gvn_uses_equal): Use it.
3954         (same_succ_hash): Likewise.
3955         (gimple_equal_p): Likewise.
3957 2018-08-02  Nick Clifton  <nickc@redhat.com>
3959         PR target/86813
3960         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3961         Define to speculation_safe_value_not_needed.
3963         PR target/86810
3964         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3965         Define to speculation_safe_value_not_needed.
3967         PR target/86810
3968         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3969         Define to speculation_safe_value_not_needed.
3971         PR target/86803
3972         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3973         Define to speculation_safe_value_not_needed.
3975         PR target/86797
3976         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3977         Define to speculation_safe_value_not_needed.
3979         PR target/86791
3980         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3981         Define to speculation_safe_value_not_needed.
3983         PR target/86789
3984         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3985         Define to speculation_safe_value_not_needed.
3987         PR target/86787
3988         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3989         Define to speculation_safe_value_not_needed.
3991         PR target/86782
3992         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3993         speculation_safe_value_not_needed.
3995         PR target/86781
3996         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3997         to speculation_safe_value_not_needed.
3999 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
4001         * doc/md.texi: Expand the documentation of instruction names
4002         to mention port-local uses.  Document '@' in pattern names.
4003         * read-md.h (overloaded_instance, overloaded_name): New structs.
4004         (mapping): Declare.
4005         (md_reader::handle_overloaded_name): New member function.
4006         (md_reader::get_overloads): Likewise.
4007         (md_reader::m_first_overload): New member variable.
4008         (md_reader::m_next_overload_ptr): Likewise.
4009         (md_reader::m_overloads_htab): Likewise.
4010         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
4011         m_next_overload_ptr and m_overloads_htab.
4012         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
4013         (get_mode_token, get_code_token, get_int_token): New functions.
4014         (map_attr_string): Add an optional argument that passes back
4015         the associated iterator.
4016         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
4017         (md_reader::handle_overloaded_name, add_overload_instance): New
4018         functions.
4019         (apply_iterators): Handle '@' names.  Report an error if '@'
4020         is used without iterators.
4021         (initialize_iterators): Initialize the new iterator_group fields.
4022         * genopinit.c (handle_overloaded_code_for)
4023         (handle_overloaded_gen): New functions.
4024         (main): Use them to print declarations of maybe_code_for_* and
4025         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
4026         * genemit.c (print_overload_arguments, print_overload_test)
4027         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
4028         (main): Use it to print definitions of maybe_code_for_* and
4029         maybe_gen_* functions.
4030         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
4031         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
4032         instead of explicit mode checks.
4033         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
4034         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
4035         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
4036         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
4037         (aarch64_expand_compare_and_swap): Likewise
4038         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
4039         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
4040         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
4041         (aarch64_constant_pool_reload_icode): Delete.
4042         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
4043         instead of aarch64_constant_pool_reload_icode.  Use
4044         code_for_aarch64_reload_mov instead of explicit mode checks.
4045         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
4046         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
4047         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
4048         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
4049         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
4050         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
4051         (aarch64_atomic_load_op_code): Delete.
4052         (aarch64_emit_atomic_load_op): Likewise.
4053         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
4054         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
4055         instead of aarch64_emit_atomic_load_op.
4056         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
4057         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
4058         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
4059         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
4060         character before the pattern name.
4061         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
4062         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
4063         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
4064         (aarch64_frecps<mode>): Likewise.
4065         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
4066         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
4067         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
4068         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
4069         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
4071 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
4073         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
4074         Allow HFmode constants if TARGET_FP_F16INST.
4076 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
4078         PR target/86014
4079         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
4080         No longer check last store for clobber of address register.
4082 2018-08-02  Martin Liska  <mliska@suse.cz>
4084         PR gcov-profile/86817
4085         * gcov.c (process_all_functions): New function.
4086         (main): Call it.
4087         (process_file): Move functions processing to
4088         process_all_functions.
4090 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
4092         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
4093         "const" to the "gimple *" and "rtx_insn *" parameters.
4094         * dumpfile.h (dump_user_location_t::dump_user_location_t):
4095         Likewise.
4096         (dump_location_t::dump_location_t): Likewise.
4098 2018-08-01  Martin Sebor  <msebor@redhat.com>
4100         PR tree-optimization/86650
4101         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
4102         rather than a "gcall *".  Directly pass the data of interest
4103         to percent_K_format, rather than building a temporary CALL_EXPR
4104         to hold it.
4105         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
4106         (gimple_fold_builtin_strncat): Adjust.
4107         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
4108         gcall* argument with gimple*.
4109         * gimple-ssa-warn-restrict.c (check_call): Same.
4110         (wrestrict_dom_walker::before_dom_children): Same.
4111         (builtin_access::builtin_access): Same.
4112         (check_bounds_or_overlap): Same
4113         (maybe_diag_overlap): Same.
4114         (maybe_diag_offset_bounds): Same.
4115         * tree-diagnostic.c (default_tree_printer): Move usage of
4116         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
4117         to this callsite.
4118         * tree-pretty-print.c (percent_K_format): Add argument.
4119         * tree-pretty-print.h: Add argument.
4120         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
4121         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
4122         (maybe_diag_stxncpy_trunc): Same.
4123         (handle_builtin_stxncpy): Same.
4124         (handle_builtin_strcat): Same.
4126 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4128         * match.pd: Optimise pointer range checks.
4130 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4132         PR tree-optimization/86758
4133         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
4134         to remove pattern statements.
4136 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4138         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
4139         result of dfs_enumerate_from when constructing stmt_vec_infos,
4140         instead of additionally calling get_loop_body.
4142 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4144         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
4145         parameter.
4146         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
4147         When creating an iv, assert that the step is not known to be zero.
4148         (vect_setup_realignment): Update call accordingly.
4149         * tree-vect-stmts.c (vectorizable_store): Likewise.
4150         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
4152 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4154         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
4155         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
4156         (vectorizable_reduction): Likewise.
4157         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4158         (vect_detect_hybrid_slp_stmts): Likewise.
4159         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4161 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
4163         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
4164         (wide_int_set_zero_nonzero_bits): ...this.
4165         (zero_nonzero_bits_from_vr): Rename to...
4166         (vrp_set_zero_nonzero_bits): ...this.
4167         (extract_range_from_multiplicative_op_1): Abstract wide int
4168         code...
4169         (wide_int_range_multiplicative_op): ...here.
4170         (extract_range_from_binary_expr_1): Extract wide int binary
4171         operations into their own functions.
4172         (wide_int_range_lshift): New.
4173         (wide_int_range_can_optimize_bit_op): New.
4174         (wide_int_range_shift_undefined_p): New.
4175         (wide_int_range_bit_xor): New.
4176         (wide_int_range_bit_ior): New.
4177         (wide_int_range_bit_and): New.
4178         (wide_int_range_trunc_mod): New.
4179         (extract_range_into_wide_ints): New.
4180         (vrp_shift_undefined_p): New.
4181         (extract_range_from_multiplicative_op): New.
4182         (vrp_can_optimize_bit_op): New.
4183         * tree-vrp.h (value_range::dump): New.
4184         (wide_int_range_multiplicative_op): New.
4185         (wide_int_range_lshift):New.
4186         (wide_int_range_shift_undefined_p): New.
4187         (wide_int_range_bit_xor): New.
4188         (wide_int_range_bit_ior): New.
4189         (wide_int_range_bit_and): New.
4190         (wide_int_range_trunc_mod): New.
4191         (zero_nonzero_bits_from_bounds): Rename to...
4192         (wide_int_set_zero_nonzero_bits): ...this.
4193         (zero_nonzero_bits_from_vr): Rename to...
4194         (vrp_set_zero_nonzero_bits): ...this.
4195         (range_easy_mask_min_max): Rename to...
4196         (wide_int_range_can_optimize_bit_op): this.
4197         * vr-values.c (simplify_bit_ops_using_ranges): Rename
4198         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4200 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4202         * tree-vectorizer.h (vect_orig_stmt): New function.
4203         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4204         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4205         (vect_create_epilog_for_reduction): Likewise.
4206         (vectorizable_live_operation): Likewise.
4207         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4208         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4209         * tree-vect-stmts.c (vectorizable_call): Likewise.
4210         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4212 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4214         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4215         argument.
4216         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4217         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4218         (vect_transform_loop): Likewise.
4219         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4221 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4223         * tree-vectorizer.h (vect_schedule_slp): Return void.
4224         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4225         (vect_schedule_slp): Likewise.
4227 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4229         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4230         argument.
4231         (vect_transform_loop): Update calls accordingly.  Schedule SLP
4232         instances before the main loop, if any exist.
4234 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4236         PR tree-optimization/86749
4237         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4238         If the lhs is used in a COND_EXPR, check that it is being used
4239         as the "then" or "else" value.
4241 2018-08-01  Tom de Vries  <tdevries@suse.de>
4243         PR target/86800
4244         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4245         speculation_safe_value_not_needed.
4247 2018-08-01  Richard Biener  <rguenther@suse.de>
4249         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4250         as base and offset.
4252 2018-08-01  Martin Liska  <mliska@suse.cz>
4254         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4255         format how successful transformation is dumped.
4256         (gimple_mod_pow2_value_transform): Likewise.
4257         (gimple_mod_subtract_transform): Likewise.
4258         (gimple_stringops_transform): Likewise.
4260 2018-08-01  Martin Liska  <mliska@suse.cz>
4262         PR value-prof/35543
4263         * value-prof.c (interesting_stringop_to_profile_p):
4264         Simplify the code and add BUILT_IN_MEMMOVE.
4265         (gimple_stringops_transform): Likewise.
4267 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
4269         * config/aarch64/aarch64-simd.md
4270         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4271         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4272         use GPI iterator instead of SI mode.
4274 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
4276         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4277         rs6000_speculation_barrier.
4278         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4279         new barrier pattern name.
4281 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
4283         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4284         (speculation_barrier): New insn.
4286 2018-08-01  Richard Biener  <rguenther@suse.de>
4288         PR bootstrap/86724
4289         * graphite.h: Include isl/id.h and isl/space.h to allow build
4290         with ISL 0.20.
4292 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
4294         PR target/86651
4295         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4296         mode for COFF targets.
4297         * defaults.h (TARGET_COFF): Define.
4298         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4299         TARGET_COFF): Define.
4300         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4301         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4302         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4304 2018-07-31  Alexandre Oliva <oliva@adacore.com>
4306         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4307         * gimple-streamer-out.c (output_bb): Save it.
4308         * lto-streamer-in.c (input_struct_function_base): Restore
4309         instance discriminator if available.  Create map on demand.
4310         * lto-streamer-out.c (output_struct_function_base): Save it if
4311         available.
4312         * final.c (decl_to_instance_map): Document LTO strategy.
4314 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
4315             Olivier Hainque  <hainque@adacore.com>
4317         * debug.h (decl_to_instance_map_t): New type.
4318         (decl_to_instance_map): Declare.
4319         (maybe_create_decl_to_instance_map): New inline function.
4320         * final.c (bb_discriminator, last_bb_discriminator): New statics,
4321         to track basic block discriminators.
4322         (final_start_function_1): Initialize them.
4323         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4324         bb_discriminator.
4325         (decl_to_instance_map): New variable.
4326         (map_decl_to_instance, maybe_set_discriminator): New functions.
4327         (notice_source_line): Set discriminator.
4329 2018-07-31  Ian Lance Taylor  <iant@golang.org>
4331         * targhooks.c (default_have_speculation_safe_value): Add
4332         ATTRIBUTE_UNUSED.
4334 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
4336         * dump-context.h: Include "pretty-print.h".
4337         (dump_context::refresh_dumps_are_enabled): New decl.
4338         (dump_context::emit_item): New decl.
4339         (class dump_context): Add fields "m_test_pp" and
4340         "m_test_pp_flags".
4341         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4342         (temp_dump_context::get_dumped_text): New decl.
4343         (class temp_dump_context): Add field "m_pp".
4344         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4345         (dump_context::refresh_dumps_are_enabled): ...and add a test for
4346         m_test_pp.
4347         (set_dump_file): Update for above change.
4348         (set_alt_dump_file): Likewise.
4349         (dump_loc): New overload, taking a pretty_printer *.
4350         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
4351         to any test pretty-printer.
4352         (make_item_for_dump_gimple_stmt): New function, adapted from
4353         optinfo::add_gimple_stmt.
4354         (dump_context::dump_gimple_stmt): Call it, and use the result,
4355         eliminating the direct usage of dump_file and alt_dump_file in
4356         favor of indirectly using them via emit_item.
4357         (make_item_for_dump_gimple_expr): New function, adapted from
4358         optinfo::add_gimple_expr.
4359         (dump_context::dump_gimple_expr): Call it, and use the result,
4360         eliminating the direct usage of dump_file and alt_dump_file in
4361         favor of indirectly using them via emit_item.
4362         (make_item_for_dump_generic_expr): New function, adapted from
4363         optinfo::add_tree.
4364         (dump_context::dump_generic_expr): Call it, and use the result,
4365         eliminating the direct usage of dump_file and alt_dump_file in
4366         favor of indirectly using them via emit_item.
4367         (make_item_for_dump_printf_va): New function, adapted from
4368         optinfo::add_printf_va.
4369         (make_item_for_dump_printf): New function.
4370         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4371         and use the result, eliminating the direct usage of dump_file and
4372         alt_dump_file in favor of indirectly using them via emit_item.
4373         (make_item_for_dump_dec): New function.
4374         (dump_context::dump_dec): Call it, and use the result,
4375         eliminating the direct usage of dump_file and alt_dump_file in
4376         favor of indirectly using them via emit_item.
4377         (make_item_for_dump_symtab_node): New function, adapted from
4378         optinfo::add_symtab_node.
4379         (dump_context::dump_symtab_node): Call it, and use the result,
4380         eliminating the direct usage of dump_file and alt_dump_file in
4381         favor of indirectly using them via emit_item.
4382         (dump_context::begin_scope): Reimplement, avoiding direct usage
4383         of dump_file and alt_dump_file in favor of indirectly using them
4384         via emit_item.
4385         (dump_context::emit_item): New member function.
4386         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4387         Set up test pretty-printer on the underlying context.  Call
4388         refresh_dumps_are_enabled.
4389         (temp_dump_context::~temp_dump_context): Call
4390         refresh_dumps_are_enabled.
4391         (temp_dump_context::get_dumped_text): New member function.
4392         (selftest::verify_dumped_text): New function.
4393         (ASSERT_DUMPED_TEXT_EQ): New macro.
4394         (selftest::test_capture_of_dump_calls): Run all tests twice, with
4395         and then without optinfo enabled.  Add uses of
4396         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
4397         * dumpfile.h: Update comment for the dump_* API.
4398         * optinfo-emit-json.cc
4399         (selftest::test_building_json_from_dump_calls): Update for new
4400         param for temp_dump_context ctor.
4401         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
4402         and "m_owned" field.
4403         (optinfo_item::~optinfo_item): Likewise.
4404         (optinfo::add_item): New member function.
4405         (optinfo::emit): Update comment.
4406         (optinfo::add_string): Delete.
4407         (optinfo::add_printf): Delete.
4408         (optinfo::add_printf_va): Delete.
4409         (optinfo::add_gimple_stmt): Delete.
4410         (optinfo::add_gimple_expr): Delete.
4411         (optinfo::add_tree): Delete.
4412         (optinfo::add_symtab_node): Delete.
4413         (optinfo::add_dec): Delete.
4414         * optinfo.h (class dump_context): New forward decl.
4415         (optinfo::add_item): New decl.
4416         (optinfo::add_string): Delete.
4417         (optinfo::add_printf): Delete.
4418         (optinfo::add_printf_va): Delete.
4419         (optinfo::add_gimple_stmt): Delete.
4420         (optinfo::add_gimple_expr): Delete.
4421         (optinfo::add_tree): Delete.
4422         (optinfo::add_symtab_node): Delete.
4423         (optinfo::add_dec): Delete.
4424         (optinfo::add_poly_int): Delete.
4425         (optinfo_item::optinfo_item): Remove "owned" param.
4426         (class optinfo_item): Remove field "m_owned".
4428 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4430         PR middle-end/86705
4431         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
4432         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
4433         requested variable alignment.
4434         (expand_one_ssa_partition): Likewise.
4435         (expand_one_var): Likewise.
4437 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4439         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4440         to speculation_safe_value_not_needed.
4442 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4444         * targhooks.h (speculation_safe_value_not_needed): New prototype.
4445         * targhooks.c (speculation_safe_value_not_needed): New function.
4446         * target.def (have_speculation_safe_value): Update documentation.
4447         * doc/tm.texi: Regenerated.
4449 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4451         * config/aarch64/iterators.md (ALLI_TI): New iterator.
4452         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4453         expand.
4454         (despeculate_copy<ALLI:mode>_insn): New insn.
4455         (despeculate_copyti_insn): New insn.
4456         (despeculate_simple<ALLI:mode>): New insn
4457         (despeculate_simpleti): New insn.
4458         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4459         function.
4460         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4461         aarch64_speculation_safe_value.
4462         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4464 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4466         * config/aarch64/aarch64-speculation.cc: New file.
4467         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4468         before pass_reorder_blocks.
4469         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4470         prototype.
4471         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4472         X14 and X15 when tracking speculation.
4473         * config/aarch64/aarch64.md (register name constants): Add
4474         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4475         (unspec): Add UNSPEC_SPECULATION_TRACKER.
4476         (speculation_barrier): New insn attribute.
4477         (cmp<mode>): Allow SP in comparisons.
4478         (speculation_tracker): New insn.
4479         (speculation_barrier): Add speculation_barrier attribute.
4480         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4481         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4482         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4484 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4486         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4487         aarch64_track_speculation is true.
4488         (tb<optab><mode>1): Likewise.
4489         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4490         generate CB[N]Z when tracking speculation.
4491         (aarch64_split_compare_and_swap): Likewise.
4492         (aarch64_split_atomic_op): Likewise.
4494 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4496         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4498 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4500         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4501         (speculation_barrier): New insn.
4503 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4505         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4506         * config/arm/arm.md (speculation_barrier): New expand.
4507         (speculation_barrier_insn): New pattern.
4509 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4511         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4512         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4513         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4514         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4515         list.
4516         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4517         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4518         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4519         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4520         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4521         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4522         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4523         * builtins.c (expand_speculation_safe_value): New function.
4524         (expand_builtin): Call it.
4525         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4526         * doc/extend.texi: Document __builtin_speculation_safe_value.
4527         * doc/md.texi: Document "speculation_barrier" pattern.
4528         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4529         TARGET_HAVE_SPECULATION_SAFE_VALUE.
4530         * doc/tm.texi: Regenerated.
4531         * target.def (have_speculation_safe_value, speculation_safe_value): New
4532         hooks.
4533         * targhooks.c (default_have_speculation_safe_value): New function.
4534         (default_speculation_safe_value): New function.
4535         * targhooks.h (default_have_speculation_safe_value): Add prototype.
4536         (default_speculation_safe_value): Add prototype.
4538 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
4540         * dump-context.h (dump_context::dump_loc): New decl.
4541         * dumpfile.c (dump_context::dump_loc): New member function.
4542         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4543         and dump_gimple_stmt.
4544         (dump_context::dump_gimple_expr_loc): Likewise, using
4545         dump_gimple_expr.
4546         (dump_context::dump_generic_expr_loc): Likewise, using
4547         dump_generic_expr.
4548         (dump_context::dump_printf_loc_va): Likewise, using
4549         dump_printf_va.
4550         (dump_context::begin_scope): Explicitly using the global function
4551         "dump_loc", rather than the member function.
4553 2018-07-31  Martin Sebor  <msebor@redhat.com>
4555         PR tree-optimization/86741
4556         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4558 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
4560         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4561         depend on whether prefetch instructions will be emitted or not.
4562         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4563         will be emitted or not.
4564         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4565         (TARGET_SETMEM_PFD): New macros.
4567 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4569         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4570         (NULL_STMT_VEC_INFO): Delete.
4571         (stmt_vec_info::operator*): Likewise.
4572         (stmt_vec_info::operator gimple *): Likewise.
4573         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4574         of NULL_STMT_VEC_INFO.
4575         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4576         (vect_reassociating_reduction_p): Likewise.
4577         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4578         (vectorizable_store): Likewise.
4579         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4580         (vec_info::free_stmt_vec_infos): Likewise.
4582 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4584         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4585         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4587 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4589         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4590         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4591         (vec_info::free_stmt_vec_info): New private member functions.
4592         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4593         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4594         * tree-parloops.c (gather_scalar_reductions): Remove calls to
4595         set_stmt_vec_info_vec and free_stmt_vec_infos.
4596         * tree-vect-loop.c (_loop_vec_info): Remove call to
4597         set_stmt_vec_info_vec.
4598         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4599         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4600         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4601         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4602         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
4603         assignments in {vec_info::,}new_stmt_vec_info that are redundant
4604         with the clearing in the xcalloc.
4605         (stmt_vec_info_vec): Delete.
4606         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4607         (vectorize_loops): Likewise.
4608         (vec_info::~vec_info): Remove argument from call to
4609         free_stmt_vec_infos.
4610         (vec_info::add_stmt): Remove vinfo argument from call to
4611         new_stmt_vec_info.
4613 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4615         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4616         rather than a gimple stmt.
4617         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
4618         information for pattern statements when passed the original
4619         statement; instead wait to be passed the pattern statement itself.
4620         Don't call set_vinfo_for_stmt here.
4621         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4622         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4623         stmt_vec_infos here.
4624         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4625         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4626         stmt_vec_infos entry.
4628 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4630         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4631         * tree-vectorizer.c (vec_info::replace_stmt): New function.
4632         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4633         * tree-vect-stmts.c (vectorizable_call): Likewise.
4634         (vectorizable_simd_clone_call): Likewise.
4636 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4638         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4639         * tree-vectorizer.c (vec_info::remove_stmt): New function.
4640         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4641         * tree-vect-loop.c (vect_transform_loop): Likewise.
4642         * tree-vect-slp.c (vect_schedule_slp): Likewise.
4643         * tree-vect-stmts.c (vect_remove_stores): Likewise.
4645 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4647         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4648         (vect_dr_stmt): Delete.
4649         * tree-vectorizer.c (vec_info::lookup_dr): New function.
4650         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4651         of DR_VECT_AUX.
4652         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4653         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4654         (vect_verify_datarefs_alignment, vect_peeling_supportable)
4655         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4656         (vect_analyze_data_refs): Likewise.
4657         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
4658         argument.
4659         (vect_find_same_alignment_drs): Likewise.
4660         (vect_slp_analyze_node_dependences): Update calls accordingly.
4661         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
4662         instead of DR_VECT_AUX.
4663         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4664         of a vector data references.  Use vec_info::lookup_dr instead of
4665         DR_VECT_AUX.
4666         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4667         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
4668         instead of DR_VECT_AUX.
4670 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4672         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4673         dr_vec_info.
4674         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4675         accordingly.
4676         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4677         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4678         (vect_gen_prolog_loop_niters): Likewise.
4680 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4682         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4683         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4684         (vect_known_alignment_in_bytes, vect_dr_behavior)
4685         (vect_get_scalar_dr_size): Take references as dr_vec_infos
4686         instead of data_references.  Update calls to other routines for
4687         which the same change has been made.
4688         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4689         dr_vec_infos instead of stmt_vec_infos.
4690         (vect_analyze_data_ref_dependence): Update call accordingly.
4691         (vect_slp_analyze_data_ref_dependence)
4692         (vect_record_base_alignments): Use DR_VECT_AUX.
4693         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4694         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4695         (vector_alignment_reachable_p, vect_get_data_access_cost)
4696         (vect_peeling_supportable, vect_analyze_group_access_1)
4697         (vect_analyze_group_access, vect_analyze_data_ref_access)
4698         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4699         (vect_compile_time_alias, vect_small_gap_p)
4700         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4701         (vect_supportable_dr_alignment): Take references as dr_vec_infos
4702         instead of data_references.  Update calls to other routines for
4703         which the same change has been made.
4704         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4705         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4706         (vect_slp_analyze_and_verify_node_alignment)
4707         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4708         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4709         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
4710         above changes.
4711         (_vect_peel_info::dr): Replace with...
4712         (_vect_peel_info::dr_info): ...this new field.
4713         (vect_peeling_hash_get_most_frequent)
4714         (vect_peeling_hash_choose_best_peeling): Update accordingly.
4715         (vect_peeling_hash_get_lowest_cost):
4716         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
4717         routines for which the same change has been made.
4718         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
4719         data_reference.
4720         * tree-vect-loop-manip.c (get_misalign_in_elems)
4721         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
4722         above changes.
4723         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4724         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4725         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4726         (get_group_load_store_type, get_negative_load_store_type)
4727         (vect_get_data_ptr_increment, vectorizable_store)
4728         (vectorizable_load): Likewise.
4729         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4730         Update calls to other routines for which the same change has been made.
4732 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4734         * tree-vectorizer.h (vec_info::move_dr): New member function.
4735         (dataref_aux): Rename to...
4736         (dr_vec_info): ...this and add "dr" and "stmt" fields.
4737         (_stmt_vec_info::dr_aux): Update accordingly.
4738         (_stmt_vec_info::data_ref_info): Delete.
4739         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4740         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4741         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4742         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4743         of data_ref.
4744         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
4745         (STMT_VINFO_DR_INFO): New macro.
4746         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4747         (set_dr_misalignment): Update after rename of dataref_aux.
4748         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
4749         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4750         initialization of STMT_VINFO_DATA_REF.
4751         * tree-vectorizer.c (vec_info::move_dr): New function.
4752         * tree-vect-patterns.c (vect_recog_bool_pattern)
4753         (vect_recog_mask_conversion_pattern)
4754         (vect_recog_gather_scatter_pattern): Use it.
4755         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4756         the "dr" and "stmt" fields of dr_vec_info instead of
4757         STMT_VINFO_DATA_REF.
4759 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4761         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4762         (is_pattern_stmt_p): Use it.
4763         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4764         on pattern statements.
4766 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4768         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4769         original stmt as a stmt_vec_info rather than a gimple stmt.
4770         (vect_pattern_recog_1): Take the statement directly as a
4771         stmt_vec_info, rather than via a gimple_stmt_iterator.
4772         Update call to vect_mark_pattern_stmts.
4773         (vect_pattern_recog): Update calls accordingly.
4775 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4777         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4778         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4779         a vect_def_type for the first argument.
4780         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4781         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
4782         operand if it isn't defined by a vectorized statement.
4783         (vect_build_gather_load_calls): Remove the mask_dt argument and
4784         update calls to vect_get_vec_def_for_stmt_copy.
4785         (vectorizable_bswap): Likewise the dt argument.
4786         (vectorizable_call): Update calls to vectorizable_bswap and
4787         vect_get_vec_def_for_stmt_copy.
4788         (vectorizable_simd_clone_call, vectorizable_assignment)
4789         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4790         (vectorizable_comparison): Update calls to
4791         vect_get_vec_def_for_stmt_copy.
4792         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
4793         vect_is_simple_use.
4794         (vect_get_loop_based_defs): Remove dt argument and update call
4795         to vect_get_vec_def_for_stmt_copy.
4796         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4797         and vect_get_vec_def_for_stmt_copy.
4798         (vectorizable_load): Update calls to vect_build_gather_load_calls
4799         and vect_get_vec_def_for_stmt_copy.
4800         * tree-vect-loop.c (vect_create_epilog_for_reduction)
4801         (vectorizable_reduction, vectorizable_live_operation): Update calls
4802         to vect_get_vec_def_for_stmt_copy.
4804 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4806         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4807         and gimple stmt arguments with a stmt_vec_info.
4808         (vect_record_base_alignments): Update calls accordingly.
4809         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4810         and gimple stmt arguments with a stmt_vec_info.
4811         (vect_build_slp_tree_1): Remove vinfo argument and update call
4812         to vect_record_max_nunits.
4813         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4814         and vect_record_max_nunits.
4816 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4818         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4819         file and take a stmt_vec_info instead of a gimple stmt.
4820         (supportable_widening_operation, vect_finish_replace_stmt)
4821         (vect_finish_stmt_generation, vect_get_store_rhs)
4822         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4823         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4824         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4825         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4826         (vect_create_data_ref_ptr, bump_vector_ptr)
4827         (vect_permute_store_chain, vect_setup_realignment)
4828         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4829         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4830         (vectorizable_reduction, vectorizable_induction)
4831         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4832         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4833         than gimple stmts as arguments.
4834         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4835         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4836         (can_group_stmts_p, vect_check_gather_scatter)
4837         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4838         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4839         (vect_permute_load_chain, vect_shift_permute_load_chain)
4840         (vect_transform_grouped_load)
4841         (vect_record_grouped_load_vectors): Likewise.
4842         * tree-vect-loop.c (vect_fixup_reduc_chain)
4843         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4844         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4845         (vectorizable_reduction, vectorizable_induction)
4846         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4847         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4848         (vect_get_load_store_mask): Likewise.
4849         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4850         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4851         * tree-vect-stmts.c (vect_mark_relevant)
4852         (is_simple_and_all_uses_invariant)
4853         (exist_non_indexing_operands_for_use_p, process_use)
4854         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4855         (vect_get_vec_def_for_operand, vect_get_vec_defs)
4856         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4857         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4858         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4859         (get_negative_load_store_type, get_load_store_type)
4860         (vect_check_load_store_mask, vect_check_store_rhs)
4861         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4862         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4863         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4864         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4865         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4866         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4867         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4868         (supportable_widening_operation): Likewise.
4870 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4872         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4873         a stmt_vec_info instead of a gcall.
4874         (vect_check_gather_scatter): Update call accordingly.
4875         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4876         of a gphi.
4877         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4878         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4879         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4880         instead of a gimple stmt.
4881         (vect_transform_loop): Update calls accordingly.
4882         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4883         stmt_vec_infos instead of gimple stmts.
4884         (vect_analyze_slp_instance): Update use accordingly.
4885         * tree-vect-stmts.c (read_vector_array, write_vector_array)
4886         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4887         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4888         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4889         (vect_gen_widened_results_half, vect_get_loop_based_defs)
4890         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4891         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4892         down to subroutines.
4894 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4896         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4897         of the worklist from a vector of gimple stmts to a vector of
4898         stmt_vec_infos.
4899         * tree-vect-stmts.c (vect_mark_relevant, process_use)
4900         (vect_mark_stmts_to_be_vectorized): Likewise
4902 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4904         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4905         statement before passing it to vect_analyze_stmt.
4906         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4907         the chain of phi vector definitions.  Track the exit phi via its
4908         stmt_vec_info.
4909         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4910         STMT_VINFO_REDUC_DEF.
4911         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4912         stmt_vec_infos to handle the statement chains.
4913         (vect_get_slp_defs): Record the first statement in the node
4914         using a stmt_vec_info.
4915         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4916         statements here and pass their stmt_vec_info down to subroutines.
4917         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4918         down to vect_finish_stmt_generation.
4919         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4920         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4921         stmt_vec_infos to subroutines.
4922         (vect_remove_stores): Use stmt_vec_infos to handle the statement
4923         chains.
4925 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4927         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4928         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4929         (vect_permute_store_chain, vect_setup_realignment)
4930         (vect_permute_load_chain, vect_shift_permute_load_chain)
4931         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4932         stmts internally, and when passing values to other vectorizer routines.
4933         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4934         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4935         (vect_analyze_loop_operations, get_initial_def_for_reduction)
4936         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4937         (vectorizable_reduction, vectorizable_induction)
4938         (vectorizable_live_operation, vect_transform_loop_stmt)
4939         (vect_transform_loop): Likewise.
4940         * tree-vect-patterns.c (vect_reassociating_reduction_p)
4941         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4942         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4943         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4944         (vect_slp_analyze_node_operations_1): Likewise.
4945         * tree-vect-stmts.c (vect_mark_relevant, process_use)
4946         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4947         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4948         (vect_finish_stmt_generation_1, get_group_load_store_type)
4949         (get_load_store_type, vect_build_gather_load_calls)
4950         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4951         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4952         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4953         (vectorizable_store, vectorizable_load, vectorizable_condition)
4954         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4955         (supportable_widening_operation): Likewise.
4956         (vect_get_vector_types_for_stmt): Likewise.
4957         * tree-vectorizer.h (vect_dr_behavior): Likewise.
4959 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4961         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4962         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4963         (vect_permute_store_chain, vect_permute_load_chain)
4964         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4965         repeated stmt_vec_info lookups.
4966         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4967         (vect_update_ivs_after_vectorizer): Likewise.
4968         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4969         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4970         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4971         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4972         (vect_bb_slp_scalar_cost): Likewise.
4973         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4975 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4977         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4978         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4979         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4980         to get gassigns and gcalls, rather than operating on generc gimple
4981         stmts.
4982         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4983         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4984         (vectorizable_load, vect_analyze_stmt): Likewise.
4985         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4987 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4989         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4990         return stmt_vec_infos rather than gimple stmts.  Do not accept
4991         null arguments.
4992         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4993         of a gimple stmt.
4994         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4995         Update use of get_later_stmt.
4996         (vect_get_constant_vectors): Update call accordingly.
4997         (vect_schedule_slp_instance): Likewise
4998         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4999         (vect_slp_analyze_instance_dependence): Likewise.
5000         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
5002 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5004         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
5005         (stmt_info_for_cost::stmt_info): ...this new field.
5006         (add_stmt_costs): Update accordingly.
5007         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5008         (vect_get_known_peeling_cost): Likewise.
5009         (vect_estimate_min_profitable_iters): Likewise.
5010         * tree-vect-stmts.c (record_stmt_cost): Likewise.
5012 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5014         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
5015         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5016         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5017         accordingly.
5018         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
5020 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5022         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
5023         a gimple stmt to a stmt_vec_info.
5024         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5026 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5028         * tree-vectorizer.h (vec_info::grouped_stores): Change from
5029         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5030         (_loop_vec_info::reduction_chains): Likewise.
5031         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
5032         accordingly.
5033         * tree-vect-slp.c (vect_analyze_slp): Likewise.
5035 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5037         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
5038         a gimple stmt to a stmt_vec_info.
5039         (_stmt_vec_info::next_element): Likewise.
5040         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
5041         (vect_slp_analyze_and_verify_node_alignment)
5042         (vect_analyze_group_access_1, vect_analyze_group_access)
5043         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
5044         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
5045         (vect_supportable_dr_alignment): Update accordingly.
5046         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
5047         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
5048         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
5049         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5050         * tree-vect-slp.c (vect_build_slp_tree_1)
5051         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
5052         (vect_split_slp_store_group, vect_analyze_slp_instance)
5053         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
5054         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
5055         (get_group_load_store_type, get_load_store_type)
5056         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
5057         (vect_transform_stmt, vect_remove_stores): Likewise.
5059 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5061         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
5062         than a gimple stmt.
5063         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5064         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
5065         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
5066         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
5067         (vector_alignment_reachable_p, vect_get_data_access_cost)
5068         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
5069         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
5070         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5071         (vect_analyze_group_access_1, vect_analyze_group_access)
5072         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
5073         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
5074         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
5075         result of vect_dr_stmt and use the stmt_vec_info instead of
5076         the associated gimple stmt.
5077         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5078         (vect_gen_prolog_loop_niters): Likewise.
5079         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5081 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5083         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
5084         to a vec<stmt_vec_info>.
5085         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
5086         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
5087         vec<stmt_vec_info>.
5088         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
5089         to a vec<stmt_vec_info>.
5090         (bst_traits::value_type, bst_traits::value_type): Likewise.
5091         (bst_traits::hash): Update accordingly.
5092         (vect_get_and_check_slp_defs): Change the stmts parameter from
5093         a vec<gimple *> to a vec<stmt_vec_info>.
5094         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
5095         (vect_build_slp_tree): Likewise.
5096         (vect_build_slp_tree_2): Likewise.  Update uses of
5097         SLP_TREE_SCALAR_STMTS.
5098         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
5099         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
5100         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
5101         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
5102         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
5103         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
5104         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
5105         (vect_get_constant_vectors, vect_get_slp_defs)
5106         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
5107         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
5108         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
5109         instead of gimple stmts.
5110         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
5111         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
5112         (vect_slp_analyze_instance_dependence): Update uses of
5113         SLP_TREE_SCALAR_STMTS.
5114         (vect_slp_analyze_and_verify_node_alignment): Likewise.
5115         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
5116         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
5117         (get_initial_defs_for_reduction): Likewise.
5118         (vect_create_epilog_for_reduction): Likewise.
5119         (vectorize_fold_left_reduction): Likewise.
5120         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
5121         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
5122         (can_vectorize_live_stmts): Likewise.
5124 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5126         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
5127         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5128         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
5129         than gimple stmts.
5130         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
5131         of a gimple stmt.
5132         (gather_scalar_reductions): Update after above interface changes.
5133         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
5134         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
5135         than gimple stmts.
5136         (vect_force_simple_reduction): Likewise.
5137         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
5138         LOOP_VINFO_REDUCTIONS.
5139         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5141 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5143         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
5144         a gimple stmt to a stmt_vec_info.
5145         * tree-vect-loop.c (vect_active_double_reduction_p)
5146         (vect_force_simple_reduction, vectorizable_reduction): Update
5147         accordingly.
5149 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5151         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
5152         vec<gimple *> to a vec<stmt_vec_info>.
5153         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
5154         the reduction_phis argument from a vec<gimple *> to a
5155         vec<stmt_vec_info>.
5156         (vectorizable_reduction): Likewise the phis local variable that
5157         is passed to vect_create_epilog_for_reduction.  Update for new type
5158         of SLP_TREE_VEC_STMTS.
5159         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
5160         (vectorizable_live_operation): Likewise.
5161         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
5162         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
5164 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5166         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
5167         a gimple stmt to a stmt_vec_info.
5168         (vectorizable_condition, vectorizable_live_operation)
5169         (vectorizable_reduction, vectorizable_induction): Pass back the
5170         vectorized statement as a stmt_vec_info.
5171         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5172         use of STMT_VINFO_VEC_STMT.
5173         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5174         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5175         as stmt_vec_infos rather than gimple stmts.
5176         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5177         to a stmt_vec_info.
5178         (vectorizable_live_operation): Likewise.
5179         (vectorizable_reduction, vectorizable_induction): Likewise,
5180         updating use of STMT_VINFO_VEC_STMT.
5181         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5182         of STMT_VINFO_VEC_STMT.
5183         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5184         (vectorizable_simd_clone_call, vectorizable_conversion)
5185         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5186         (vectorizable_store, vectorizable_load, vectorizable_condition)
5187         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5188         from a gimple stmt to a stmt_vec_info.
5189         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
5190         pointer to a stmt_vec_info to the vectorizable_* routines.
5192 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5194         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5195         a gimple stmt to a stmt_vec_info.
5196         (is_pattern_stmt_p): Update accordingly.
5197         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5198         (vect_record_grouped_load_vectors): Likewise.
5199         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5200         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5201         (vect_model_reduction_cost): Likewise.
5202         (vect_create_epilog_for_reduction): Likewise.
5203         (vectorizable_reduction, vectorizable_induction): Likewise.
5204         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5205         Return the stmt_vec_info for the pattern statement.
5206         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5207         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5208         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5209         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5210         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5211         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5212         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5213         (free_stmt_vec_info, vect_is_simple_use): Likewise.
5215 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5217         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5218         (vect_finish_stmt_generation): Likewise.
5219         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5220         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5221         (vect_build_gather_load_calls): Use the return value of the above
5222         functions instead of a separate call to vinfo_for_stmt.  Use narrow
5223         scopes for the input gimple stmt and wider scopes for the associated
5224         stmt_vec_info.  Use vec_info::lookup_def when setting these
5225         stmt_vec_infos from an SSA_NAME definition.
5226         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5227         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5228         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5229         (vectorizable_store, vectorizable_load, vectorizable_condition)
5230         (vectorizable_comparison): Likewise.
5231         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5232         (vectorizable_reduction): Likewise.
5234 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5236         * tree-vectorizer.h (vect_is_simple_use): Add an optional
5237         stmt_vec_info * parameter before the optional gimple **.
5238         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5239         (process_use, vect_get_vec_def_for_operand_1): Update callers.
5240         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5241         * tree-vect-loop.c (vectorizable_reduction): Likewise.
5242         (vectorizable_live_operation): Likewise.
5243         * tree-vect-patterns.c (type_conversion_p): Likewise.
5244         (vect_look_through_possible_promotion): Likewise.
5245         (vect_recog_rotate_pattern): Likewise.
5246         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5248 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5250         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5251         a typedef to a wrapper class.
5252         (NULL_STMT_VEC_INFO): New macro.
5253         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5254         (stmt_vec_info::operator*): New function.
5255         (stmt_vec_info::operator gimple *): Likewise.
5256         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5257         (add_stmt_costs): Likewise.
5258         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5259         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5260         (vect_get_known_peeling_cost): Likewise.
5261         (vect_estimate_min_profitable_iters): Likewise.
5262         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5263         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5264         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5265         (vectorizable_store, free_stmt_vec_infos): Likewise.
5266         (new_stmt_vec_info): Change return type of xcalloc to
5267         _stmt_vec_info *.
5269 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5271         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5272         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5273         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5274         a single_imm_use-based sequence.
5275         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5277 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5279         * tree-vectorizer.h (vec_info::lookup_def): Declare.
5280         * tree-vectorizer.c (vec_info::lookup_def): New function.
5281         * tree-vect-patterns.c (vect_get_internal_def): Use it.
5282         (vect_widened_op_tree): Likewise.
5283         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5284         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5285         (vectorizable_reduction): Likewise.
5286         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5287         of a gimple *.
5288         (vect_is_slp_reduction): Update calls accordingly.  Use
5289         vec_info::lookup_def.
5290         (vect_is_simple_reduction): Likewise
5291         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5293 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5295         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5296         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5297         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5298         of vinfo_for_stmt.
5299         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5300         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5301         (vect_update_vf_for_slp, vect_analyze_loop_operations)
5302         (vect_is_slp_reduction, vectorizable_induction)
5303         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5304         * tree-vect-patterns.c (vect_init_pattern_stmt):
5305         (vect_determine_min_output_precision_1, vect_determine_precisions)
5306         (vect_pattern_recog): Likewise.
5307         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5308         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5309         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5310         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5311         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5312         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
5313         info field from a loop to a loop_vec_info.
5315 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5317         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5318         (vec_info::add_stmt): Declare.
5319         * tree-vectorizer.c (vec_info::add_stmt): New function.
5320         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5321         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5322         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5323         (vectorizable_induction): Likewise.
5324         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5325         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5326         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5327         (vectorizable_load): Likewise.
5328         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5329         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5330         (vect_recog_gather_scatter_pattern): Likewise.
5331         (append_pattern_def_seq): Likewise.  Remove a check that is
5332         performed by add_stmt itself.
5334 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5336         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5337         which make_ssa_name was called with new_stmt before new_stmt
5338         had been created.
5340 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5342         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5343         split out from...
5344         (vect_is_slp_reduction): ...here...
5345         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
5346         that are already known to be false.
5348 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5350         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5351         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
5352         STMT_VINFO_NUM_SLP_USES when it's true.
5353         (vect_free_slp_instance): Add a final_p parameter and pass it to
5354         vect_free_slp_tree.
5355         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5356         (vect_analyze_slp_instance): Likewise.
5357         (vect_slp_analyze_operations): Likewise.
5358         (vect_slp_analyze_bb_1): Likewise.
5359         * tree-vectorizer.c (vec_info): Likewise.
5360         * tree-vect-loop.c (vect_transform_loop): Likewise.
5362 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5364         * tree-vect-loop.c (vectorizable_reduction): Assert that the
5365         function is not called for second and subsequent members of
5366         a reduction group.
5368 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5370         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5371         cases for nested loops from here to ...
5372         (vect_create_epilog_for_reduction): ...here.  Only call
5373         vect_is_simple_use for inner-loop reductions.
5375 2018-07-31  Martin Liska  <mliska@suse.cz>
5377         PR gcov-profile/85338
5378         PR gcov-profile/85350
5379         PR gcov-profile/85372
5380         * profile.c (struct location_triplet): New.
5381         (struct location_triplet_hash): Likewise.
5382         (output_location): Do not output a BB that
5383         is already recorded for a line.
5384         (branch_prob): Use streamed_locations.
5386 2018-07-31  Martin Liska  <mliska@suse.cz>
5388         PR gcov-profile/85370
5389         * coverage.c (coverage_begin_function): Do not mark target
5390         clones as artificial functions.
5392 2018-07-31  Martin Liska  <mliska@suse.cz>
5394         PR gcov-profile/83813
5395         PR gcov-profile/84758
5396         PR gcov-profile/85217
5397         PR gcov-profile/85332
5398         * profile.c (branch_prob): Do not record GOTO expressions
5399         for GIMPLE statements which locations are already streamed.
5401 2018-07-31  Olivier Hainque  <hainque@adacore.com>
5403         * gcc.c (handle_spec_function): Accept a soft_matched_part
5404         argument, as do_spec_1.  Pass it down to ...
5405         (eval_spec_function): Accept a soft_matched_part argument,
5406         and pass it down to ...
5407         (do_spec_2): Accept a soft_matched_part argument, and pass
5408         it down to do_spec_1.
5409         (do_spec_1): Pass soft_matched_part to handle_spec_function.
5410         (handle_braces): Update call to handle_spec_function.
5411         (driver::set_up_specs): Update calls to do_spec_2.
5412         (compare_debug_dump_opt_spec_function): Likewise.
5413         (compare_debug_self_opt_spec_function): Likewise.
5415 2018-07-31  Olivier Hainque  <hainque@adacore.com>
5417         * common.opt (nolibc): New option.
5418         * doc/invoke.texi (Link Options): Document it.
5419         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
5420         * config/alpha/linux.h: Likewise.
5421         * config/arc/elf.h: Likewise.
5422         * config/arm/uclinux-elf.h: Likewise.
5423         * config/arm/unknown-elf.h: Likewise.
5424         * config/avr/avrlibc.h: Likewise.
5425         * config/bfin/bfin.h: Likewise.
5426         * config/bfin/linux.h: Likewise.
5427         * config/bfin/uclinux.h: Likewise.
5428         * config/darwin.h: Likewise.
5429         * config/darwin10.h: Likewise.
5430         * config/darwin12.h: Likewise.
5431         * config/gnu-user.h: Likewise.
5432         * config/lm32/uclinux-elf.h: Likewise.
5433         * config/pa/pa-hpux11.h: Likewise.
5434         * config/pa/pa64-hpux.h: Likewise.
5435         * config/sparc/sparc.h: Likewise.
5437 2018-07-31  Olivier Hainque  <hainque@adacore.com>
5439         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
5440         undefined variables.
5442 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
5444         PR target/86640
5445         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
5446         instead of GEN_INT.
5448 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5450         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5451         terminated string literal.
5453 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
5455         PR rtl-optimization/85160
5456         * combine.c (is_just_move): New function.
5457         (try_combine): Allow combining two instructions into two if neither of
5458         the original instructions was a move.
5460 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
5462         PR target/86673
5463         * doc/extend.texi (Global Register Variables): Discourage use of type
5464         qualifiers.
5465         (Local Register Variables): Likewise.
5467 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
5469         PR tree-optimization/86506
5470         * hwint.h (ceil_log2): Resync with hwint.c implementation.
5472 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
5474         PR target/86547
5475         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5476         hard_regno, make sure no insn between `from` and `to` clobbers it.
5478 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
5479             Tom de Vries  <tdevries@suse.de>
5481         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5482         (PTX_DEFAULT_RUNTIME_DIM): ... this.
5483         (nvptx_goacc_validate_dims): Set default worker and gang dims to
5484         PTX_DEFAULT_RUNTIME_DIM.
5485         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5487 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
5489         * config/pa/pa.c (pa_output_addr_vec): Align address table.
5490         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5491         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5493 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
5495         * config/rs6000/constraints.md (wG constraint): Delete, no longer
5496         used.
5497         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5498         predicate to reflect toc fusion has been deleted.
5499         (toc_fusion_mem_raw): Delete, no longer used.
5500         (toc_fusion_mem_wrapped): Likewise.
5501         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5502         fusion mask bit.
5503         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5504         Delete, no longer used.
5505         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5506         meant to be used for toc fusion.
5507         (rs6000_debug_print_mode): Delete toc fusion debugging.
5508         (rs6000_debug_reg_global): Likewise.
5509         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5510         fusion and secondary reload support that were never used.
5511         (rs6000_option_override_internal): Delete TOC fusion, that was only
5512         partially defined, and it did not work unless you also used the
5513         -mcmodel= switch.
5514         (rs6000_legitimate_address_p): Delete TOC fusion support.
5515         (rs6000_opt_masks): Likewise.
5516         (fusion_wrap_memory_address): Delete function, no longer used.
5517         (fusion_split_address); Delete TOC fusion support.
5518         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5519         longer used with toc fusion being deleted.
5520         (TARGET_TOC_FUSION_FP): Likewise.
5521         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5522         UNSPEC.
5523         (toc fusion spliter): Delete TOC fusion support.
5524         (toc_fusionload_<mode>): Likewise.
5525         (toc_fusionload_di): Likewise.
5526         (fusion_gpr_load_<mode>): Delete generator function, this insn no
5527         longer needs to be named.  Rename predicate to delete TOC fusion.
5528         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5529         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5530         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5531         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5532         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5534 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5536         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5537         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
5538         __int128 in built-in function prototypes.
5539         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5540         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5542 2018-07-27  Martin Sebor  <msebor@redhat.com>
5544         PR tree-optimization/86696
5545         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5546         types, including enums.
5547         (handle_char_store): Be prepared for the above function to fail.
5549 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
5551         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5552         when optimization level is lower than 2 or optimize for size.
5554 2018-07-26  Martin Sebor  <msebor@redhat.com>
5556         PR tree-optimization/86043
5557         PR tree-optimization/86042
5558         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5559         (get_string_cst_length): Rename...
5560         (get_min_string_length): ...to this.  Add argument.
5561         (handle_char_store): Extend to handle multi-character stores by
5562         MEM_REF.
5563         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
5564         * tree.h (initializer_zerop): Add argument.
5566 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
5568         PR middle-end/86660
5569         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5570         declare target to variables if they have always,{to,from,tofrom} map
5571         kinds.
5573 2018-07-26  Martin Liska  <mliska@suse.cz>
5575         PR lto/86548
5576         * lto-wrapper.c: Add linker_output as prefix
5577         for ltrans_output_file.
5579 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
5581         PR rtl-optimization/85805
5582         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5583         value for hard registers if that was written in the same mode.
5585 2018-07-26  Martin Liska  <mliska@suse.cz>
5587         PR gcov-profile/86536
5588         * gcov.c (format_gcov): Use printf format %.*f directly
5589         and do not handle special values.
5591 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5593         * common/config/arc/arc-common.c (arc_option_optimization_table):
5594         Update default optimizations for size.
5596 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5598         * config/arc/arc.md (movsf_insn): Add short instruction selection.
5599         * config/arc/constraints.md (CfZ): New constraint.
5600         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5601         (subsf3_fpu): Likewise.
5602         (cmpsf_fpu): Likewise.
5603         (cmpsf_fpu_uneq): Likewise.
5605 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5607         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5608         accesses as well.
5609         (arc_is_uncached_mem_p): uncached applies to both the variable and
5610         the pointer.
5612 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5614         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5615         register names.
5617 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
5619         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5620         field "m_scopes" from vec to auto_vec.
5622 2018-07-25  Martin Liska  <mliska@suse.cz>
5624         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5625         return type.
5627 2018-07-25  Richard Biener  <rguenther@suse.de>
5629         PR debug/86654
5630         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5631         special wrt context_die late.
5632         (gen_subprogram_die): Re-use DIEs in local scope.
5634 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
5636         PR tree-optimization/86644
5637         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
5639 2018-07-25  Martin Liska  <mliska@suse.cz>
5641         PR middle-end/86645
5642         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5643         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5645 2018-07-25  Martin Liska  <mliska@suse.cz>
5647         PR sanitizer/79635
5648         * params.def: Explain ASan abbreviation and provide
5649         a documentation link.
5651 2018-07-24  Martin Sebor  <msebor@redhat.com>
5653         PR tree-optimization/86622
5654         PR tree-optimization/86532
5655         * builtins.h (string_length): Declare.
5656         * builtins.c (c_strlen): Correct handling of non-constant offsets.
5657         (check_access): Be prepared for non-constant length ranges.
5658         (string_length): Make extern.
5659         * expr.c (string_constant): Only handle the minor non-constant
5660         array index.  Use string_constant to compute the length of
5661         a generic string constant.
5663 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
5665         PR tree-optimization/86618
5666         * tree-vect-stmts.c (vectorizable_call): Don't take the address
5667         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5669 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
5671         PR tree-optimization/86636
5672         * json.cc (json::object::set): Fix comment.  Add assertions.
5673         (json::array::append): Move here from json.h.  Add comment and an
5674         assertion.
5675         (json::string::string): Likewise.
5676         * json.h (json::array::append): Move to json.cc.
5677         (json::string::string): Likewise.
5678         * optinfo-emit-json.cc
5679         (optrecord_json_writer::impl_location_to_json): Assert that we
5680         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5681         wrapper around it.  Expand the location once, rather than three
5682         times.
5683         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5684         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5685         wrappers.
5686         (optrecord_json_writer::optinfo_to_json): Likewise, in four
5687         places.  Fix some overlong lines.
5689 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
5691         * config/aarch64/aarch64-simd.md
5692         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5693         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5694         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5695         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5696         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5697         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5698         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5699         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5700         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5702 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
5704         PR middle-end/86627
5705         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5706         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
5707         and abs_d == d, do the power of two handling if profitable.
5709 2018-07-24  Richard Biener  <rguenther@suse.de>
5711         * match.pd: Add BIT_FIELD_REF canonicalizations.
5713 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5715         PR c/86617
5716         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5718 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5720         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5721         terminated STRING_CST object.
5723 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5725         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5727 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
5729         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5730         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5731         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5732         the elements into a register.
5733         (rs6000_split_v4si_init_di_reg): Delete.
5734         (rs6000_split_v4si_init): Delete.
5735         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5736         (vsx_init_v4si): Rewrite as a define_expand.
5738 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
5740         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5741         zero_extend argument from memory): New.
5743 2018-07-22  Martin Sebor  <msebor@redhat.com>
5745         PR bootstrap/86621
5746         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5747         diagnosing calls with unknown arguments unless -Walloca-larger-than
5748         is restricted to less than PTRDIFF_MAX bytes.
5750 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
5752         * doc/gcov.texi (Invoking Gcov): Editorial changes.
5754 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
5756         * pretty-print.c (text_info::set_location): Remove redundant
5757         "line_table" parameter from call to rich_location::set_range.
5759 2018-07-20  Martin Sebor  <msebor@redhat.com>
5761         PR middle-end/82063
5762         * builtins.c (expand_builtin_alloca): Adjust.
5763         * calls.c (alloc_max_size): Simplify.
5764         * cgraphunit.c (cgraph_node::expand): Adjust.
5765         * common.opt (larger_than_size, warn_frame_larger_than): Remove
5766         variables.
5767         (frame_larger_than_size): Same.
5768         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5769         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5770         Initialize.
5771         * doc/invoke.texi (GCC Command Options): Document option arguments.
5772         Explain byte-size arguments and suffixes.
5773         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5774         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5775         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5776         * doc/options.texi (UInteger): Expand.
5777         (Host_Wide_Int, ByteSize): Document new properties.
5778         * final.c (final_start_function_1): Include sizes in an error message.
5779         * function.c (frame_offset_overflow): Same.
5780         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5781         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5782         Diagnose unbounded alloca calls only for limits of less than
5783         PTRDIFF_MAX.
5784         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
5785         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
5786         for alloca(0).
5787         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
5788         only for limits of less than PTRDIFF_MAX.
5789         * langhooks-def.h (lhd_handle_option): Change function argument
5790         to HOST_WIDE_INT.
5791         * langhooks.c (lhd_handle_option): Same.
5792         * langhooks.h (handle_option): Same.
5793         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5794         ByteSize flags.
5795         (var_type, var_type_struct): Same.
5796         (var_set): Handle ByteSize flag.
5797         * optc-gen.awk: Add comments to output to ease debugging.  Make
5798         use of HOST_WIDE_INT where appropriate.
5799         * opts-gen-save.awk:  Use %lx to format unsigned long.
5800         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5801         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5802         arguments.  Parse bytes-size suffixes.
5803         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5804         (enum_value_to_arg): Same.
5805         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
5806         (handle_option): Adjust.
5807         (generate_option): Change function argument to HOST_WIDE_INT.
5808         (cmdline_handle_error): Adjust.
5809         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5810         (set_option): Change function argument to HOST_WIDE_INT.
5811         (option_enabled): Handle cl_host_wide_int.
5812         (get_option_state): Handle CLVC_SIZE.
5813         (control_warning_option): Same.
5814         * opts.c (common_handle_option): Change function argument to
5815         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
5816         OPT_Wvla_larger_than_.
5817         * opts.h (enum cl_var_type): Add an enumerator.
5818         * stor-layout.c (layout_decl): Print a more meaningful warning.
5819         * toplev.c (output_stack_usage): Adjust.
5821 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
5823         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5824         call to inline_expand_builtin_string_cmp.
5825         (expand_builtin_strcmp): Likewise.
5826         (expand_builtin_strncmp): Likewise.
5827         (inline_string_cmp): Delete the last parameter, change char_type_node
5828         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5829         two operands.
5830         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5831         the inlining expansion on target where the type of the call has same or
5832         narrower precision than unsigned char.
5834 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
5836         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5837         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5838         * common.opt (fsave-optimization-record): New option.
5839         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5840         * doc/invoke.texi (-fsave-optimization-record): New option.
5841         * dumpfile.c: Include "optinfo-emit-json.h".
5842         (struct kv_pair): Move to coretypes.h.
5843         (optgroup_options): Make non-static.
5844         (dump_context::end_scope): Call
5845         optimization_records_maybe_pop_dump_scope.
5846         * dumpfile.h (optgroup_options): New decl.
5847         * json.cc: New file.
5848         * json.h: New file.
5849         * optinfo-emit-json.cc: New file.
5850         * optinfo-emit-json.h: New file.
5851         * optinfo.cc: Include "optinfo-emit-json.h".
5852         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5853         (optinfo_enabled_p): Check optimization_records_enabled_p.
5854         (optinfo_wants_inlining_info_p): Likewise.
5855         * optinfo.h: Update comment.
5856         * profile-count.c (profile_quality_as_string): New function.
5857         * profile-count.h (profile_quality_as_string): New decl.
5858         (profile_count::quality): New accessor.
5859         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5860         and optinfo_emit_json_cc_tests.
5861         * selftest.h (selftest::json_cc_tests): New decl.
5862         (selftest::optinfo_emit_json_cc_tests): New decl.
5863         * toplev.c: Include "optinfo-emit-json.h".
5864         (compile_file): Call optimization_records_finish.
5865         (do_compile): Call optimization_records_start.
5866         * tree-ssa-live.c: Include optinfo.h.
5867         (remove_unused_scope_block_p): Retain inlining information if
5868         optinfo_wants_inlining_info_p returns true.
5870 2018-07-20  Richard Biener  <rguenther@suse.de>
5872         PR debug/86585
5873         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5874         to cover -flto-partition=none.
5876 2018-07-20  Martin Liska  <mliska@suse.cz>
5878         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5879         (get_decl_source_range): Remove unused function.
5881 2018-07-20  Richard Biener  <rguenther@suse.de>
5883         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5884         (struct vn_phi_s): Likewise.
5885         (struct vn_reference_s): Likewise.
5886         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5887         for searching the slot of an entry known to be in the hash itself.
5888         (vn_phi_hasher::equal): Likewise.
5889         (vn_reference_hasher::equal): Likewise.
5890         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5891         globals.
5892         (optimistic_info, current_info): Remove, keeping only valid_info.
5893         (vn_reference_lookup_1): Remove fallback lookup.
5894         (vn_reference_lookup_2): Likewise.
5895         (vn_nary_op_lookup_1): Likewise.
5896         (vn_phi_lookup): Likewise.
5897         (vn_nary_build_or_lookup_1): Make sure to not chain the built
5898         hash element.
5899         (vn_reference_insert): Adjust, chain the inserted hash element
5900         at last_inserted_ref.
5901         (vn_reference_insert_pieces): Likewise.
5902         (visit_reference_op_call): Likewise.
5903         (vn_nary_op_insert_into): Chain the inserted hash element at
5904         last_inserted_nary.
5905         (vn_nary_op_insert_pieces): Adjust.
5906         (vn_nary_op_insert): Likewise.
5907         (vn_nary_op_insert_stmt): Likewise.
5908         (vn_phi_insert): Adjust, chain the inserted hash element at
5909         last_inserted_phi.
5910         (process_scc): Remove clearing and copying the optimistic
5911         table.  Instead remove elements inserted during an optimistic
5912         iteration from the single table we maintain.
5913         (init_scc_vn): Adjust.
5914         (free_scc_vn): Likewise.
5915         (sccvn_dom_walker::record_cond): Likewise.
5916         (sccvn_dom_walker::after_dom_children): Likewise.
5918 2018-07-19  Martin Sebor  <msebor@redhat.com>
5920         PR tree-optimization/84047
5921         PR tree-optimization/83776
5922         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5923         (check_array_bounds): Call it.
5925 2018-07-19  Martin Sebor  <msebor@redhat.com>
5927         * align.h (align_flags): Use member initialization.
5929 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
5931         * Makefile.in (OBJS): Add optinfo.o.
5932         * coretypes.h (class symtab_node): New forward decl.
5933         (struct cgraph_node): New forward decl.
5934         (class varpool_node): New forward decl.
5935         * dump-context.h: New file.
5936         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5937         "tree-pass.h".
5938         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5939         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5940         (set_alt_dump_file): Likewise.
5941         (dump_context::~dump_context): New dtor.
5942         (dump_gimple_stmt): Move implementation to...
5943         (dump_context::dump_gimple_stmt): ...this new member function.
5944         Add the stmt to any pending optinfo, creating one if need be.
5945         (dump_gimple_stmt_loc): Move implementation to...
5946         (dump_context::dump_gimple_stmt_loc): ...this new member function.
5947         Start a new optinfo and add the stmt to it.
5948         (dump_gimple_expr): Move implementation to...
5949         (dump_context::dump_gimple_expr): ...this new member function.
5950         Add the stmt to any pending optinfo, creating one if need be.
5951         (dump_gimple_expr_loc): Move implementation to...
5952         (dump_context::dump_gimple_expr_loc): ...this new member function.
5953         Start a new optinfo and add the stmt to it.
5954         (dump_generic_expr): Move implementation to...
5955         (dump_context::dump_generic_expr): ...this new member function.
5956         Add the tree to any pending optinfo, creating one if need be.
5957         (dump_generic_expr_loc): Move implementation to...
5958         (dump_context::dump_generic_expr_loc): ...this new member
5959         function.  Add the tree to any pending optinfo, creating one if
5960         need be.
5961         (dump_printf): Move implementation to...
5962         (dump_context::dump_printf_va): ...this new member function.  Add
5963         the text to any pending optinfo, creating one if need be.
5964         (dump_printf_loc): Move implementation to...
5965         (dump_context::dump_printf_loc_va): ...this new member function.
5966         Start a new optinfo and add the stmt to it.
5967         (dump_dec): Move implementation to...
5968         (dump_context::dump_dec): ...this new member function.  Add the
5969         value to any pending optinfo, creating one if need be.
5970         (dump_context::dump_symtab_node): New member function.
5971         (dump_context::get_scope_depth): New member function.
5972         (dump_context::begin_scope): New member function.
5973         (dump_context::end_scope): New member function.
5974         (dump_context::ensure_pending_optinfo): New member function.
5975         (dump_context::begin_next_optinfo): New member function.
5976         (dump_context::end_any_optinfo): New member function.
5977         (dump_context::s_current): New global.
5978         (dump_context::s_default): New global.
5979         (dump_scope_depth): Delete global.
5980         (dumpfile_ensure_any_optinfo_are_flushed): New function.
5981         (dump_symtab_node): New function.
5982         (get_dump_scope_depth): Reimplement in terms of dump_context.
5983         (dump_begin_scope): Likewise.
5984         (dump_end_scope): Likewise.
5985         (selftest::temp_dump_context::temp_dump_context): New ctor.
5986         (selftest::temp_dump_context::~temp_dump_context): New dtor.
5987         (selftest::verify_item): New function.
5988         (ASSERT_IS_TEXT): New macro.
5989         (ASSERT_IS_TREE): New macro.
5990         (ASSERT_IS_GIMPLE): New macro.
5991         (selftest::test_capture_of_dump_calls): New test.
5992         (selftest::dumpfile_c_tests): Call it.
5993         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5994         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5995         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5996         descriptive comment.
5997         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5998         (dump_node, dump_bb): Move these unrelated decls.
5999         (class dump_manager): Add leading comment.
6000         * optinfo.cc: New file.
6001         * optinfo.h: New file.
6003 2018-07-19  Michael Collison  <michael.collison@arm.com>
6004             Richard Henderson <rth@redhat.com>
6006         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
6007         (subti): Handle op1 zero.
6008         (subvti4, usub4ti4): New.
6009         (*sub<GPI>3_compare1_imm): New.
6010         (sub<GPI>3_carryinCV): New.
6011         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
6012         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
6014 2018-07-19  Michael Collison  <michael.collison@arm.com>
6015             Richard Henderson <rth@redhat.com>
6017         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
6018         (addti3): Create simpler code if low part is already known to be 0.
6019         (addvti4, uaddvti4): New.
6020         (*add<GPI>3_compareC_cconly_imm): New.
6021         (*add<GPI>3_compareC_cconly): New.
6022         (*add<GPI>3_compareC_imm): New.
6023         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
6024         handle constants within this pattern..
6025         (*add<GPI>3_compareV_cconly_imm): New.
6026         (*add<GPI>3_compareV_cconly): New.
6027         (*add<GPI>3_compareV_imm): New.
6028         (add<GPI>3_compareV): New.
6029         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
6030         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
6031         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
6032         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
6033         with 'comparison' operator.
6034         (*add<GPI>3_compareV_cconly_imm): Ditto.
6035         (*add<GPI>3_compareV_cconly): Ditto.
6036         (*add<GPI>3_compareV_imm): Ditto.
6037         (add<GPI>3_compareV): Ditto.
6038         (add<mode>3_carryinC): Ditto.
6039         (*add<mode>3_carryinC_zero): Ditto.
6040         (*add<mode>3_carryinC): Ditto.
6041         (add<mode>3_carryinV): Ditto.
6042         (*add<mode>3_carryinV_zero): Ditto.
6043         (*add<mode>3_carryinV): Ditto.
6045 2018-07-19  Michael Collison  <michael.collison@arm.com>
6046             Richard Henderson <rth@redhat.com>
6048         * config/aarch64/aarch64-modes.def (CC_V): New.
6049         * config/aarch64/aarch64-protos.h
6050         (aarch64_addti_scratch_regs): Declare
6051         (aarch64_subvti_scratch_regs): Declare.
6052         (aarch64_expand_subvti): Declare.
6053         (aarch64_gen_unlikely_cbranch): Declare
6054         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
6055         for signed overflow using CC_Vmode.
6056         (aarch64_get_condition_code_1): Handle CC_Vmode.
6057         (aarch64_gen_unlikely_cbranch): New function.
6058         (aarch64_addti_scratch_regs): New function.
6059         (aarch64_subvti_scratch_regs): New function.
6060         (aarch64_expand_subvti): New function.
6062 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
6064         * config/aarch64/aarch64-option-extensions.def: New entry for profile
6065         extension.
6066         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
6067         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
6068         extension.
6070 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
6072         PR target/83009
6073         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6074         address check not strict.
6076 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
6078         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
6079         Umq with Umn.
6080         (store_pair_lanes<mode>): Likewise.
6081         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
6082         enum value 'ADDR_QUERY_LDP_STP_N'.
6083         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
6084         (aarch64_print_address_internal): Add declaration.
6085         (aarch64_print_ldpstp_address): Remove.
6086         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
6087         (aarch64_print_operand): Change printing of 'y'.
6088         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
6089         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
6090         'true' rather than '1'.
6091         * config/aarch64/constraints.md (Uml): Likewise.
6092         (Uml): Rename to Umn.
6093         (Umq): Remove.
6095 2018-07-19  Richard Biener  <rguenther@suse.de>
6097         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
6098         a trailing array.
6099         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
6100         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
6101         (vn_reference_hasher): Likewise.
6102         (struct vn_tables_s): Remove obstack and alloc-pool members.
6103         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
6104         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
6105         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
6106         (vn_reference_insert_pieces): Likewise.
6107         (alloc_vn_nary_op_noinit): Adjust.
6108         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
6109         (vn_phi_eq): Adjust.
6110         (shared_lookup_phiargs): Remove.
6111         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
6112         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
6113         (visit_reference_op_call): Likewise.
6114         (copy_nary, copy_phi, copy_reference): Remove.
6115         (process_scc): Rewind the obstack when iterating.  Do not
6116         copy the elements to valid_info but just move them from one
6117         hashtable to the other.
6118         (allocate_vn_table): Adjust.
6119         (free_vn_table): Likewise.
6120         (init_scc_vn): Likewise.
6121         (free_scc_vn): Likewise.
6123 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
6125         PR target/86560
6126         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
6127         indirect_return as function type attribute.
6128         (ix86_attribute_table): Change indirect_return to function
6129         type attribute.
6130         * doc/extend.texi: Update indirect_return attribute.
6132 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
6134         * wide-int.h (widest2_int): New.
6135         * gimple-fold.c (arith_overflowed_p): Use it.
6136         * tree.h (widest2_int_cst): New.
6137         * tree-vrp.c (wide_int_binop_overflow): Rename from
6138         vrp_int_const_binop.
6139         Rewrite to work on trees.
6140         (extract_range_from_multiplicative_op_1): Abstract code to...
6141         (wide_int_range_min_max): ...here.
6142         (wide_int_range_cross_product): ...and here.
6143         (extract_range_from_binary_expr_1): Abstract overflow code to...
6144         (wide_int_range_mult_wrapping): ...here.
6145         * tree-vrp.h (wide_int_range_cross_product): New.
6146         (wide_int_range_mult_wrapping): New.
6148 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
6149             Julia Koval  <julia.koval@intel.com>
6151         * config/i386/x86-tune-costs.h (skylake_memcpy,
6152         skylake_memset): Replace rep_prefix with unrolling for size 512.
6154 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
6156         PR middle-end/86544
6157         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
6158         comparision with EQ_EXPR in last stmt.
6160 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6162         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
6163         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
6164         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
6165         previously known as "PowerPC AltiVec Built-in Functions".  Move
6166         some material to new subsubsections "PowerPC AltiVec Built-in
6167         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
6168         ISA 2.07".
6169         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
6170         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
6171         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6172         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6174 2018-07-18  Richard Biener  <rguenther@suse.de>
6176         PR tree-optimization/86557
6177         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
6178         EXACT_DIV_EXPR.
6180 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
6182         * config/s390/s390.c (s390_function_profiler): Generate CFI.
6184 2018-07-17  Jeff Law  <law@redhat.com>
6186         * config/arm/arm.c (get_label_padding): Update for recent
6187         changes to label_to_alignment.
6189         PR tree-optimization/86010
6190         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6192         * config/mips/mips.c (vr4130_align_insns): Update for recent
6193         changes to label_to_alignment.
6195         * config/frv/frv.c (frv_label_align): Update for recent changes
6196         to label_to_alignment.
6198         * config/nios2/nios2.c (nios2_label_align): Update for recent
6199         changes which dropped ALIGN_LABELS_LOG.
6201 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
6203         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6204         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6206 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
6208         * config/arc/arc.c (arc_label_align): Use align_labels instead of
6209         deprecated align_labels_log.
6211 2018-07-17  Richard Biener  <rguenther@suse.de>
6213         PR lto/86456
6214         * dwarf2out.c (init_sections_and_labels): Always generate
6215         a debug_line_str_section for early LTO debug.
6216         (dwarf2out_finish): Reset debug_line_str_hash output early.
6217         Bump counter for extra dwarf5 .debug_loc labels to not conflict
6218         with fat LTO part.
6219         (dwarf2out_early_finish): Output debug_line_str.
6221 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
6223         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6224         index register on z196 or later.
6226 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
6228         * config/s390/s390.c (s390_default_align): Set default function
6229         alignment to 16.
6230         (s390_override_options_after_change): Call s390_default align.
6231         (s390_option_override_internal): Call s390_default_align.
6232         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6234 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
6236         PR middle-end/86542
6237         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6238         to _looptemp_ clauses, other than the first two.
6240 2018-07-17  Martin Liska  <mliska@suse.cz>
6242         * opts.c: Do not enable OPT_falign_* for -Os.
6244 2018-07-17  Martin Liska  <mliska@suse.cz>
6246         * align.h (MAX_CODE_ALIGN): New.
6247         (MAX_CODE_ALIGN_VALUE): New.
6248         * common/config/i386/i386-common.c (ix86_handle_option):
6249         (MAX_CODE_ALIGN): Moved to align.h.
6250         * final.c (MAX_CODE_ALIGN): Likewise.
6251         * opts.c (parse_and_check_align_values):
6252         (MAX_CODE_ALIGN): Likewise.
6253         (MAX_CODE_ALIGN_VALUE): Likewise.
6255 2018-07-17  Martin Liska  <mliska@suse.cz>
6257         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6258         in order to fulfil coding style.
6259         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6260         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6261         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6262         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6263         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6264         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6265         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6266         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6267         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6269 2018-07-17  Martin Liska  <mliska@suse.cz>
6271         * align.h: New file.
6272         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6273         directly.
6274         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6275         align_flags of label_to_alignment.
6276         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6277         align_flags class.
6278         * config/m68k/m68k.c: Do not use removed align_labels_value and
6279         align_loops_value.
6280         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6281         (LOOP_ALIGN): Likewise.
6282         (LABEL_ALIGN): Likewise.
6283         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6284         Remove not used macro.
6285         (rs6000_loop_align): Change return type to align_flags.
6286         (rs6000_loop_align_max_skip): Remove.
6287         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6288         Change return type to align_flags.
6289         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6290         Remove not used macro.
6291         (rs6000_loop_align):  Change return type to align_flags.
6292         (rs6000_loop_align_max_skip): Remove.
6293         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6294         * config/rx/rx-protos.h (rx_align_for_label): Make it
6295         static function.
6296         * config/rx/rx.c (rx_align_for_label): Change return type
6297         to align_flags.
6298         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6299         macro definitions.
6300         into align_flags class.
6301         (LABEL_ALIGN): Likewise.
6302         (LOOP_ALIGN): Likewise.
6303         * config/s390/s390.c (s390_label_align): Use align_flags
6304         class member.
6305         (s390_asm_output_function_label): Likewise.
6306         * config/sh/sh.c (sh_override_options_after_change):
6307         Use align_flags class directly without macros.
6308         (find_barrier): Likewise.
6309         (barrier_align): Likewise.
6310         (sh_loop_align): Likewise.
6311         * config/spu/spu.c (spu_option_override):
6312         Use align_flags_tuple::get_value instead of removed macros.
6313         (spu_sched_init): Likewise.
6314         * config/spu/spu.h (GTY): Likewise.
6315         * config/visium/visium.c (visium_option_override):
6316         Set "8" as default secondary alignment.
6317         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6318         in order to guarantee secondary alignment of 8.
6319         * coretypes.h: Include align.h header file.
6320         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6321         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6322         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6323         * doc/tm.texi.in: Likewise.
6324         * final.c (struct label_alignment): Remove not used structure.
6325         (LABEL_ALIGN): Change type to align_flags.
6326         (LOOP_ALIGN): Likewise.
6327         (JUMP_ALIGN): Likewise.
6328         (default_loop_align_max_skip): Remove.
6329         (default_label_align_max_skip): Likewise.
6330         (default_jump_align_max_skip): Likewise.
6331         (default_label_align_after_barrier_max_skip):
6332         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6333         (LABEL_TO_MAX_SKIP): Remove.
6334         (label_to_alignment): Return align_flags type instead of integer.
6335         (label_to_max_skip): Remove.
6336         (align_fuzz): Use align_flags type.
6337         (compute_alignments): Use align_flags type and use align_flags::max
6338         to combine multiple alignments.
6339         (grow_label_align): Grow vec instead of C array.
6340         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6341         (shorten_branches):  Use align_flags type and use align_flags::max
6342         to combine multiple alignments.
6343         (final_scan_insn_1): Remove usage of secondary alignment that comes
6344         from label alignment, but instead use proper secondary alignment
6345         which is computed in grow_label_align.
6346         * flags.h (struct align_flags_tuple): Move to align.h.
6347         (struct align_flags): Likewise.
6348         (state_align_loops): Rename to align_loops.
6349         (state_align_jumps): Rename to align_jumps.
6350         (state_align_labels): Rename to align_labels.
6351         (state_align_functions): Rename to align_functions.
6352         (align_loops_log): Remove.
6353         (align_jumps_log): Remove.
6354         (align_labels_log): Remove.
6355         (align_functions_log): Remove.
6356         (align_loops_max_skip): Remove.
6357         (align_jumps_max_skip): Remove.
6358         (align_labels_max_skip): Remove.
6359         (align_functions_max_skip): Remove.
6360         (align_loops_value): Remove.
6361         (align_jumps_value): Remove.
6362         (align_labels_value): Remove.
6363         (align_functions_value): Remove.
6364         * output.h (label_to_alignment): Change return type to align_flags.
6365         (label_to_max_skip): Remove.
6366         * target.def: Remove loop_align_max_skip, label_align_max_skip,
6367         jump_align_max_skip macros.
6368         * targhooks.h (default_loop_align_max_skip): Remove.
6369         (default_label_align_max_skip): Likewise.
6370         (default_jump_align_max_skip): Likewise.
6371         (default_label_align_after_barrier_max_skip): Remove.
6372         * toplev.c (read_log_maxskip): Use ::normalize function.
6373         (parse_N_M): Remove not used argument and also call ::normalize.
6374         (parse_alignment_opts): Do not pass unused arguments.
6375         * varasm.c (assemble_start_function): Use directly align_functions
6376         instead of removed macros.
6377         * system.h: Do not poison removed macros.
6379 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
6381         PR middle-end/86539
6382         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
6383         and cond temporaries don't have reference type if iterator has
6384         pointer type.  For init use &for_pre_body instead of pre_p if
6385         for_pre_body is non-empty.
6387 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
6389         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
6390         double-double modes to SFmode directly directly.
6391         (trunc<mode>sf2_fprs): Delete.
6393 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
6395         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
6396         for conversions between IFmode and the decimal floating point modes.
6397         (init_float128_ieee): Use the correct names for conversions between
6398         KFmode and the decimal floating point modes.
6400 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
6402         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
6403         for the conversions between TDmode and IFmode.
6404         (init_float128_ieee): Use more correct names for the conversions
6405         between TDmode and KFmode.
6407 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
6409         PR tree-optimization/86526
6410         * builtins.c (expand_builtin_memcmp): Formatting fixes.
6411         (inline_expand_builtin_string_cmp): Likewise.
6412         (inline_string_cmp): Likewise.  Use c_readstr instead of
6413         builtin_memcpy_read_str.  Add unit_mode temporary.
6415 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6417         PR middle-end/86528
6418         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
6419         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
6421 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6423         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
6424         Alphabetize prototypes of built-in functions, separating out
6425         built-in functions that are listed in this section but should be
6426         described elsewhere.
6428 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
6430         PR target/86511
6431         * expmed.c (emit_store_flag): Do not emit setcc followed by a
6432         conditional move when trapping comparison was split to a
6433         non-trapping one (and vice versa).
6435 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
6437         * config/s390/s390.c (s390_function_profiler): Generate nops
6438         instead of profiler call sequences.
6439         * config/s390/s390.opt: Add the new option.
6441 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
6443         * config/s390/s390.c (s390_function_profiler): Generate
6444         __mcount_loc section.
6445         * config/s390/s390.opt: Add the new option.
6447 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
6449         * common.opt: Add the new warning.
6450         * config/s390/s390.c (s390_function_profiler): Emit "brasl
6451         %r0,__fentry__" when -mfentry is specified.
6452         (s390_option_override_internal): Disallow -mfentry for 31-bit
6453         CPUs.
6454         * config/s390/s390.opt: Add the new option.
6456 2018-07-16  Richard Biener  <rguenther@suse.de>
6458         PR lto/86523
6459         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6460         for function-local FUNCTION_DECL and RESULT_DECL immediately.
6462 2018-07-16  Martin Liska  <mliska@suse.cz>
6464         PR ipa/86529
6465         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6466         to ::get_create.
6468 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
6470         * config/arc/arcHS.md: Update ARCHS scheduling rules.
6472 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
6474         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6475         for ARCHS4x.
6476         * config/arc/arc-cpus.def (hs4x): New cpu.
6477         (hs4xd): Likewise.
6478         * config/arc/arc-tables.opt: Regenerate.
6479         * config/arc/arc.c (arc_sched_issue_rate): New function.
6480         (TARGET_SCHED_ISSUE_RATE): Define.
6481         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6482         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6483         fpu_cvt.
6484         (attr tune): Add ARCHS4x tune values.
6485         (attr tune_dspmpy): Define.
6486         (*tst): Correct instruction type.
6487         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6488         * config/arc/arcHS4x.md: New file.
6489         * config/arc/fpu.md: Update instruction type attributes.
6490         * config/arc/t-multilib: Regenerate.
6492 2018-07-16  Tom de Vries  <tdevries@suse.de>
6494         PR debug/86455
6495         * var-tracking.c (vt_initialize): Fix pre_dec handling.
6497 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6499         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6500         early clobber.
6502 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
6504         PR tree-optimization/86514
6505         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6506         conversion to a boolean type from a type with greater precision.
6508 2018-07-16  Tom de Vries  <tdevries@suse.de>
6510         * var-tracking.c (vt_initialize): Print adjusted insn slim if
6511         dump_flags request TDF_SLIM.
6513 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
6515         * fold-const.c (int_const_binop_1): Abstract...
6516         (wide_int_binop): ...wide int code here.
6517         (poly_int_binop): ...poly int code here.
6518         Abstract the rest of int_const_binop_1 into int_const_binop.
6519         * fold-const.h (wide_int_binop): New.
6520         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6521         Remove useless PLUS/MINUS_EXPR case.
6522         (zero_nonzero_bits_from_vr): Move wide int code...
6523         (zero_nonzero_bits_from_bounds): ...here.
6524         (extract_range_from_binary_expr_1): Move mask optimization code...
6525         (range_easy_mask_min_max): ...here.
6526         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6527         (range_easy_mask_min_max): New.
6529 2018-07-15  Jeff Law  <law@redhat.com>
6531         PR target/85993
6532         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6533         block.
6535 2018-07-14  Jim Wilson  <jimw@sifive.com>
6537         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6539 2018-07-14  Paul Koning  <ni1d@arrl.net>
6541         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6543 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
6545         * lto-streamer-out.c (copy_function_or_variable): Dump info about
6546         copying section.
6548 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
6549             Steve Munroe  <munroesj52@gmail.com>
6551         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6552         (_mm_andnot_si128): Likewise.
6553         (_mm_or_si128): Likewise.
6554         (_mm_xor_si128): Likewise.
6556 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
6558         PR middle-end/78809
6559         * builtins.c (expand_builtin_memcmp): Inline the calls first
6560         when result_eq is false.
6561         (expand_builtin_strcmp): Inline the calls first.
6562         (expand_builtin_strncmp): Likewise.
6563         (inline_string_cmp): New routine. Expand a string compare
6564         call by using a sequence of char comparison.
6565         (inline_expand_builtin_string_cmp): New routine. Inline expansion
6566         a call to str(n)cmp/memcmp.
6567         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6568         New option.
6569         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6571 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
6573         * config/arm/driver-arm.c: Include arm-native.h.
6574         (host_detect_local_cpu): Use auto-generated data tables.
6575         (vendors, arm_cpu_table): Delete.  Move part information to ...
6576         * config/arm/arm-cpus.in: ... here.
6577         * config/arm/parsecpu.awk (gen_native): New function.
6578         (vendor, part): New CPU fields.
6579         (END): Add support for building the native CPU detection tables.
6580         * config/arm/t-arm (arm-native.h): Add build rule.
6581         (driver-arm.o): Add dependency on arm-native.h.
6583 2018-07-13  Richard Biener  <rguenther@suse.de>
6585         PR middle-end/85974
6586         * match.pd (addr1 - addr2): Allow either of the operand to
6587         have a conversion.
6589 2018-07-13  Tom de Vries  <tdevries@suse.de>
6591         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6592         in remap_ssa_name.
6594 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
6596         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6597         arrays instead of numbered variables.
6599 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
6601         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6602         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6604 2018-07-13  Richard Biener  <rguenther@suse.de>
6606         PR debug/86452
6607         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6608         instead of get_context_die.
6610 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
6611             Richard Biener  <rguenther@suse.de>
6613         PR middle-end/86489
6614         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6615         that the loop latch destination where phi is defined.
6617 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
6619         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6620         (riscv_expand_epilogue): Add assertion to check interrupt mode.
6621         (riscv_set_current_function): Extract getting interrupt type to new
6622         function.
6623         (riscv_get_interrupt_type): New function.
6624         (riscv_merge_decl_attributes): New function, checking interrupt type is
6625         same.
6626         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6628 2018-07-12  Paul Koning  <ni1d@arrl.net>
6630         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6631         directive.
6633 2018-07-12  Paul Koning  <ni1d@arrl.net>
6635         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6636         zero reference, add doloop_end instead.
6637         * doc/md.texi (decrement_and_branch_until_zero): Remove.
6638         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
6639         detail for doloop_end.
6641 2018-07-12  Martin Sebor  <msebor@redhat.com>
6643         PR c/86453
6644         * attribs.c (decl_attributes): Reject conflicting attributes before
6645         calling attribute handlers.
6647 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
6649         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6650          parameter.
6651         (gcc::dump_manager::get_dump_file_name): likewise.
6652         (dump_begin): Likewise.
6653         * dumpfile.h (dump_begin): Update prototype.
6654         (gcc::dump_manager::get_dump_file_name,
6655         gcc::dump_manager::get_dump_file_name): Update prototype.
6657 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6659         * internal-fn.h (vectorizable_internal_fn_p): New function.
6660         * tree-vect-slp.c (compatible_calls_p): Likewise.
6661         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
6662         to internal functions.
6663         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6665 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6667         * fold-const.h (inverse_conditions_p): Declare.
6668         * fold-const.c (inverse_conditions_p): New function.
6669         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
6670         that test the inverse condition of a conditional internal function.
6671         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6672         * internal-fn.c (internal_fn_mask_index): Handle conditional
6673         internal functions.
6674         (vectorized_internal_fn_supported_p): New function.
6675         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6676         (any_pred_load_store): Replace with...
6677         (need_to_predicate): ...this new variable.
6678         (redundant_ssa_names): New variable.
6679         (ifcvt_can_use_mask_load_store): Move initial checks to...
6680         (ifcvt_can_predicate): ...this new function.  Handle tree codes
6681         for which a conditional internal function exists.
6682         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6683         instead of ifcvt_can_use_mask_load_store.  Update after variable
6684         name change.
6685         (predicate_load_or_store): New function, split out from
6686         predicate_mem_writes.
6687         (check_redundant_cond_expr): New function.
6688         (value_available_p): Likewise.
6689         (predicate_rhs_code): Likewise.
6690         (predicate_mem_writes): Rename to...
6691         (predicate_statements): ...this.  Use predicate_load_or_store
6692         and predicate_rhs_code.
6693         (combine_blocks, tree_if_conversion): Update after above name changes.
6694         (ifcvt_local_dce): Handle redundant_ssa_names.
6695         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6696         general conditional functions.
6697         * tree-vect-stmts.c (vectorizable_call): Likewise.
6699 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6700             Alan Hayward  <alan.hayward@arm.com>
6701             David Sherwood  <david.sherwood@arm.com>
6703         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6704         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6705         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6706         plus and minus and convert them into IFN_COND_FMA-based sequences.
6707         (convert_mult_to_fma): Handle conditional plus and minus.
6709 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6711         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6712         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6713         (cond_fnms_optab): New optabs.
6714         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6715         internal functions.
6716         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6717         * internal-fn.h (get_conditional_internal_fn): Declare.
6718         (get_unconditional_internal_fn): Likewise.
6719         * internal-fn.c (cond_ternary_direct): New macro.
6720         (expand_cond_ternary_optab_fn): Likewise.
6721         (direct_cond_ternary_optab_supported_p): Likewise.
6722         (FOR_EACH_COND_FN_PAIR): Likewise.
6723         (get_conditional_internal_fn): New function.
6724         (get_unconditional_internal_fn): Likewise.
6725         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6726         (gimple_match_op::gimple_match_op): Add a new overload for 5
6727         operands.
6728         (gimple_match_op::set_op): Likewise.
6729         (gimple_resimplify5): Declare.
6730         * genmatch.c (decision_tree::gen): Generate simplifications for
6731         5 operands.
6732         * gimple-match-head.c (gimple_simplify): Define an overload for
6733         5 operands.  Handle calls with 5 arguments in the top-level overload.
6734         (convert_conditional_op): Handle conversions from unconditional
6735         internal functions to conditional ones.
6736         (gimple_resimplify5): New function.
6737         (build_call_internal): Pass a fifth operand.
6738         (maybe_push_res_to_seq): Likewise.
6739         (try_conditional_simplification): Try converting conditional
6740         internal functions to unconditional internal functions.
6741         Handle 3-operand unconditional forms.
6742         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6743         Define ternary equivalents of the current rules for binary conditional
6744         internal functions.
6745         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6746         ternary operations.
6747         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6748         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6749         (optab): Handle them.
6750         (SVE_COND_FP_TERNARY): New int iterator.
6751         (sve_fmla_op, sve_fmad_op): New int attributes.
6752         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6753         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6754         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6756 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6758         * target.def (preferred_else_value): New target hook.
6759         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6760         * doc/tm.texi: Regenerate.
6761         * targhooks.h (default_preferred_else_value): Declare.
6762         * targhooks.c (default_preferred_else_value): New function.
6763         * internal-fn.h (conditional_internal_fn_code): Declare.
6764         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6765         (get_conditional_internal_fn): Use it.
6766         (conditional_internal_fn_code): New function.
6767         * gimple-match.h (gimple_match_cond): New struct.
6768         (gimple_match_op): Add a cond member function.
6769         (gimple_match_op::gimple_match_op): Update all forms to take a
6770         gimple_match_cond.
6771         * genmatch.c (expr::gen_transform): Use the same condition as res_op
6772         for the suboperation, but don't specify a particular else_value.
6773         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6774         (visit_nary_op, visit_reference_op_load): Pass
6775         gimple_match_cond::UNCOND to the gimple_match_op constructor.
6776         * gimple-match-head.c: Include tree-eh.h
6777         (convert_conditional_op): New function.
6778         (maybe_resimplify_conditional_op): Likewise.
6779         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6780         (gimple_resimplify2): Likewise.
6781         (gimple_resimplify3): Likewise.
6782         (gimple_resimplify4): Likewise.
6783         (maybe_push_res_to_seq): Return null for conditional operations.
6784         (try_conditional_simplification): New function.
6785         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
6786         constructor.
6787         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6788         IFN_COND_* call.
6789         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6790         function.
6791         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6793 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
6795         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6796         DECL_FCONTEXT
6797         (hash_tree): Do not hash DECL_FCONTEXT
6798         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6799         Do not stream DECL_FCONTEXT.
6800         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6801         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6803 2018-07-12  Richard Biener  <rguenther@suse.de>
6805         PR debug/86462
6806         * dwarf2out.c (gen_block_die): Only output blocks when they have
6807         at least one !DECL_IGNORED_P variable.
6809 2018-07-12  Richard Biener  <rguenther@suse.de>
6811         PR target/84829
6812         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6813         Remove -mieee-fp handling.
6815 2018-07-12  Richard Biener  <rguenther@suse.de>
6817         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6818         left-over from last patch.
6820 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
6822         PR tree-optimization/86492
6823         * gimple-ssa-store-merging.c
6824         (imm_store_chain_info::coalesce_immediate_stores): Call
6825         check_no_overlap even for the merge_overlapping case.  Formatting fix.
6827 2018-07-12  Richard Biener  <rguenther@suse.de>
6829         PR middle-end/86479
6830         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6831         move possibly trapping operations into the conditional.
6833 2018-07-12  Richard Biener  <rguenther@suse.de>
6835         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6836         (vn_lookup_simplify_result): Remove recursion limit applied
6837         here.
6838         (vn_nary_build_or_lookup_1): Adjust.
6839         (try_to_simplify): Likewise.
6840         * gimple-match-head.c (gimple_resimplify1): Instead apply one
6841         here.
6842         (gimple_resimplify2): Likewise.
6843         (gimple_resimplify3): Likewise.
6844         (gimple_resimplify4): Likewise.
6846 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
6848         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6849         Use __mmask64 type instead of __mmask8 for __M argument.
6850         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6851         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6852         __U argument.
6853         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6854         __mmask16 for __M argument.
6855         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6856         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6857         to __mmask16 instead of __mmask8.
6858         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6859         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6860         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6861         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6862         instead of __mmask16 for __U argument.
6863         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6864         __mmask16 instead of __mmask8 for __U argument.
6865         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6866         __U argument.
6867         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6868         __mmask16.
6869         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6870         argument.
6871         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6872         __U argument.
6873         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6874         __mmask16.
6875         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6876         of __mmask16.
6877         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6878         __U argument.
6879         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6880         __U argument.
6881         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6882         __U argument.
6883         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6884         __U argument.
6885         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6886         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6887         return type as well as __M argument type and all casts from __mmask8
6888         to __mmask32.
6889         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6890         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6891         return type as well as __M argument type and all casts from __mmask8
6892         to __mmask16.
6893         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6894         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6895         return type as well as __M argument type and all casts from __mmask8
6896         to __mmask32.
6897         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6898         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6899         return type as well as __M argument type and all casts from __mmask8
6900         to __mmask16.
6901         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6902         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6903         __mmask16.
6905 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
6907         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6908         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6909         for __U argument.
6911 2018-07-11  Paul Koning  <ni1d@arrl.net>
6913         * doc/md.texi (define_subst): Document how multiple occurrences of
6914         the same argument in the replacement pattern are handled.
6916 2018-07-11  Paul Koning  <ni1d@arrl.net>
6918         * doc/extend.texi (Common Variable Attributes): Move "mode" into
6919         alphabetical order.
6920         (Common Type Attributes): Add "mode" attribute.
6922 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
6924         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6925         stream DECL_ORIGINAL_TYPE.
6926         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6927         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6928         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6929         Do not walk original type.
6930         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6931         external decls.
6932         (write_ts_decl_non_common_tree_pointers): Do not stream
6933         DECL_ORIGINAL_TYPE
6934         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6935         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6937 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
6939         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6940         thread twice from the same starting edge.
6942 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
6944         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6945         * gimple.c (gimple_call_nonnull_result_p): ...here...
6946         (gimple_call_nonnull_arg): ...and here.
6947         * gimple.h (gimple_call_nonnull_result_p): New.
6948         (gimple_call_nonnull_arg): New.
6950 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
6952         * config/arm/arm-cpus.in: Move information from fpu field of each
6953         cpu definition to the isa field.
6954         * config/arm/parsecpu.awk (fpu): Delete match rule.
6955         (gen_comm_data): Don't add bits from the CPU's FPU entry.
6957 2018-07-11  Richard Biener  <rguenther@suse.de>
6959         PR debug/86457
6960         * dwarf2out.c (init_sections_and_labels): Use
6961         output_asm_line_debug_info consistently.
6962         (dwarf2out_early_finish): Likewise.
6963         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6964         type units.
6966 2018-07-11  Richard Biener  <rguenther@suse.de>
6968         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6969         Rework father_bb setting in a way to avoid propagating constants
6970         multiple times on a loop body.
6972 2018-07-10  Mark Wielaard  <mark@klomp.org>
6974         PR debug/86459
6975         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6977 2018-07-10  Richard Biener  <rguenther@suse.de>
6979         * hash-map.h (hash_map::iterator::operator*): Return
6980         references to key and value.
6982 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
6984         PR c++/86443
6985         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6986         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6987         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6988         (gimplify_omp_for): For composite loops, move outer
6989         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6990         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
6991         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
6992         TREE_LIST for both the original class iterator and the "last" helper
6993         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
6994         loop, remember has_decl_expr from outer composite loops for the
6995         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6997 2018-07-09  Martin Sebor  <msebor@redhat.com>
6999         PR middle-end/77357
7000         PR middle-end/86428
7001         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
7002         accessing implicitly initialized array elements.
7003         * expr.c (string_constant): Handle string initializers of
7004         character arrays within aggregates.
7005         * gimple-fold.c (fold_array_ctor_reference): Add argument.
7006         Store element offset.  As a special case, handle zero size.
7007         (fold_nonarray_ctor_reference): Same.
7008         (fold_ctor_reference): Add argument.  Store subobject offset.
7009         * gimple-fold.h (fold_ctor_reference): Add argument.
7011 2018-07-09  Paul Koning  <ni1d@arrl.net>
7013         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
7014         (pdp11_insn_cost): New function.
7015         (pdp11_md_asm_adjust): New function.
7016         (TARGET_INVALID_WITHIN_DOLOOP): Define.
7017         (pdp11_rtx_costs): Update to match machine better.
7018         (output_addr_const_pdp11): Correct format mismatch warnings.
7019         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
7020         * config/pdp11/pdp11.md: General change to add base_cost and/or
7021         length attributes for use by new pdp11_insn_cost function.
7022         (MIN_BRANCH): Correct definition.
7023         (MIN_SOB): Ditto.
7024         (doloop_end): Use standard pattern name for looping pattern.
7025         (doloop_end_nocc): New.
7026         (movsf): Add another constraint alternative.
7027         (zero_extendqihi2): Add constraint alternatives for not in place
7028         extend.
7029         (zero_extendhisi2): Remove.
7030         (shift patterns): Add CC handling variants.
7031         (bswaphi2): New.
7032         (bswapsi2): New.
7033         (rothi3): New.
7034         (define_peephole2): New peephole to recognize mov that sets CC for
7035         subsequent test.
7037 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7039         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
7040         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
7041         wi::add.
7043 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
7045         PR c/86420
7046         * real.c (real_nextafter): Return true if result is denormal.
7048 2018-07-09  Martin Liska  <mliska@suse.cz>
7050         * common.opt: Add back wrongly removed attribute.
7052 2018-07-09  Richard Biener  <rguenther@suse.de>
7054         PR debug/86413
7055         * dwarf2out.c (gen_block_die): For an early generated DIE
7056         always output high/low PC attributes.
7058 2018-07-09  Tom de Vries  <tdevries@suse.de>
7060         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
7061         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
7062         onto VAR_DECL with abstract origin.
7064 2018-07-07  Jim Wilson  <jimw@sifive.com>
7066         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
7068 2018-07-07  Tom de Vries  <tdevries@suse.de>
7070         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
7072 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
7074         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
7075         overflow_type.
7076         (combine_bound): Use wide-int overflow calculation instead of
7077         rolling our own.
7078         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
7079         overflow_type.
7080         * fold-const.c (int_const_binop_2): Same.
7081         (extract_muldiv_1): Same.
7082         (fold_div_compare): Same.
7083         (fold_abs_const): Same.
7084         * match.pd: Same.
7085         * poly-int.h (add): Same.
7086         (sub): Same.
7087         (neg): Same.
7088         (mul): Same.
7089         * predict.c (predict_iv_comparison): Same.
7090         * profile-count.c (slow_safe_scale_64bit): Same.
7091         * simplify-rtx.c (simplify_const_binary_operation): Same.
7092         * tree-chrec.c (tree_fold_binomial): Same.
7093         * tree-data-ref.c (split_constant_offset_1): Same.
7094         * tree-if-conv.c (idx_within_array_bound): Same.
7095         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
7096         * tree-ssa-phiopt.c (minmax_replacement): Same.
7097         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
7098         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
7099         * vr-values.c (vr_values::adjust_range_with_scev): Same.
7100         * wide-int.cc (wi::add_large): Same.
7101         (wi::mul_internal): Same.
7102         (wi::sub_large): Same.
7103         (wi::divmod_internal): Same.
7104         * wide-int.h: Change overflow type to overflow_type for neg, add,
7105         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
7106         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
7107         mul_internal, divmod_internal.
7108         (overflow_type): New enum.
7109         (accumulate_overflow): New.
7111 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7113         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
7114         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
7116 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
7118         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
7119         argument is checked for zero before entering loop, avoid checking again.
7121 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
7123         * gimplify.h (generic_expr_could_trap_p): Set as global function.
7124         * gimplify.h (generic_expr_could_trap_p): Likwise.
7125         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
7127 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
7129         PR tree-optimization/86401
7130         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
7131         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
7132         (fold_bit_and_mask): ... here.  New helper function for match.pd.
7133         * fold-const.h (fold_bit_and_mask): Declare.
7134         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
7136 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
7138         PR target/86324
7139         * target.def (translate_mode_attribute): New hook.
7140         * targhooks.h (default_translate_mode_attribute): Declare.
7141         * targhooks.c (default_translate_mode_attribute): New function.
7142         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
7143         * doc/tm.texi: Regenerate.
7144         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
7145         (rs6000_translate_mode_attribute): New function.
7147 2018-07-06  Paul Koning  <ni1d@arrl.net>
7149         * doc/md.texi (define_split): Document DONE and FAIL.
7150         (define_peephole2): Ditto.
7152 2018-07-05  Jeff Law  <law@redhat.com>
7154         PR tree-optimization/86010
7155         * tree-ssa-dse.c (compute_trims): More aggressively trim at
7156         both the head and tail of mem* and str* calls.
7158 2018-07-05  Jim Wilson  <jimw@sifive.com>
7160         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
7162 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
7164         * config/aarch64/aarch64-simd.md: correct flags text for
7165         MIN_EXPR replacement.
7167 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
7169         * configure: Regenerated.
7171 2018-07-05  Carl Love  <cel@us.ibm.com>
7173         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7174         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7175         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7176         VSX_BUILTIN_DOUBLEL_V4SF.
7178 2018-07-05  Martin Sebor  <msebor@redhat.com>
7180         PR c++/86400
7181         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7182         than its domain to compute its the upper bound of a char array.
7184 2018-07-05  Nathan Sidwell  <nathan@acm.org>
7186         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7187         * doc/cpp.texi: Update comment.
7188         * doc/tm.texi: Rebuilt.
7189         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7190         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7191         * doc/extend.texi (Backwards Compatibility): Clarify it is system
7192         headers affected by extern "C".
7193         * system.h: Poison NO_IMPLICIT_EXTERN_C.
7194         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7195         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7196         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7197         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7198         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7199         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7200         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7201         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7202         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7203         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7204         config/sparc/sp64-elf.h, config/spu/spu.h,
7205         config/stormy16/stormy16.h, config/v850/v850.h,
7206         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7207         define NO_IMPLICIT_EXTERN_C.
7208         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7210 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
7212         PR target/84711
7213         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7214         instead of GET_MODE_SIZE when comparing Units.
7216 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
7218         PR target/84711
7219         * rtlanal.c (set_noop_p): Constrain on mode change,
7220         include hard-reg-set.h
7222 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
7224         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7226 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
7228         Revert
7229         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
7231         PR sanitizer/84250
7232         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7233         libasan.
7234         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7236 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
7238         PR sanitizer/84250
7239         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7240         libasan.
7241         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7243 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
7245         PR middle-end/86380
7246         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7248 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
7250         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7251         neg_*_op* variables.
7253 2018-07-04  Martin Liska  <mliska@suse.cz>
7255         * tree-switch-conversion.c: Define
7256         max_ratio_for_speed and max_ratio_for_size constants.
7258 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
7259             Martin Liska  <mliska@suse.cz>
7261         PR middle-end/66240
7262         PR target/45996
7263         PR c/84100
7264         * common.opt: Rename align options with 'str_' prefix.
7265         * common/config/i386/i386-common.c (set_malign_value): New
7266         function.
7267         (ix86_handle_option): Use it to set -falign-* options/
7268         * config/aarch64/aarch64-protos.h (struct tune_params): Change
7269         type from int to string.
7270         * config/aarch64/aarch64.c: Update default values from int
7271         to string.
7272         * config/alpha/alpha.c (alpha_override_options_after_change):
7273         Likewise.
7274         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7275         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7276         max skip conditionally.
7277         * config/i386/freebsd.h (SUBALIGN_LOG): New.
7278         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7279         max skip conditionally.
7280         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7281         max skip conditionally.
7282         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7283         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7284         max skip conditionally.
7285         * config/i386/i386.c (struct ptt): Change type from int to
7286         string.
7287         (ix86_default_align): Set default values.
7288         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7289         max skip conditionally.
7290         * config/i386/iamcu.h (SUBALIGN_LOG): New.
7291         (ASM_OUTPUT_MAX_SKIP_ALIGN):
7292         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7293         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7294         max skip conditionally.
7295         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7296         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7297         * config/i386/x86-64.h (SUBALIGN_LOG): New.
7298         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7299         max skip conditionally.
7300         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7301         * config/ia64/ia64.c (ia64_option_override): Set default values
7302         for alignment options.
7303         * config/m68k/m68k.c: Handle new str_align_* options.
7304         * config/mips/mips.c (mips_set_compression_mode): Change
7305         type of constants.
7306         (mips_option_override): Set default values for options.
7307         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7308         Likewise.
7309         * config/rs6000/rs6000.c (rs6000_option_override_internal):
7310         Likewise.
7311         * config/rx/rx.c (rx_option_override): Likewise.
7312         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7313         (LABEL_ALIGN): Use align_labels_log.
7314         (LOOP_ALIGN): Use align_loops_align.
7315         * config/s390/s390.c (s390_asm_output_function_label): Use new
7316         macros.
7317         * config/sh/sh.c (sh_override_options_after_change):
7318         Change type of constants.
7319         * config/spu/spu.c (spu_sched_init): Likewise.
7320         * config/sparc/sparc.c (sparc_option_override): Set default
7321         values for options.
7322         * config/visium/visium.c (visium_option_override): Likewise.
7323         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7324         emit p2align format with last argument if it's not needed.
7325         * doc/invoke.texi: Document extended format of -falign-*.
7326         * final.c: Use align_labels alignment.
7327         * flags.h (struct target_flag_state): Change type to use
7328         align_flags.
7329         (struct align_flags_tuple): New.
7330         (struct align_flags): Likewise.
7331         (align_loops_log): Redefine macro to use new types.
7332         (align_loops_max_skip): Redefine macro to use new types.
7333         (align_jumps_log): Redefine macro to use new types.
7334         (align_jumps_max_skip): Redefine macro to use new types.
7335         (align_labels_log): Redefine macro to use new types.
7336         (align_labels_max_skip): Redefine macro to use new types.
7337         (align_functions_log): Redefine macro to use new types.
7338         (align_loops): Redefine macro to use new types.
7339         (align_jumps): Redefine macro to use new types.
7340         (align_labels): Redefine macro to use new types.
7341         (align_functions): Redefine macro to use new types.
7342         (align_functions_max_skip): Redefine macro to use new types.
7343         (align_loops_value): New macro.
7344         (align_jumps_value): New macro.
7345         (align_labels_value): New macro.
7346         (align_functions_value): New macro.
7347         * function.c (invoke_set_current_function_hook): Propagate
7348         alignment values from flags to global variables default in
7349         topleev.h.
7350         * ipa-icf.c (sem_function::equals_wpa): Use
7351         cl_optimization_option_eq instead of memcmp.
7352         * lto-streamer.h (cl_optimization_stream_out): Support streaming
7353         of string types.
7354         (cl_optimization_stream_in): Likewise.
7355         * optc-save-gen.awk: Support strings in cl_optimization.
7356         * opth-gen.awk: Likewise.
7357         * opts.c (finish_options): Remove error checking of invalid
7358         value ranges.
7359         (MAX_CODE_ALIGN): Remove.
7360         (MAX_CODE_ALIGN_VALUE): Likewise.
7361         (parse_and_check_align_values): New function.
7362         (check_alignment_argument): Likewise.
7363         (common_handle_option): Use check_alignment_argument.
7364         * opts.h (parse_and_check_align_values): Declare.
7365         * toplev.c (init_alignments): Remove.
7366         (read_log_maxskip): New.
7367         (parse_N_M): Likewise.
7368         (parse_alignment_opts): Likewise.
7369         (backend_init_target): Remove usage of init_alignments.
7370         * toplev.h (parse_alignment_opts): Declare.
7371         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7372         argument.
7373         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7374         * tree.c (cl_option_hasher::equal): New.
7375         * varasm.c: Use new global macros.
7377 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
7379         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7380         Use a simpler align directive also if MAXSKIP = ALIGN-1.
7381         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7382         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7383         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7384         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7385         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
7386         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
7387         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
7388         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7389         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7390         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7391         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7393 2018-07-04  Martin Liska  <mliska@suse.cz>
7394             Jonathan Wakely  <jwakely@redhat.com>
7396         * coverage.c: Use correct type.
7397         * doc/invoke.texi: Language correction.
7399 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
7401         PR target/85620
7402         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
7403         ENDBRANCH for non-tail call which may return via indirect branch.
7404         * doc/extend.texi: Document indirect_return attribute.
7406 2018-07-03  Martin Sebor  <msebor@redhat.com>
7408         PR tree-optimization/86274
7409         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
7410         precondition.
7411         (format_floating): Correct handling of infinities and NaNs.
7413 2018-07-03  Martin Sebor  <msebor@redhat.com>
7415         * print-tree.c (print_real_cst): New function.
7416         (print_node_brief): Call it.
7417         (print_node): Ditto.
7419 2018-07-03  Jeff Law  <law@redhat.com>
7421         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
7422         into a single pattern.
7424         * config/h8300/h8300.md (ors code_iterator): New.
7425         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
7426         a single pattern and single splitter.
7427         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
7428         (iorqi3_1, xorqi3_1): Likewise.
7429         (iorqi3, xorqi3 expanders): Similarly.
7431         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
7432         (movmd_internal) into a single pattern using the P mode iterator.
7433         (movmd splitters): Similarly.
7434         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
7435         (movsd splitters): Similarly.
7437         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
7438         ADDB, ADDW and ADDL into a single ADD attribute which selects the
7439         right table based on the size of the operand.
7440         * config/h8300/h8300.md (length_table): Corresponding changes. All
7441         references to "addb", "addw" and "addl" changed to "add".
7442         (btst patterns): Merge two variants into a single pattern.
7443         (tstqi, tsthi): Likewise.
7444         (addhi3_incdec, addsi3_incdec): Likewise.
7445         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
7446         (mulhi3, mulsi3): Likewise.
7447         (udivhi3, udivsi3): Likewise.
7448         (divhi3, divsi3): Likewise.
7449         (andorqi3, andorhi3, andorsi3): Likewise.
7451 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
7453         PR target/85694
7454         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7455         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7457 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7459         PR tree-optimization/85694
7460         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7461         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7462         UNSPEC_URHADD.
7463         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7464         (<u>avg<mode>3_ceil): New patterns.
7466 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
7468         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7469         scan-tree-dump directive.
7470         * gcc.dg/vect/slp-perm-2.c: Likewise.
7471         * gcc.dg/vect/slp-perm-3.c: Likewise.
7472         * gcc.dg/vect/slp-perm-5.c: Likewise.
7473         * gcc.dg/vect/slp-perm-6.c: Likewise.
7474         * gcc.dg/vect/slp-perm-7.c: Likewise.
7475         * gcc.dg/vect/slp-perm-8.c: Likewise.
7477 2018-07-03  Marek Polacek  <polacek@redhat.com>
7479         PR middle-end/86202
7480         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7482 2018-07-03  Richard Biener  <rguenther@suse.de>
7484         PR ipa/86389
7485         * tree-ssa-structalias.c (find_func_clobbers): Properly
7486         handle indirect calls.
7488 2018-07-03  Jeff Law  <law@redhat.com>
7490         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7491         (shifts): New code iterator.
7492         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7493         expander.  Fix HImode handling on H8/SX.
7494         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7495         (subqi3, subhi3, subsi3 expanders): Likewise.
7496         (andqi3, andhi3, andsi3 expanders): Likewise.
7497         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7498         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7499         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7500         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7501         (zero_extendqihi2, zero_extendqisi2): Likewise.
7502         (extendqihi2, extendqisi2): Likewise.
7503         (rotlqi3, rotlhi3, rotlsi3): Likewise.
7504         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7505         (rotlqi3_1, rotlhi3_1): Likewise.
7506         (logicalhi3_sn, logicalsi3_sn): Likewise.
7507         (logicalhi3, logicalsi3): Likewise.
7509 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7511         * tree-vect-patterns.c (vect_recog_rotate_pattern)
7512         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7513         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7514         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7515         type to append_pattern_def_seq instead of creating a stmt_vec_info
7516         directly.
7517         (build_mask_conversion): Likewise.  Remove vinfo argument.
7518         (vect_add_conversion_to_patterm): Likewise, renaming to...
7519         (vect_add_conversion_to_pattern): ...this.
7520         (vect_recog_mask_conversion_pattern): Update call to
7521         build_mask_conversion.  Pass the vector type to
7522         append_pattern_def_seq here too.
7523         (vect_recog_gather_scatter_pattern): Update call to
7524         vect_add_conversion_to_pattern.
7526 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7528         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7529         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7530         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7531         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7532         STMT_VINFO_PATTERN_DEF_SEQ to null here.
7533         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7534         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7535         append_pattern_def_seq instead of new_pattern_def_seq.
7536         (vect_recog_divmod_pattern): Do both of the above.
7537         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7538         is null.
7540 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7542         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7543         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7544         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7545         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7546         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7547         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7548         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7549         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7550         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7551         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7552         parameter with a single stmt_vec_info.
7553         (vect_recog_func_ptr): Likewise.
7554         (vect_recog_gather_scatter_pattern): Likewise, folding in...
7555         (vect_try_gather_scatter_pattern): ...this.
7556         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7557         the stmt_vec_info of the statement to be matched.  Don't clear
7558         STMT_VINFO_RELATED_STMT.
7559         (vect_pattern_recog): Update call accordingly.
7561 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7563         PR tree-optimization/85694
7564         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7565         (uavgM3_ceil): Document new optabs.
7566         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7567         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7568         functions.
7569         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7570         (savg_ceil_optab): New optabs.
7571         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7572         (vect_vect_recog_func_ptrs): Add it.
7573         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7574         constant directly from the associated lhs.
7576 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7578         * tree-vect-patterns.c (vect_split_statement): New function.
7579         (vect_convert_input): Use it to try to split an existing cast.
7581 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7583         * poly-int.h (print_hex): New function.
7584         * dumpfile.h (dump_dec, dump_hex): Declare.
7585         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7586         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7587         min_input_precision, operation_precision and operation_sign.
7588         * tree-vect-patterns.c (vect_get_range_info): New function.
7589         (vect_same_loop_or_bb_p, vect_single_imm_use)
7590         (vect_operation_fits_smaller_type): Delete.
7591         (vect_look_through_possible_promotion): Add an optional
7592         single_use_p parameter.
7593         (vect_recog_over_widening_pattern): Rewrite to use new
7594         stmt_vec_info infomration.  Handle one operation at a time.
7595         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7596         (vect_truncatable_operation_p, vect_set_operation_type)
7597         (vect_set_min_input_precision): New functions.
7598         (vect_determine_min_output_precision_1): Likewise.
7599         (vect_determine_min_output_precision): Likewise.
7600         (vect_determine_precisions_from_range): Likewise.
7601         (vect_determine_precisions_from_users): Likewise.
7602         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7603         (vect_vect_recog_func_ptrs): Put over_widening first.
7604         Add cast_forwprop.
7605         (vect_pattern_recog): Call vect_determine_precisions.
7607 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7609         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7610         statements that have been replaced by further pattern statements.
7611         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7613 2018-07-03  Richard Biener  <rguenther@suse.de>
7615         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7616         always set *dt.  Dump vectype in vectype overload.
7617         * dumpfile.h (dump_gimple_expr): New function.
7618         (dump_gimple_expr_loc): Likewise.
7619         * dumpfile.c (dump_gimple_expr): New function.
7620         (dump_gimple_expr_loc): Likewise.
7622 2018-07-02  Jeff Law  <law@redhat.com>
7624         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7625         the H8/300, H8/300H and H8/S variants into a single pattern.
7626         (movhi_h8300, movqi_h8300hs): Similarly.
7627         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7628         (QHI mode iterator): New.
7630         * config/h8300/h8300.md: Remove trailing whitespace.
7632 2018-07-02  Jim Wilson  <jimw@sifive.com>
7634         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7635         instead of emit_insn for interrupt returns.
7636         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7637         (riscv_sret, riscv_uret): Likewise.
7639 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
7641         * pretty-print.c (selftest::test_pp_format): Move save and restore
7642         of quotes to class auto_fix_quotes, and add an instance.
7643         * selftest.c: Include "intl.h".
7644         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7645         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7646         * selftest.h (selftest::auto_fix_quotes): New class.
7648 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
7650         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7651         (aarch64_sve_prepare_conditional_op): Remove.
7652         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7653         Allow aarch64_simd_reg_or_zero as select operand; remove
7654         the aarch64_sve_prepare_conditional_op call.
7655         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7656         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7657         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7658         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7659         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7660         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7661         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7662         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7663         and a splitters to match all of the *_any patterns.
7664         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7666         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7667         (SVE_COND_FP_BINARY_REV): Remove.
7668         (sve_int_op_rev, sve_fp_op_rev): New.
7669         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7670         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7671         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7672         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7673         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7674         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7675         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7676         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7677         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7679         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7680         Remove match_dup 1 from the inner unspec.
7681         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7683         * config/aarch64/aarch64.md (movprfx): New attr.
7684         (length): Default movprfx to 8.
7685         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7686         (*madd<SVE_I>, *msub<SVE_I): Likewise.
7687         (*<su>mul<SVE_I>3_highpart): Likewise.
7688         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7689         (*v<ASHIFT><SVE_I>3): Likewise.
7690         (*<su><MAXMIN><SVE_I>3): Likewise.
7691         (*<su><MAXMIN><SVE_F>3): Likewise.
7692         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7693         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7694         (*div<SVE_F>4): Likewise.
7696 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
7698         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7699         in dump string.
7701 2018-07-02  Richard Biener  <rguenther@suse.de>
7703         PR tree-optimization/86363
7704         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7705         memset argument refers to a non-variable address.
7707 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
7709         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7710         {PLUS,MINUS}_EXPR code to...
7711         (adjust_symbolic_bound): ...here,
7712         (combine_bound): ...here,
7713         (set_value_range_with_overflow): ...and here.
7715 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
7717         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7718         code...
7719         (extract_range_from_abs_expr): ...here.
7721 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
7723         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7724         -fno-omit-frame-pointer when not optimizing.
7726 2018-07-02  Martin Liska  <mliska@suse.cz>
7728         PR ipa/86279
7729         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7730         (propagate_nothrow): Likewise.
7732 2018-07-02  Martin Liska  <mliska@suse.cz>
7734         PR ipa/86323
7735         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7737 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
7739         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7740         function in r262149, changing "loc" param from source_location to
7741         const dump_location_t &.
7742         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7743         declaration, as above.
7745 2018-07-01  Paul Koning  <ni1d@arrl.net>
7747         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7748         -munit-asm, -mgnu-asm, -mdec-asm.
7749         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7750         (pdp11_output_labelref): New.
7751         (pdp11_output_def): New.
7752         (pdp11_output_addr_vec_elt): New.
7753         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
7754         %# and %@ format codes.
7755         (pdp11_option_override): New.
7756         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7757         (pdp11_output_ident): New.
7758         (pdp11_asm_named_section): New.
7759         (pdp11_asm_init_sections): New.
7760         (pdp11_file_start): New.
7761         (pdp11_file_end): New.
7762         (output_ascii): Use .ascii/.asciz for -mdec-asm.
7763         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
7764         %o, like %c but octal.
7765         (pdp11_option_override): New.
7766         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7767         -mdec-asm.
7768         (DATA_SECTION_ASM_OP): Ditto.
7769         (READONLY_DATA_SECTION_ASM_OP): New.
7770         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7771         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7772         (ASM_OUTPUT_LABELREF): Ditto.
7773         (ASM_OUTPUT_DEF): Ditto.
7774         (ASM_OUTPUT_EXTERNAL): New.
7775         (ASM_OUTPUT_SOURCE_FILENAME): New.
7776         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7777         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7778         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
7779         %# and %@ format codes.
7780         * config/pdp11/pdp11.opt (mgnu-asm): New.
7781         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7782         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7783         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7785 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
7787         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7788         dereferencing path[] beyond its length.
7789         (debug_path): New.
7790         (debug_all_paths): New.
7791         (rewire_first_differing_edge): New.
7792         (adjust_paths_after_duplication): New.
7793         (duplicate_thread_path): Call adjust_paths_after_duplication.
7794         Add new argument.
7795         (thread_through_all_blocks): Add new argument to
7796         duplicate_thread_path.
7798 2018-06-30  Jim Wilson  <jimw@sifive.com>
7800         * config/riscv/predicates.md (p2m1_shift_operand): New.
7801         (high_mask_shift_operand): New.
7802         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7803         pattern using p2m1_shift_operand.
7804         (lshsi3_zero_extend_3+2): New combiner pattern using
7805         high_mask_shift_operand.
7807 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7809         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7810         split out from...
7811         (vect_recog_rotate_pattern): ...here.
7812         (vect_convert_input): Try to insert casts of invariants in the
7813         preheader.
7814         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7815         preheader to be empty.
7817 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7819         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7820         vector type.  If given, install it in the new statement's
7821         STMT_VINFO_VECTYPE.
7822         (vect_element_precision): New function.
7823         (vect_unpromoted_value): New struct.
7824         (vect_unpromoted_value::vect_unpromoted_value): New function.
7825         (vect_unpromoted_value::set_op): Likewise.
7826         (vect_look_through_possible_promotion): Likewise.
7827         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7828         (vect_widened_op_tree, vect_convert_input): Likewise.
7829         (vect_convert_inputs, vect_convert_output): Likewise.
7830         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7831         to handle the optional cast of the multiplication result and
7832         vect_widened_op_tree to detect the widened multiplication itself.
7833         Do not require the input and output of promotion casts to have
7834         the same sign, but base the signedness of the operation on the
7835         input rather than the result.  If the pattern includes two
7836         promotions, check that those promotions have the same sign.
7837         Do not restrict the MULT_EXPR handling to a double-width result;
7838         handle quadruple-width results and wider.  Use vect_convert_inputs
7839         to convert the inputs to the common type.
7840         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
7841         to handle the optional cast of the ABS result.  Also allow a sign
7842         change or a sign extension between the ABS and MINUS.
7843         Use vect_widened_op_tree to detect the widened subtraction and use
7844         vect_convert_inputs to convert the inputs to the common type.
7845         (vect_handle_widen_op_by_const): Delete.
7846         (vect_recog_widen_op_pattern): New function.
7847         (vect_recog_widen_mult_pattern): Use it.
7848         (vect_recog_widen_shift_pattern): Likewise.
7849         (vect_recog_widen_sum_pattern): Use
7850         vect_look_through_possible_promotion to handle the promoted
7851         PLUS_EXPR operand.
7853 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7855         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7856         the containing gimple_seq *.
7857         * gimple-iterator.h (gsi_for_stmt): Declare it.
7858         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7859         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7860         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7861         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7862         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7863         checks.
7864         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7865         split out from...
7866         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
7867         statement being replaced is part of an existing pattern
7868         definition sequence, inserting the new pattern statements before
7869         the original one.
7870         (vect_pattern_recog_1): Don't return a bool.  If the statement
7871         is already part of a pattern, instead apply pattern matching
7872         to the pattern definition statements.  Don't clear the
7873         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7874         (vect_pattern_recog): Don't break after the first match;
7875         continue processing the pattern definition statements instead.
7876         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7878 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7880         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7881         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7882         (vect_recog_widen_sum_pattern): Use it.
7884 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7886         * tree-vect-loop.c (vectorizable_reduction): Assert that the
7887         phi is not a pattern statement and has not been replaced by
7888         a pattern statement.
7889         * tree-vect-patterns.c (type_conversion_p): Don't check
7890         STMT_VINFO_IN_PATTERN_P.
7891         (vect_recog_vector_vector_shift_pattern): Likewise.
7892         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7893         the pattern statement rather than the original statement; check
7894         directly for a WIDEN_MULT_EXPR here.
7895         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7896         vect_is_simple_use to return the pattern statement rather
7897         than the original statement; use is_pattern_stmt_p to check
7898         for such a pattern statement.
7899         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7900         to return the pattern statement rather than the original statement;
7901         don't do the same transformation here.
7902         (vect_is_simple_use): If the defining statement has been replaced
7903         by a pattern statement, return the pattern statement instead.
7904         Remove the corresponding (local) transformation from the vectype
7905         overload.
7907 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7909         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7910         end and default to null.
7911         * tree-vect-loop.c (vect_create_epilog_for_reduction)
7912         (vectorizable_reduction): Update calls accordingly, dropping the
7913         gimple ** argument if the passed-back statement isn't needed.
7914         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7915         (vect_recog_rotate_pattern): Likewise.
7916         (vect_recog_mask_conversion_pattern): Likewise.
7917         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7918         (vect_mask_constant_operand_p): Likewise.
7919         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7920         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7921         (get_group_load_store_type, get_load_store_type): Likewise.
7922         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7923         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7924         (vectorizable_conversion, vectorizable_assignment): Likewise.
7925         (vectorizable_shift, vectorizable_operation): Likewise.
7926         (vectorizable_store, vect_is_simple_cond): Likewise.
7927         (vectorizable_condition, vectorizable_comparison): Likewise.
7928         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7929         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7930         and move it to the end.  Cope with null def_stmt_outs.
7932 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7934         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7936 2018-06-29  Jeff Law  <law@redhat.com>
7938         * config/v850/v850.c (v850_legitimate_address_p): Handle large
7939         displacements for TARGET_V850E2V3 and newer.
7940         (TARGET_LRA_P): Remove.  Defaults to LRA now.
7941         * config/v850/v850.md (sign23byte_load): Remove.
7942         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7943         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7945 2018-06-29  Martin Liska  <mliska@suse.cz>
7947         PR lto/85759
7948         * coverage.c (coverage_init): Mangle full path name.
7949         * doc/invoke.texi: Document the change.
7950         * gcov-io.c (mangle_path): New.
7951         * gcov-io.h (mangle_path): Likewise.
7952         * gcov.c (mangle_name): Use mangle_path for path mangling.
7954 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7956         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7957         if starting source register is not even.
7959 2018-06-29  Martin Liska  <mliska@suse.cz>
7961         PR tree-optimization/86263
7962         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7963         Make edge redirection.
7965 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
7967         * dumpfile.c (dump_loc): Add indentation based on scope depth.
7968         (dump_scope_depth): New variable.
7969         (get_dump_scope_depth): New function.
7970         (dump_begin_scope): New function.
7971         (dump_end_scope): New function.
7972         * dumpfile.h (get_dump_scope_depth): New declaration.
7973         (dump_begin_scope): New declaration.
7974         (dump_end_scope): New declaration.
7975         (class auto_dump_scope): New class.
7976         (AUTO_DUMP_SCOPE): New macro.
7977         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7978         AUTO_DUMP_SCOPE.
7980 2018-06-29  Richard Biener  <rguenther@suse.de>
7982         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7983         compute_all_dependences succeeds.
7984         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7985         exceed --param loop-max-datarefs-for-datadeps.
7987 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
7989         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7991 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
7993         PR target/86348
7994         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7995         alternative 0 in preferred_for_speed attribute.
7997 2018-06-28  Paul Koning  <ni1d@arrl.net>
7999         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
8000         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
8001         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
8002         * config/pdp11/pdp11.md: Correct "length" attribute calculation
8003         for shift insn patterns.
8005 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
8007         * cgraph.c (cgraph_node::get_body): Replace assignments to
8008         "dump_file" with calls to set_dump_file.
8009         * dumpfile.c (alt_dump_file): Make static, and group with...
8010         (alt_flags): ...this definition.
8011         (dumps_are_enabled): New variable.
8012         (refresh_dumps_are_enabled): New function.
8013         (set_dump_file): New function.
8014         (set_alt_dump_file): New function.
8015         (gcc::dump_manager::dump_start): Replace assignments to
8016         "dump_file" and "alt_dump_file" with calls to set_dump_file and
8017         set_alt_dump_file.
8018         (gcc::dump_manager::dump_finish): Likewise.
8019         * dumpfile.h (alt_dump_file): Delete decl.
8020         (dumps_are_enabled): New variable decl.
8021         (set_dump_file): New function decl.
8022         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
8023         global.
8024         * tree-nested.c (lower_nested_functions): Replace assignments to
8025         "dump_file" with calls to set_dump_file.
8027 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
8029         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
8030         goto_locus of each outgoing edge of each basic block.
8032 2018-06-28  Richard Biener  <rguenther@suse.de>
8034         * dwarf2out.c (decl_scope_table): Remove.
8035         (push_decl_scope): Likewise.
8036         (pop_decl_scope): Likewise.
8037         (gen_type_die_for_member): Do not call push/pop_decl_scope.
8038         (gen_struct_or_union_type_die): Likewise.
8039         (gen_tagged_type_die): Likewise.
8040         (dwarf2out_init): Do not initialize decl_scope_table.
8041         (dwarf2out_c_finalize): Do not free it.
8043 2018-06-28  Richard Biener  <rguenther@suse.de>
8045         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
8046         deciding whether to not re-use a DIE.
8048 2018-06-28  Richard Biener  <rguenther@suse.de>
8050         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
8051         DW_AT_abstract_origin attribute.
8053 2018-06-28  Martin Liska  <mliska@suse.cz>
8055         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8056         Use newly introduced constants.
8057         * tree-switch-conversion.h (struct jump_table_cluster):
8058         Define max_ratio_for_size and max_ratio_for_speed.
8060 2018-06-28  Martin Liska  <mliska@suse.cz>
8062         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8063         Add new checking assert to catch invalid state.
8064         (jump_table_cluster::can_be_handled): Handle single case
8065         clusters.
8066         (jump_table_cluster::is_beneficial): Bail out for such case.
8067         (bit_test_cluster::find_bit_tests):
8068         Add new checking assert to catch invalid state.
8069         (bit_test_cluster::can_be_handled): Handle single case
8070         clusters.
8071         (bit_test_cluster::is_beneficial): Bail out for such case.
8072         (switch_decision_tree::analyze_switch_statement):
8073         Fix comment.
8075 2018-06-28  Martin Liska  <mliska@suse.cz>
8077         * common.opt: Introduce -completion option.
8078         * gcc.c (driver_handle_option): Handle it.
8079         (driver::main): Print completions if completion
8080         is set.
8081         * opt-suggestions.c (option_proposer::get_completions):
8082         New function.
8083         (option_proposer::suggest_completion): Likewise.
8084         (option_proposer::find_param_completions): Likewise.
8085         (verify_autocompletions): Likewise.
8086         (test_completion_valid_options): Likewise.
8087         (test_completion_valid_params): Likewise.
8088         (in_completion_p): Likewise.
8089         (empty_completion_p): Likewise.
8090         (test_completion_partial_match): Likewise.
8091         (test_completion_garbage): Likewise.
8092         (opt_proposer_c_tests): Likewise.
8093         * opt-suggestions.h: Declare new functions.
8094         * opts.c (common_handle_option): Handle OPT__completion_.
8095         * selftest-run-tests.c (selftest::run_tests): Add
8096         opt_proposer_c_tests.
8097         * selftest.c (assert_str_startswith): New.
8098         * selftest.h (assert_str_startswith): Likewise.
8099         (opt_proposer_c_tests): New.
8100         (ASSERT_STR_STARTSWITH): Likewise.
8102 2018-06-28  Martin Liska  <mliska@suse.cz>
8104         * Makefile.in: Add opt-suggestions.o.
8105         * gcc-main.c: Include opt-suggestions.h.
8106         * gcc.c (driver::driver): Likewise.
8107         (driver::~driver): Remove m_option_suggestions.
8108         (driver::build_option_suggestions): Moved to option_proposer.
8109         (driver::suggest_option): Likewise.
8110         (driver::handle_unrecognized_options): Use option_proposer.
8111         * gcc.h (class driver): Add new memver m_option_proposer.
8112         * opt-suggestions.c: New file.
8113         * opt-suggestions.h: New file.
8115 2018-06-28  Martin Liska  <mliska@suse.cz>
8117         * vec.h (class auto_string_vec): New (moved from auto_argvec).
8118         (auto_string_vec::~auto_string_vec): Likewise.
8120 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
8122         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
8123         prevent_decl_creation_for_types fields up and add reset_location field.
8124         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
8125         statement if id->reset_location is true.
8126         (copy_edges_for_bb): Do not set goto_locus on the new edges if
8127         id->reset_location is true.
8128         (copy_phis_for_bb): Force input_location on the arguments if
8129         id->reset_location is true.
8130         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
8131         is set on the function to be inlined.
8133 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
8135         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
8137 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
8139         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
8140         registers for Pmode.
8141         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
8142         hard registers for the clobbered pseudo.
8144 2018-06-27  Paul Koning  <ni1d@arrl.net>
8146         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8147         mutually exclusive options.
8148         * config/pdp11/constraints.md (h): New constraint.
8149         (O): Update definition to match shift code generation.
8150         (D): New constraint.
8151         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
8152         (CCFP): Remove.
8153         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
8154         function.
8155         (output_jump): Change arguments.
8156         (pdp11_fixed_cc_regs): New function.
8157         (pdp11_cc_mode): Ditto.
8158         (pdp11_expand_shift): Ditto.
8159         (pdp11_assemble_shift): Ditto.
8160         (pdp11_small_shift): Ditto.
8161         (pdp11_branch_cost): Remove.
8162         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
8163         from output.
8164         (pdp11_register_move_cost): Update for CC registers.
8165         (pdp11_rtx_costs): Add case for LSHIFTRT.
8166         (pdp11_output_jump): Add CCNZ mode conditional branches.
8167         (notice_update_cc_on_set): Remove.
8168         (pdp11_cc_mode): New function.
8169         (simple_memory_operand): Correct pre/post decrement case.
8170         (no_side_effect_operand): New function.
8171         (pdp11_regno_reg_class): Add CC_REGS class.
8172         (pdp11_fixed_cc_regs): New function.
8173         (pdp11_small_shift): New function.
8174         (pdp11_expand_shift): New function to expand shift insns.
8175         (pdp11_assemble_shift): New function to output shifts.
8176         (pdp11_branch_cost): Remove.
8177         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
8178         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
8179         (WCHAR_TYPE): Ditto.
8180         (PTRDIFF_TYPE): Ditto.
8181         (ADJUST_INSN_LENGTH): New macro.
8182         (FIXED_REGISTERS): Add CC registers.
8183         (CALL_USED_REGISTERS): Ditto.
8184         (reg_class): Ditto.
8185         (REG_CLASS_NAMES): Ditto.
8186         (REG_CLASS_CONTENTS): Ditto.
8187         (SELECT_CC_MODE): Use new function.
8188         (TARGET_FLAGS_REGNUM): New macro.
8189         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8190         (cc0_reg_rtx): Remove.
8191         (CC_STATUS_MDEP): Remove.
8192         (CC_STATUS_MDEFP_INIT): Remove.
8193         (CC_IN_FPU): Remove.
8194         (NOTICE_UPDATE_CC): Remove.
8195         (REGISTER_NAMES): Add CC registers.
8196         (BRANCH_COST): Change to constant 1.
8197         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8198         handling.
8199         * config/pdp11/pdp11.opt (mbcopy): Remove.
8200         (mbcopy-builtin): Remove.
8201         (mbranch-cheap): Remove.
8202         (mbranch-expensive): Remove.
8203         * config/pdp11/predicates.md (expand_shift_operand): Update to
8204         match shift code generation.
8205         (ccnz_operator): New predicate.
8206         * doc/invoke.texi (PDP-11 Options): Remove deleted options
8207         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8208         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
8209         exclusive options.
8210         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
8211         description of O constraint.
8213 2018-06-27  Jeff Law  <law@redhat.com>
8214             Austin Law  <austinklaw@gmail.com>
8216         * config/v850/v850.md (addsi3_set_flags): New pattern.
8217         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8218         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8219         (zero_extendhisi2_v850_set_flags): Likewise.
8220         (zero_extendqisi2_v850_set_flags): Likewise.
8221         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8222         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8223         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8225         * config/v850/v850-protos.h (notice_update_cc): Remove.
8226         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8227         (v850_print_operand): Handle 'D' and "d".
8228         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8229         Add handling of arithmetic/logical operations compared against zero.
8230         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8231         Do not look at v850_compare_op, instead get mode from last argument.
8232         (v850_gen_compare): Remove
8233         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8234         after reload for prologue insns.
8235         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8236         patterns.
8237         (construct_save_jarl): Likewise.
8238         (TARGET_FLAGS_REGNUM): Define.
8239         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8240         (NOTICE_UPDATE_CC): Remove.
8241         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8242         than cc0.  Conditionalize on reload_completed.
8243         (cmpsi_insn, setfcc_insn): Likewise.
8244         (tst1 splitter): Turn into define_and_split which sets the flags
8245         after reload.
8246         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8247         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8248         (cstoresf4, cstoredf4): Clobber the flags.
8249         (cmpsi, cmpsf, cmpdf): Remove expanders.
8250         (setf_insn): Remove pattern.
8251         (addsi3): Turn into define_and_split which clobbers the flags after
8252         reload and a suitable pattern (addsi3_clobber_flags) for use after
8253         reload.
8254         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8255         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8256         (ashrsi3, ashrsi3_v850e2): Likewise.
8257         (bins): Clobber the flags.
8258         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8259         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8260         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8261         (call_value_internal_short, call_value_internal_long): Likewise.
8262         (callt_save_interrupt, callt_return_interrupt): Likewise.
8263         (save_interrupt, return_interrupt): Likewise.
8264         (callt_save_all_interrupt, save_all_interrupt): Likewise.
8265         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8266         (restore_all_interrupt, _restore_all_interrupt): Likewise.
8267         (All FP comparisons): Only allow after reload has completed.
8268         (trfsr): Likewise.
8269         (divh, divhu): Tweak output template.
8270         (branch_z_normal, branch_z_invert): Remove
8271         (branch_nz_normal, branch_nz_invert): Likewise.
8272         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8274         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8275         * config/v850/v850.c (notice_update_cc): Remove.
8276         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
8277         (CC_NO_CARRY): Likewise.
8278         (NOTICE_UPDATE_CC): Define to nothing.
8279         * config/v850/v850.md: Remove block comment on cc0 handling
8280         Remove "cc" attribute from all patterns.  Remove cc_status handling
8281         from all patterns.  Minor formatting fixes.
8283 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8285         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8286         (cortex-a76.cortex-a55): Likewise.
8287         * config/aarch64/aarch64-tune.md: Regenerate.
8288         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8289         cortex-a76.cortex-a55.
8291 2018-06-27  Jeff Law  <law@redhat.com>
8293         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8294         (MULTILIB_DIRNAMES): Similarly.
8296 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
8298         * gimple.h (gimple_return_retbnd): Delete.
8299         (gimple_return_set_retbnd): Likewise.
8300         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8301         gimple_return_set_retbnd.
8302         * gimple-pretty-print.c (dump_gimple_return): Remove call to
8303         gimple_return_retbnd and adjust.
8304         * tree-inline.h (struct copy_body_data): Remove retbnd field.
8305         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8306         Explicitly return NULL in a couple more cases.  Move assertion
8307         on debug statements and remove unreachable code.
8308         (reset_debug_binding): Do not test id->retbnd.
8309         (expand_call_inline): Do not set it.
8311 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8313         * configure.ac: Add --disable-gcov option.
8314         * configure: Regenerate.
8315         * Makefile.in: Honour @enable_gcov@.
8316         * doc/install.texi: Document --disable-gcov.
8318 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8320         * config/arm/arm-cpus.in (cortex-a76): New entry.
8321         (cortex-a76.cortex-a55): Likewise.
8322         * config/arm/arm-tables.opt: Regenerate.
8323         * config/arm/arm-tune.md: Likewise.
8324         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8325         * doc/invoke.texi (ARM Options): Document cortex-a76 and
8326         cortex-a76.cortex-a55.
8328 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
8330         PR target/85769
8331         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8333 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
8335         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8336         comment.
8337         (EPILOGUE_USES): Likewise.
8339 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
8341         * tree-inline.c (remap_location): New function extracted from...
8342         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
8343         (copy_phis_for_bb): ...here.  Call remap_location.
8344         (copy_cfg_body): Adjust call to copy_edges_for_bb.
8346 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
8348         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8349         unaligned vsx for 16B memset.
8351 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
8353         PR target/86285
8354         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8355         ieee128_float_type_node to long_double_type_node unless
8356         TARGET_LONG_DOUBLE_128 is set.
8358 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
8360         * cfgloop.c (get_loop_location): Convert return type from
8361         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8362         by implicit construction from rtx_insn *, and using
8363         dump_user_location_t::from_function_decl for the fallback case.
8364         * cfgloop.h (get_loop_location): Convert return type from
8365         location_t to dump_user_location_t.
8366         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8367         dump_printf_loc to pass in a dump_location_t rather than a
8368         location_t, via the gimple stmt.
8369         * coverage.c (get_coverage_counts): Update calls to
8370         dump_printf_loc to pass in dump_location_t rather than a
8371         location_t.
8372         * doc/optinfo.texi (Dump types): Convert example of
8373         dump_printf_loc from taking "locus" to taking "insn".  Update
8374         description of the "_loc" calls to cover dump_location_t.
8375         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8376         "selftest.h".
8377         (dump_user_location_t::dump_user_location_t): New constructors,
8378         from gimple *stmt and rtx_insn *.
8379         (dump_user_location_t::from_function_decl): New function.
8380         (dump_loc): Make static.
8381         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
8382         const dump_location_t &.
8383         (dump_generic_expr_loc): Delete.
8384         (dump_printf_loc): Convert param "loc" from location_t to
8385         const dump_location_t &.
8386         (selftest::test_impl_location): New function.
8387         (selftest::dumpfile_c_tests): New function.
8388         * dumpfile.h: Include "profile-count.h".
8389         (class dump_user_location_t): New class.
8390         (struct dump_impl_location_t): New struct.
8391         (class dump_location_t): New class.
8392         (dump_printf_loc): Convert 2nd param from source_location to
8393         const dump_location_t &.
8394         (dump_generic_expr_loc): Delete.
8395         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
8396         const dump_location_t &.
8397         * gimple-fold.c (fold_gimple_assign): Update call to
8398         dump_printf_loc to pass in a dump_location_t rather than a
8399         location_t, via the gimple stmt.
8400         (gimple_fold_call): Likewise.
8401         * gimple-loop-interchange.cc
8402         (loop_cand::analyze_iloop_reduction_var): Update for change to
8403         check_reduction_path.
8404         (tree_loop_interchange::interchange): Update for change to
8405         find_loop_location.
8406         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
8407         change in return-type of find_loop_location.
8408         (graphite_regenerate_ast_isl): Likewise.
8409         * graphite-optimize-isl.c (optimize_isl): Likewise.
8410         * graphite.c (graphite_transform_loops): Likewise.
8411         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
8412         pass in a dump_location_t rather than a location_t, via the
8413         gimple stmt.
8414         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8415         * ipa.c (walk_polymorphic_call_targets): Likewise.
8416         * loop-unroll.c (report_unroll): Convert "locus" param from
8417         location_t to dump_location_t.
8418         (decide_unrolling): Update for change to get_loop_location's
8419         return type.
8420         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
8421         location_t to dump_user_location_t.
8422         (grid_find_single_omp_among_assignments_1): Updates calls to
8423         dump_printf_loc to pass in a dump_location_t rather than a
8424         location_t, via the gimple stmt.
8425         (grid_parallel_clauses_gridifiable): Convert "tloc" from
8426         location_t to dump_location_t.  Updates calls to dump_printf_loc
8427         to pass in a dump_location_t rather than a location_t, via the
8428         gimple stmt.
8429         (grid_inner_loop_gridifiable_p): Likewise.
8430         (grid_dist_follows_simple_pattern): Likewise.
8431         (grid_gfor_follows_tiling_pattern): Likewise.
8432         (grid_target_follows_gridifiable_pattern): Likewise.
8433         (grid_attempt_target_gridification): Convert initialization
8434         of local "grid" from memset to zero-initialization; FIXME: does
8435         this require C++11?  Update call to dump_printf_loc to pass in a
8436         optinfo_location rather than a location_t, via the gimple stmt.
8437         * profile.c (read_profile_edge_counts): Updates call to
8438         dump_printf_loc to pass in a dump_location_t rather than a
8439         location_t
8440         (compute_branch_probabilities): Likewise.
8441         * selftest-run-tests.c (selftest::run_tests): Call
8442         dumpfile_c_tests.
8443         * selftest.h (dumpfile_c_tests): New decl.
8444         * tree-loop-distribution.c (pass_loop_distribution::execute):
8445         Update for change in return type of find_loop_location.
8446         * tree-parloops.c (parallelize_loops): Likewise.
8447         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
8448         "locus" from location_t to dump_user_location_t.
8449         (canonicalize_loop_induction_variables): Likewise.
8450         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8451         for change in return type of find_loop_location.
8452         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8453         to dump_printf_loc to pass in a dump_location_t rather than a
8454         location_t, via the stmt.
8455         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8456         Likewise.
8457         * tree-vect-loop-manip.c (find_loop_location): Convert return
8458         type from source_location to dump_user_location_t.
8459         (vect_do_peeling): Update for above change.
8460         (vect_loop_versioning): Update for change in type of
8461         vect_location.
8462         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8463         from location_t to dump_user_location_t.
8464         (vect_estimate_min_profitable_iters): Update for change in type
8465         of vect_location.
8466         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8467         location_t to dump_location_t.
8468         (vect_slp_bb): Update for change in type of vect_location.
8469         * tree-vectorizer.c (vect_location): Convert from source_location
8470         to dump_user_location_t.
8471         (try_vectorize_loop_1): Update for change in vect_location's type.
8472         (vectorize_loops): Likewise.
8473         (increase_alignment): Likewise.
8474         * tree-vectorizer.h (vect_location): Convert from source_location
8475         to dump_user_location_t.
8476         (find_loop_location): Convert return type from source_location to
8477         dump_user_location_t.
8478         (check_reduction_path): Convert 1st param from location_t to
8479         dump_user_location_t.
8480         * value-prof.c (check_counter): Update call to dump_printf_loc to
8481         pass in a dump_user_location_t rather than a location_t; update
8482         call to error_at for change in type of "locus".
8483         (check_ic_target): Update call to dump_printf_loc to
8484         pass in a dump_user_location_t rather than a location_t, via the
8485         call_stmt.
8487 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
8489         * config/s390/s390.h (enum processor_flags): Do not use
8490         default tune parameter when -march was specified.
8492 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
8494         PR target/86314
8495         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8496         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8497         operands.
8499 2018-06-26  Richard Biener  <rguenther@suse.de>
8501         PR tree-optimization/86287
8502         PR bootstrap/86316
8503         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8504         (vect_analyze_loop): Initialize n_stmts.
8506 2018-06-26  Richard Biener  <rguenther@suse.de>
8508         PR middle-end/86271
8509         * fold-const.c (fold_convertible_p): Pointer extension
8510         isn't valid.
8512 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
8514         PR debug/86064
8515         * dwarf2out.c (loc_list_has_views): Adjust comments.
8516         (dw_loc_list): Split single cross-partition range with
8517         nonzero locview.
8519 2018-06-25  Jeff Law  <law@redhat.com>
8521         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8522         on -mbig-switch by default.
8524         * config/v850/predicates.md (const_float_1_operand): Fix match_code
8525         test.
8526         (const_float_0_operand): Remove unused predicate.
8527         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8528         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8529         (recipsf2): New expander.  Original pattern now called
8530         (recipsf2_insn).
8531         (recipdf2, recipdf2_insn): Similarly.
8532         (rsqrtsf2, rsqrtsf2_insn): Similarly
8533         (rsqrtdf2, rsqrtdf2_insn): Similarly
8535 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
8537         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8538         Simplify logic for FreeBSD (twice).
8540 2018-06-25  Martin Sebor  <msebor@redhat.com>
8542         PR tree-optimization/86204
8543         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8544         a strnlen result if it's less than the length of the string.
8546 2018-06-25  Martin Sebor  <msebor@redhat.com>
8548         PR tree-optimization/85700
8549         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8550         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8551         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8553 2018-06-25  Martin Sebor  <msebor@redhat.com>
8555         * doc/extend.texi (Zero-length arrays): Update and clarify.
8557 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
8559         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8560         added IEEE/IBM long double multilib support on PowerPC little
8561         endian Linux systems.
8562         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8563         (MULTILIB_DEFAULTS): Likewise.
8564         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8565         Likewise.
8566         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8567         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8568         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8570 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
8572         PR middle-end/86311
8573         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8574         (REORDER_45): Likewise.
8576 2018-06-25  Jeff Law  <law@redhat.com>
8578         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8579         dividend to 32 bits.  Adjust length.
8580         (udivmodhi4): Cleanup output template.  Fix length.
8582 2018-06-25  Carl Love  <cel@us.ibm.com>
8584         * config/rs6000/vsx.md: Change word selector to prefered location.
8586 2018-06-25  Richard Biener  <rguenther@suse.de>
8588         PR tree-optimization/86304
8589         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8590         epilogue-if-converted loops as well.
8592 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
8594         * lto-section-out.c (lto_begin_section): Do not print section
8595         name for noaddr and unnumbered dumps.
8597 2018-06-25  Richard Biener  <rguenther@suse.de>
8599         * tree-vectorizer.h (struct vec_info_shared): New structure
8600         with parts split out from struct vec_info and loop_nest from
8601         struct _loop_vec_info.
8602         (struct vec_info): Adjust accordingly.
8603         (struct _loop_vec_info): Likewise.
8604         (LOOP_VINFO_LOOP_NEST): Adjust.
8605         (LOOP_VINFO_DATAREFS): Likewise.
8606         (LOOP_VINFO_DDRS): Likewise.
8607         (struct _bb_vec_info): Likewise.
8608         (BB_VINFO_DATAREFS): Likewise.
8609         (BB_VINFO_DDRS): Likewise.
8610         (struct _stmt_vec_info): Add dr_aux member.
8611         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8612         (DR_MISALIGNMENT_UNINITIALIZED): New.
8613         (set_dr_misalignment): Adjust.
8614         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8615         (vect_analyze_loop): Adjust prototype.
8616         (vect_analyze_loop_form): Likewise.
8617         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8618         Compute dependences lazily.
8619         (vect_record_base_alignments): Use shared datarefs/ddrs.
8620         (vect_verify_datarefs_alignment): Likewise.
8621         (vect_analyze_data_refs_alignment): Likewise.
8622         (vect_analyze_data_ref_accesses): Likewise.
8623         (vect_analyze_data_refs): Likewise.
8624         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8625         constructor parameter for shared part.
8626         (vect_analyze_loop_form): Pass in shared part and adjust.
8627         (vect_analyze_loop_2): Pass in storage for the number of
8628         stmts.  Move loop nest finding to the caller.  Compute
8629         datarefs lazily.
8630         (vect_analyze_loop): Pass in shared part.
8631         (vect_transform_loop): Verify shared datarefs are unchanged.
8632         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8633         constructor parameter for shared part.
8634         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8635         (vect_slp_bb): Verify shared datarefs are unchanged before
8636         transform.
8637         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8638         change.
8639         (new_stmt_vec_info): Initialize DR_AUX misalignment to
8640         DR_MISALIGNMENT_UNINITIALIZED.
8641         * tree-vectorizer.c (vec_info::vec_info): Add constructor
8642         parameter for shared part.
8643         (vec_info::~vec_info): Adjust.
8644         (vec_info_shared::vec_info_shared): New.
8645         (vec_info_shared::~vec_info_shared): Likewise.
8646         (vec_info_shared::save_datarefs): Likewise.
8647         (vec_info_shared::check_datarefs): Likewise.
8648         (try_vectorize_loop_1): Construct shared part live for analyses
8649         of a single loop for multiple vector sizes.
8650         * tree-parloops.c (gather_scalar_reductions): Adjust.
8652 2018-06-25  Richard Biener  <rguenther@suse.de>
8654         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8655         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8656         (vect_analyze_data_refs): Remove similar code from here and
8657         simplify accordingly.
8659 2018-06-25  Richard Biener  <rguenther@suse.de>
8661         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8662         for reverse storage order accesses rather than asserting
8663         they cannot happen here.
8665 2018-06-25  Tom de Vries  <tdevries@suse.de>
8667         PR debug/86257
8668         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8669         Use data16 instead of .byte for insn prefix.
8671 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
8673         PR C++/86082
8674         * parser.c (make_char_string_pack): Pass this literal chars
8675         through cpp_interpret_string.
8676         (cp_parser_userdef_numeric_literal): Check the result of
8677         make_char_string_pack.
8679 2018-06-24  Maya Rashish  <coypu@sdf.org>
8681         * ginclude/stddef.h: Simplify conditions around avoiding
8682         re-definition of __size_t.
8684 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8686         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8687         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8689 2018-06-22  Maya Rashish  <coypu@sdf.org>
8691         * doc/invoke.texi (mno-fancy-math-387): Update for changes
8692         made to OpenBSD and NetBSD through the years.
8694 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8696         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8697         behavior of vec_pack (vector double, vector double) to match
8698         behavior of vec_float2 (vector double, vector double).
8700 2018-06-22  Olivier Hainque  <hainque@adacore.com>
8702         * gimplify.c (gimplify_function_tree): Prevent creation
8703         of a trampoline for the address of the current function
8704         passed to entry/exit instrumentation hooks.
8706 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
8708         PR target/86222
8709         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8710         correctly.
8712 2018-06-22  Martin Liska  <mliska@suse.cz>
8714         PR tree-optimization/86263
8715         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8716         Bail out if is_enabled is false.
8717         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8718         New declaration.
8719         (jump_table_cluster::is_enabled): New function.
8721 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8723         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8724         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8725         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8726         (lto_input_ts_binfo_tree_pointers): Likewise.
8727         * tree-streamer-out.c (streamer_write_tree_bitfields,
8728         write_ts_binfo_tree_pointers): Likewise.
8729         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8731 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8733         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8735 2018-06-22  Martin Liska  <mliska@suse.cz>
8737         * symbol-summary.h (get): Make it pure and inline move
8738         functionality from ::get function.
8739         (get): Remove and inline into ::get and ::get_create.
8740         (get_create): Move code from ::get function.
8742 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8744         PR target/85994
8745         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8746         -x assembler-with-cpp.
8748 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8750         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8751         _FILE_OFFSET_BITS=64 for C++.
8753 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
8755         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8756         conversion insn that shows up when pr85657-3.c is compiled using
8757         IEEE 128-bit long double.
8758         (neg<mode>2_internal): Use the correct mode to check whether the
8759         mode is IBM extended.
8760         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8761         multiply and divide external functions from being created more
8762         than once.
8764 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8766         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8767         functions.
8768         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
8769         the edge can be forwarded.
8770         (cfg_layout_merge_blocks): Likewise.
8772 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8774         * except.c (finish_eh_generation): Commit edge insertions only after
8775         the EH edges have been redirected from post-landing to landing pads.
8777 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8779         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8780         create_tmp_var_for to create the FRAME decl.
8781         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8783 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8785         * tree-inline.c (copy_edges_for_bb): Minor tweak.
8786         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8787         debug statement when resetting its value.
8788         (expand_call_inline): Copy the locus of the call onto the assignment
8789         of the return value, if any.  Use local variable in more cases.
8791 2018-06-21  Martin Liska  <mliska@suse.cz>
8793         * ipa-pure-const.c (propagate_nothrow): Use
8794         funct_state_summaries->get.
8795         (dump_malloc_lattice): Likewise.
8796         (propagate_malloc): Likewise.
8798 2018-06-21  Richard Biener  <rguenther@suse.de>
8800         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8801         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8802         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8803         comment.
8804         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8805         BLOCK_ABSTRACT_ORIGIN unconditionally.
8807 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
8809         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8810         deleting it.
8811         * ipa-reference.c (ipa_reference_c_finalize): Delete
8812         ipa_ref_opt_sum_summaries and set it to NULL.
8814 2018-06-21  Tom de Vries  <tdevries@suse.de>
8816         PR tree-optimization/85859
8817         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8818         test with comment from bb_no_side_effects_p.
8820 2018-06-21  Richard Biener  <rguenther@suse.de>
8822         PR tree-optimization/86232
8823         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8824         max for constant niter.
8826 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8828         * config/aarch64/aarch64-simd.md
8829         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8831 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8833         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8834         Make opernads of the unspec commutative.
8836 2018-06-21  Richard Biener  <rguenther@suse.de>
8838         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8839         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8840         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8841         (vect_analyze_data_ref_dependence): Re-order checks to deal with
8842         NULL DR_STEP.
8843         (vect_record_base_alignments): Do not record base alignment
8844         for gathers or scatters.
8845         (vect_compute_data_ref_alignment): Drop return value that is always
8846         true.  Bail out early for gathers or scatters.
8847         (vect_enhance_data_refs_alignment): Bail out early for gathers
8848         or scatters.
8849         (vect_find_same_alignment_drs): Likewise.
8850         (vect_analyze_data_refs_alignment): Remove dead code.
8851         (vect_slp_analyze_and_verify_node_alignment): Likewise.
8852         (vect_analyze_data_refs): For possible gathers or scatters do
8853         not create an alternate DR, just check their possible validity
8854         and mark them.  Adjust DECL_NONALIASED handling to not rely
8855         on DR_BASE_ADDRESS.
8856         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8857         update inits of gathers or scatters.
8858         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8859         Also copy gather/scatter flag to pattern vinfo.
8861 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8863         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8864         behavior of vec_packsu (vector unsigned long long, vector unsigned
8865         long long) to match behavior of vec_packs with same signature.
8867 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
8868             Thomas Schwinge <thomas@codesourcery.com>
8869             Cesar Philippidis  <cesar@codesourcery.com>
8871         * gimplify.c (gimplify_scan_omp_clauses): Add support for
8872         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8873         (gimplify_adjust_omp_clauses): Likewise.
8874         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8875         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8876         (gimplify_omp_target_update): Update handling of acc update and
8877         enter/exit data.
8878         * omp-low.c (install_var_field): Remove unused parameter
8879         base_pointers_restrict.
8880         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8881         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8882         FINALIZE}
8883         (omp_target_base_pointers_restrict_p): Delete.
8884         (scan_omp_target): Update call to scan_sharing_clauses.
8885         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8886         FINALIZE}.
8887         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8888         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8889         (convert_local_omp_clauses): Likewise.
8890         * tree-pretty-print.c (dump_omp_clause): Likewise.
8891         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
8892         FINALIZE}.
8893         (omp_clause_code_name): Likewise.
8895 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
8897         PR debug/86194
8898         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8899         be narrowed.
8901         PR tree-optimization/86231
8902         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
8903         anti-range don't overwrite *vr0min before using it to compute *vr0max.
8905 2018-06-20  Tom de Vries  <tdevries@suse.de>
8907         PR tree-optimization/86097
8908         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8909         iv type if signedness of iv type is not the same as that of *nit.
8911 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
8913         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
8914         EDGE_EH edges, verify they are all EDGE_EH.
8916 2018-06-20  Maya Rashish  <coypu@sdf.org>
8918         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8920 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8922         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8923         * config/aarch64/aarch64.c (xgene1_tunings): Add
8924         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8925         (aarch64_mode_valid_for_sched_fusion_p):
8926         Allow 16-byte modes.
8927         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8928         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8929         128-bit modes.
8930         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8931         New pattern.
8932         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8933         * config/aarch64/iterators.md (VQ2): New mode iterator.
8935 2018-06-20  Martin Liska  <mliska@suse.cz>
8937         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8938         Change default ratio from 10 to 8.
8940 2018-06-20  Martin Liska  <mliska@suse.cz>
8942         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8943         New.
8944         (bit_test_cluster::find_bit_tests): Likewise.
8945         (switch_decision_tree::analyze_switch_statement): Find clusters.
8946         * tree-switch-conversion.h (struct jump_table_cluster): Document
8947         hierarchy.
8949 2018-06-20  Martin Liska  <mliska@suse.cz>
8951         * tree-switch-conversion.c (switch_conversion::collect):
8952         Record m_uniq property.
8953         (switch_conversion::expand): Bail out for special conditions.
8954         (group_cluster::~group_cluster): New.
8955         (group_cluster::group_cluster): Likewise.
8956         (group_cluster::dump): Likewise.
8957         (jump_table_cluster::emit): New.
8958         (switch_decision_tree::fix_phi_operands_for_edges): New.
8959         (struct case_node): Remove struct.
8960         (jump_table_cluster::can_be_handled): New.
8961         (case_values_threshold): Moved to header.
8962         (reset_out_edges_aux): Likewise.
8963         (jump_table_cluster::is_beneficial): New.
8964         (bit_test_cluster::can_be_handled): Likewise.
8965         (add_case_node): Remove.
8966         (bit_test_cluster::is_beneficial): New.
8967         (case_bit_test::cmp): New.
8968         (bit_test_cluster::emit): New.
8969         (expand_switch_as_decision_tree_p): Remove.
8970         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8971         (fix_phi_operands_for_edge): Likewise.
8972         (switch_decision_tree::analyze_switch_statement): New.
8973         (compute_cases_per_edge): Move ...
8974         (switch_decision_tree::compute_cases_per_edge): ... here.
8975         (try_switch_expansion): Likewise.
8976         (switch_decision_tree::try_switch_expansion): Likewise.
8977         (record_phi_operand_mapping): Likewise.
8978         (switch_decision_tree::record_phi_operand_mapping): Likewise.
8979         (emit_case_decision_tree): Likewise.
8980         (switch_decision_tree::emit): Likewise.
8981         (balance_case_nodes): Likewise.
8982         (switch_decision_tree::balance_case_nodes): Likewise.
8983         (dump_case_nodes): Likewise.
8984         (switch_decision_tree::dump_case_nodes): Likewise.
8985         (emit_jump): Likewise.
8986         (switch_decision_tree::emit_jump): Likewise.
8987         (emit_cmp_and_jump_insns): Likewise.
8988         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8989         (emit_case_nodes): Likewise.
8990         (switch_decision_tree::emit_case_nodes): Likewise.
8991         (conditional_probability): Remove.
8992         * tree-switch-conversion.h (enum cluster_type): New.
8993         (PRINT_CASE): New.
8994         (struct cluster): Likewise.
8995         (cluster::cluster): Likewise.
8996         (struct simple_cluster): Likewise.
8997         (simple_cluster::simple_cluster): Likewise.
8998         (struct group_cluster): Likewise.
8999         (struct jump_table_cluster): Likewise.
9000         (struct bit_test_cluster): Likewise.
9001         (struct min_cluster_item): Likewise.
9002         (struct case_tree_node): Likewise.
9003         (case_tree_node::case_tree_node): Likewise.
9004         (jump_table_cluster::case_values_threshold): Likewise.
9005         (struct case_bit_test): Likewise.
9006         (struct switch_decision_tree): Likewise.
9007         (struct switch_conversion): Likewise.
9008         (switch_decision_tree::reset_out_edges_aux): Likewise.
9010 2018-06-20  Martin Liska  <mliska@suse.cz>
9012         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
9013         (hoist_edge_and_branch_if_true): Likewise.
9014         (expand_switch_using_bit_tests_p): Likewise.
9015         (struct case_bit_test): Likewise.
9016         (case_bit_test_cmp): Likewise.
9017         (emit_case_bit_tests): Likewise.
9018         (switch_conversion::switch_conversion): New class.
9019         (struct switch_conv_info): Remove old struct.
9020         (collect_switch_conv_info): More to ...
9021         (switch_conversion::collect): ... this.
9022         (check_range): Likewise.
9023         (switch_conversion::check_range): Likewise.
9024         (check_all_empty_except_final): Likewise.
9025         (switch_conversion::check_all_empty_except_final): Likewise.
9026         (check_final_bb): Likewise.
9027         (switch_conversion::check_final_bb): Likewise.
9028         (create_temp_arrays): Likewise.
9029         (switch_conversion::create_temp_arrays): Likewise.
9030         (free_temp_arrays): Likewise.
9031         (gather_default_values): Likewise.
9032         (switch_conversion::gather_default_values): Likewise.
9033         (build_constructors): Likewise.
9034         (switch_conversion::build_constructors): Likewise.
9035         (constructor_contains_same_values_p): Likewise.
9036         (switch_conversion::contains_same_values_p): Likewise.
9037         (array_value_type): Likewise.
9038         (switch_conversion::array_value_type): Likewise.
9039         (build_one_array): Likewise.
9040         (switch_conversion::build_one_array): Likewise.
9041         (build_arrays): Likewise.
9042         (switch_conversion::build_arrays): Likewise.
9043         (gen_def_assigns): Likewise.
9044         (switch_conversion::gen_def_assigns): Likewise.
9045         (prune_bbs): Likewise.
9046         (switch_conversion::prune_bbs): Likewise.
9047         (fix_phi_nodes): Likewise.
9048         (switch_conversion::fix_phi_nodes): Likewise.
9049         (gen_inbound_check): Likewise.
9050         (switch_conversion::gen_inbound_check): Likewise.
9051         (process_switch): Use the newly created class.
9052         (switch_conversion::expand): New.
9053         (switch_conversion::~switch_conversion): New.
9054         * tree-switch-conversion.h: New file.
9056 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9058         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
9059         tree-vect-patterns.c.
9060         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
9061         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
9062         (vect_recog_sad_pattern): Likewise.
9063         (vect_recog_widen_sum_pattern): Likewise.
9064         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
9065         (vect_recog_widen_shift_pattern): Remove the type_in argument.
9066         (vect_recog_rotate_pattern): Likewise.
9067         (vect_recog_mult_pattern): Likewise.
9068         (vect_recog_vector_vector_shift_pattern): Likewise.
9069         (vect_recog_divmod_pattern): Likewise.
9070         (vect_recog_mixed_size_cond_pattern): Likewise.
9071         (vect_recog_bool_pattern): Likewise.
9072         (vect_recog_mask_conversion_pattern): Likewise.
9073         (vect_try_gather_scatter_pattern): Likewise.
9074         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
9075         (vect_recog_over_widening_pattern): Likewise.
9076         (vect_recog_gather_scatter_pattern): Likewise.
9077         (vect_recog_func_ptr): Move from tree-vectorizer.h
9078         (vect_vect_recog_func_ptrs): Move further down the file.
9079         (vect_recog_func): Likewise.  Remove the third argument.
9080         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
9081         (vect_pattern_recog_1): Expect the pattern function to do any
9082         necessary target tests.  Also expect it to provide a vector type.
9083         Remove the type_in handling.
9085 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9087         * tree-vect-patterns.c (vect_pattern_detected): New function.
9088         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
9089         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
9090         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
9091         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
9092         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9093         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9094         (vect_recog_mask_conversion_pattern)
9095         (vect_try_gather_scatter_pattern): Likewise.
9097 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9099         * tree-vect-patterns.c (vect_get_internal_def): New function.
9100         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9101         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
9102         (search_type_for_mask_1): Use it.
9104 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9106         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9107         redundant WIDEN_SUM_EXPR handling.
9108         (vect_recog_sad_pattern): Likewise.
9110 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9112         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9113         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
9114         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
9115         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
9116         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
9118 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9120         * tree-vect-stmts.c (vectorizable_call): Make sure that we
9121         use the stmt_vec_info of the original bb statement for the
9122         new zero assignment, even if the call is part of a pattern.
9124 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9126         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
9127         that the sequence is attached to the original statement rather
9128         than the pattern statement.
9129         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
9130         PATTERN_DEF_SEQ from the original statement rather than
9131         the main pattern statement.
9132         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
9133         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
9134         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
9136 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9138         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
9139         definition statements before the early exit for statements that aren't
9140         live or relevant.
9141         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
9142         split out from...
9143         (vect_transform_loop): ...here.  Process pattern definition
9144         statements without first checking whether the main pattern
9145         statement is live or relevant.
9147 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
9149         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
9150         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
9152 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
9154         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
9155         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
9156         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
9157         (expand_block_compare): Change select_block_compare_mode call.
9158         (expand_strncmp_align_check): Use new functions, fix comment.
9159         (emit_final_str_compare_gpr): New function.
9160         (expand_strn_compare): Refactor and clean up code.
9161         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
9163 2018-06-19  Tony Reix  <tony.reix@atos.com>
9164             Damien Bergamini  <damien.bergamini@atos.com>
9165             David Edelsohn  <dje.gcc@gmail.com>
9167         * collect2.c (static_obj): New variable.
9168         (static_libs): New variable.
9169         (is_in_list): Uncomment declaration.
9170         (main): Track AIX libraries linked statically.
9171         (is_in_list): Uncomment definition.
9172         (scan_prog_file): Don't add AIX shared libraries initializer
9173         to constructor list if linking statically.
9175 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
9177         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
9178         constant.
9179         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9181 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
9183         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9184         blocks.
9186 2018-06-19  Martin Liska  <mliska@suse.cz>
9188         * config/i386/i386.c (ix86_can_inline_p): Do not use
9189         ipa_fn_summaries::get_create.
9190         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9191         get.
9192         (devirtualization_time_bonus): Likewise.
9193         (ipcp_propagate_stage): Likewise.
9194         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9195         (edge_set_predicate): Likewise.
9196         (evaluate_conditions_for_known_args): Likewise.
9197         (evaluate_properties_for_edge): Likewise.
9198         (ipa_call_summary::reset): Tranform to ...
9199         (ipa_call_summary::~ipa_call_summary): ... this.
9200         (ipa_fn_summary::reset): Transform to ...
9201         (ipa_fn_summary::~ipa_fn_summary): ... this.
9202         (ipa_fn_summary_t::remove): Rename to ...
9203         (ipa_fn_summary_t::remove_callees): ... this.
9204         (ipa_fn_summary_t::duplicate): Use placement new
9205         instead of memory copy.
9206         (ipa_call_summary_t::duplicate): Likewise.
9207         (ipa_call_summary_t::remove): Remove.
9208         (dump_ipa_call_summary): Change get_create to get.
9209         (ipa_dump_fn_summary): Dump only when summary exists.
9210         (analyze_function_body): Use symbol_summary::get instead
9211         of get_create.
9212         (compute_fn_summary): Likewise.
9213         (estimate_edge_devirt_benefit): Likewise.
9214         (estimate_edge_size_and_time): Likewise.
9215         (inline_update_callee_summaries): Likewise.
9216         (remap_edge_change_prob): Likewise.
9217         (remap_edge_summaries): Likewise.
9218         (ipa_merge_fn_summary_after_inlining): Likewise.
9219         (write_ipa_call_summary): Likewise.
9220         (ipa_fn_summary_write): Likewise.
9221         (ipa_free_fn_summary): Likewise.
9222         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9223         (struct ipa_call_summary): Likewise.
9224         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9225         of get_create.
9226         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9227         (estimate_size_after_inlining): Likewise.
9228         (estimate_growth): Likewise.
9229         (growth_likely_positive): Likewise.
9230         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9231         (inline_call): Likewise.
9232         * ipa-inline.c (caller_growth_limits): Likewise.
9233         (can_inline_edge_p): Likewise.
9234         (can_inline_edge_by_limits_p): Likewise.
9235         (compute_uninlined_call_time): Likewise.
9236         (compute_inlined_call_time): Likewise.
9237         (want_inline_small_function_p): Likewise.
9238         (edge_badness): Likewise.
9239         (update_caller_keys): Likewise.
9240         (update_callee_keys): Likewise.
9241         (inline_small_functions): Likewise.
9242         (inline_to_all_callers_1): Likewise.
9243         (dump_overall_stats): Likewise.
9244         (early_inline_small_functions): Likewise.
9245         (early_inliner): Likewise.
9246         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9247         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9248         * ipa-pure-const.c (malloc_candidate_p): Likewise.
9249         * ipa-split.c (execute_split_functions): Likewise.
9250         * symbol-summary.h: Likewise.
9251         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9253 2018-06-19  Richard Biener  <rguenther@suse.de>
9255         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9256         (vectorize_loops): ... here.  Fix dbgcnt handling.
9257         (try_vectorize_loop): Wrap try_vectorize_loop_1.
9259 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
9261         PR target/86197
9262         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9263         ieee128 argument takes up only one (vector) register, not two (floating
9264         point) registers.
9266 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
9268         * gimplify.c (gimplify_init_constructor): Really never clear for an
9269         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9271 2018-06-19  Richard Biener  <rguenther@suse.de>
9273         PR tree-optimization/86179
9274         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9275         after failed recognition.
9277 2018-06-18  Martin Sebor  <msebor@redhat.com>
9279         PR middle-end/85602
9280         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9281         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9282         Handle integer subtraction.
9283         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9284         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9286 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
9288         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9289         param from rtx to rtx_insn *.
9290         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9291         param.
9292         (frv_ifcvt_modify_insn): Likwise.
9293         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9294         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
9295         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9296         as_a <rtx_insn *> cast to local "unprotected_region" once
9297         it's been established that it's not NULL or pc_rtx.
9298         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9299         param "sethi" from rtx to rtx_insn *.
9300         (nds32_group_float_insns): Likewise for param "insn".
9301         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9302         param.
9303         (vax_output_int_subtract): Likewise.
9304         * config/vax/vax.c (vax_output_int_add): Likewise for param
9305         "insn".
9306         (vax_output_int_subtract): Likewise.
9307         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9308         (emit_pattern_after): Likewise for param "after".
9309         (emit_insn_after): Likewise.
9310         (emit_jump_insn_after): Likewise.
9311         (emit_call_insn_after): Likewise.
9312         (emit_debug_insn_after): Likewise.
9313         (emit_pattern_before): Likewise for param "before".
9314         (emit_insn_before): Likewise.
9315         (emit_jump_insn_before): Likewise.
9316         * final.c (get_insn_template): Likewise for param "insn", removing
9317         a cast.
9318         * output.h (get_insn_template): Likewise for 2nd param.
9319         * rtl.h (emit_insn_before): Likewise.
9320         (emit_jump_insn_before): Likewise.
9321         (emit_debug_insn_before_noloc): Likewise.
9322         (emit_insn_after): Likewise.
9323         (emit_jump_insn_after): Likewise.
9324         (emit_call_insn_after): Likewise.
9325         (emit_debug_insn_after): Likewise.
9326         (set_insn_deleted): Likewise for param.
9328 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
9330         PR target/85358
9331         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9332         floating point modes, so that IFmode is numerically greater than
9333         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9334         to declare the ordering.  This prevents IFmode from being
9335         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9336         machine.  Include rs6000-modes.h to share the fractional values
9337         between genmodes* and the rest of the compiler.
9338         (IFmode): Likewise.
9339         (KFmode): Likewise.
9340         (TFmode): Likewise.
9341         * config/rs6000/rs6000-modes.h: New file.
9342         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9343         meaning of rs6000_long_double_size so that 126..128 selects an
9344         appropriate 128-bit floating point type.
9345         (rs6000_option_override_internal): Likewise.
9346         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9347         (TARGET_LONG_DOUBLE_128): Change the meaning of
9348         rs6000_long_double_size so that 126..128 selects an appropriate
9349         128-bit floating point type.
9350         (LONG_DOUBLE_TYPE_SIZE): Update comment.
9351         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9352         source and destination to match the standard usage.
9353         (truncifkf2): Likewise.
9354         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9355         ISA 2.07 to use an explicit clobber, instead of passing in a
9356         temporary.
9357         (copysign<mode>3_soft): Likewise.
9359 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
9361         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9362         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9363         (vect_slp_analyze_instance_dependence): Likewise.
9364         (vect_enhance_data_refs_alignment): Likewise.
9365         (vect_analyze_data_refs_alignment): Likewise.
9366         (vect_slp_analyze_and_verify_instance_alignment
9367         (vect_analyze_data_ref_accesses): Likewise.
9368         (vect_prune_runtime_alias_test_list): Likewise.
9369         (vect_analyze_data_refs): Likewise.
9370         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9371         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9372         (vect_analyze_scalar_cycles_1): Likewise.
9373         (vect_get_loop_niters): Likewise.
9374         (vect_analyze_loop_form_1): Likewise.
9375         (vect_update_vf_for_slp): Likewise.
9376         (vect_analyze_loop_operations): Likewise.
9377         (vect_analyze_loop): Likewise.
9378         (vectorizable_induction): Likewise.
9379         (vect_transform_loop): Likewise.
9380         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9381         * tree-vect-slp.c (vect_analyze_slp): Likewise.
9382         (vect_make_slp_decision): Likewise.
9383         (vect_detect_hybrid_slp): Likewise.
9384         (vect_slp_analyze_operations): Likewise.
9385         (vect_slp_bb): Likewise.
9386         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
9387         (vectorizable_bswap): Likewise.
9388         (vectorizable_call): Likewise.
9389         (vectorizable_simd_clone_call): Likewise.
9390         (vectorizable_conversion): Likewise.
9391         (vectorizable_assignment): Likewise.
9392         (vectorizable_shift): Likewise.
9393         (vectorizable_operation): Likewise.
9394         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
9396 2018-06-18  Martin Sebor  <msebor@redhat.com>
9398         PR tree-optimization/81384
9399         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
9400         * builtins.c (expand_builtin_strnlen): New function.
9401         (expand_builtin): Call it.
9402         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
9403         * builtins.def (BUILT_IN_STRNLEN): New.
9404         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
9405         Warn for bounds in excess of maximum object size.
9406         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
9407         single-value ranges.  Handle strnlen.
9408         (handle_builtin_strlen): Handle strnlen.
9409         (strlen_check_and_optimize_stmt): Same.
9410         * doc/extend.texi (Other Builtins): Document strnlen.
9412 2018-06-18  Maya Rashish  <coypu@sdf.org>
9414         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
9415         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
9416         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
9418         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
9419         here to ...
9420         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
9422 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
9424         * tree.c (escaped_string::escape): Replace cast to char * by
9425         const_cast<char *> (unescaped).
9427 2018-06-18  Nick Clifton  <nickc@redhat.com>
9429         PR 84195
9430         * tree.c (escaped_string): New class.  Converts an unescaped
9431         string into its escaped equivalent.
9432         (warn_deprecated_use): Use the new class to convert the
9433         deprecation message, if present.
9434         (test_escaped_strings): New self test.
9435         (test_c_tests): Add test_escaped_strings.
9436         * doc/extend.texi (deprecated): Add a note that the
9437         deprecation message is affected by the -fmessage-length
9438         option, and that control characters will be escaped.
9439         (#pragma GCC error): Document this pragma.
9440         (#pragma GCC warning): Likewise.
9441         * doc/invoke.texi (-fmessage-length): Document this option's
9442         effect on the #warning and #error preprocessor directives and
9443         the deprecated attribute.
9445 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
9447         * tree.c (decl_value_expr_lookup): Revert latest change.
9448         (decl_value_expr_insert): Likewise.
9450 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
9452         * gimplify.c (nonlocal_vlas): Delete.
9453         (nonlocal_vla_vars): Likewise.
9454         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9455         referenced VLAs.
9456         (gimplify_body): Do not create and destroy nonlocal_vlas.
9457         * tree-nested.c: Include diagnostic.h.
9458         (use_pointer_in_frame): Tweak.
9459         (lookup_field_for_decl): Add assertion and declare the transformation.
9460         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9461         internal error when the reference is in a wrong context.  Do not
9462         create a debug decl by default.
9463         (note_nonlocal_block_vlas): Delete.
9464         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9465         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
9466         create a debug decl by default.
9467         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9468         call is in a wrong context.
9469         (fixup_vla_decls): New function.
9470         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
9471         debug variables were created.
9472         * tree.c (decl_value_expr_lookup): Add checking assertion.
9473         (decl_value_expr_insert): Likewise.
9475 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
9477         PR middle-end/82479
9478         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9479         * tree-scalar-evolution.c (interpret_expr): Likewise.
9480         (expression_expensive_p): Likewise.
9481         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9482         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9483         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9484         (ssa_defined_by_minus_one_stmt_p): New.
9486 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
9488         PR middle-end/64946
9489         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9490         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9491         * dojump.c (do_jump): Likewise.
9492         * expr.c (expand_expr_real_2): Check operand type's sign.
9493         * fold-const.c (const_unop): Handle ABSU_EXPR.
9494         (fold_abs_const): Likewise.
9495         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9496         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9497         (strip_sign_op_1): Likesise.
9498         * match.pd: Add new pattern to generate ABSU_EXPR.
9499         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9500         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9501         * tree-eh.c (operation_could_trap_helper_p): Likewise.
9502         * tree-inline.c (estimate_operator_cost): Likewise.
9503         * tree-pretty-print.c (dump_generic_node): Likewise.
9504         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9505         * tree.def (ABSU_EXPR): New.
9507 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
9509         PR middle-end/86095
9510         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9511         documented as preserved for backward compatibility only.
9512         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9514         PR rtl-optimization/86108
9515         * bb-reorder.c (create_forwarder_block): Renamed to ...
9516         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
9517         jump from new landing pad to the second part.
9518         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9519         Adjust callers.
9521 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
9523         PR middle-end/85878
9524         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9525         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9526         Only call store_expr for halves if the mode is the same.
9528         PR middle-end/86123
9529         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9530         Fix up comment formatting.
9532 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9534         * typed-splay-tree.h (typed_splay_tree::remove): New function.
9535         (typed_splay_tree::closure,
9536         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9537         (typed_splay_tree::typed_splay_tree,
9538         typed_splay_tree::operator =): Declared private.
9539         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9540         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9541         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9542         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9543         typed_splay_tree::splay_tree_splay,
9544         typed_splay_tree::splay_tree_foreach_helper,
9545         typed_splay_tree::splay_tree_insert,
9546         typed_splay_tree::splay_tree_remove,
9547         typed_splay_tree::splay_tree_lookup,
9548         typed_splay_tree::splay_tree_predecessor,
9549         typed_splay_tree::splay_tree_successor,
9550         typed_splay_tree::splay_tree_min,
9551         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9552         (typed_splay_tree::root, typed_splay_tree::comp,
9553         typed_splay_tree::delete_key,
9554         typed_splay_tree::delete_value): New data members.
9555         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9556         typed_splay_tree::remove.
9558 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
9560         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9561         -mginv and -mno-ginv to the assembler.
9562         * config/mips/mips.opt (-mcrc): New option.
9563         (-mginv): Likewise.
9564         * doc/invoke.text (-mcrc): Document.
9565         (-mginv): Likewise.
9567 2018-06-15  Nick Clifton  <nickc@redhat.com>
9569         PR 84195
9570         * tree.c (escaped_string): New class.  Converts an unescaped
9571         string into its escaped equivalent.
9572         (warn_deprecated_use): Use the new class to convert the
9573         deprecation message, if present.
9574         (test_escaped_strings): New self test.
9575         (test_c_tests): Add test_escaped_strings.
9576         * doc/extend.texi (deprecated): Add a note that the
9577         deprecation message is affected by the -fmessage-length
9578         option, and that control characters will be escaped.
9579         (#pragma GCC error): Document this pragma.
9580         (#pragma GCC warning): Likewise.
9581         * doc/invoke.texi (-fmessage-length): Document this option's
9582         effect on the #warning and #error preprocessor directives and
9583         the deprecated attribute.
9585 2018-06-15  Richard Biener  <rguenther@suse.de>
9587         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9588         here, also noting vector size used.
9589         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
9590         size used in MSG_OPTIMIZED_LOCATIONS dump.
9591         (pass_slp_vectorize::execute): Adjust.
9593 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
9595         PR target/85968
9596         * config/arc/arc.c (arc_return_address_register): Fix
9597         if-condition.
9599 2018-06-15  Richard Biener  <rguenther@suse.de>
9601         PR middle-end/86159
9602         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9603         leave useless conversion stripping to force_gimple_operand_gsi.
9604         (gimplify_build2): Likewise.
9605         (gimplify_build1): Likewise.
9607 2018-06-15  Richard Biener  <rguenther@suse.de>
9609         PR middle-end/86076
9610         * tree-cfg.c (move_stmt_op): unshare invariant addresses
9611         before adjusting their block.
9613 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9615         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9616         multilibs for *-*-rtems*.
9617         * config/riscv/t-rtems: New file.
9619 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
9621         PR middle-end/86122
9622         * match.pd ((A +- CST1) +- CST2): Punt if last resort
9623         unsigned_type_for returns NULL.
9625         PR target/85945
9626         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9627         subregs of multi-word pseudos unless the float mode has word size.
9629 2018-06-14  Richard Biener  <rguenther@suse.de>
9631         PR middle-end/86139
9632         * tree-vect-generic.c (build_word_mode_vector_type): Remove
9633         duplicate and harmful type_hash_canon.
9634         * tree.c (type_hash_canon): Assert we didn't find ourselves.
9636 2018-06-14  Richard Biener  <rguenther@suse.de>
9638         PR ipa/86124
9639         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9640         NULL cgraph_node.
9642 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9644         * config/rtems.h (STDINT_LONG32): Define.
9646 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
9647             Prachi Godbole  <prachi.godbole@imgtec.com>
9649         * config/mips/mips-cpus.def: Define P6600.
9650         * config/mips/mips-tables.opt: Regenerate.
9651         * config/mips/mips.c (mips_ucbranch_type): New enum.
9652         (mips_rtx_cost_data): Add support for P6600.
9653         (mips_issue_rate): Likewise.
9654         (mips_multipass_dfa_lookahead): Likewise.
9655         (mips_avoid_hazard): Likewise.
9656         (mips_reorg_process_insns): Likewise.
9657         (mips_classify_branch_p6600): New function.
9658         * config/mips/mips.h (TUNE_P6600): New define.
9659         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9660         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9661         * config/mips/mips.md: Include p6600.md.
9662         (processor): Add p6600.
9663         * config/mips/p6600.md: New file.
9664         * doc/invoke.texi: Add p6600 to supported architectures.
9666 2018-06-13  Martin Sebor  <msebor@redhat.com>
9668         PR tree-optimization/86114
9669         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9670         of integer types.
9671         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9673 2018-06-13  Richard Biener  <rguenther@suse.de>
9675         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9676         Properly set vector type of the intermediate stmt.
9677         * tree-vect-stmts.c (vectorizable_operation): The destination
9678         var always has vectype_out type.
9680 2018-06-13  Jeff Law  <law@redhat.com>
9682         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9683         integer 0 for argument to print_rtl_with_bb.
9684         (rl78_reorg): Likewise.
9686 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
9688         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9689         from rtx to rtx_insn *.
9690         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9691         "label".
9692         (add_sched_insns_for_speculation): Likewise for local "target",
9693         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9694         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9695         from rtx_insn ** to rtx_code_label **.
9696         (reorg_emit_nops): Likewise.
9697         (c6x_reorg): Likewise for local "call_labels".
9698         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9699         rtx to rtx_insn *.
9700         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9701         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9702         the loops over LABEL_REFS.
9703         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9704         braf_label.
9705         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9706         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9707         (split_branches): Strengthen local "olabel" from rtx to
9708         rtx_insn *, adding a safe_as_a cast.
9709         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9710         to "rtx_insn *".
9711         (add_insn_after): Likewise for first two params.
9712         (add_insn_before): Likewise.
9713         (remove_insn): Likewise for param.
9714         (emit_pattern_before_noloc): Likewise for second and third params.
9715         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9716         (emit_call_insn_before_noloc): Likewise.
9717         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9718         to "rtx_insn *".
9719         (emit_barrier_before): Likewise.
9720         (emit_label_before): Strengthen "label" param from "rtx" to
9721         "rtx_code_label *".  Strengthen "before" param from "rtx" to
9722         "rtx_insn *".
9723         (emit_insn_after_1): Strengthen "after" param from "rtx" to
9724         "rtx_insn *".
9725         (emit_pattern_after_noloc): Likewise.
9726         (emit_insn_after_noloc): Likewise.
9727         (emit_jump_insn_after_noloc): Likewise.
9728         (emit_call_insn_after_noloc): Likewise.
9729         (emit_debug_insn_after_noloc): Likewise.
9730         (emit_barrier_after): Likewise.
9731         (emit_label_after): Likewise for both params.
9732         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
9733         "loc" param from "int" to "location_t".
9734         (emit_insn_after_setloc): Likewise.
9735         (emit_jump_insn_after_setloc): Likewise.
9736         (emit_call_insn_after_setloc): Likewise.
9737         (emit_debug_insn_after_setloc): Likewise.
9738         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
9739         "loc" param from "int" to "location_t".
9740         (emit_pattern_before): Convert NULL_RTX to NULL.
9741         (emit_insn_before_setloc): Convert "loc" param from "int" to
9742         "location_t".
9743         (emit_jump_insn_before_setloc): Likewise.
9744         (emit_call_insn_before_setloc): Likewise.
9745         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9746         rtx_insn *.  Convert "loc" param from "int" to "location_t".
9747         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9748         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9749         Convert 3rd param from "int" to "location_t".
9750         (emit_barrier_before, emit_barrier_after, next_real_insn):
9751         Strengthen param from rtx to rtx_insn *.
9752         (emit_label_before): Strengthen 1st param from "rtx" to
9753         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
9754         "rtx_insn *".
9755         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9756         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9757         Strengthen 2nd param from "rtx" to "rtx_insn *".
9758         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9759         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9760         Likewise. Convert 3rd param from "int" to "location_t".
9761         (emit_label_after): Strengthen 1st param from "rtx" to
9762         "rtx_code_label *".
9763         (next_real_insn, remove_insn): Strengthen param from "rtx" to
9764         "rtx_insn *".
9765         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9766         from "rtx" to "rtx_insn *".
9768 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
9770         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9771         bodies streamed in with -Q.
9772         * dumpfile.c (dump_files): Add lto-stream-out dump file.
9773         * dumpfile.h (tree_dump_index): Add lto_stream_out.
9774         * gimple-streamer-out.c: Include gimple-pretty-print.h
9775         (output_bb): Dump stmts streamed.
9776         * lto-section-out.c: Include print-tree.h
9777         (lto_begin_section): Dump sections created.
9778         (lto_output_decl_index): Dump decl encoded.
9779         * lto-streamer-out.c: Include print-tree.h
9780         (create_output_block): Dump output block created.
9781         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9782         (output_function): Dump function output.
9783         (output_constructor): Dump constructor streamed.
9784         (write_global_stream): Output indexes encoded.
9785         (produce_asm_for_decls): Dump streams encoded.
9786         * lto-streamer.c (streamer_dump_file): New global var.
9787         * lto-streamer.h (streamer_dump_file): Declare.
9788         * passes.c (ipa_write_summaries): Initialize streamer dump.
9789         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9790         in.
9792 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
9794         PR target/86048
9795         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9796         offsets for register save directives.  Emit a second batch of save
9797         directives, if need be, when the function accesses prior frames.
9799 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9801         * config/arc/fpu.md (fmasf4): Force operand to register.
9802         (fnmasf4): Likewise.
9804 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9806         * config/arc/arc-protos.h (arc_pad_return): Remove.
9807         * config/arc/arc.c (machine_function): Remove force_short_suffix
9808         and size_reason.
9809         (arc_print_operand): Adjust printing of '&'.
9810         (arc_verify_short): Remove conditional printing of short suffix.
9811         (arc_final_prescan_insn): Remove reference to size_reason.
9812         (pad_return): New function.
9813         (arc_reorg): Call pad_return.
9814         (arc_pad_return): Remove.
9815         (arc_init_machine_status): Remove reference to force_short_suffix.
9816         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9817         (attr length): When attribute iscompact is true force to 2
9818         regardless; in the case of maybe check if we want to force the
9819         instruction to have 4 bytes length.
9820         (nopv): Change it to generate 4 byte long nop as well.
9821         (blockage): New pattern.
9822         (simple_return): Remove call to arc_pad_return.
9823         (p_return_i): Likewise.
9825 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9827         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9829 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9831         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9832         ARC cores.
9834 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9836         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9837         for ARC700 and ARCv2.
9839 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
9841         PR target/86076
9842         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9843         operands[2] instead of operands[1].
9846 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
9848         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9849         case, check whether the outer register overlaps an unallocatable
9850         register, not just whether it fits the required class.
9852 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
9854         * poly-int.h (can_div_trunc_p): Add new overload in which all values
9855         are poly_ints.
9856         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9857         (memrefs_conflict_p): Likewise.
9858         (init_alias_analysis): Likewise.
9859         * cfgexpand.c (expand_debug_expr): Likewise.
9860         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9861         * cse.c (fold_rtx): Likewise.
9862         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9863         * expr.c (emit_block_move_hints): Likewise.
9864         (clear_storage_hints, push_block, emit_push_insn): Likewise.
9865         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9866         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9867         (emit_group_store): Likewise.
9868         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
9869         to read the PRE/POST_MODIFY increment.
9870         * calls.c (store_one_arg): Use strip_offset.
9871         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9872         poly_int_rtx_p.
9873         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9874         by a VEC_SELECT.
9875         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9876         (simplify_binary_operation_1): Extend CONST_INT handling to
9877         poly_int_rtx_p.
9878         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9879         than a HOST_WIDE_INT.
9880         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9881         poly_int64.
9882         (adjust_mems, add_stores): Update accodingly.
9883         (vt_canonicalize_addr): Track polynomial offsets.
9884         (emit_note_insn_var_location): Likewise.
9885         (vt_add_function_parameter): Likewise.
9886         (vt_initialize): Likewise.
9888 2018-06-12  Jeff Law  <law@redhat.com>
9890         * config.gcc (alpha*-*-freebsd*): Remove.
9891         * config/alpha/freebsd.h: Remove.
9893 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
9895         PR other/69968
9896         * spellcheck-tree.c (levenshtein_distance): Rename to...
9897         (get_edit_distance): ...this, and update for underlying renaming.
9898         * spellcheck-tree.h (levenshtein_distance): Rename to...
9899         (get_edit_distance): ...this.
9900         * spellcheck.c (levenshtein_distance): Rename to...
9901         (get_edit_distance): ...this.  Convert from Levenshtein distance
9902         to Damerau-Levenshtein distance by supporting transpositions of
9903         adjacent characters.  Rename "v1" to "v_next" and "v0" to
9904         "v_one_ago".
9905         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9906         (selftest::test_edit_distance_unit_test_oneway): ...this, and
9907         update for underlying renaming.
9908         (selftest::levenshtein_distance_unit_test): Rename to...
9909         (selftest::test_get_edit_distance_unit): ...this, and update for
9910         underlying renaming.
9911         (selftest::test_find_closest_string): Add example from PR 69968
9912         where transposition helps
9913         (selftest::test_metric_conditions): Update for renaming.
9914         (selftest::test_metric_conditions): Likewise.
9915         (selftest::spellcheck_c_tests): Likewise.
9916         * spellcheck.h (levenshtein_distance): Rename both overloads to...
9917         (get_edit_distance): ...this.
9918         (best_match::consider): Update for renaming.
9920 2018-06-12  Martin Sebor  <msebor@redhat.com>
9922         PR tree-optimization/85259
9923         * builtins.c (compute_objsize): Handle constant offsets.
9924         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9925         true iff a warning has been issued.
9926         * gimple.h (gimple_nonartificial_location): New function.
9927         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9928         gimple_nonartificial_location and handle -Wno-system-headers.
9929         (handle_builtin_stxncpy): Same.
9931 2018-06-12  Martin Sebor  <msebor@redhat.com>
9933         PR c/85931
9934         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9936 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
9938         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9939         BUILTIN_VEC_XST entries for pointer to double and long long.
9941 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
9943         PR target/85990
9944         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9945         Update comments.
9946         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9947         Likewise.
9949 2018-06-12  Martin Liska  <mliska@suse.cz>
9951         * doc/options.texi: Document IntegerRange.
9953 2018-06-12  Martin Liska  <mliska@suse.cz>
9955         * config/i386/i386.opt: Make MPX-related options as Deprecated.
9956         * opt-functions.awk: Handle Deprecated flag.
9957         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9958         and report error.
9959         (read_cmdline_option): Report warning for a deprecated option.
9960         * opts.h (struct cl_option): Add new field cl_deprecated.
9961         (CL_ERR_DEPRECATED): New.
9963 2018-06-12  Martin Liska  <mliska@suse.cz>
9965         * doc/options.texi: Document Deprecated option flag.
9967 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9969         * config/arc/arc-arch.h (arc_extras): New enum.
9970         (arc_cpu_t):Add field extra.
9971         (arc_cpu_types): Consider the extras.
9972         * config/arc/arc-cpus.def: Add extras info.
9973         * config/arc/arc-opts.h (processor_type): Consider extra field.
9974         * config/arc/arc.c (arc_override_options): Handle extra field.
9976 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9978         * config/arc/arc-arch.h: Update ARC_OPTX macro.
9979         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9980         field.
9981         * config/arc/arc.c (arc_init): Update pic warning.
9982         (irq_range): Update irq range parsing warnings.
9983         (arc_override_options): Update various warning messages.
9984         (arc_handle_aux_attribute): Likewise.
9986 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
9988         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9990 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9992         * doc/sourcebuild.texi: Document usage of line number 0 in verify
9993         compiler messages directives.
9995 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
9997         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9998         * config/mips/mips-tables.opt: Regenerate.
9999         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
10000         mips64r6.
10001         * doc/invoke.texi: Document -march=i6500.
10003 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
10005         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
10006         (i6400_gpmul): Add cpu_unit.
10007         (i6400_gpdiv): Likewise.
10008         (i6400_msa_add_d): Update reservations.
10009         (i6400_msa_int_add) Likewise.
10010         (i6400_msa_short_logic3) Likewise.
10011         (i6400_msa_short_logic2) Likewise.
10012         (i6400_msa_short_logic) Likewise.
10013         (i6400_msa_move) Likewise.
10014         (i6400_msa_cmp) Likewise.
10015         (i6400_msa_short_float2) Likewise.
10016         (i6400_msa_div_d) Likewise.
10017         (i6400_msa_long_logic1) Likewise.
10018         (i6400_msa_long_logic2) Likewise.
10019         (i6400_msa_mult) Likewise.
10020         (i6400_msa_long_float2) Likewise.
10021         (i6400_msa_long_float4) Likewise.
10022         (i6400_msa_long_float5) Likewise.
10023         (i6400_msa_long_float8) Likewise.
10024         (i6400_fpu_fadd): Include frint type.
10025         (i6400_fpu_store): New define_insn_reservation.
10026         (i6400_fpu_load): Likewise.
10027         (i6400_fpu_move): Likewise.
10028         (i6400_fpu_fcmp): Likewise.
10029         (i6400_fpu_fmadd): Likewise.
10030         (i6400_int_mult): Include imul3nc type and update reservation.
10031         (i6400_int_div): Include idiv3 type and update reservation.
10032         (i6400_int_load): Update to check type not move_type.
10033         (i6400_int_store): Likewise.
10034         (i6400_int_prefetch): Set zero latency.
10036 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
10038         * gcc.c: Document new %@{...} sequence.
10039         (LINK_COMMAND_SPEC): Use it for the -L switches.
10040         (cpp_unique_options): Use it for the -I switches.
10041         (at_file_argbuf): New global variable.
10042         (in_at_file): Likewise.
10043         (alloc_args): Create at_file_argbuf.
10044         (clear_args): Truncate at_file_argbuf.
10045         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
10046         (open_at_file): New function.
10047         (close_at_file): Likewise.
10048         (create_at_file): Delete.
10049         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
10050         <'o'>: Likewise.
10051         <'@'>: New case.
10052         (validate_switches_from_spec): Deal with %@{...} sequence.
10053         (validate_switches): Likewise.
10054         (driver::finalize): Call clear_args.
10056 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
10058         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
10060 2018-06-11  Martin Sebor  <msebor@redhat.com>
10062         * doc/invoke.texi (-Wall): List -Wc++17-compat.
10063         (Wno-class-memaccess): Add @opindex.
10064         (Wno-templates, Wno-multiple-inheritance): Same.
10065         (Wno-virtual-inheritance, Wno-namespaces): Same.
10066         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
10067         (Wno-format-overflow, Wno-format-truncation): Same.
10068         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
10069         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
10070         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
10071         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
10072         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
10073         (Wno-misspelled-isr): Same.
10075 2018-06-11  Martin Sebor  <msebor@redhat.com>
10077         * PR tree-optimization/86083
10078         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
10080 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
10082         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
10084 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
10086         PR target/85755
10087         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
10088         on the correct operand.
10089         (*movdi_internal64): Ditto.
10091 2018-06-11  Martin Liska  <mliska@suse.cz>
10093         PR tree-optimization/86089
10094         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
10096 2018-06-11  Julia Koval  <julia.koval@intel.com>
10098         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
10099         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
10100         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
10102 2018-06-11  Olivier Hainque  <hainque@adacore.com>
10104         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
10105         for Ada with strict dwarf2.
10107 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
10109         PR target/85755
10110         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
10111         addresses.
10113 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
10115         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
10117 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
10119         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
10120         TARGET_ELF.
10122 2018-06-08  Martin Liska  <mliska@suse.cz>
10124         * tree-cfg.h (debug_function): Fix argument type to match
10125         implementation.
10127 2018-06-08  Martin Liska  <mliska@suse.cz>
10129         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
10130         Remove usage of MPX-related (and removed) fields.
10131         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
10133 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
10135         * cfg.c (debug): Use TDF_NONE rather than 0.
10136         * cfghooks.c (debug): Likewise.
10137         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
10138         (struct dump_option_value_info): Convert to...
10139         (struct kv_pair): ...this template type.
10140         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
10141         rather than 0.
10142         (optinfo_verbosity_options): Likewise.
10143         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
10144         OPTGROUP_NONE.
10145         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
10146         than int for "optgroup_flags" param.
10147         (dump_generic_expr_loc): Use dump_flags_t rather than int for
10148         "dump_kind" param.
10149         (dump_dec): Likewise.
10150         (dump_finish): Use TDF_NONE rather than 0.
10151         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
10152         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
10153         than 0.  Update for change to option_ptr.
10154         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
10155         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
10156         0.  Update for changes to optinfo_verbosity_options and
10157         optgroup_options.
10158         (opt_info_switch_p): Convert optgroup_flags from int to
10159         optgroup_flags_t.
10160         (dump_basic_block): Use dump_flags_t rather than int
10161         for "dump_kind" param.
10162         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
10163         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
10164         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
10165         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
10166         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
10167         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
10168         TDF_NONE): Convert from macros to...
10169         (enum dump_flag): ...this new enum.
10170         (dump_flags_t): Update to use enum.
10171         (operator|, operator&, operator~, operator|=, operator&=):
10172         Implement for dump_flags_t.
10173         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10174         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10175         Convert from macros to...
10176         (enum optgroup_flag): ...this new enum.
10177         (optgroup_flags_t): New typedef.
10178         (operator|, operator|=): Implement for optgroup_flags_t.
10179         (struct dump_file_info): Convert field "alt_flags" to
10180         dump_flags_t.  Convert field "optgroup_flags" to
10181         optgroup_flags_t.
10182         (dump_basic_block): Use dump_flags_t rather than int for param.
10183         (dump_generic_expr_loc): Likewise.
10184         (dump_dec): Likewise.
10185         (dump_register): Convert param "optgroup_flags" to
10186         optgroup_flags_t.
10187         (opt_info_enable_passes): Likewise.
10188         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10189         than 0.
10190         * gimple-pretty-print.c (debug): Likewise.
10191         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10192         (merged_store_group::apply_stores): Likewise.
10193         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10194         * gimple.c (verify_gimple_pp): Likewise.
10195         * graphite-poly.c (print_pbb_body): Likewise.
10196         * passes.c (pass_manager::register_one_dump_file): Convert
10197         local "optgroup_flags" to optgroup_flags_t.
10198         * print-tree.c (print_node): Use TDF_NONE rather than 0.
10199         (debug): Likewise.
10200         (debug_body): Likewise.
10201         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10202         to optgroup_flags_t.
10203         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10204         than 0.
10205         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10206         (convert_mult_to_fma): Likewise.
10207         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10208         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10209         * tree-vect-data-refs.c (dump_lower_bound): Convert param
10210         "dump_kind" to dump_flags_t.
10212 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
10214         * config/rs6000/rs6000.c (min, max): Delete.
10216 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
10218         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10219         -mabi=spe and -mabi=no-spe.
10221 2018-06-08  Martin Liska  <mliska@suse.cz>
10223         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10224         where we expect an existing summary.
10226 2018-06-08  Martin Liska  <mliska@suse.cz>
10228         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10229         * ipa-inline.h (estimate_edge_growth): Likewise.
10231 2018-06-08  Martin Liska  <mliska@suse.cz>
10233         * cgraph.c (function_version_hasher::hash): Use
10234         cgraph_node::get_uid ().
10235         (function_version_hasher::equal):
10236         * cgraph.h (cgraph_node::get_uid): New method.
10237         * ipa-inline.c (update_caller_keys): Use
10238         cgraph_node::get_uid ().
10239         (update_callee_keys): Likewise.
10240         * ipa-utils.c (searchc): Likewise.
10241         (ipa_reduced_postorder): Likewise.
10242         * lto-cgraph.c (input_node): Likewise.
10243         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10244         * symbol-summary.h (symtab_insertion): Likewise.
10245         (symtab_removal): Likewise.
10246         (symtab_duplication): Likewise.
10247         * tree-pretty-print.c (dump_function_header): Likewise.
10248         * tree-sra.c (convert_callers_for_node): Likewise.
10250 2018-06-08  Martin Liska  <mliska@suse.cz>
10252         * cgraph.c (symbol_table::create_edge): Always assign a new
10253         unique number.
10254         (symbol_table::free_edge): Do not recycle numbers.
10255         * cgraph.h (cgraph_edge::get): New method.
10256         * symbol-summary.h (symtab_removal): Use it.
10257         (symtab_duplication): Likewise.
10258         (call_summary::hashable_uid): Remove.
10260 2018-06-08  Martin Liska  <mliska@suse.cz>
10262         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10263         (initialize_growth_caches): Remove.
10264         (free_growth_caches): Likewise.
10265         (do_estimate_edge_time): Use edge_growth_cache.
10266         (do_estimate_edge_size): Likewise.
10267         (do_estimate_edge_hints): Likewise.
10268         * ipa-inline.c (reset_edge_caches): Likewise.
10269         (recursive_inlining): Likewise.
10270         (inline_small_functions): Likewise.
10271         * ipa-inline.h (initialize_growth_caches): Remove.
10272         (estimate_edge_size): Likewise.
10273         (estimate_edge_time): Likewise.
10274         (estimate_edge_hints): Likewise.
10275         (reset_edge_growth_cache): Likewise.
10276         * symbol-summary.h (call_summary::remove): New method.
10278 2018-06-08  Martin Liska  <mliska@suse.cz>
10280         * ipa-cp.c (class edge_clone_summary): New summary.
10281         (grow_edge_clone_vectors): Remove.
10282         (ipcp_edge_duplication_hook): Remove.
10283         (class edge_clone_summary_t): New call_summary class.
10284         (ipcp_edge_removal_hook): Remove.
10285         (edge_clone_summary_t::duplicate): New function.
10286         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10287         (create_specialized_node): Likewise.
10288         (ipcp_driver): Initialize edge_clone_summaries and do not
10289         register hooks.
10291 2018-06-08  Martin Liska  <mliska@suse.cz>
10293         * symbol-summary.h (get): New function.
10294         (call_summary::m_initialize_when_cloning): New class member.
10296 2018-06-08  Martin Liska  <mliska@suse.cz>
10298         * cgraph.c (cgraph_node::remove): Do not recycle uid.
10299         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10300         (symbol_table::allocate_cgraph_symbol): Do not set uid.
10301         * passes.c (uid_hash_t): Record removed_nodes by their uids.
10302         (remove_cgraph_node_from_order): Use the removed_nodes set.
10303         (do_per_function_toporder): Likwise.
10304         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10305         instead of summary_uid.
10306         (symtab_removal): Likewise.
10307         (symtab_duplication): Likewise.
10309 2018-06-08  Martin Liska  <mliska@suse.cz>
10311         * ipa-cp.c (ipcp_store_bits_results): Use
10312         ipcp_transformation_sum.
10313         (ipcp_store_vr_results): Likewise.
10314         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10315         to ...
10316         (ipcp_transformation_initialize): ... this.
10317         (ipa_set_node_agg_value_chain):
10318         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10319         (write_ipcp_transformation_info): Likewise.
10320         (read_ipcp_transformation_info): Likewise.
10321         (ipcp_update_bits): Likewise.
10322         (ipcp_update_vr): Likewise.
10323         (ipcp_transform_function): Likewise.
10324         * ipa-prop.h: Rename ipcp_transformation_summary to
10325         ipcp_transformation.
10326         (class ipcp_transformation_t): New function summary.
10327         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10328         (ipa_get_agg_replacements_for_node): Likewise.
10330 2018-06-08  Martin Liska  <mliska@suse.cz>
10332         * ipa-pure-const.c (struct funct_state_d): Do it class instead
10333         of struct.
10334         (class funct_state_summary_t): New function_summary class.
10335         (has_function_state): Remove.
10336         (get_function_state): Likewise.
10337         (set_function_state): Likewise.
10338         (add_new_function): Likewise.
10339         (funct_state_summary_t::insert): New function.
10340         (duplicate_node_data): Remove.
10341         (remove_node_data): Remove.
10342         (funct_state_summary_t::duplicate): New function.
10343         (register_hooks): Create new funct_state_summaries.
10344         (pure_const_generate_summary): Use it.
10345         (pure_const_write_summary): Likewise.
10346         (pure_const_read_summary): Likewise.
10347         (propagate_pure_const): Likewise.
10348         (propagate_nothrow): Likewise.
10349         (dump_malloc_lattice): Likewise.
10350         (propagate_malloc): Likewise.
10351         (execute): Do not register hooks, just remove summary
10352         instead.
10353         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10354         constructor.
10356 2018-06-08  Martin Liska  <mliska@suse.cz>
10358         * ipa-reference.c (remove_node_data): Remove.
10359         (duplicate_node_data): Likewise.
10360         (class ipa_ref_var_info_summary_t): New class.
10361         (class ipa_ref_opt_summary_t): Likewise.
10362         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10363         (get_reference_optimization_summary): Use
10364         ipa_ref_opt_sum_summaries.
10365         (set_reference_vars_info): Remove.
10366         (set_reference_optimization_summary): Likewise.
10367         (ipa_init): Create summaries.
10368         (init_function_info): Use function summary.
10369         (ipa_ref_opt_summary_t::duplicate): New function.
10370         (ipa_ref_opt_summary_t::remove): New function.
10371         (get_read_write_all_from_node): Fix GNU coding style.
10372         (propagate): Use function summary.
10373         (write_node_summary_p): Fix GNU coding style.
10374         (stream_out_bitmap): Likewise.
10375         (ipa_reference_read_optimization_summary): Use function summary.
10376         (ipa_reference_c_finalize): Do not release hooks.
10378 2018-06-08  Martin Liska  <mliska@suse.cz>
10380         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10381         (analyze_function_body): Extract multiple calls of get_create.
10382         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10383         * ipa-inline.c (recursive_inlining): Use ::get method.
10384         * ipa-inline.h (estimate_edge_growth): Likewise.
10386 2018-06-08  Martin Liska  <mliska@suse.cz>
10388         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
10389         HSA_INVALID.
10390         (hsa_function_summary::hsa_function_summary): Use the new enum
10391         value.
10392         (hsa_gpu_implementation_p): Use hsa_summaries::get.
10393         * hsa-gen.c (hsa_get_host_function): Likewise.
10394         (get_brig_function_name): Likewise.
10395         * ipa-hsa.c (process_hsa_functions): Likewise.
10396         (ipa_hsa_write_summary): Likewise.
10397         * symbol-summary.h (symtab_duplication): Use ::get function/
10398         (get): New function.
10400 2018-06-08  Martin Liska  <mliska@suse.cz>
10402         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
10403         of get.
10404         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
10405         (hsa_register_kernel): Likewise.
10406         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
10407         * hsa-gen.c (hsa_get_host_function): Likewise.
10408         (get_brig_function_name): Likewise.
10409         (generate_hsa): Likewise.
10410         (pass_gen_hsail::execute): Likewise.
10411         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
10412         (devirtualization_time_bonus): Likewise.
10413         (ipcp_propagate_stage): Likewise.
10414         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10415         (edge_set_predicate): Likewise.
10416         (evaluate_conditions_for_known_args): Likewise.
10417         (evaluate_properties_for_edge): Likewise.
10418         (ipa_fn_summary::reset): Likewise.
10419         (ipa_fn_summary_t::duplicate): Likewise.
10420         (dump_ipa_call_summary): Likewise.
10421         (ipa_dump_fn_summary): Likewise.
10422         (analyze_function_body): Likewise.
10423         (compute_fn_summary): Likewise.
10424         (estimate_edge_devirt_benefit): Likewise.
10425         (estimate_edge_size_and_time): Likewise.
10426         (estimate_calls_size_and_time): Likewise.
10427         (estimate_node_size_and_time): Likewise.
10428         (inline_update_callee_summaries): Likewise.
10429         (remap_edge_change_prob): Likewise.
10430         (remap_edge_summaries): Likewise.
10431         (ipa_merge_fn_summary_after_inlining): Likewise.
10432         (ipa_update_overall_fn_summary): Likewise.
10433         (read_ipa_call_summary): Likewise.
10434         (inline_read_section): Likewise.
10435         (write_ipa_call_summary): Likewise.
10436         (ipa_fn_summary_write): Likewise.
10437         (ipa_free_fn_summary): Likewise.
10438         * ipa-hsa.c (process_hsa_functions): Likewise.
10439         (ipa_hsa_write_summary): Likewise.
10440         (ipa_hsa_read_section): Likewise.
10441         * ipa-icf.c (sem_function::merge): Likewise.
10442         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10443         (do_estimate_edge_time): Likewise.
10444         (estimate_size_after_inlining): Likewise.
10445         (estimate_growth): Likewise.
10446         (growth_likely_positive): Likewise.
10447         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10448         (inline_call): Likewise.
10449         * ipa-inline.c (caller_growth_limits): Likewise.
10450         (can_inline_edge_p): Likewise.
10451         (can_inline_edge_by_limits_p): Likewise.
10452         (compute_uninlined_call_time): Likewise.
10453         (compute_inlined_call_time): Likewise.
10454         (want_inline_small_function_p): Likewise.
10455         (edge_badness): Likewise.
10456         (update_caller_keys): Likewise.
10457         (update_callee_keys): Likewise.
10458         (recursive_inlining): Likewise.
10459         (inline_small_functions): Likewise.
10460         (inline_to_all_callers_1): Likewise.
10461         (dump_overall_stats): Likewise.
10462         (early_inline_small_functions): Likewise.
10463         (early_inliner): Likewise.
10464         * ipa-inline.h (estimate_edge_growth): Likewise.
10465         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10466         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10467         * ipa-prop.h (IPA_NODE_REF): Likewise.
10468         (IPA_EDGE_REF): Likewise.
10469         * ipa-pure-const.c (malloc_candidate_p): Likewise.
10470         (propagate_malloc): Likewise.
10471         * ipa-split.c (execute_split_functions): Likewise.
10472         * symbol-summary.h: Rename get to get_create.
10473         (get): Likewise.
10474         (get_create): Likewise.
10475         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10477 2018-06-08  Martin Liska  <mliska@suse.cz>
10479         * symbol-summary.h (release): Move definition out of class
10480         declaration.
10481         (symtab_removal): Likewise.
10482         (symtab_duplication): Likewise.
10484 2018-06-08  Martin Liska  <mliska@suse.cz>
10486         * symbol-summary.h (function_summary): Move constructor
10487         implementation out of class declaration.
10488         (release): Likewise.
10489         (symtab_insertion): Likewise.
10490         (symtab_removal): Likewise.
10491         (symtab_duplication): Likewise.
10492         (get): Likewise.
10494 2018-06-08  Martin Liska  <mliska@suse.cz>
10496         * Makefile.in: Remove support for MPX (macros, related functions,
10497         fields in cgraph_node, ...).
10498         * builtin-types.def (BT_BND): Likewise.
10499         (BT_FN_BND_CONST_PTR): Likewise.
10500         (BT_FN_CONST_PTR_BND): Likewise.
10501         (BT_FN_VOID_PTR_BND): Likewise.
10502         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10503         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10504         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10505         (expand_builtin_mempcpy_with_bounds): Likewise.
10506         (expand_builtin_memset_with_bounds): Likewise.
10507         (expand_builtin_memset_args): Likewise.
10508         (std_expand_builtin_va_start): Likewise.
10509         (expand_builtin): Likewise.
10510         (expand_builtin_with_bounds): Likewise.
10511         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10512         (DEF_LIB_BUILTIN_CHKP): Likewise.
10513         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10514         (DEF_CHKP_BUILTIN): Likewise.
10515         (BUILT_IN_MEMCPY): Likewise.
10516         (BUILT_IN_MEMMOVE): Likewise.
10517         (BUILT_IN_MEMPCPY): Likewise.
10518         (BUILT_IN_MEMSET): Likewise.
10519         (BUILT_IN_STPCPY): Likewise.
10520         (BUILT_IN_STRCAT): Likewise.
10521         (BUILT_IN_STRCHR): Likewise.
10522         (BUILT_IN_STRCPY): Likewise.
10523         (BUILT_IN_STRLEN): Likewise.
10524         (BUILT_IN_MEMCPY_CHK): Likewise.
10525         (BUILT_IN_MEMMOVE_CHK): Likewise.
10526         (BUILT_IN_MEMPCPY_CHK): Likewise.
10527         (BUILT_IN_MEMSET_CHK): Likewise.
10528         (BUILT_IN_STPCPY_CHK): Likewise.
10529         (BUILT_IN_STRCAT_CHK): Likewise.
10530         (BUILT_IN_STRCPY_CHK): Likewise.
10531         * calls.c (store_bounds): Likewise.
10532         (emit_call_1): Likewise.
10533         (special_function_p): Likewise.
10534         (maybe_warn_nonstring_arg): Likewise.
10535         (initialize_argument_information): Likewise.
10536         (finalize_must_preallocate): Likewise.
10537         (compute_argument_addresses): Likewise.
10538         (expand_call): Likewise.
10539         * cfgexpand.c (expand_call_stmt): Likewise.
10540         (expand_return): Likewise.
10541         (expand_gimple_stmt_1): Likewise.
10542         (pass_expand::execute): Likewise.
10543         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10544         (cgraph_node::remove): Likewise.
10545         (cgraph_node::dump): Likewise.
10546         (cgraph_node::verify_node): Likewise.
10547         * cgraph.h (chkp_function_instrumented_p): Likewise.
10548         (symtab_node::get_alias_target): Likewise.
10549         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10550         (cgraph_local_p): Likewise.
10551         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10552         (cgraph_edge::rebuild_references): Likewise.
10553         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10554         (walk_polymorphic_call_targets): Likewise.
10555         (cgraph_node::expand_thunk): Likewise.
10556         (symbol_table::output_weakrefs): Likewise.
10557         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10558         (ix86_handle_option): Likewise.
10559         * config/i386/constraints.md: Likewise.
10560         * config/i386/i386-builtin-types.def (BND): Likewise.
10561         (VOID): Likewise.
10562         (PVOID): Likewise.
10563         (ULONG): Likewise.
10564         * config/i386/i386-builtin.def (BDESC_END): Likewise.
10565         (BDESC_FIRST): Likewise.
10566         (BDESC): Likewise.
10567         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10568         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10569         * config/i386/i386.c (enum reg_class): Likewise.
10570         (ix86_target_string): Likewise.
10571         (ix86_option_override_internal): Likewise.
10572         (ix86_conditional_register_usage): Likewise.
10573         (ix86_valid_target_attribute_inner_p): Likewise.
10574         (ix86_set_indirect_branch_type): Likewise.
10575         (ix86_set_current_function): Likewise.
10576         (ix86_function_arg_regno_p): Likewise.
10577         (init_cumulative_args): Likewise.
10578         (ix86_function_arg_advance): Likewise.
10579         (ix86_function_arg): Likewise.
10580         (ix86_pass_by_reference): Likewise.
10581         (ix86_function_value_regno_p): Likewise.
10582         (ix86_function_value_1): Likewise.
10583         (ix86_function_value_bounds): Likewise.
10584         (ix86_return_in_memory): Likewise.
10585         (ix86_setup_incoming_vararg_bounds): Likewise.
10586         (ix86_va_start): Likewise.
10587         (indirect_thunk_need_prefix): Likewise.
10588         (print_reg): Likewise.
10589         (ix86_print_operand): Likewise.
10590         (ix86_expand_call): Likewise.
10591         (ix86_output_function_return): Likewise.
10592         (reg_encoded_number): Likewise.
10593         (BDESC_VERIFYS): Likewise.
10594         (ix86_init_mpx_builtins): Likewise.
10595         (ix86_init_builtins): Likewise.
10596         (ix86_emit_cmove): Likewise.
10597         (ix86_emit_move_max): Likewise.
10598         (ix86_expand_builtin): Likewise.
10599         (ix86_builtin_mpx_function): Likewise.
10600         (ix86_get_arg_address_for_bt): Likewise.
10601         (ix86_load_bounds): Likewise.
10602         (ix86_store_bounds): Likewise.
10603         (ix86_load_returned_bounds): Likewise.
10604         (ix86_store_returned_bounds): Likewise.
10605         (ix86_class_likely_spilled_p): Likewise.
10606         (ix86_hard_regno_mode_ok): Likewise.
10607         (x86_order_regs_for_local_alloc): Likewise.
10608         (ix86_mitigate_rop): Likewise.
10609         (ix86_bnd_prefixed_insn_p): Likewise.
10610         (ix86_mpx_bound_mode): Likewise.
10611         (ix86_make_bounds_constant): Likewise.
10612         (ix86_initialize_bounds): Likewise.
10613         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10614         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10615         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10616         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10617         (TARGET_CHKP_BOUND_MODE): Likewise.
10618         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10619         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10620         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10621         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10622         * config/i386/i386.h (TARGET_MPX): Likewise.
10623         (TARGET_MPX_P): Likewise.
10624         (VALID_BND_REG_MODE): Likewise.
10625         (FIRST_BND_REG): Likewise.
10626         (LAST_BND_REG): Likewise.
10627         (enum reg_class): Likewise.
10628         (BND_REG_P): Likewise.
10629         (BND_REGNO_P): Likewise.
10630         (BNDmode): Likewise.
10631         (ADJUST_INSN_LENGTH): Likewise.
10632         * config/i386/i386.md: Likewise.
10633         * config/i386/i386.opt: Likewise.
10634         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10635         (defined): Likewise.
10636         (LINK_MPX): Likewise.
10637         (MPX_SPEC): Likewise.
10638         (LIBMPX_SPEC): Likewise.
10639         (LIBMPXWRAPPERS_SPEC): Likewise.
10640         (CHKP_SPEC): Likewise.
10641         * config/i386/predicates.md: Likewise.
10642         * dbxout.c (dbxout_type): Likewise.
10643         * doc/extend.texi: Likewise.
10644         * doc/invoke.texi: Likewise.
10645         * doc/md.texi: Likewise.
10646         * doc/tm.texi: Likewise.
10647         * doc/tm.texi.in: Likewise.
10648         * dwarf2out.c (is_base_type): Likewise.
10649         (gen_formal_types_die): Likewise.
10650         (gen_subprogram_die): Likewise.
10651         (gen_type_die_with_usage): Likewise.
10652         (gen_decl_die): Likewise.
10653         (dwarf2out_late_global_decl): Likewise.
10654         * expr.c (expand_assignment): Likewise.
10655         (emit_storent_insn): Likewise.
10656         (store_expr_with_bounds): Likewise.
10657         (store_expr): Likewise.
10658         (expand_expr_real_1): Likewise.
10659         * expr.h (store_expr_with_bounds): Likewise.
10660         * function.c (use_register_for_decl): Likewise.
10661         (struct bounds_parm_data): Likewise.
10662         (assign_parms_augmented_arg_list): Likewise.
10663         (assign_parm_find_entry_rtl): Likewise.
10664         (assign_parm_is_stack_parm): Likewise.
10665         (assign_parm_load_bounds): Likewise.
10666         (assign_bounds): Likewise.
10667         (assign_parms): Likewise.
10668         (expand_function_start): Likewise.
10669         * gcc.c (CHKP_SPEC): Likewise.
10670         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10671         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10672         (wrestrict_dom_walker::check_call): Likewise.
10673         * gimple.c (gimple_build_call_from_tree): Likewise.
10674         * gimple.h (enum gf_mask): Likewise.
10675         (gimple_call_with_bounds_p): Likewise.
10676         (gimple_call_set_with_bounds): Likewise.
10677         * gimplify.c (gimplify_init_constructor): Likewise.
10678         * ipa-cp.c (initialize_node_lattices): Likewise.
10679         (propagate_constants_across_call): Likewise.
10680         (find_more_scalar_values_for_callers_subset): Likewise.
10681         * ipa-hsa.c (process_hsa_functions): Likewise.
10682         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10683         * ipa-icf.c (sem_function::merge): Likewise.
10684         * ipa-inline.c (early_inliner): Likewise.
10685         * ipa-pure-const.c (warn_function_noreturn): Likewise.
10686         (warn_function_cold): Likewise.
10687         (propagate_pure_const): Likewise.
10688         * ipa-ref.h (enum GTY): Likewise.
10689         * ipa-split.c (find_retbnd): Likewise.
10690         (consider_split): Likewise.
10691         (split_function): Likewise.
10692         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10693         * ipa.c (walk_polymorphic_call_targets): Likewise.
10694         (symbol_table::remove_unreachable_nodes): Likewise.
10695         (process_references): Likewise.
10696         (cgraph_build_static_cdtor_1): Likewise.
10697         * lto-cgraph.c (lto_output_node): Likewise.
10698         (output_refs): Likewise.
10699         (compute_ltrans_boundary): Likewise.
10700         (input_overwrite_node): Likewise.
10701         (input_node): Likewise.
10702         (input_cgraph_1): Likewise.
10703         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10704         * passes.c (pass_manager::execute_early_local_passes): Likewise.
10705         (class pass_chkp_instrumentation_passes): Likewise.
10706         (make_pass_chkp_instrumentation_passes): Likewise.
10707         * passes.def: Likewise.
10708         * rtl.h (struct GTY): Likewise.
10709         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10710         * stor-layout.c (layout_type): Likewise.
10711         * symtab.c: Likewise.
10712         * target.def: Likewise.
10713         * targhooks.c (default_chkp_bound_type): Likewise.
10714         (default_chkp_bound_mode): Likewise.
10715         (default_builtin_chkp_function): Likewise.
10716         (default_chkp_function_value_bounds): Likewise.
10717         (default_chkp_make_bounds_constant): Likewise.
10718         (default_chkp_initialize_bounds): Likewise.
10719         * targhooks.h (default_chkp_bound_type): Likewise.
10720         (default_chkp_bound_mode): Likewise.
10721         (default_builtin_chkp_function): Likewise.
10722         (default_chkp_function_value_bounds): Likewise.
10723         (default_chkp_make_bounds_constant): Likewise.
10724         (default_chkp_initialize_bounds): Likewise.
10725         * toplev.c (compile_file): Likewise.
10726         (process_options): Likewise.
10727         * tree-core.h (DEF_BUILTIN): Likewise.
10728         (DEF_BUILTIN_CHKP): Likewise.
10729         * tree-inline.c (declare_return_variable): Likewise.
10730         (remap_gimple_stmt): Likewise.
10731         (copy_bb): Likewise.
10732         (initialize_inlined_parameters): Likewise.
10733         (expand_call_inline): Likewise.
10734         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10735         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10736         (make_pass_ipa_chkp_produce_thunks): Likewise.
10737         (make_pass_chkp): Likewise.
10738         (make_pass_chkp_opt): Likewise.
10739         (make_pass_chkp_instrumentation_passes): Likewise.
10740         * tree-pretty-print.c (dump_generic_node): Likewise.
10741         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10742         * tree-ssa-dce.c (propagate_necessity): Likewise.
10743         (eliminate_unnecessary_stmts): Likewise.
10744         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10745         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10746         * tree-ssa-sccvn.h: Likewise.
10747         * tree-ssa-strlen.c (get_string_length): Likewise.
10748         (valid_builtin_call): Likewise.
10749         (adjust_last_stmt): Likewise.
10750         (handle_builtin_strchr): Likewise.
10751         (handle_builtin_strcpy): Likewise.
10752         (handle_builtin_stxncpy): Likewise.
10753         (handle_builtin_memcpy): Likewise.
10754         (handle_builtin_strcat): Likewise.
10755         (strlen_check_and_optimize_stmt): Likewise.
10756         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10757         * tree-streamer-in.c: Likewise.
10758         * tree-streamer.c (record_common_node): Likewise.
10759         * tree.c (tree_code_size): Likewise.
10760         (wide_int_to_tree_1): Likewise.
10761         (type_contains_placeholder_1): Likewise.
10762         (build_common_tree_nodes): Likewise.
10763         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10764         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10765         (POINTER_BOUNDS_P): Likewise.
10766         (BOUNDED_TYPE_P): Likewise.
10767         (BOUNDED_P): Likewise.
10768         (CALL_WITH_BOUNDS_P): Likewise.
10769         (pointer_bounds_type_node): Likewise.
10770         * value-prof.c (gimple_ic): Likewise.
10771         * var-tracking.c (vt_add_function_parameters): Likewise.
10772         * varasm.c (make_decl_rtl): Likewise.
10773         (assemble_start_function): Likewise.
10774         (output_constant): Likewise.
10775         (maybe_assemble_visibility): Likewise.
10776         * varpool.c (ctor_for_folding): Likewise.
10777         * chkp-builtins.def: Remove.
10778         * ipa-chkp.c: Remove.
10779         * ipa-chkp.h: Remove.
10780         * rtl-chkp.c: Remove.
10781         * rtl-chkp.h: Remove.
10782         * tree-chkp-opt.c: Remove.
10783         * tree-chkp.c: Remove.
10784         * tree-chkp.h: Remove.
10786 2018-06-07  Carl Love  <cel@us.ibm.com>
10788         * config/rs6000/vsx.md (vextract_fp_from_shorth,
10789         vextract_fp_from_shortl): Add BE support.
10791 2018-06-07  Paul Koning  <ni1d@arrl.net>
10793         * compare-elim.c (try_merge_compare): Don't merge compare if
10794         address contains a side effect.
10795         (try_eliminate_compare): Likewise.
10797 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
10799         * config.gcc: Support "tremont".
10800         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10801         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10802         PROCESSOR_TREMONT.
10803         * config/i386/i386.c (m_TREMONT): Define.
10804         (processor_target_table): Add "tremont".
10805         (PTA_TREMONT): Define.
10806         (ix86_lea_outperforms): Add TARGET_TREMONT.
10807         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10808         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10809         and M_INTEL_GOLDMONT_PLUS.
10810         (fold_builtin_cpu): Add "tremont".
10811         (ix86_add_stmt_cost): Add TARGET_TREMONT.
10812         (ix86_option_override_internal): Add "tremont".
10813         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10814         (processor_type): Add PROCESSOR_TREMONT.
10815         * config/i386/x86-tune.def: Add m_TREMONT.
10816         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10818 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10820         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10821         symbol defined for msp430i* devices to be lower case.
10823 2018-06-07  Richard Biener  <rguenther@suse.de>
10825         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10826         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
10827         Properly wrap signed arithmetic if overflow wraps.
10829 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
10831         PR tree-optimization/69615
10832         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10833         of a cast from a same precision integral SSA_NAME in a bb dominated
10834         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
10835         cast to utype if rhs2 has already a compatible type.
10837 2018-06-07  Richard Biener  <rguenther@suse.de>
10839         PR tree-optimization/85935
10840         * graphite-scop-detection.c (find_params_in_bb): Analyze
10841         condition operands with respect to the correct loop.  Assert
10842         the analysis doesn't fail.
10844 2018-06-04  Carl Love  <cel@us.ibm.com>
10846         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10847         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
10848         as it is slightly cheaper.
10849         (first_match_or_eos_index_<mode>):
10850         Calculate index using natural element order.
10851         (first_match_index_<mode>):
10852         Calculate index using natural element order.
10853         (first_match_or_eos_index_<mode>):
10854         Calculate index using natural order.
10855         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10856         for BE and LE modes.
10857         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10858         P9V_BUILTIN_VCLZLSBB_V16QI.
10859         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10860         specific.
10862 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10864         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10865         indentation and line wrap for many prototypes.  Add missing
10866         @smallexample directives around block of prototypes for vec_xl and
10867         vec_xst.
10869 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
10871         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10872         track if we pass or return IEEE 128-bit floating point.
10873         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10874         C++ mangling that is compatible with GCC 8.1.
10875         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10876         (init_cumulative_args): Note if we pass or return IEEE 128-bit
10877         floating point types.
10878         (rs6000_function_arg_advance_1): Likewise.
10879         (rs6000_mangle_type): Optionally generate mangled names that match
10880         what GCC 8.1 generated for IEEE 128-bit floating point types.
10881         (rs6000_globalize_decl_name): If we have an external function that
10882         passes or returns IEEE 128-bit types, generate a weak reference
10883         from the mangled name used in GCC 8.1 to the current mangled
10884         name.
10885         (rs6000_init_builtins): Make __ibm128 use the long double type if
10886         long double is IBM extended double.  Make __float128 use the long
10887         double type if long double is IEEE 128-bit.
10889         PR target/85657
10890         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10891         macro for __ibm128 built-in functions.
10892         (PACK_IF): Add __ibm128 pack/unpack functions.
10893         (UNPACK_IF): Likewise.
10894         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10895         enable long double built-in functions if long double is IEEE
10896         128-bit floating point.
10897         (rs6000_invalid_builtin): Update long double built-in function
10898         error message.
10899         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10900         functions, adjust the built-in function to use the long double
10901         built-in function if __ibm128 and long double are the same type.
10902         * doc/extend.texi (PowerPC builtins): Update documention for
10903         __builtin_{,un}pack_longdouble.  Add documentation for
10904         __builtin_{,un}pack_ibm128.
10906 2018-06-06  Jim Wilson  <jimw@sifive.com>
10908         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10909         (struct machine_function): New field interrupt_mode.
10910         (riscv_handle_type_attribute): New function.  Add forward declaration.
10911         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10912         (riscv_expand_epilogue): Check interrupt_mode field.
10913         (riscv_set_current_function): Check interrupt attribute args and
10914         set interrupt_mode field.
10915         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10916         (riscv_sret, riscv_uret): New.
10917         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10918         new arguments to interrupt attribute.
10920 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
10922         PR target/63177
10923         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10924         Don't handle -mcpu=power8 if -mpower9-vector is also used.
10926 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10928         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10929         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10930         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10931         several redundant entries.
10933 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
10935         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10936         type from "rtx" to "rtx_insn *".
10937         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10938         for local "call_insn", removing cast.
10939         (ix86_expand_call): Likewise, introducing a "call_insn" local.
10941 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
10943         PR tree-optimization/86066
10944         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10945         for BIT_INSERT_EXPR stores.
10947 2018-06-06  Richard Biener  <rguenther@suse.de>
10949         PR tree-optimization/86062
10950         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10951         component refs ontop
10952         of to be offsetted base.
10954 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10956         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10957         to be static and remove check on interrupt attribute name.
10959 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10961         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10962         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10964 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
10966         PR target/79924
10967         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10968         second argument.
10969         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10970         Remove second argument, change how error is called.
10971         (aarch64_layout_arg): Remove second argument from
10972         aarch64_err_no_fpadvsimd call.
10973         (aarch64_init_cumulative_args): Ditto.
10974         (aarch64_gimplify_va_arg_expr): Ditto.
10975         * config/aarch64/aarch64.md (mov<mode>): Ditto.
10977 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
10979         * config/i386/i386.md (simple_return_indirect_internal): New expander.
10980         (*simple_return_indirect_internal<mode>): Rename from
10981         simple_return_indirect_internal.  Use W mode iterator.
10982         (rstorssp): New expander.
10983         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
10984         (clrssbsy): New expander.
10985         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
10987 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10989         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10990         __typeof__.
10991         (cmse_check_pointed_object): Likewise.
10993 2018-06-05  Martin Liska  <mliska@suse.cz>
10995         PR gcov-profile/47618
10996         * doc/invoke.texi: Document how -fprofile-dir format
10997         is extended.
10999 2018-06-05  Richard Biener  <rguenther@suse.de>
11001         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
11002         removal pretend DOM info isn't available so we do not update
11003         it and only remove edges, not dominated blocks.  Actually free
11004         DOM info in case we removed something.  Remove unreachable blocks.
11005         (mfb_keep_latches): Work with either DOM info or marked backedges.
11006         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
11007         first.  Mark backedges if DOM info isn't available.
11008         (Re-)compute DOM info after cleanup_control_flow_pre.
11010 2018-06-05  Richard Biener  <rguenther@suse.de>
11012         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
11013         (locus_discrim_hasher::hash): Adjust.
11014         (locus_discrim_hasher::equal): Likewise.
11015         (next_discriminator_for_locus): Work on line directly.
11016         (same_line_p): Pass in expanded locus1 as well.
11017         (assign_discriminators): Avoid redundant location expansions.
11019 2018-06-05  Richard Biener  <rguenther@suse.de>
11021         PR tree-optimization/86046
11022         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
11023         if required after clearing TREE_ADDRESSABLE.
11025 2018-06-05  Richard Biener  <rguenther@suse.de>
11027         PR tree-optimization/86047
11028         * tree-ssa-loop.c (for_each_index): Glob handling of all
11029         decls and constants and really handle all of them.
11031 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11033         PR target/81497
11034         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
11035         qualifier_void_pointer and qualifier_const_void_pointer.
11036         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
11037         (arm_init_builtins): Handle the above.
11038         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
11039         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
11040         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
11041         void intrinsics.
11043 2018-06-05  Martin Liska  <mliska@suse.cz>
11045         * auto-profile.c (read_autofdo_file): Do not use
11046         gcov_ctr_summary struct.
11047         (afdo_callsite_hot_enough_for_early_inline): Likewise.
11048         * coverage.c (struct counts_entry): Likewise.
11049         (read_counts_file): Read just single summary entry.
11050         (get_coverage_counts): Use gcov_summary struct.
11051         * coverage.h (get_coverage_counts): Likewise.
11052         * gcov-dump.c (dump_working_sets): Likewise.
11053         (tag_summary): Dump just single summary.
11054         * gcov-io.c (gcov_write_summary): Write just histogram
11055         summary.
11056         (gcov_read_summary): Read just single summary.
11057         (compute_working_sets): Use gcov_summary struct.
11058         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
11059         of GCOV_COUNTERS_SUMMABLE.
11060         (GCOV_COUNTERS_SUMMABLE): Remove.
11061         (GCOV_FIRST_VALUE_COUNTER): Replace with
11062         GCOV_COUNTER_V_INTERVAL.
11063         (struct gcov_ctr_summary): Remove.
11064         (struct gcov_summary): Directly use fields of former
11065         gcov_ctr_summary.
11066         (compute_working_sets): Use gcov_summary struct.
11067         * gcov.c (read_count_file): Do not use ctrs fields.
11068         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
11069         struct.
11070         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
11071         struct.
11072         * profile.c: Likewise.
11073         * profile.h: Likewise.
11075 2018-06-05  Martin Liska  <mliska@suse.cz>
11077         PR gcov-profile/84846
11078         * gcov.c (output_lines): Print working directory only
11079         in intermediate format.
11081 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
11083         * config/s390/s390-builtin-types.def: Add void function type.
11084         * config/s390/s390-builtins.def: Use the function type for the
11085         tbeginc builtin.
11087 2018-06-04  Jim Wilson  <jimw@sifive.com>
11089         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
11090         to int.
11091         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
11092         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
11093         handle EH_RETURN_DATA_REGNO registers properly.
11094         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
11095         (riscv_expand_epilogue): Update comment.  Change argument name and
11096         type.  Update code to use new name and type.  Pass new args to
11097         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
11098         EXCEPTION_RETURN.
11099         * config/riscv/riscv.md (NORMAL_RETURN): New.
11100         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
11101         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
11102         (eh_return): Call gen_eh_return_internal and emit barrier.
11103         (eh_return_internal): Call riscv_expand_epilogue.
11105 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
11107         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
11108         bit_insertion field and declare can_be_merged_into method.
11109         (merged_store_group::can_be_merged_into): New method.
11110         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
11111         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
11112         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
11114 2018-06-04  Richard Biener  <rguenther@suse.de>
11116         PR tree-optimization/85955
11117         * builtins.c (fold_builtin_sincos): Convert pointers to
11118         destination to appropriate type before dereferencing.
11120 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
11122         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
11124 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
11126         * expr.c (expand_expr_real_1): Force the operand into memory if
11127         its TYPE_MODE is BLKmode and if there is no integer mode for
11128         the number of bits being extracted.
11130 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
11132         PR target/85832
11133         PR target/86036
11134         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11135         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
11137 2018-06-04  Richard Biener  <rguenther@suse.de>
11139         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
11140         (cleanup_tree_cfg_noloop): ... single caller.  Do
11141         start_recording_case_labels later.
11143 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
11145         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
11146         to _IMMINTRIN_H_INCLUDED.
11147         * config/i386/pconfigintrin.h: Ditto.
11148         * config/i386/waitpkgintrin.h: Ditto.
11149         * config/i386/immintrin.h: Add includes for sgxintrin.h,
11150         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
11151         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
11152         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
11153         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
11154         waitpkgintrin.h and cldemoteintrin.h.
11156 2018-06-04  Richard Biener  <rguenther@suse.de>
11158         PR tree-optimization/86038
11159         * tracer.c (find_best_successor): Check probability for
11160         being initialized, bail out if not.
11162 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
11164         PR target/86003
11165         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
11166         of bits to ignore when comparing architectures.
11168 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
11170         PR tree-optimization/69615
11171         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11172         maximum or minimum of the type, try to merge it also as if
11173         range1 is + [-, x - 1] or + [x + 1, -].
11175         PR c++/86025
11176         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
11178 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
11180         PR tree-optimization/86034
11181         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11182         the unsigned bitfield type in a bit insertion sequence if it does not
11183         have a larger precision than the bitfield size.
11184         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11186 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
11188         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11190 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
11192         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11193         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11194         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11195         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11197 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
11199         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11200         Disable -fdelete-null-pointer-checks for ELF toolchain.
11202 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
11203             Kito Cheng  <kito.cheng@gmail.com>
11205         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11206         (nds32le-*-*, nds32be-*-*): Integrate checking process.
11207         (nds32*-*-*): Add glibc and uclibc conditions.
11208         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11209         (TARGET_EXCEPT_UNWIND_INFO): Define.
11210         * config/nds32/elf.h: New file.
11211         * config/nds32/linux.h: New file.
11212         * config/nds32/nds32-elf.opt: New file.
11213         * config/nds32/nds32-linux.opt: New file.
11214         * config/nds32/nds32-fp-as-gp.c
11215         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11216         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11217         TARGET_LINUX_ABI.
11218         (nds32_asm_file_end): Ditto.
11219         (nds32_print_operand): Ditto.
11220         (nds32_insert_attributes): Ditto.
11221         (nds32_init_libfuncs): New function.
11222         (TARGET_HAVE_TLS): Define.
11223         (TARGET_INIT_LIBFUNCS): Define.
11224         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11225         spec content.
11226         (TARGET_ELF): Apply different mcmodel setting.
11227         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11228         been migrated into elf.h and linux.h files.
11229         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11230         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11231         (mcmodel): The content has been migrated into nds32-elf.opt and
11232         nds32-linux.opt files.
11233         * config/nds32/t-elf: New file.
11234         * config/nds32/t-linux: New file.
11236 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
11237             Shiva Chen  <shiva0217@gmail.com>
11239         * config/nds32/constants.md (unspec_volatile_element): Add
11240         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11241         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11242         optimization.
11243         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11244         (make_pass_nds32_fp_as_gp): Declare.
11245         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11246         optmization pass.
11247         (nds32_asm_function_end_prologue): Remove unused asm output.
11248         (nds32_asm_function_begin_epilogue): Remove unused asm output.
11249         (nds32_asm_file_start): Output necessary fp_as_gp information.
11250         (nds32_option_override): Adjust register usage.
11251         (nds32_expand_prologue): Consider fp_as_gp situation.
11252         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11253         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11254         (epilogue): Ditto.
11255         (return): Ditto.
11256         (simple_return): Ditto.
11257         (omit_fp_begin): Output special directive for fp_as_gp.
11258         (omit_fp_end): Output special directive for fp_as_gp.
11259         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11260         mforbid-fp-as-gp): New options.
11262 2018-06-01  Mark Wielaard  <mark@klomp.org>
11264         * dwarf2out.c (dwarf2out_finish): Remove generation of
11265         DW_AT_loclists_base.
11267 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
11269         * gimple-ssa-store-merging.c: Include gimple-fold.h.
11270         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11271         (struct merged_store_group): Add bit_insertion field.
11272         (dump_char_array): Use standard hexadecimal format.
11273         (merged_store_group::merged_store_group): Set bit_insertion to false.
11274         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
11275         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
11276         also print the mask in the dump file.
11277         (pass_store_merging::gate): Minor tweak.
11278         (imm_store_chain_info::coalesce_immediate): Fix wrong association
11279         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
11280         stores with INTEGER_CST stores.
11281         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11282         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11283         and use it throughout.  Generate bit insertion sequences if need be.
11284         (pass_store_merging::process_store): Remove redundant condition.
11285         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11287 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
11289         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11290         the 128-bit floating point types.  Fix function comment.
11292 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11294         * config/aarch64/aarch64-simd.md
11295         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11296         mnemonics.
11297         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11298         mnemonics.
11300 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
11302         PR tree-optimization/85989
11303         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11304         variable.
11305         (backprop::intersect_uses): Check it when deciding whether this
11306         is a backedge reference.
11307         (backprop::process_block): Add each phi to m_visited_phis
11308         after visiting it, then clear it at the end.
11310 2018-06-01  Richard Biener  <rguenther@suse.de>
11312         * tree-vectorizer.h (vect_dr_stmt): New function.
11313         (vect_get_load_cost): Adjust.
11314         (vect_get_store_cost): Likewise.
11315         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11316         Use vect_dr_stmt instead of DR_SMTT.
11317         (vect_record_base_alignments): Likewise.
11318         (vect_calculate_target_alignment): Likewise.
11319         (vect_compute_data_ref_alignment): Likewise and make static.
11320         (vect_update_misalignment_for_peel): Likewise.
11321         (vect_verify_datarefs_alignment): Likewise.
11322         (vector_alignment_reachable_p): Likewise.
11323         (vect_get_data_access_cost): Likewise.  Pass down
11324         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11325         (vect_get_peeling_costs_all_drs): Likewise.
11326         (vect_peeling_hash_get_lowest_cost): Likewise.
11327         (vect_enhance_data_refs_alignment): Likewise.
11328         (vect_find_same_alignment_drs): Likewise.
11329         (vect_analyze_data_refs_alignment): Likewise.
11330         (vect_analyze_group_access_1): Likewise.
11331         (vect_analyze_group_access): Likewise.
11332         (vect_analyze_data_ref_access): Likewise.
11333         (vect_analyze_data_ref_accesses): Likewise.
11334         (vect_vfa_segment_size): Likewise.
11335         (vect_small_gap_p): Likewise.
11336         (vectorizable_with_step_bound_p): Likewise.
11337         (vect_prune_runtime_alias_test_list): Likewise.
11338         (vect_analyze_data_refs): Likewise.
11339         (vect_supportable_dr_alignment): Likewise.
11340         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11341         (vect_gen_prolog_loop_niters): Likewise.
11342         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11343         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11344         modify DR_STMT.
11345         (vect_recog_mask_conversion_pattern): Likewise.
11346         (vect_try_gather_scatter_pattern): Likewise.
11347         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11348         to vect_get_store_cost.
11349         (vect_get_store_cost): Get stmt_info instead of DR.
11350         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11351         (vect_get_load_cost): Get stmt_info instead of DR.
11353 2018-06-01  Richard Biener  <rguenther@suse.de>
11355         PR middle-end/86017
11356         * gimple-fold.c (var_decl_component_p): Also allow offsetted
11357         vars wrapped in MEM_REFs.
11359 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
11361         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11362         Fix subreg tests so that we only return a choice between
11363         GENERAL_REGS and FP_REGS if the original classes included both.
11365 2018-06-01  Richard Biener  <rguenther@suse.de>
11367         PR ipa/85960
11368         * tree-ssa-structalias.c (get_function_part_constraint):
11369         Handle NULL fi->decl.
11370         (find_func_aliases_for_call): Properly handle indirect
11371         fi from direct call.
11372         (find_func_clobbers): Likewise.
11373         (ipa_pta_execute): Likewise.
11374         (create_variable_info_for): For functions that are ifunc_resolver
11375         resolve to a varinfo that contains the result of the resolver call.
11376         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11377         aliases.
11379 2018-05-31  Michael Collison  <michael.collison@arm.com>
11381         * config/aarch64/aarch64.md:
11382         (*fix_to_zero_extenddfdi2): New pattern.
11383         * gcc.target/aarch64/fix_extend1.c: New testcase.
11385 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
11387         PR middle-end/78809
11388         PR middle-end/83026
11389         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
11390         and BUILT_IN_STRNCMP_EQ.
11391         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
11392         BUILT_IN_STRNCMP_EQ.
11393         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
11394         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11395         (gimple_fold_builtin): Likewise.
11396         * tree-ssa-strlen.c (compute_string_length): New function.
11397         (determine_min_obsize): New function.
11398         (handle_builtin_string_cmp): New function to handle calls to
11399         string compare functions.
11400         (strlen_optimize_stmt): Add handling to builtin string compare
11401         calls.
11402         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11403         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11404         * tree.c (build_common_builtin_nodes): Add new defines of
11405         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
11407 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
11409         PR target/85984
11410         * bb-reorder.c (pass_partition_blocks::gate): Return false for
11411         functions with naked attribute.
11413 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
11415         * config/i386/sse.md (avx_vec_concat<mode>):
11416         Substitute concat_tg_mode mode attribute with xtg_mode.
11417         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
11418         (concat_tg_mode): Remove mode attribute.
11420 2018-05-31  Martin Sebor  <msebor@redhat.com>
11422         PR c/82063
11423         * calls.c (alloc_max_size): Correct a logic error/typo.
11424         Treat excessive arguments as infinite.  Warn for invalid arguments.
11425         * doc/invoke.texi (-Walloc-size-larger-than): Update.
11427 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
11429         PR target/85829
11430         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
11431         and movx for Haswell.
11433 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
11434             Cesar Philippidis  <cesar@codesourcery.com>
11436         PR middle-end/85879
11437         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
11438         when emitting error on private/firstprivate reductions.
11439         * omp-low.c (lower_omp_target): Avoid reference-type processing
11440         on pointers for firstprivate clause.
11442 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
11444         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
11445         (st1x2): Likewise.
11446         (st1x3): Likewise.
11447         * config/aarch64/aarch64-simd.md
11448         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11449         (aarch64_ld1_x3_<mode>): Likewise
11450         (aarch64_st1x2<VALLDIF:mode>): Likewise
11451         (aarch64_st1_x2_<mode>): Likewise
11452         (aarch64_st1x3<VALLDIF:mode>): Likewise
11453         (aarch64_st1_x3_<mode>): Likewise
11454         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11455         (vld1_s8_x3): Likewise.
11456         (vld1_u16_x3): Likewise.
11457         (vld1_s16_x3): Likewise.
11458         (vld1_u32_x3): Likewise.
11459         (vld1_s32_x3): Likewise.
11460         (vld1_u64_x3): Likewise.
11461         (vld1_s64_x3): Likewise.
11462         (vld1_f16_x3): Likewise.
11463         (vld1_f32_x3): Likewise.
11464         (vld1_f64_x3): Likewise.
11465         (vld1_p8_x3): Likewise.
11466         (vld1_p16_x3): Likewise.
11467         (vld1_p64_x3): Likewise.
11468         (vld1q_u8_x3): Likewise.
11469         (vld1q_s8_x3): Likewise.
11470         (vld1q_u16_x3): Likewise.
11471         (vld1q_s16_x3): Likewise.
11472         (vld1q_u32_x3): Likewise.
11473         (vld1q_s32_x3): Likewise.
11474         (vld1q_u64_x3): Likewise.
11475         (vld1q_s64_x3): Likewise.
11476         (vld1q_f16_x3): Likewise.
11477         (vld1q_f32_x3): Likewise.
11478         (vld1q_f64_x3): Likewise.
11479         (vld1q_p8_x3): Likewise.
11480         (vld1q_p16_x3): Likewise.
11481         (vld1q_p64_x3): Likewise.
11482         (vst1_s64_x2): Likewise.
11483         (vst1_u64_x2): Likewise.
11484         (vst1_f64_x2): Likewise.
11485         (vst1_s8_x2): Likewise.
11486         (vst1_p8_x2): Likewise.
11487         (vst1_s16_x2): Likewise.
11488         (vst1_p16_x2): Likewise.
11489         (vst1_s32_x2): Likewise.
11490         (vst1_u8_x2): Likewise.
11491         (vst1_u16_x2): Likewise.
11492         (vst1_u32_x2): Likewise.
11493         (vst1_f16_x2): Likewise.
11494         (vst1_f32_x2): Likewise.
11495         (vst1_p64_x2): Likewise.
11496         (vst1q_s8_x2): Likewise.
11497         (vst1q_p8_x2): Likewise.
11498         (vst1q_s16_x2): Likewise.
11499         (vst1q_p16_x2): Likewise.
11500         (vst1q_s32_x2): Likewise.
11501         (vst1q_s64_x2): Likewise.
11502         (vst1q_u8_x2): Likewise.
11503         (vst1q_u16_x2): Likewise.
11504         (vst1q_u32_x2): Likewise.
11505         (vst1q_u64_x2): Likewise.
11506         (vst1q_f16_x2): Likewise.
11507         (vst1q_f32_x2): Likewise.
11508         (vst1q_f64_x2): Likewise.
11509         (vst1q_p64_x2): Likewise.
11510         (vst1_s64_x3): Likewise.
11511         (vst1_u64_x3): Likewise.
11512         (vst1_f64_x3): Likewise.
11513         (vst1_s8_x3): Likewise.
11514         (vst1_p8_x3): Likewise.
11515         (vst1_s16_x3): Likewise.
11516         (vst1_p16_x3): Likewise.
11517         (vst1_s32_x3): Likewise.
11518         (vst1_u8_x3): Likewise.
11519         (vst1_u16_x3): Likewise.
11520         (vst1_u32_x3): Likewise.
11521         (vst1_f16_x3): Likewise.
11522         (vst1_f32_x3): Likewise.
11523         (vst1_p64_x3): Likewise.
11524         (vst1q_s8_x3): Likewise.
11525         (vst1q_p8_x3): Likewise.
11526         (vst1q_s16_x3): Likewise.
11527         (vst1q_p16_x3): Likewise.
11528         (vst1q_s32_x3): Likewise.
11529         (vst1q_s64_x3): Likewise.
11530         (vst1q_u8_x3): Likewise.
11531         (vst1q_u16_x3): Likewise.
11532         (vst1q_u32_x3): Likewise.
11533         (vst1q_u64_x3): Likewise.
11534         (vst1q_f16_x3): Likewise.
11535         (vst1q_f32_x3): Likewise.
11536         (vst1q_f64_x3): Likewise.
11537         (vst1q_p64_x3): Likewise.
11539 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11541         * config/msp430/msp430.c (msp430_output_labelref): Prepend
11542         user_label_prefix to name.
11544         * tree-core.h: Update comment about the format of NAME string
11545         passed to handler in attribute_spec.
11547         * config/msp430/msp430.md: Remove erroneous subreg expression from
11548         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11549         zero_extend{q,h}isi2.
11551 2018-05-30  Borislav Petkov  <bp@suse.de>
11553         * doc/extend.texi: Document some architecture specific
11554         constraints and sort entries.
11556 2018-05-30  Martin Sebor  <msebor@redhat.com>
11558         PR middle-end/85369
11559         * builtins.c (expand_builtin_stpcpy_1): New function.
11560         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11561         only if the former succeeds.
11563 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
11565         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11566         in saphira.
11568 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11570         * doc/invoke.texi (-flinker-output): Document
11572 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11574         * passes.c (ipa_write_summaries): Only modify statements if body
11575         is in memory.
11576         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11577         incrementally linking.
11578         (ipa_passes): Likewise.
11579         * lto-cgraph.c (lto_output_node): When incrementally linking do not
11580         pass down resolution info.
11581         * common.opt (flag_incremental_link): Update info.
11582         * gcc.c (plugin specs): Turn flinker-output=* to
11583         -plugin-opt=-linker-output-known
11584         * toplev.c (compile_file): Also cut compilation when doing incremental
11585         link.
11586         * flag-types. (enum lto_partition_model): Add
11587         LTO_LINKER_OUTPUT_NOLTOREL.
11588         (invoke.texi): Add -flinker-output docs.
11589         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11590         link same way as WPA; do not stream in dead initializers.
11592         * dwarf2out.c (dwarf2out_die_ref_for_decl,
11593         darf2out_register_external_decl): Support incremental link.
11595 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11597         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11599 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11601         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11602         it down to simple_object_copy_lto_debug_sections.
11603         (run_gcc): Determine incremental LTO link time and configure
11604         lto1 into non-wpa mode, disable renaming of debug sections.
11606 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11608         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11609         descriptions of various incorrectly documented functions.
11611 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11613         Revert:
11614         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11615         address check not strict.
11617 2018-05-30  Richard Biener  <rguenther@suse.de>
11619         PR tree-optimization/85964
11620         * tracer.c (better_p): Drop initialized count check, we only
11621         call the function with initialized counts now.
11622         (find_best_successor): Do find a best edge if one
11623         has uninitialized count.
11624         (find_best_predecessor): Likewise.  Do BB frequency check only
11625         if count is initialized.
11627 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
11629         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11630         (aarch64_ldrstr_offset_compare): New.
11631         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11632         load/store orderings.
11633         (aarch64_gen_adjusted_ldpstp): Likewise.
11635 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
11637         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11638         Check for subset of GENERAL_REGS and FP_REGS.
11639         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11640         r=w alternative.
11642 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
11644         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11645         and wi::to_poly_offset.  Add the current offset and then check
11646         whether the sum fits, rather than using an unchecked addition of
11647         a checked term.  Check for a shwi rather than a uhwi.
11648         * expr.c (get_bit_range): Use tree_to_poly_uint64.
11649         (store_constructor): Use poly_int_tree_p.
11650         (expand_expr_real_1): Likewise.
11651         * function.c (assign_temp): Likewise.
11652         * fold-const.c (const_binop): Use poly_int_tree_p and
11653         wi::to_poly_offset.
11654         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
11655         division.
11656         * ipa-icf-gimple.c (func_checker::compare_operand): Use
11657         to_poly_offset for MEM offsets.
11658         * ipa-icf.c (sem_variable::equals): Likewise.
11659         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11660         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11661         wi::to_poly_offset for BIT_FIELD_REF offsets.
11662         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11663         wi::to_poly_offset.
11664         * var-tracking.c (emit_note_insn_var_location): Use
11665         tree_to_poly_uint64.
11667 2018-05-29  Jim Wilson  <jimw@sifive.com>
11669         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11671 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
11673         PR target/85950
11674         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11675         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11676         sequence.
11677         (sse4_1_round<mode>2): Use nonimmediate_operand
11678         for operand 1 predicate.
11680 2018-05-29  Martin Sebor  <msebor@redhat.com>
11681             Richard Biener  <rguenther@suse.de>
11683         PR testsuite/85888
11684         * calls.c (get_size_range): Call determine_value_range instead
11685         of get_value_range..
11686         * tree-vrp.h (determine_value_range): Declared new function.
11687         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11689 2018-05-29  Richard Biener  <rguenther@suse.de>
11691         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11692         sure to use non-pattern stmts for get_earlier_stmt arguments.
11693         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11694         called on pattern stmts.
11695         (get_later_stmt): Likewise.
11697 2018-05-29  Martin Liska  <mliska@suse.cz>
11699         PR gcov-profile/85759
11700         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11701         env variables.
11703 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
11705         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11706         VEC_UNPACK_*_EXPR.
11707         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11708         VEC_PACK_*_EXPR.
11710         PR target/85918
11711         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11712         VEC_PACK_FLOAT_EXPR): New tree codes.
11713         * tree-pretty-print.c (op_code_prio): Handle
11714         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11715         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11716         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11717         * tree-inline.c (estimate_operator_cost): Likewise.
11718         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11719         * fold-const.c (const_binop): Likewise.
11720         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11721         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11722         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11723         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11724         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11725         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11726         * expr.c (expand_expr_real_2): Likewise.
11727         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11728         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11729         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11730         optabs.
11731         * optabs.c (expand_widen_pattern_expr): For
11732         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11733         sign from result type rather than operand's type.
11734         (expand_binop_directly): For vec_packu_float_optab and
11735         vec_packs_float_optab allow result type to be different from operand's
11736         type.
11737         * optabs-tree.c (optab_for_tree_code): Handle
11738         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11739         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
11740         * tree-vect-generic.c (expand_vector_operations_1):  Handle
11741         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11742         VEC_PACK_FLOAT_EXPR.
11743         * tree-vect-stmts.c (supportable_widening_operation): Handle
11744         FIX_TRUNC_EXPR.
11745         (supportable_narrowing_operation): Handle FLOAT_EXPR.
11746         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11747         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11748         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
11749         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11750         mode attributes.
11751         (vec_pack<floatprefix>_float_<mode>): New expander.
11752         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11753         attributes.
11754         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11755         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11756         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11757         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11758         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11759         Document.
11760         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11761         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11762         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11763         VEC_PACK_FLOAT_EXPR): Document.
11765 2018-05-29  Richard Biener  <rguenther@suse.de>
11767         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11768         member.
11769         (stmt_vec_info_vec): Make pointer.
11770         (init_stmt_vec_info_vec): Remove.
11771         (free_stmt_vec_info_vec): Likewise.
11772         (set_stmt_vec_info_vec): New function.
11773         (free_stmt_vec_infos): Likewise.
11774         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11775         (set_vinfo_for_stmt): Likewise.
11776         (get_earlier_stmt): Likewise.
11777         (get_later_stmt): Likewise.
11778         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11779         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11780         (vec_info::~vec_info): Free stmt_vec_infos.
11781         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11782         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11783         (pass_slp_vectorize::execute): Likewise.
11784         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11785         (free_stmt_vec_info_vec): Likewise.
11786         (set_stmt_vec_info_vec): New function.
11787         (free_stmt_vec_infos): Likewise.
11788         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11789         the global stmt_vec_info_vec.
11790         * tree-parloops.c (gather_scalar_reductions): Use
11791         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11792         vector.
11794 2018-05-29  Richard Biener  <rguenther@suse.de>
11796         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11798 2018-05-29  Martin Liska  <mliska@suse.cz>
11799             David Malcolm  <dmalcolm@redhat.com>
11801         * vec.c (test_reverse): New.
11802         (vec_c_tests): Add new test.
11803         * vec.h (vl_ptr>::reverse): New function.
11805 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
11807         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11809         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11810         and later.
11812 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11814         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11816 2018-05-28  Richard Biener  <rguenther@suse.de>
11818         PR tree-optimization/85933
11819         * tree-vect-data-refs.c (vect_record_base_alignments): Only
11820         look at stmts marked as vectorizable.
11822 2018-05-28  Richard Biener  <rguenther@suse.de>
11824         PR tree-optimization/85934
11825         * tree-vect-generic.c (expand_vector_operations_1): Hoist
11826         vector boolean check before scalar optimization.
11828 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
11830         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11831         for armv5te.
11833 2018-05-28  Mark Wielaard  <mark@klomp.org>
11835         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11836         if it is an expression containing a minus sign.
11838 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
11840         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11842 2018-05-27  Paul Koning  <ni1d@arrl.net>
11844         * config/pdp11/pdp11.md (truncsihi2): Remove.
11846 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
11847             Chung-Ju Wu  <jasonwucj@gmail.com>
11849         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11850         implementation.
11851         (unaligned_store_dw): Ditto.
11852         * config/nds32/nds32-memory-manipulation.c
11853         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11854         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11855         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11856         (emit_setmem_word_loop): Rename to ...
11857         (emit_setmem_doubleword_loop): ... this.
11858         (nds32_gen_dup_4_byte_to_word_value): New function.
11859         (nds32_gen_dup_8_byte_to_double_word_value): New function.
11860         (nds32_expand_setmem_loop): Refine implementation.
11861         (nds32_expand_setmem_loop_v3m): Ditto.
11862         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11863         pattern.
11865 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
11867         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11869 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
11871         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11872         (nds32_init_machine_status): Initialize machine->attr_naked_p and
11873         machine->attr_no_prologue_p.
11874         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11875         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11876         (nds32_expand_epilogue): Consider attr_naked_p.
11877         (nds32_expand_epilogue_v3pop): Likewise.
11878         (nds32_can_use_return_insn): Likewise.
11879         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11880         attr_no_prologue_p fields.
11881         * config/nds32/nds32.opt (mret-in-naked-func): New option.
11883 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
11885         PR target/85918
11886         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11887         attributes.
11888         * config/i386/sse.md
11889         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11890         Rename to ...
11891         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11892         ... this.
11893         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11894         Rename to ...
11895         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11896         ... this.
11897         (*<floatsuffix>floatv2div2sf2): Rename to ...
11898         (*float<floatunssuffix>v2div2sf2): ... this.
11899         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11900         (float<floatunssuffix>v2div2sf2_mask): ... this.
11901         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11902         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11903         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11904         to ...
11905         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11906         ... this.
11907         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11908         Rename to ...
11909         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11910         ... this.
11911         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11912         Rename to ...
11913         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11914         ... this.
11915         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11916         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11917         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11918         gen_ufix_truncv8dfv8si2.
11919         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11920         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11921         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11922         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11923         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11924         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11925         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11926         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11928 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
11930         PR target/85900
11931         PR target/85345
11932         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11934 2018-05-25  Jim Wilson  <jimw@sifive.com>
11936         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11937         * config/riscv/riscv.c (struct machine_function): Add
11938         interrupt_handler_p and attribute_checked_p fields.
11939         (riscv_attribute_table): Add interrupt.
11940         (riscv_interrupt_type_p): New.
11941         (riscv_save_reg_p): Save extra regs for interrupt handler.
11942         (riscv_use_save_libcall): Return false  for interrupt handler.
11943         (riscv_first_stack_step): Add forward declaration.
11944         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
11945         for interrupt handler with large frame.  Use it for saved reg list.
11946         (riscv_expand_prologue): Move flag_stack_usage_info support to
11947         eliminate duplication.
11948         (riscv_expand_epilogue): Generate mret for interrupt handler.
11949         (riscv_epilogue_uses): New.
11950         (riscv_can_use_return_insn): Return false for interrupt handler.
11951         (riscv_function_ok_for_sibcall): Likewise.
11952         (riscv_set_current_function): Add interrupt handler support.
11953         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11954         * config/riscv/riscv.md (UNSPECV_MRET): New.
11955         (GP_REGNUM): New.
11956         (riscv_frflags, riscv_fsflags): Use tab after opcode.
11957         (riscv_mret): New.
11958         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11960 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
11962         PR tree-optimization/85712
11963         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11964         this candidate has already been replaced in-situ by a copy.
11966 2018-05-25  Jason Merrill  <jason@redhat.com>
11968         PR c++/80485 - inline function non-zero address.
11969         * symtab.c (nonzero_address): Check DECL_COMDAT.
11971 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
11973         PR target/83628
11974         * config/alpha/alpha.md (ashlsi3): New insn pattern.
11975         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
11976         extension of SImode operation.  Use const123_operand predicate.
11977         (*saddsi_1): Remove.
11978         (*saddl_se_1): Ditto.
11979         (*ssubsi_1): Ditto.
11980         (*ssubl_se_1): Ditto.
11981         * config/alpha/predicates.md (const123_operand): New predicate.
11982         * config/alpha/constraints.md (P): Use IN_RANGE.
11984 2018-05-25  Richard Biener  <rguenther@suse.de>
11986         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11987         defaulted to true.
11988         (ref_maybe_used_by_stmt_p): Likewise.
11989         (stmt_may_clobber_ref_p): Likewise.
11990         (stmt_may_clobber_ref_p_1): Likewise.
11991         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11992         and pass it along.
11993         (ref_maybe_used_by_stmt_p): Likewise.
11994         (stmt_may_clobber_ref_p): Likewise.
11995         (stmt_may_clobber_ref_p_1): Likewise.
11996         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11997         the alias oracle to disambiguate DRs with stmts DR analysis
11998         couldn't handle.
11999         (vect_analyze_data_refs): Do not give up on not analyzable
12000         DRs for BB vectorization.  Remove code truncating the dataref
12001         vector.
12003 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
12005         PR target/85832
12006         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12007         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
12008         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
12010 2018-05-25  Richard Biener  <rguenther@suse.de>
12012         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
12013         function, combining stmt data ref gathering and fatal analysis
12014         parts.
12015         (vect_analyze_data_refs): Remove now redudnant code and simplify.
12016         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
12017         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
12018         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
12019         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
12021 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
12023         PR tree-optimization/85720
12024         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
12025         SCC if all partitions are builtins.
12026         (version_loop_by_alias_check): New parameter.  Generate cancelable
12027         runtime alias check if all partitions are builtins.
12028         (distribute_loop): Update call to above function.
12030 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
12032         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
12033         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
12034         (parm_default_def_partition_arg): Ditto.
12035         (set_parm_default_def_partition): Ditto.
12036         (get_parm_default_def_partitions): Ditto and make it static.
12037         (get_undefined_value_partitions): Ditto and make it static.
12038         (remove_ssa_form): Refactor call to init_var_map here.
12039         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
12040         computation for loop region.
12041         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
12042         (register_default_def): Delete.
12043         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
12044         (parm_default_def_partition_arg): Ditto.
12045         (set_parm_default_def_partition): Ditto.
12046         (get_parm_default_def_partitions): Ditto and make it static.
12047         (get_undefined_value_partitions): Ditto and make it static.
12048         (coalesce_with_default, coalesce_with_default): Update comment.
12049         (create_coalesce_list_for_region): New func factored out from
12050         create_outofssa_var_map.
12051         (populate_coalesce_list_for_outofssa): New func factored out from
12052         create_outofssa_var_map and coalesce_ssa_name.
12053         (create_outofssa_var_map): Delete.
12054         (coalesce_ssa_name): Refactor to support live range computation.
12055         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
12056         (get_parm_default_def_partitions): Delete.
12057         (get_undefined_value_partitions): Ditto.
12058         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
12059         computation for loop region.
12060         (new_tree_live_info, loe_visit_block): Ditto.
12061         (live_worklist, set_var_live_on_entry): Ditto.
12062         (calculate_live_on_exit, verify_live_on_entry): Ditto.
12063         * tree-ssa-live.h (struct _var_map): New fields.
12064         (init_var_map): Change decl.
12065         (region_contains_p): New.
12067 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
12069         * tree-ssa-live.h (live_merge_and_clear): Delete.
12071 2018-05-25  Richard Biener  <rguenther@suse.de>
12073         PR c++/85912
12074         * tree-dump.c (dequeue_and_dump): Remove access to removed
12075         operand 2 of a SWITCH_EXPR.
12077 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12079         * doc/sourcebuild.texi (vect_double_cond_arith): Include
12080         multiplication and division.
12081         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
12082         (cond_udiv@var{m}, cond_umod@var{m}): Document.
12083         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
12084         (cond_udiv_optab, cond_umod_optab): New optabs.
12085         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
12086         (IFN_COND_RDIV): New internal functions.
12087         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
12088         TRUNC_MOD_EXPR and RDIV_EXPR.
12089         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
12090         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
12091         New unspecs.
12092         (SVE_INT_BINARY): Include mult.
12093         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
12094         (optab, sve_int_op): Handle mult.
12095         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
12096         UNSPEC_COND_DIV.
12097         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
12098         for SVE_INT_BINARY_SD.
12100 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12102         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
12103         (optab, sve_int_op): Handle div and udiv.
12104         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
12105         for SVE_INT_BINARY_SD.
12106         (*<optab><mode>3): New insn for the same.
12108 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12110         * tree-vect-patterns.c: Include predict.h.
12111         (vect_recog_divmod_pattern): Restrict check for division support
12112         to when optimizing for size.
12114 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12116         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
12117         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
12118         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
12119         (gimple_match_op::set_op): Likewise.
12120         (gimple_resimplify4): Declare.
12121         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
12122         (expr::gen_transform): Likewise.
12123         (decision_tree::gen): Generate a simplification routine for 4 operands.
12124         * gimple-match-head.c (gimple_simplify): Add an overload for
12125         4 operands.  In the top-level function, handle up to 4 call
12126         arguments and call gimple_resimplify4.
12127         (gimple_resimplify4): New function.
12128         (build_call_internal): Pass a fourth operand.
12129         (maybe_push_to_seq): Likewise.
12130         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
12131         Fold VEC_COND_EXPRs of an operation and a default value into
12132         an IFN_COND_* function if possible.
12133         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
12134         New unspecs.
12135         (SVE_COND_FP_BINARY): Include them.
12136         (optab, sve_fp_op): Handle them.
12137         (SVE_INT_BINARY_REV): New code iterator.
12138         (SVE_COND_FP_BINARY_REV): New int iterator.
12139         (commutative): New int attribute.
12140         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
12141         Declare.
12142         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
12143         function.
12144         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
12145         (*cond_<optab><mode>): New patterns for reversed operands.
12147 2018-05-25  Richard Biener  <rguenther@suse.de>
12149         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
12150         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
12151         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
12152         (STMT_VINFO_GROUPED_ACCESS): Adjust.
12153         * tree-vect-data-refs.c (everywhere): Adjust users.
12154         * tree-vect-loop.c (everywhere): Likewise.
12155         * tree-vect-slp.c (everywhere): Likewise.
12156         * tree-vect-stmts.c (everywhere): Likewise.
12157         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12159 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12161         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
12162         Rename to...
12163         (gcc_cv_as_section_exclude): ... this.
12164         Try Solaris as #exclude syntax.
12165         * configure: Regenerate.
12166         * config.in: Regenerate.
12167         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
12168         SECTION_EXCLUDE.
12169         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
12170         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12172         * varasm.c (default_elf_asm_named_section): Don't check if
12173         HAVE_GAS_SECTION_EXCLUDE is defined.
12175 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12177         * doc/md.texi: Update the documentation of the cond_* optabs
12178         to mention the new final operand.  Fix GET_MODE_NUNITS call.
12179         Describe the scalar case too.
12180         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12181         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12182         instead of 2.
12183         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12184         (get_conditional_internal_fn): Update comment.
12185         * tree-vect-loop.c (vectorizable_reduction): Pass the original
12186         accumulator value as a final argument to conditional functions.
12187         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12188         a define_expand and add an "else" operand.  Assert for now that
12189         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
12190         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12191         (*cond_<optab><mode>): New patterns.
12192         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12193         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12194         (UNSPEC_COND_EOR): Delete.
12195         (optab): Remove associated mappings.
12196         (SVE_INT_BINARY): New code iterator.
12197         (sve_int_op): Remove int attribute and add "minus" to the code
12198         attribute.
12199         (SVE_COND_INT_OP): Delete.
12200         (SVE_COND_FP_OP): Rename to...
12201         (SVE_COND_FP_BINARY): ...this.
12203 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12205         * optabs.c (can_reuse_operands_p): New function.
12206         (maybe_legitimize_operands): Try to reuse the results for
12207         earlier operands.
12209 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
12211         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12212         Add {q} suffix to insn mnemonic.
12214 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12216         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12217         (msp430_warn_func_return): New.
12219 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
12221         * fold-const.c (tree_nonzero_bits): New function.
12222         * fold-const.h (tree_nonzero_bits): Likewise.
12223         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12224         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12226 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
12228         PR target/85900
12229         PR target/85345
12230         * varasm.c (assemble_alias): Check ifunc_resolver only on
12231         FUNCTION_DECL.
12233 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
12235         PR target/85903
12236         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12237         when memory input operand is handled.
12239 2018-05-24  Luis Machado  <luis.machado@linaro.org>
12241         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12242         global.
12243         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12245 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
12247         * match.pd: Delay FMA folds until after vectorization.
12249 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12251         PR target/83009
12252         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12253         address check not strict.
12255 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
12257         * gimple-match.h (gimple_match_op): New class.
12258         (mprts_hook): Replace parameters with a gimple_match_op *.
12259         (maybe_build_generic_op): Likewise.
12260         (gimple_simplified_result_is_gimple_val): Replace parameters with
12261         a const gimple_match_op *.
12262         (gimple_simplify): Replace code_helper * and tree * parameters with
12263         a gimple_match_op * parameter.
12264         (gimple_resimplify1): Replace code_helper *, tree and tree *
12265         parameters with a gimple_match_op * parameter.
12266         (gimple_resimplify2): Likewise.
12267         (gimple_resimplify3): Likewise.
12268         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12269         parameters with a gimple_match_op * parameter.
12270         * gimple-match-head.c (gimple_simplify): Change prototypes of
12271         auto-generated functions to take a gimple_match_op * instead of
12272         separate code_helper * and tree * parameters.  Make the same
12273         change in the top-level overload and update calls to the
12274         gimple_resimplify routines.  Update calls to the auto-generated
12275         functions and to maybe_push_res_to_seq in the publicly-facing
12276         operation-specific gimple_simplify overloads.
12277         (gimple_match_op::MAX_NUM_OPS): Define.
12278         (gimple_resimplify1): Replace rcode and ops with a single res_op
12279         parameter.  Update call to gimple_simplify.
12280         (gimple_resimplify2): Likewise.
12281         (gimple_resimplify3): Likewise.
12282         (mprts_hook): Replace parameters with a gimple_match_op *.
12283         (maybe_build_generic_op): Likewise.
12284         (build_call_internal): Replace type, nargs and ops with
12285         a gimple_match_op *.
12286         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12287         with a single gimple_match_op *.  Update calls to mprts_hook,
12288         build_call_internal and gimple_simplified_result_is_gimple_val.
12289         Factor out code that is common to the tree_code and combined_fn cases.
12290         * genmatch.c (expr::gen_transform): Replace tem_code and
12291         tem_ops with a gimple_match_op called tem_op.  Update calls
12292         to the gimple_resimplify functions and maybe_push_res_to_seq.
12293         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12294         res_ops.  Update call to the gimple_resimplify functions.
12295         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12296         (decision_tree::gen): Make the functions take a gimple_match_op *
12297         called res_op instead of separate res_code and res_ops parameters.
12298         Update call accordingly.
12299         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12300         and ops with a single res_op parameter.  Update calls to
12301         maybe_build_generic_op and maybe_push_res_to_seq.
12302         (fold_stmt_1): Update calls to gimple_simplify and
12303         replace_stmt_with_simplification.
12304         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12305         and gimple_simplified_result_is_gimple_val.
12306         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12307         gimple_simplify.
12308         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12309         with a gimple_match_op *.
12310         (vn_nary_build_or_lookup): Likewise.  Update call to
12311         vn_nary_build_or_lookup_1.
12312         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12313         gimple_match_op *.  Update calls to the gimple_resimplify routines
12314         and to gimple_simplified_result_is_gimple_val.
12315         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12316         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12317         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12318         (visit_nary_op): Likewise.
12319         (visit_reference_op_load): Likewise.
12321 2018-05-23  Luis Machado  <luis.machado@linaro.org>
12323         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12324         modifier for printing the step amount.
12326 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
12328         PR target/78849
12329         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12330         types.
12332 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
12334         * doc/sourcebuild.texi (Endianness): New subsubsection.
12336 2018-05-23  Luis Machado  <luis.machado@linaro.org>
12338         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12339         <prefetch_dynamic_strides>: New const bool field.
12340         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12341         prefetch_dynamic_strides.
12342         (exynosm1_prefetch_tune): Likewise.
12343         (thunderxt88_prefetch_tune): Likewise.
12344         (thunderx_prefetch_tune): Likewise.
12345         (thunderx2t99_prefetch_tune): Likewise.
12346         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12347         false.
12348         (aarch64_override_options_internal): Update to set
12349         PARAM_PREFETCH_DYNAMIC_STRIDES.
12350         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12351         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12352         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12353         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12354         prefetch-dynamic-strides setting.
12356 2018-05-23  Luis Machado  <luis.machado@linaro.org>
12358         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12359         <minimum_stride>: New const int field.
12360         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12361         minimum_stride field defaulting to -1.
12362         (exynosm1_prefetch_tune): Likewise.
12363         (thunderxt88_prefetch_tune): Likewise.
12364         (thunderx_prefetch_tune): Likewise.
12365         (thunderx2t99_prefetch_tune): Likewise.
12366         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12367         <default_opt_level>: Set to 3.
12368         (aarch64_override_options_internal): Update to set
12369         PARAM_PREFETCH_MINIMUM_STRIDE.
12370         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12371         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12372         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12373         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12374         stride is constant and is below the minimum stride threshold.
12376 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12378         * config/arm/arm-cpus.in (mode26): Delete.
12379         (armv4): Delete mode26 reference.
12380         * config/arm/arm.c (arm_configure_build_target): Delete use of
12381         isa_bit_mode26.
12383 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
12385         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12386         New insn pattern.
12387         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
12388         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
12389         for non-SSE modes.
12390         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
12391         (floatunsdidf2): Ditto.
12393 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
12395         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
12396         (fixuns_trunc<mode>si2_avx512f): Ditto.
12397         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
12398         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
12399         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
12401 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
12403         PR rtl-optimization/79985
12404         * df-scan.c (df_insn_refs_collect): Remove special case for
12405         global registers and asm statements.
12407 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
12409         * extend.texi (Global Register Variables): Rewrite the bullet list.
12410         Note that the register is available for allocation. Note that access
12411         via inline asm must use constraints. Add note about async-signal
12412         handlers. Remove paragraph about automagic register selection.
12414 2018-05-23  Richard Biener  <rguenther@suse.de>
12416         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
12417         of fixed offset from memset VN.
12419 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
12421         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
12422         first_interp field.
12423         (alloc_cand_and_find_basis): Initialize first_interp field.
12424         (slsr_process_mul): Modify first_interp field.
12425         (slsr_process_add): Likewise.
12426         (slsr_process_cast): Modify first_interp field for each new
12427         interpretation.
12428         (slsr_process_copy): Likewise.
12429         (dump_candidate): Dump first_interp field.
12430         (replace_mult_candidate): Process all interpretations, not just
12431         subsequent ones.
12432         (replace_rhs_if_not_dup): Likewise.
12433         (replace_one_candidate): Likewise.
12435 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
12437         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
12438         Add new boolean.
12439         (aarch64_needs_frame_chain): New function.
12440         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
12442 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
12444         PR target/84882
12445         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12446         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
12447         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
12448         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12449         as true for strict-align.
12450         (aarch64_can_inline_p): Perform checks even when callee has no
12451         attributes to check for strict alignment.
12452         * doc/extend.texi (AArch64 Function Attributes): Document
12453         no-strict-align.
12454         * doc/invoke.texi: (AArch64 Options): Likewise.
12456 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
12458         PR tree-optimization/85853
12459         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12460         the handling of the root of the node to...
12461         (vect_slp_analyze_node_operations_1): ...this new function,
12462         and run the whole thing with the child nodes' def types
12463         set according to their SLP node's def type.
12465 2018-05-23  Richard Biener  <rguenther@suse.de>
12467         PR middle-end/85874
12468         * tree-data-ref.c (create_runtime_alias_checks): Defer
12469         and ignore overflow warnings.
12471 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
12473         PR tree-optimization/85822
12474         * tree-vrp.c (is_masked_range_test): Fix handling of negative
12475         constants.
12477 2018-05-23  Richard Biener  <rguenther@suse.de>
12479         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12480         memset constants via native_interpret_expr.
12482 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
12484         PR target/85345
12485         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12486         attribute.
12487         (cgraph_node::create_alias): Likewise.
12488         (cgraph_node::get_availability): Check ifunc_resolver instead
12489         of looking up ifunc attribute.
12490         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12491         * varasm.c (do_assemble_alias): Likewise.
12492         (assemble_alias): Likewise.
12493         (default_binds_local_p_3): Likewise.
12494         * cgraph.h (cgraph_node): Add ifunc_resolver.
12495         (cgraph_node::only_called_directly_or_aliased_p): Return false
12496         for IFUNC resolver.
12497         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12498         attribute.
12499         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12500         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12501         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12502         instead of looking up ifunc attribute.
12504 2018-05-22  Luis Machado  <luis.machado@linaro.org>
12506         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12508 2018-05-22  Martin Sebor  <msebor@redhat.com>
12510         PR middle-end/85359
12511         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12512         only when expasion succeeds.
12513         (expand_builtin_strcmp): Same.
12514         (expand_builtin_strncmp): Same.
12516 2018-05-22  Martin Sebor  <msebor@redhat.com>
12518         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12520 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
12521             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12523         * config/aarch64/aarch64-ldpstp.md: Replace uses of
12524         aarch64_mem_pair_operand with memory_operand and delete operand swapping
12525         code.
12526         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12527         Add check for legitimate_address.
12528         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12529         (aarch64_swap_ldrstr_operands): New.
12530         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12531         Define prototype.
12533 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
12534             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12536         * config/aarch64/aarch64.md: New patterns to generate stp
12537         and ldp.
12538         (store_pair_sw, store_pair_dw): New patterns to generate stp for
12539         single words and double words.
12540         (load_pair_sw, load_pair_dw): Likewise.
12541         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12542         Delete.
12543         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12544         Delete.
12545         * config/aarch64/aarch64-ldpstp.md: Modify peephole
12546         for different mode ldpstp and add peephole for merged zero stores.
12547         Likewise for loads.
12548         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12549         Add size check.
12550         (aarch64_gen_store_pair): Rename calls to match new patterns.
12551         (aarch64_gen_load_pair): Rename calls to match new patterns.
12552         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12553         (load_pair<DREG:mode><DREG2:mode>): ... This.
12554         (store_pair<mode>): Rename to...
12555         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12556         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12557         New mode iterators.
12558         (V_INT_EQUIV): Handle SImode.
12559         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12560         New predicate.
12562 2018-05-22  Martin Sebor  <msebor@redhat.com>
12564         PR c/85623
12565         * calls.c (maybe_warn_nonstring_arg): Use string length to set
12566         or ajust the presumed bound on an operation to avoid unnecessary
12567         warnings.
12569 2018-05-22  Martin Sebor  <msebor@redhat.com>
12571         PR tree-optimization/85826
12572         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12573         assuming that a DECL necesarily has a constant size.
12575 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
12577         PR middle-end/85862
12578         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12580 2018-05-22  Richard Biener  <rguenther@suse.de>
12582         PR tree-optimization/85834
12583         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12584         non-constant and non-zero memset arguments.
12586 2018-05-22  Martin Liska  <mliska@suse.cz>
12588         PR ipa/85607
12589         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12591 2018-05-22  Richard Biener  <rguenther@suse.de>
12593         PR tree-optimization/85863
12594         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12595         comparisons when vectype is specified.
12596         (vectorizable_condition): Do not specify vectype for
12597         vect_is_simple_cond when SLP vectorizing.
12599 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
12601         PR target/85657
12602         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12603         define __ibm128 as long double.
12604         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12605         as a distinct type when IEEE 128-bit support is enabled.
12606         (init_float128_ieee): Fix up conversions between IFmode and IEEE
12607         128-bit types to use the correct functions.
12608         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12609         convert between 128-bit floating point types that have different
12610         modes but the same representation, instead of using gen_lowpart to
12611         makean alias.
12612         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12613         KFmode.
12614         (IFKF_reg): New attributes to give the register constraints for
12615         IFmode and KFmode.
12616         (extend<mode>tf2_internal): New insns to mark an explicit
12617         conversion between 128-bit floating point types that have a
12618         different mode but share the same representation.
12620 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
12622         PR tree-optimization/85814
12623         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12624         a null return from get_strinfo when unsharing the next
12625         strinfo in the chain.
12627 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
12629         PR gcc/84923
12630         * varasm.c (weak_finish): Clean up weak_decls.
12632 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12634         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12635         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12636         UNSPEC_UADALP values.
12637         * config/aarch64/iterators.md (ABAL): New int iterator.
12638         (ABDL2): Likewise.
12639         (ADALP): Likewise.
12640         (sur): Add mappings for the above.
12641         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12642         New define_insn.
12643         (aarch64_<sur>abal<mode>_4): Likewise.
12644         (aarch64_<sur>adalp<mode>_3): Likewise.
12645         (<sur>sadv16qi): New define_expand.
12647 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
12649         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12650         (*movdf_internal): Ditto.
12651         (*rcpsf2_sse): Ditto.
12652         (*rsqrtsf2_sse): Ditto.
12653         (*sqrt<mode>2_sse): Ditto.
12655 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
12657         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12658         eor3q<mode>4.
12659         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12660         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12661         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12662         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12663         vbcaxq_s64): New.
12664         * config/aarch64/arm_neon.h: Likewise.
12665         * config/aarch64/iterators.md (VQ_I): New.
12667 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
12669         * config.gcc: Add arc/t-multilib-linux to tmake_file for
12670         arc*-*-linux*.
12671         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12672         MULTILIB_DIRNAMES
12674 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
12676         * config/nds32/constraints.md (S): New constraint.
12677         * config/nds32/nds32.md (call_internal): Use constraint S.
12678         (call_value_internal): Likewise.
12679         (sibcall_internal): Likewise.
12680         (sibcall_value_internal): Likewise.
12682 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
12683             Chung-Ju Wu  <jasonwucj@gmail.com>
12685         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12686         into consideration.
12688 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
12689             Chung-Ju Wu  <jasonwucj@gmail.com>
12691         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12692         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12693         (nds32_rtx_costs_impl): Simplify.
12694         (nds32_address_cost_impl): Simplify.
12695         (nds32_init_rtx_costs): New function.
12696         (nds32_rtx_costs_speed_prefer): Likewise.
12697         (nds32_rtx_costs_size_prefer): Likewise.
12698         (nds32_address_cost_speed_prefer): Likewise.
12699         (nds32_address_cost_speed_fwprop): Likewise.
12700         (nds32_address_cost_size_prefer): Likewise.
12701         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12702         * config/nds32/nds32.c (nds32_option_override): Use
12703         nds32_init_rtx_costs function.
12705 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
12707         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12708         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12709         (TARGET_PIPELINE_N8): Likewise.
12710         (TARGET_PIPELINE_N10): Likewise.
12711         (TARGET_PIPELINE_N13): Likewise.
12712         (TARGET_PIPELINE_GRAYWOLF): Likewise.
12714 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
12716         * config/nds32/nds32-fpu.md: Update copyright year.
12718 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12720         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12722 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12724         * config/nds32/nds32.c
12725         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12726         * config/nds32/nds32.opt (minline-asm-r15): New option.
12728 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12730         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12731         MASK_HW_ABS.
12732         * config/nds32/nds32.md (abssi2): New pattern.
12734 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
12736         * config/i386/i386.md (rex64namesuffix): New mode attribute.
12737         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12738         Merge insn pattern from sse_cvtsi2ss<round_name> and
12739         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12740         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12741         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12742         using SWI48 mode iterator.
12743         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12744         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12745         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12746         pattern from sse_cvttss2si<round_saeonly_name>
12747         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12748         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12749         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12750         using SWI48 mode iterator.
12751         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12752         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12753         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12754         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12755         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12756         using SWI48 mode iterator.
12757         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12758         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12759         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12760         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12761         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12762         SWI48 mode iterator.
12763         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12764         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12765         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12766         pattern from sse_cvttsd2si<round_saeonly_name>
12767         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12769 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12771         * config/nds32/nds32-md-auxiliary.c
12772         (nds32_valid_smw_lwm_base_p): Refine.
12773         (nds32_output_smw_single_word): Refine.
12774         (nds32_output_smw_double_word): New.
12775         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12777 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12779         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12780         (nds32_output_stack_pop): Refine.
12781         (nds32_expand_unaligned_load): Refine.
12782         (nds32_expand_unaligned_store): Refine.
12784 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
12785             Chung-Ju Wu  <jasonwucj@gmail.com>
12787         * config/nds32/constants.md: Add TP_REGNUM constant.
12788         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12789         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12790         UNSPEC_ADD32.
12791         * config/nds32/nds32-doubleword.md: Consider flag_pic.
12792         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12793         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12794         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12795         and PIC code generation.
12796         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12797         code generation.
12798         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12799         optimization.
12800         * config/nds32/nds32.md: Support TLS and PIC.
12801         * config/nds32/nds32.c: Support TLS and PIC.
12802         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12803         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12804         predicate.
12806 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12808         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12809         mode with E_ prefix.
12811 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
12812             Chung-Ju Wu  <jasonwucj@gmail.com>
12814         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12815         * config/nds32/nds32-md-auxiliary.c
12816         (symbolic_reference_mentioned_p): New.
12817         (nds32_legitimize_ict_address): New.
12818         (nds32_expand_ict_move): New.
12819         (nds32_indirect_call_referenced_p): New.
12820         (nds32_symbol_binds_local_p): Delete.
12821         (nds32_long_call_p): Modify.
12822         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12823         * config/nds32/nds32-protos.h
12824         (symbolic_reference_mentioned_p): Declare.
12825         (nds32_legitimize_ict_address): Declare.
12826         (nds32_expand_ict_move): Declare.
12827         (nds32_indirect_call_referenced_p): Declare.
12828         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12829         (nds32_relax_group): Use nds32_ict_const_p as condition.
12830         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12831         (nds32_asm_file_start): Output ict_model directive in asm code.
12832         (nds32_legitimate_address_p): Consider indirect call.
12833         (nds32_print_operand): Consider indirect call.
12834         (nds32_print_operand_address): Consider indirect call.
12835         (nds32_insert_attributes): Handle "indirect_call" attribute.
12836         (TARGET_LEGITIMATE_ADDRESS_P): Define.
12837         (TARGET_LEGITIMATE_CONSTANT_P): Define.
12838         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12839         (TARGET_DELEGITIMIZE_ADDRESS): Define.
12840         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12841         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12842         (TARGET_ICT_MODEL_SMALL): Define.
12843         (TARGET_ICT_MODEL_LARGE): Define.
12844         * config/nds32/nds32.md (movsi): Consider ict model.
12845         (call, call_value): Consider ict model.
12846         (sibcall, sibcall_value): Consider ict model.
12847         * config/nds32/nds32.opt (mict-model): New option.
12848         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12849         model.
12851 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
12852             Monk Chiang  <sh.chiang04@gmail.com>
12853             Jim Wilson <jimw@sifive.com>
12855         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12856         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
12857         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12858         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12859         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
12860         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12861         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12862         compute save_libcall_adjustment properly.
12863         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12864         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12865         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12866         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12867         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12868         (ABI_SPEC): Handle mabi=ilp32e.
12869         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12870         (RVE): Add RVE mask.
12871         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12872         <-march>: Add rv32e as an example.
12874 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
12876         PR c++/82899
12877         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12878         (intra_create_variable_infos): Handle C++ constructors.
12880 2018-05-18  Martin Liska  <mliska@suse.cz>
12882         * passes.def: Remove a redundant pass.
12884 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
12886         PR bootstrap/85838
12887         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12889 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12891         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12892         (ARMv4): Update.
12893         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12894         (ARMv6m): Update.
12895         (armv2, armv2a, armv3, armv3m): Delete architectures.
12896         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12897         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12898         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12899         Delete cpus.
12900         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12901         (*mulsidi3adddi): Likewise.
12902         (mulsidi3): Likewise.
12903         (*mulsidi3_nov6): Likewise.
12904         (umulsidi3): Likewise.
12905         (umulsidi3_nov6): Likewise.
12906         (umaddsidi4): Likewise.
12907         (*umulsidi3adddi): Likewise.
12908         (smulsi3_highpart): Likewise.
12909         (*smulsi3_highpart_nov6): Likewise.
12910         (umulsi3_highpart): Likewise.
12911         (*umulsi3_highpart_nov6): Likewise.
12912         * config/arm/arm.h (arm_arch3m): Delete.
12913         * config/arm/arm.c (arm_arch3m): Delete.
12914         (arm_option_override_internal): Update armv3-related comment.
12915         (arm_configure_build_target): Delete use of isa_bit_mode32.
12916         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12917         (arm_rtx_costs_internal): Delete check of arm_arch3m.
12918         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12919         (mulsa3): Likewise.
12920         (mulusa3): Likewise.
12921         * config/arm/arm-protos.h (arm_arch3m): Delete.
12922         * config/arm/arm-tables.opt: Regenerate.
12923         * config/arm/arm-tune.md: Likewise.
12924         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12925         deleted architectures.
12927 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12929         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12930         (armv5t, armv5te): New features.
12931         (ARMv5, ARMv5e): Delete fgroups.
12932         (ARMv5t, ARMv5te): Adjust for above changes.
12933         (ARMv6m): Likewise.
12934         (armv5, armv5e): Delete arches.
12935         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12936         arm_arch5.
12937         (*call_reg_arm): Likewise.
12938         (*call_value_reg_armv5): Likewise.
12939         (*call_value_reg_arm): Likewise.
12940         (*call_symbol): Likewise.
12941         (*call_value_symbol): Likewise.
12942         (*sibcall_insn): Likewise.
12943         (*sibcall_value_insn): Likewise.
12944         (clzsi2): Likewise.
12945         (prefetch): Likewise.
12946         (define_split and define_peephole2 dependent on arm_arch5):
12947         Likewise.
12948         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12949         arm_arch5e.
12950         (TARGET_ARM_QBIT): Likewise.
12951         (TARGET_DSP_MULTIPLY): Likewise.
12952         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12953         (arm_arch5, arm_arch5e): Delete.
12954         (arm_arch5t, arm_arch5te): Declare.
12955         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12956         (arm_arch5t): Declare.
12957         (arm_option_reconfigure_globals): Update for the above.
12958         (arm_options_perform_arch_sanity_checks): Update comment, replace
12959         use of arm_arch5 with arm_arch5t.
12960         (use_return_insn): Likewise.
12961         (arm_emit_call_insn): Likewise.
12962         (output_return_instruction): Likewise.
12963         (arm_final_prescan_insn): Likewise.
12964         (arm_coproc_builtin_available): Likewise.
12965         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12966         arm_arch5e with arm_arch5t and arm_arch5te.
12967         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12968         (arm_arch5t, arm_arch5te): Declare.
12969         * config/arm/arm-tables.opt: Regenerate.
12970         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12971         * config/arm/t-multilib: Likewise.
12972         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12973         instead of arm_arch5.
12974         (*call_reg_thumb1): Likewise.
12975         (*call_value_reg_thumb1_v5): Likewise.
12976         (*call_value_reg_thumb1): Likewise.
12977         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12978         unreachable path.
12979         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12981 2018-05-18  Martin Liska  <mliska@suse.cz>
12983         PR gcov-profile/84846
12984         * doc/gcov.texi: Document -t option of gcov tool.
12986 2018-05-18  Martin Liska  <mliska@suse.cz>
12988         PR gcov-profile/84846
12989         * gcov.c (print_usage): Add new -t option.
12990         (process_args): Handle the option.
12991         (generate_results): Use stdout as output when requested by
12992         the option.
12994 2018-05-18  Martin Liska  <mliska@suse.cz>
12996         PR gcov-profile/84846
12997         * coverage.c (coverage_init): Write PWD to .gcno file.
12998         * doc/gcov.texi: Document how working directory is printed.
12999         * gcov-dump.c (dump_gcov_file): Print PWD.
13000         * gcov.c (output_intermediate_file): Likewise.
13001         (read_graph_file): Read PWD string.
13002         (output_lines): Print PWD.
13004 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13006         PR middle-end/85817
13007         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
13008         for retval and return false if all args to phi are zero.
13010 2018-05-18  Richard Biener  <rguenther@suse.de>
13012         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
13013         method.
13014         (evrp_dom_walker::before_dom_children): Call it.
13016 2018-05-18  Richard Biener  <rguenther@suse.de>
13018         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
13019         results when processing array refs with variable index.
13021 2018-05-18  Toon Moene  <toon@moene.org>
13023         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
13024         directly after that of -floop-interchange. Indicate that both
13025         options are enabled by default when specifying -O3.
13027 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13029         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
13030         iterator.  Delete separate integer-mode vec_set<mode> expander.
13031         (aarch64_simd_vec_setv2di): Delete.
13032         (vec_setv2di): Delete.
13033         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
13034         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
13035         the "w, r" alternative.
13037 2018-05-18  Martin Liska  <mliska@suse.cz>
13039         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
13040         * tree-pass.h (make_pass_lower_switch_O0): New function.
13041         * tree-switch-conversion.c (node_has_low_bound): Remove.
13042         (node_has_high_bound): Likewise.
13043         (node_is_bounded): Likewise.
13044         (class pass_lower_switch): Make it a template type and create
13045         two instances.
13046         (pass_lower_switch::execute): Add template argument.
13047         (make_pass_lower_switch): New function.
13048         (make_pass_lower_switch_O0): New function.
13049         (do_jump_if_equal): Remove.
13050         (emit_case_nodes): Simplify to just handle all 3 cases and leave
13051         all the hard work to tree optimization passes.
13053 2018-05-18  Martin Liska  <mliska@suse.cz>
13055         * dbgcnt.c (limit_low): Renamed from limit.
13056         (limit_high): New variable.
13057         (dbg_cnt_is_enabled): Check for upper limit.
13058         (dbg_cnt): Adjust dumping.
13059         (dbg_cnt_set_limit_by_index): Add new argument for high
13060         value.
13061         (dbg_cnt_set_limit_by_name): Likewise.
13062         (dbg_cnt_process_single_pair): Parse new format.
13063         (dbg_cnt_process_opt): Use strtok.
13064         (dbg_cnt_list_all_counters): Remove 'value' and add
13065         'limit_high'.
13066         * doc/invoke.texi: Document changes.
13068 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
13070         * doc/sourcebuild.texi (scalar_all_fma): Document.
13071         * tree.def (FMA_EXPR): Delete.
13072         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
13073         * internal-fn.c (ternary_direct): New macro.
13074         (expand_ternary_optab_fn): Likewise.
13075         (direct_ternary_optab_supported_p): Likewise.
13076         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
13077         * builtins.c (fold_builtin_fma): Delete.
13078         (fold_builtin_3): Don't call it.
13079         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
13080         * expr.c (expand_expr_real_2): Likewise.
13081         * fold-const.c (operand_equal_p): Likewise.
13082         (fold_ternary_loc): Likewise.
13083         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
13084         * gimple.c (DEFTREECODE): Likewise.
13085         * gimplify.c (gimplify_expr): Likewise.
13086         * optabs-tree.c (optab_for_tree_code): Likewise.
13087         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
13088         * tree-eh.c (operation_could_trap_p): Likewise.
13089         (stmt_could_throw_1_p): Likewise.
13090         * tree-inline.c (estimate_operator_cost): Likewise.
13091         * tree-pretty-print.c (dump_generic_node): Likewise.
13092         (op_code_prio): Likewise.
13093         * tree-ssa-loop-im.c (stmt_cost): Likewise.
13094         * tree-ssa-operands.c (get_expr_operands): Likewise.
13095         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
13096         * fold-const-call.h (fold_fma): Delete.
13097         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
13098         CFN_FNMA and CFN_FNMS.
13099         (fold_fma): Delete.
13100         * genmatch.c (combined_fn): New enum.
13101         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
13102         (commutative_op): New function.
13103         (commutate): Use it.  Handle more than 2 operands.
13104         (dt_operand::gen_gimple_expr): Use commutative_op.
13105         (parser::parse_expr): Allow :c to be used with non-binary
13106         operators if the commutative operand is known.
13107         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
13108         CFN_FMS, CFN_FNMA and CFN_FNMS.
13109         (backprop::process_assign_use): Remove FMA_EXPR handling.
13110         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
13111         (gen_hsa_fma): New function.
13112         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
13113         IFN_FNMA and IFN_FNMS.
13114         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
13115         * gimple-fold.h (follow_all_ssa_edges): Declare.
13116         * gimple-fold.c (follow_all_ssa_edges): New function.
13117         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
13118         gimple_build interface and use follow_all_ssa_edges to fold the result.
13119         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
13120         instead of checking for optabs directly.
13121         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
13122         rather than FMA_EXPRs.
13123         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
13124         call to IFN_FMA instead of an FMA_EXPR.
13126 2018-05-17  Jim Wilson  <jimw@sifive.com>
13128         * expr.c (do_tablejump): When converting index to Pmode, if we have a
13129         sign extended promoted subreg, and the range does not have the sign bit
13130         set, then do a sign extend.
13132         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
13133         test, check for sign extended subreg and/or constant operands, and
13134         do a sign extend in that case.
13136 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
13138         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
13139         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
13140         Add untyped.
13141         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
13142         Change logics_shift_reg to logics_shift_imm.
13143         (thunderx2t99_fp_loadpair_basic): Delete.
13144         (thunderx2t99_fp_storepair_basic): Delete.
13145         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
13146         (thunderx2t99_asimd_polynomial): Delete.
13147         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13148         and neon_fp_mul_d_scalar_q.
13149         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
13150         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
13151         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
13152         (thunderx2t99_asimd_lut): Add missing tbl types.
13153         (thunderx2t99_asimd_ext): Delete.
13154         (thunderx2t99_asimd_load1_1_mult): Delete.
13155         (thunderx2t99_asimd_load1_2_mult): Delete.
13156         (thunderx2t99_asimd_load1_ldp): New.
13157         (thunderx2t99_asimd_load1): New.
13158         (thunderx2t99_asimd_load2): Add missing *load2* types.
13159         (thunderx2t99_asimd_load3): New.
13160         (thunderx2t99_asimd_load4): New.
13161         (thunderx2t99_asimd_store1_1_mult): Delete.
13162         (thunderx2t99_asimd_store1_2_mult): Delete.
13163         (thunderx2t99_asimd_store2_mult): Delete.
13164         (thunderx2t99_asimd_store2_onelane): Delete.
13165         (thunderx2t99_asimd_store_stp): New.
13166         (thunderx2t99_asimd_store1): New.
13167         (thunderx2t99_asimd_store2): New.
13168         (thunderx2t99_asimd_store3): New.
13169         (thunderx2t99_asimd_store4): New.
13171 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
13173         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13174         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
13176 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
13177             Segher Boessenkool  <segher@kernel.crashing.org>
13179         PR target/85698
13180         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13181         operand.
13183 2018-05-17  Richard Biener  <rguenther@suse.de>
13185         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13186         for pruning loop and prune defs feeding only already visited PHIs.
13188 2018-05-17  Richard Biener  <rguenther@suse.de>
13190         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13192 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
13193             Richard Biener  <rguenther@suse.de>
13195         PR tree-optimization/85793
13196         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13197         for VMAT_ELEMENTWISE.
13199 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
13201         * internal-fn.h (lookup_internal_fn): Declare
13202         * internal-fn.c (lookup_internal_fn): New function.
13203         * gimple.c (gimple_build_call_from_tree): Handle calls to
13204         internal functions.
13205         * gimple-pretty-print.c (dump_gimple_call): Print "." before
13206         internal function names.
13207         * tree-pretty-print.c (dump_generic_node): Likewise.
13208         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13210 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
13212         * gimple-fold.h (gimple_build): Make the function forms take
13213         combined_fn rather than built_in_function.
13214         (gimple_simplify): Likewise.
13215         * gimple-match-head.c (gimple_simplify): Likewise.
13216         * gimple-fold.c (gimple_build): Likewise.
13217         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13218         rather than gimple_build_call_internal.
13219         (get_initial_defs_for_reduction): Likewise.
13220         (vect_create_epilog_for_reduction): Likewise.
13221         (vectorizable_live_operation): Likewise.
13223 2018-05-17  Martin Liska  <mliska@suse.cz>
13225         * gimple-ssa-sprintf.c (format_directive): Do not use
13226         space in between 'G_' and '('.
13228 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
13230         PR target/85323
13231         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13232         even if the mask is not all ones.
13234         PR target/85323
13235         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13236         vector.
13237         (ix86_gimple_fold_builtin): Likewise.
13239         PR target/85323
13240         * config/i386/i386.c: Include tree-vector-builder.h.
13241         (ix86_vector_shift_count): New function.
13242         (ix86_fold_builtin): Fold shift builtins by scalar count.
13243         (ix86_gimple_fold_builtin): Likewise.
13245         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13246         _mm512_setzero): New intrinsics.
13248 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
13249             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13251         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13252         code generation for cases where splatting a value is not useful.
13253         * simplify-rtx.c (simplify_ternary_operation): Simplify
13254         vec_merge across a vec_duplicate and a paradoxical subreg forming
13255         a vector mode to a vec_concat.
13257 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
13259         * config.gcc: Support "goldmont-plus".
13260         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13261         "goldmont-plus".
13262         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13263         PROCESSOR_GOLDMONT_PLUS.
13264         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13265         (processor_target_table): Add "goldmont-plus".
13266         (PTA_GOLDMONT_PLUS): Define.
13267         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13268         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13269         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13270         (fold_builtin_cpu): Add "goldmont-plus".
13271         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13272         (ix86_option_override_internal): Add "goldmont-plus".
13273         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13274         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13275         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13276         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13278 2018-05-17  Richard Biener  <rguenther@suse.de>
13280         PR tree-optimization/85757
13281         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13282         remove defs that only feed that PHI from further processing.
13284 2018-05-16  Jim Wilson  <jimw@sifive.com>
13286         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13287         asterisk to name.
13288         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13290 2018-05-16  Mark Wielaard  <mark@klomp.org>
13292         * dwarf2out.c (count_index_strings): New function.
13293         (output_indirect_strings): Call count_index_strings and generate
13294         header for dwarf_version >= 5.
13296 2018-05-16  Mark Wielaard  <mark@klomp.org>
13298         * dwarf2out.c (dwarf_FORM): New function.
13299         (set_indirect_string): Use dwarf_FORM.
13300         (reset_indirect_string): Likewise.
13301         (size_of_die): Likewise.
13302         (value_format): Likewise.
13303         (output_die): Likewise.
13304         (add_skeleton_AT_string): Likewise.
13305         (output_macinfo_op): Likewise.
13306         (index_string): Likewise.
13307         (output_index_string_offset): Likewise.
13308         (output_index_string): Likewise.
13309         (count_index_strings): Likewise.
13311 2018-05-16  Carl Love  <cel@us.ibm.com>
13313         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13314         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13316 2018-05-16  Martin Jambor  <mjambor@suse.cz>
13318         * ipa-prop.c (ipa_free_all_edge_args): Remove.
13319         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13321 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
13323         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13324         (fnma<mode>4): Likewise.
13325         (fms<mode>4): Likewise.
13326         (fnms<mode>4): Likewise.
13327         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13328         (aarch64_fnma<mode>4): Likewise.
13329         (aarch64_fms<mode>4): Likewise.
13330         (aarch64_fnms<mode>4): Likewise.
13331         (aarch64_fnmadd<mode>4): Likewise.
13333 2018-05-16  Jason Merrill  <jason@redhat.com>
13335         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
13337 2018-05-16  Richard Biener  <rguenther@suse.de>
13339         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13340         (dump_stmt_cost): Declare.
13341         (add_stmt_cost): Dump cost we add.
13342         (add_stmt_costs): New function.
13343         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13344         No longer exported.
13345         (vect_analyze_stmt): Adjust prototype.
13346         (vectorizable_condition): Likewise.
13347         (vectorizable_live_operation): Likewise.
13348         (vectorizable_reduction): Likewise.
13349         (vectorizable_induction): Likewise.
13350         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13351         cost vector to pass to vectorizable_ and record afterwards.
13352         (vect_model_reduction_cost): Take cost vector argument and adjust.
13353         (vect_model_induction_cost): Likewise.
13354         (vectorizable_reduction): Likewise.
13355         (vectorizable_induction): Likewise.
13356         (vectorizable_live_operation): Likewise.
13357         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13358         SLP_TREE_NUMBER_OF_VEC_STMTS.
13359         (vect_analyze_slp_cost_1): Remove.
13360         (vect_analyze_slp_cost): Likewise.
13361         (vect_slp_analyze_node_operations): Take visited args and
13362         a target cost vector.  Avoid processing already visited stmt sets.
13363         (vect_slp_analyze_operations): Use a local cost vector to gather
13364         costs and register those of non-discarded instances.
13365         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13366         (vect_schedule_slp_instance): Remove copying of
13367         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
13368         zero.
13369         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13370         adding cost.  Record cost entry location.
13371         (vect_prologue_cost_for_slp_op): Function to compute cost of
13372         a constant or invariant generated for SLP vect in the prologue,
13373         split out from vect_analyze_slp_cost_1.
13374         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
13375         (vect_model_promotion_demotion_cost): Likewise.
13376         (vect_model_store_cost): Likewise, make static.
13377         (vect_model_load_cost): Likewise.
13378         (vectorizable_bswap): Add cost vector arg and adjust.
13379         (vectorizable_call): Likewise.
13380         (vectorizable_simd_clone_call): Likewise.
13381         (vectorizable_conversion): Likewise.
13382         (vectorizable_assignment): Likewise.
13383         (vectorizable_shift): Likewise.
13384         (vectorizable_operation): Likewise.
13385         (vectorizable_store): Likewise.
13386         (vectorizable_load): Likewise.
13387         (vectorizable_condition): Likewise.
13388         (vectorizable_comparison): Likewise.
13389         (can_vectorize_live_stmts): Likewise.
13390         (vect_analyze_stmt): Likewise.
13391         (vect_transform_stmt): Adjust calls to vectorizable_*.
13392         * tree-vectorizer.c: Include gimple-pretty-print.h.
13393         (dump_stmt_cost): New function.
13395 2018-05-16  Richard Biener  <rguenther@suse.de>
13397         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
13398         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
13399         * tree-ssa-dse.c: Include tree-ssa-loop.h.
13400         (check_name): New callback.
13401         (dse_classify_store): Track cycles via a visited bitmap of PHI
13402         defs and simplify handling of in-loop and across loop dead stores
13403         and properly fail for loop-variant refs.  Handle byte-tracking with
13404         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13405         limiting the walk.
13407 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
13409         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
13410         (vect_get_mask_type_for_stmt): Likewise.
13411         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
13412         split out from...
13413         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
13414         to determine the statement's vector type and the vector type that
13415         should be used for calculating nunits.  Deal with cases in which
13416         the type has to be deferred.
13417         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
13418         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
13419         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
13420         (vect_determine_vf_for_stmt): New functions, split out from...
13421         (vect_determine_vectorization_factor): ...here.
13422         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
13423         (vect_get_mask_type_for_stmt): New functions, split out from
13424         vect_determine_vectorization_factor.
13426 2018-05-16  Richard Biener  <rguenther@suse.de>
13428         * tree-cfg.c (verify_gimple_assign_ternary): Properly
13429         verify the [VEC_]COND_EXPR embedded comparison.
13431 2018-05-15  Martin Sebor  <msebor@redhat.com>
13433         PR tree-optimization/85753
13434         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
13435         RECORD_TYPE in addition to ARRAY_TYPE.
13437 2018-05-15  Martin Sebor  <msebor@redhat.com>
13439         PR middle-end/85643
13440         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
13442 2018-05-15  Richard Biener  <rguenther@suse.de>
13444         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
13445         add by_clobber_p one.  Change algorithm to collect all defs
13446         representing uses we need to walk and try reducing them to
13447         a single one before failing.
13448         (dse_dom_walker::dse_optimize_stmt): Adjust.
13450 2018-05-13  Mark Wielaard  <mark@klomp.org>
13452         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13453         (size_of_loc_descr): Likewise.
13454         (output_loc_operands): Likewise.
13455         (output_loc_operands_raw): Likewise.
13456         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13457         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13458         (hash_loc_operands): Likewise.
13459         (compare_loc_operands): Likewise.
13461 2018-05-14  Mark Wielaard  <mark@klomp.org>
13463         * dwarf2out.c (count_index_addrs): New function.
13464         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13466 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13468         PR tree-optimization/83648
13469         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13470         return value as malloc candidate.
13472 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13474         PR ipa/85734
13475         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13476         param as true in call to suggest_attribute.
13478 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
13480         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13481         -mreadonly-in-sdata.
13483 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13485         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13486         New pattern.
13487         (aarch64_crypto_aesd_fused): Likewise.
13489 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
13491         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13492         (movsi_aarch64): Likewise.
13493         (load_pairsi): Likewise.
13494         (load_pairdi): Likewise.
13495         (store_pairsi): Likewise.
13496         (store_pairdi): Likewise.
13497         (load_pairsf): Likewise.
13498         (load_pairdf): Likewise.
13499         (store_pairsf): Likewise.
13500         (store_pairdf): Likewise.
13501         (zero_extend): Likewise.
13502         (trunc): Swap alternatives.
13503         (fcvt_target): Add '?' to prefer w over r.
13505 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
13507         PR target/85756
13508         * config/i386/i386.md: Disallow non-commutative arithmetics in
13509         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13510         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
13511         in the peephole2 before it.
13513 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
13515         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13516         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13517         (ix86_handle_option): Handle -mcldemote.
13518         * config.gcc: New header.
13519         * config/i386/cldemoteintrin.h: New file.
13520         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13521         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13522         -mcldemote.
13523         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13524         OPTION_MASK_ISA_CLDEMOTE.
13525         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13526         (ix86_valid_target_attribute_inner_p): Ditto.
13527         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13528         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13529         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13530         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13531         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13532         (cldemote): New.
13533         * config/i386/i386.opt: Add -mcldemote.
13534         * config/i386/x86intrin.h: New header.
13535         * doc/invoke.texi: Add -mcldemote.
13537 2018-05-14  Richard Biener  <rguenther@suse.de>
13539         * doc/match-and-simplify.texi: Adjust :s documentation.
13541 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
13543         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13544         intended memcpy size.
13545         (REORDER_45): Likewise.
13547 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
13549         * sort.cc: New file.
13550         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13551         * vec.c (qsort_chk): Use gcc_qsort.
13552         * Makefile.in (OBJS-libcommon): Add sort.o.
13553         (build/sort.o): New target.  Use it...
13554         (BUILD_RTL): ... here, and...
13555         (build/gencfn-macros): ... here, and...
13556         (build/genmatch): ... here.
13558 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13559             Chung-Ju Wu  <jasonwucj@gmail.com>
13561         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13562         * config/nds32/nds32-graywolf.md: New file.
13563         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13564         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13565         pipeline.
13566         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13567         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13568         * config/nds32/nds32.md (pipeline_model): Add graywolf.
13569         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13570         * config/nds32/pipelines.md: Include n15 settings.
13572 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13573             Chung-Ju Wu  <jasonwucj@gmail.com>
13575         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13576         * config/nds32/nds32-n13.md: New file.
13577         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13578         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13579         pipeline.
13580         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13581         * config/nds32/nds32.md (pipeline_model): Add n13.
13582         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13583         * config/nds32/pipelines.md: Include n13 settings.
13585 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13586             Chung-Ju Wu  <jasonwucj@gmail.com>
13588         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13589         * config/nds32/nds32-n10.md: New file.
13590         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13591         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13592         pipeline.
13593         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13594         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13595         * config/nds32/nds32.md (pipeline_model): Add n10.
13596         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13597         * config/nds32/pipelines.md: Include n10 settings.
13599 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
13600             Kito Cheng  <kito.cheng@gmail.com>
13601             Chung-Ju Wu  <jasonwucj@gmail.com>
13603         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13604         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13605         Add enum values for DSP extension instructions.
13606         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13607         New constraints.
13608         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13609         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13610         New code iterators.
13611         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13612         * config/nds32/nds32-dspext.md: New file for DSP implementation.
13613         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13614         * config/nds32/nds32-intrinsic.md: Likewise.
13615         * config/nds32/nds32_intrinsic.h: Likewise.
13616         * config/nds32/nds32-md-auxiliary.c: Likewise.
13617         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13618         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13619         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13620         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13621         * config/nds32/nds32-protos.h: New declarations for DSP extension.
13622         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13623         TYPE_DMAC in switch statement.
13624         * config/nds32/nds32.c: New checking and implementation for DSP
13625         extension instructions.
13626         * config/nds32/nds32.h: Likewise.
13627         * config/nds32/nds32.md: Likewise.
13628         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13629         * config/nds32/predicates.md: Implement new predicates for DSP
13630         extension.
13632 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
13634         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13635         Reformat alternatives and attributes so it is easier to identify
13636         which constraints/attributes go with which instruction.
13637         (mov<mode>_hardfloat32, FMOVE64): Likewise.
13638         (mov<mode>_softfloat32, FMOVE64): Likewise.
13639         (mov<mode>_hardfloat64, FMOVE64): Likewise.
13640         (mov<mode>_softfloat64, FMOVE64): Likewise.
13642 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13644         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13645         subsection.
13646         (Basic PowerPC Built-in Functions): The new name of the
13647         subsection previously known as "PowerPC Built-in Functions".
13648         (Basic PowerPC Built-in Functions Available on all Configurations):
13649         New subsubsection.
13650         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13651         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13652         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13653         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13655 2018-05-11  Martin Jambor  <mjambor@suse.cz>
13657         PR ipa/85655
13658         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13659         single const.
13661 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
13663         PR target/85733
13664         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13666 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
13668         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13669         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13670         (ix86_handle_option): Handle -mwaitpkg.
13671         * config.gcc: New header.
13672         * config/i386/cpuid.h (bit_WAITPKG): New bit.
13673         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13674         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13675         function type.
13676         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13677         OPTION_MASK_ISA_WAITPKG.
13678         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13679         (ix86_option_override_internal): Add PTA_WAITPKG.
13680         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13681         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13682         IX86_BUILTIN_TPAUSE.
13683         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13684         __builtin_ia32_umwait and __builtin_ia32_tpause.
13685         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13686         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13687         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13688         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13689         UNSPECV_TPAUSE): New.
13690         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13691         * config/i386/i386.opt: Add -mwaitpkg.
13692         * config/i386/waitpkgintrin.h: New file.
13693         * config/i386/x86intrin.h: New header.
13694         * doc/invoke.texi: Add -mwaitpkg.
13696 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
13698         PR target/85606
13699         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13700         equivalent.
13701         (cortex-m0): Use armv6s-m isa.
13702         (cortex-m0plus): Likewise.
13703         (cortex-m1): Likewise.
13704         (cortex-m0.small-multiply): Likewise.
13705         (cortex-m0plus.small-multiply): Likewise.
13706         (cortex-m1.small-multiply): Likewise.
13708 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
13709             Jakub Jelinek  <jakub@redhat.com>
13711         PR tree-optimization/85692
13712         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13713         source permute as well.
13715 2018-05-11  Martin Liska  <mliska@suse.cz>
13717         PR sanitizer/85556
13718         * doc/extend.texi: Document LLVM style format for no_sanitize
13719         attribute.
13721 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
13723         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13724         mode_supports_vsx_dform_quad to mode_supports_dq_form.
13725         (mode_supports_vsx_dform_quad): Likewise.
13726         (mode_supports_vmx_dform): Move these functions to be next to the
13727         other mode_supports functions.
13728         (mode_supports_dq_form): Likewise.
13729         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13730         mode_supports_dq_form.
13731         (reg_offset_addressing_ok_p): Likewise.
13732         (offsettable_ok_by_alignment): Likewise.
13733         (rs6000_legitimate_offset_address_p): Likewise.
13734         (legitimate_lo_sum_address_p): Likewise.
13735         (rs6000_legitimize_address): Likewise.
13736         (rs6000_legitimize_reload_address): Likewise.
13737         (rs6000_secondary_reload_inner): Likewise.
13738         (rs6000_preferred_reload_class): Likewise.
13739         (rs6000_output_move_128bit): Likewise.
13741 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
13743         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13744         Generate SImode target register for null target.
13745         <case IX86_BUILTIN_XGETBV>: Ditto.
13746         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13747         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13749 2018-05-10  Carl Love  <cel@us.ibm.com>
13751         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13752         dcbtt and dcbtstt if operands[2] is 0.
13754 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
13756         PR target/85693
13757         * config/i386/sse.md (usadv64qi): New expander.
13759 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
13761         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13762         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13763         -maltivec=be support.
13764         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13765         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13766         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13767         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13768         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13769         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13770         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13771         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13772         altivec_vsumsws): Adjust.
13773         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13774         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13775         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13776         support.
13777         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13778         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13779         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13780         (altivec_lve<VI_char>x): Delete expand.
13781         (*altivec_lve<VI_char>x_internal): Rename to...
13782         (altivec_lve<VI_char>x): ... this.
13783         (altivec_lvxl_<mode>): Delete expand.
13784         (*altivec_lvxl_<mode>_internal): Rename to ...
13785         (altivec_lvxl_<mode>): ... this.
13786         (altivec_stvxl_<mode>): Delete expand.
13787         (*altivec_stvxl_<mode>_internal): Rename to ...
13788         (altivec_stvxl_<mode>): ... this.
13789         (altivec_stve<VI_char>x): Delete expand.
13790         (*altivec_stve<VI_char>x_internal): Rename to ...
13791         (altivec_stve<VI_char>x): ... this.
13792         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13793         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13794         reduc_plus_scal_<mode>): Adjust.
13795         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13796         comment.
13797         (rs6000_cpu_cpp_builtins): Adjust.
13798         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13799         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13800         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13801         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13802         -maltivec=be support.
13803         (rs6000_split_vec_extract_var): Adjust.
13804         (rs6000_split_v4si_init): Adjust.
13805         (swap_selector_for_mode): Delete.
13806         (altivec_expand_lvx_be, altivec_expand_stvx_be,
13807         altivec_expand_stvex_be): Delete.
13808         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13809         -maltivec=be support.
13810         (rs6000_gimple_fold_builtin): Ditto.
13811         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13812         Adjust.
13813         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13814         (TARGET_DIRECT_MOVE_64BIT): Adjust.
13815         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13816         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13817         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13818         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13819         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13820         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13821         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13822         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13823         anonymous split): Adjust.
13824         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13825         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13827 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
13829         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13830         when --with-gxx-include-dir is also specified.
13831         * configure: Regenerate.
13833 2018-05-09  Jim Wilson  <jimw@sifive.com>
13835         PR target/84797
13836         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13837         * config/riscv/t-withmultilib: New.
13838         * config/riscv/withmultilib.h: New.
13839         * doc/install.texi: Document RISC-V --with-multilib-list support.
13841 2018-05-09  Richard Biener  <rguenther@suse.de>
13843         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13844         vector.
13845         (vect_bb_vectorization_profitable_p): Adjust.  Compute
13846         actual scalar cost using the cost vector and the add_stmt_cost
13847         machinery.
13849 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13851         PR rtl-optimization/85645
13852         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13853         in the REG_CFA_REGISTER note for LR, don't leave it empty.
13855 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13857         PR rtl-optimization/85645
13858         * shrink-wrap.c (spread_components): Return a boolean saying if
13859         anything was changed.
13860         (try_shrink_wrapping_separate): Iterate spread_components until
13861         nothing changes anymore.
13863 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13865         PR rtl-optimization/85645
13866         * regrename.c (build_def_use): Also kill the chains that include the
13867         destination of a REG_CFA_REGISTER note.
13869 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13871         PR rtl-optimization/85645
13872         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13873         insn that has a REG_CFA_REGISTER note.
13875 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
13877         * cfgexpand.c (expand_clobber): New function.
13878         (expand_gimple_stmt_1): Use it.
13879         * tree-vect-stmts.c (vect_clobber_variable): New function,
13880         split out from...
13881         (vectorizable_simd_clone_call): ...here.
13882         (vectorizable_store): Emit a clobber either side of an
13883         IFN_STORE_LANES sequence.
13884         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13886 2018-05-09  Tom de Vries  <tom@codesourcery.com>
13888         PR target/85626
13889         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13890         (define_insn "trap_if_false"): Add exit after trap.
13892 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
13894         PR rtl-optimization/85638
13895         * bb-reorder.c: Include common/common-target.h.
13896         (create_forwarder_block): New function extracted from...
13897         (fix_up_crossing_landing_pad): ...here.  Rename into...
13898         (dw2_fix_up_crossing_landing_pad): ...this.
13899         (sjlj_fix_up_crossing_landing_pad): New function.
13900         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13901         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13902         from both partitions and exit the loop after one iteration.
13904 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13906         Revert:
13907         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13908         subsection.
13909         (Basic PowerPC Built-in Functions): The new name of the
13910         subsection previously known as "PowerPC Built-in Functions".
13911         (Basic PowerPC Built-in Functions Available on all Configurations):
13912         New subsubsection.
13913         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13914         subsubsection.
13915         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13916         subsubsection.
13917         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13918         subsubsection.
13919         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13920         subsubsection.
13922 2018-05-08  Jim Wilson  <jimw@sifive.com>
13924         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13925         (LD_EMUL_SUFFIX): New.
13926         (LINK_SPEC): Use it.
13928 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13930         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13931         subsection.
13932         (Basic PowerPC Built-in Functions): The new name of the
13933         subsection previously known as "PowerPC Built-in Functions".
13934         (Basic PowerPC Built-in Functions Available on all Configurations):
13935         New subsubsection.
13936         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13937         subsubsection.
13938         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13939         subsubsection.
13940         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13941         subsubsection.
13942         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13943         subsubsection.
13945 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
13947         PR target/85683
13948         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13949         after cmpelim optimization.
13951 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
13953         * config.gcc: Support "goldmont".
13954         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13955         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13956         PROCESSOR_GOLDMONT.
13957         * config/i386/i386.c (m_GOLDMONT): Define.
13958         (processor_target_table): Add "goldmont".
13959         (PTA_GOLDMONT): Define.
13960         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13961         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13962         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13963         (fold_builtin_cpu): Add "goldmont".
13964         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13965         (ix86_option_override_internal): Add "goldmont".
13966         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13967         (processor_type): Add PROCESSOR_GOLDMONT.
13968         * config/i386/i386.md: Add CPU "glm".
13969         * config/i386/glm.md: New file.
13970         * config/i386/x86-tune.def: Add m_GOLDMONT.
13971         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13973 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
13975         PR target/85572
13976         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13977         E_V4DImode.
13978         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13979         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
13980         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
13982         PR target/85317
13983         * config/i386/i386.c (ix86_fold_builtin): Handle
13984         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13986         PR target/85480
13987         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13988         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13990 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
13992         PR target/85658
13993         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13994         (check_arch): Likewise.
13995         (check_fpu): Return the result rather than printing it.
13996         (end arch): Fix operator precedence.
13997         (end cpu): Likewise.
13998         (END): Print the result from check_fpu.
14000 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
14001             Alan Hayward  <alan.hayward@arm.com>
14002             David Sherwood  <david.sherwood@arm.com>
14004         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
14005         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
14006         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
14007         (*fcmuo<mode>_and): New patterns.
14009 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
14011         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
14012         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
14013         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
14014         (cmp_op, sve_imm_con): New code attributes.
14015         (SVE_COND_INT_CMP, imm_con): Delete.
14016         (cmp_op): Remove above unspecs from int attribute.
14017         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
14018         to...
14019         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
14020         comparison-specific unspecs.
14021         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
14022         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
14023         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
14024         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
14025         (*vec_fcm<cmp_op><mode>): Rename to...
14026         (*fcm<cmp_op><mode>): ...this and adjust likewise.
14027         (*vec_fcmuo<mode>): Rename to...
14028         (*fcmuo<mode>): ...this and adjust likewise.
14029         (*pred_fcm<cmp_op><mode>): New pattern.
14030         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
14031         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
14032         functions.
14033         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
14034         and UNORDERED.
14035         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
14036         (aarch64_emit_sve_predicated_cond): New function.
14037         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
14038         (aarch64_emit_unspec_cond_or): Replace with...
14039         (aarch64_emit_sve_or_conds): ...this new function.  Use
14040         aarch64_emit_sve_ptrue_op for the individual comparisons and
14041         aarch64_emit_binop to OR them together.
14042         (aarch64_emit_inverted_unspec_cond): Replace with...
14043         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
14044         aarch64_emit_sve_ptrue_op for the comparison and
14045         aarch64_emit_unop to invert the result.
14046         (aarch64_expand_sve_vec_cmp_float): Update after the above
14047         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
14049 2018-05-07  Nathan Sidwell  <nathan@acm.org>
14051         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
14052         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
14053         (Backwards Compatibility): Likewise.
14055 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14057         PR bootstrap/85681
14058         Revert:
14059         2018-05-07  Luis Machado  <luis.machado@linaro.org>
14061         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14062         <prefetch_dynamic_strides>: New const bool field.
14063         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14064         prefetch_dynamic_strides.
14065         (exynosm1_prefetch_tune): Likewise.
14066         (thunderxt88_prefetch_tune): Likewise.
14067         (thunderx_prefetch_tune): Likewise.
14068         (thunderx2t99_prefetch_tune): Likewise.
14069         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14070         to false.
14071         (aarch64_override_options_internal): Update to set
14072         PARAM_PREFETCH_DYNAMIC_STRIDES.
14073         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14074         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14075         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14076         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14077         prefetch-dynamic-strides setting.
14079         2018-05-07  Luis Machado  <luis.machado@linaro.org>
14081         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14082         <minimum_stride>: New const int field.
14083         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14084         minimum_stride field.
14085         (exynosm1_prefetch_tune): Likewise.
14086         (thunderxt88_prefetch_tune): Likewise.
14087         (thunderx_prefetch_tune): Likewise.
14088         (thunderx2t99_prefetch_tune): Likewise.
14089         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14090         (aarch64_override_options_internal): Update to set
14091         PARAM_PREFETCH_MINIMUM_STRIDE.
14092         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14093         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14094         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14095         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14096         stride is constant and is below the minimum stride threshold.
14098 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14100         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
14101         to 512.
14103 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14105         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14106         <prefetch_dynamic_strides>: New const bool field.
14107         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14108         prefetch_dynamic_strides.
14109         (exynosm1_prefetch_tune): Likewise.
14110         (thunderxt88_prefetch_tune): Likewise.
14111         (thunderx_prefetch_tune): Likewise.
14112         (thunderx2t99_prefetch_tune): Likewise.
14113         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14114         to false.
14115         (aarch64_override_options_internal): Update to set
14116         PARAM_PREFETCH_DYNAMIC_STRIDES.
14117         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14118         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14119         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14120         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14121         prefetch-dynamic-strides setting.
14123 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14125         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14126         <minimum_stride>: New const int field.
14127         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14128         minimum_stride field.
14129         (exynosm1_prefetch_tune): Likewise.
14130         (thunderxt88_prefetch_tune): Likewise.
14131         (thunderx_prefetch_tune): Likewise.
14132         (thunderx2t99_prefetch_tune): Likewise.
14133         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14134         (aarch64_override_options_internal): Update to set
14135         PARAM_PREFETCH_MINIMUM_STRIDE.
14136         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14137         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14138         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14139         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14140         stride is constant and is below the minimum stride threshold.
14142 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
14144         PR c++/85659
14145         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
14146         the type is addressable.  Don't force op into register if it has
14147         BLKmode.
14149 2018-05-05  Roland McGrath  <mcgrathr@google.com>
14151         PR other/77609
14152         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
14153         any section for which we don't know a specific type it should have,
14154         regardless of name.  Previously this was done only for the exact
14155         names ".init_array", ".fini_array", and ".preinit_array".
14156         (default_elf_asm_named_section): Add comment about
14157         relationship with default_section_type_flags and SECTION_NOTYPE.
14158         (get_section): Don't consider it a type conflict if one side has
14159         SECTION_NOTYPE and the other doesn't, as long as neither has the
14160         SECTION_BSS et al used in the default_section_type_flags logic.
14162 2018-05-05  Tom de Vries  <tom@codesourcery.com>
14164         PR target/85653
14165         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
14166         (workaround_barsyncs): New function.
14167         (nvptx_reorg): Use workaround_barsyncs.
14168         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14169         (define_expand "nvptx_membar_cta"): New define_expand.
14170         (define_insn "*nvptx_membar_cta"): New insn.
14172 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
14174         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14175         To improve optimization opportunities.
14176         * builtin-types.def: The new needed builtin types for the above.
14178 2018-05-04  Richard Biener  <rguenther@suse.de>
14180         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14181         * gimple-ssa-store-merging.c
14182         (imm_store_chain_info::output_merged_store): Remove redundant create,
14183         release split_store vector contents on failure.
14184         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14185         scalar stmt vector on cache hit.
14187 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
14189         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14190         Xilinx FP support.
14191         * config.gcc (powerpc-xilinx-eabi*): Remove.
14192         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14193         support.
14194         (fusion_addis_mem_combo_load): Ditto.
14195         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14196         FP support.
14197         (rs6000_cpu_cpp_builtins): Ditto.
14198         * config/rs6000/rs6000-linux.c
14199         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14200         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14201         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14202         support.
14203         (rs6000_setup_reg_addr_masks): Ditto.
14204         (rs6000_init_hard_regno_mode_ok): Ditto.
14205         (rs6000_option_override_internal): Ditto.
14206         (legitimate_lo_sum_address_p): Ditto.
14207         (rs6000_legitimize_address): Ditto.
14208         (rs6000_legitimize_reload_address): Ditto.
14209         (rs6000_legitimate_address_p): Ditto.
14210         (abi_v4_pass_in_fpr): Ditto.
14211         (setup_incoming_varargs): Ditto.
14212         (rs6000_gimplify_va_arg): Ditto.
14213         (rs6000_split_multireg_move): Ditto.
14214         (rs6000_savres_strategy): Ditto.
14215         (rs6000_emit_prologue_components): Ditto.
14216         (rs6000_emit_epilogue_components): Ditto.
14217         (rs6000_emit_prologue): Ditto.
14218         (rs6000_emit_epilogue): Ditto.
14219         (rs6000_elf_file_end): Ditto.
14220         (rs6000_function_value): Ditto.
14221         (rs6000_libcall_value): Ditto.
14222         * config/rs6000/rs6000.h: Ditto.
14223         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14224         (TARGET_MINMAX): ... this.  New.
14225         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14226         * config/rs6000/rs6000.md: Remove Xilinx FP support.
14227         (*movsi_internal1_single): Delete.
14228         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14229         mfpu=, mxilinx-fpu): Delete.
14230         * config/rs6000/singlefp.h: Delete.
14231         * config/rs6000/sysv4.h: Remove Xilinx FP support.
14232         * config/rs6000/t-rs6000: Ditto.
14233         * config/rs6000/t-xilinx: Delete.
14234         * config/rs6000/titan.md: Adjust for fp_type removal.
14235         * config/rs6000/vsx.md: Remove Xilinx FP support.
14236         (VStype_simple): Delete.
14237         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14238         * config/rs6000/xfpu.h: Delete.
14239         * config/rs6000/xfpu.md: Delete.
14240         * config/rs6000/xilinx.h: Delete.
14241         * config/rs6000/xilinx.opt: Delete.
14242         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14243         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14245 2018-05-04  Tom de Vries  <tom@codesourcery.com>
14247         PR libgomp/85639
14248         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14249         if ignore == 0.
14251 2018-05-04  Richard Biener  <rguenther@suse.de>
14253         PR middle-end/85627
14254         * tree-complex.c (update_complex_assignment): We are always in SSA form.
14255         (expand_complex_div_wide): Likewise.
14256         (expand_complex_operations_1): Likewise.
14257         (expand_complex_libcall): Preserve EH info of the original stmt.
14258         (tree_lower_complex): Handle removed blocks.
14259         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14260         on complex multiplication and division libcall builtins.
14262 2018-05-04  Richard Biener  <rguenther@suse.de>
14264         PR middle-end/85574
14265         * fold-const.c (negate_expr_p): Restrict negation of operand
14266         zero of a division to when we know that can happen without
14267         overflow.
14268         (fold_negate_expr_1): Likewise.
14270 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
14272         PR libstdc++/85466
14273         * real.h (real_nextafter): Declare.
14274         * real.c (real_nextafter): New function.
14275         * fold-const-call.c (fold_const_nextafter): New function.
14276         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14277         CASE_CFN_NEXTTOWARD.
14278         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14279         even when arg1_mode is different from arg0_mode.
14281 2018-05-03  Nathan Sidwell  <nathan@acm.org>
14283         * doc/extend.texi (Deprecated Features): Remove
14284         -ffriend-injection.
14285         (Backwards Compatibility): Likewise.
14286         * doc/invoke.texi (C++ Language Options): Likewise.
14287         (C++ Dialect Options): Likewise.
14289 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
14291         PR target/85530
14292         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14293         _mm512_mask_mullox_epi64): New intrinsics.
14295 2018-05-03  Tom de Vries  <tom@codesourcery.com>
14297         PR testsuite/85106
14298         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14299         dump files): Add offload-tree.
14301 2018-05-03  Richard Biener  <rguenther@suse.de>
14303         PR tree-optimization/85615
14304         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14305         to loops not nested in BBs loop father to avoid creating multi-entry
14306         loops.
14308 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14310         PR tree-optimization/70291
14311         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14312         arguments.  Change return type to tree.  Emit libcall as a new
14313         statement rather than replacing existing one when inplace_p is true.
14314         (expand_complex_multiplication_components): New function.
14315         (expand_complex_multiplication): Expand floating-point complex
14316         multiplication using the above.
14317         (expand_complex_division): Rename inner_type parameter to type.
14318         Update expand_complex_libcall call-site.
14319         (expand_complex_operations_1): Update expand_complex_multiplication
14320         and expand_complex_division call-sites.
14322 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
14324         PR target/85582
14325         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14326         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14327         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14328         the highest significant bit of the shift count mask is clear.  In
14329         check whether and[sq]i3 is needed verify that all significant bits
14330         of the shift count other than the highest are set.
14332 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14334         PR libgomp/82428
14335         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14336         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14337         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14338         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14339         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14340         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14341         __builtin_goacc_parlevel_size.
14343 2018-05-02  Richard Biener  <rguenther@suse.de>
14345         PR tree-optimization/85597
14346         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14347         do not use split vect_get_vec_defs call but call vect_get_slp_defs
14348         directly.
14350 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14352         PR testsuite/85106
14353         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14354         dump files): Add ltrans-tree.
14356 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14358         PR testsuite/85106
14359         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14360         dump files): Add wpa-ipa.
14362 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
14364         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
14365         powerpc*-*-linux*paired* target.
14366         * config/rs6000/750cl.h: Delete.
14367         * config/rs6000/paired.h: Delete.
14368         * config/rs6000/paired.md: Delete.
14369         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14370         float support.
14371         * config/rs6000/rs6000-builtin.def: Remove paired float support.
14372         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14373         comment.  Remove paired float support.
14374         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14375         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14376         VECTOR_PAIRED.
14377         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14378         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14379         declarations.
14380         * config/rs6000/rs6000.c: Remove paired float support.
14381         (paired_expand_vector_init, paired_expand_vector_move,
14382         paired_emit_vector_compare, paired_emit_vector_cond_expr,
14383         (paired_expand_lv_builtin, paired_expand_stv_builtin,
14384         paired_expand_builtin, paired_expand_predicate_builtin,
14385         paired_init_builtins): Delete.
14386         * config/rs6000/rs6000.h: Remove paired float support.
14387         * config/rs6000/rs6000.md: Remove paired float support.
14388         (move_from_CR_ov_bit): Delete.
14389         * config/rs6000/rs6000.opt (mpaired): Delete.
14390         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
14391         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
14393 2018-05-02  Richard Biener  <rguenther@suse.de>
14395         PR middle-end/85567
14396         * gimplify.c (gimplify_save_expr): When in SSA form allow
14397         SAVE_EXPRs to compute to SSA vars.
14399 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
14401         PR target/85582
14402         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14403         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14404         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
14405         clobber operands[2], instead use a new pseudo.  Formatting fixes.
14407 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
14409         PR tree-optimization/85586
14410         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
14411         exit early for statements in the same group if the accesses are
14412         not strided.
14414 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14416         PR lto/85451
14417         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
14418         error message.
14420 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
14422         PR tree-optimization/85143
14423         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
14425 2018-05-01  Tom de Vries  <tom@codesourcery.com>
14427         PR lto/85451
14428         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
14429         not found" error message.
14431 2018-05-01  Tom de Vries  <tom@codesourcery.com>
14433         PR other/83786
14434         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
14435         * vec.c (test_ordered_remove_if): New function.
14436         (vec_c_tests): Call test_ordered_remove_if.
14437         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
14438         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
14439         * tree-vect-patterns.c (vect_pattern_recog_1): Use
14440         VEC_ORDERED_REMOVE_IF.
14442 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14444         PR tree-optimization/82665
14445         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
14446         pointer subtraction where arguments come from a memchr call.
14448 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
14450         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14451         --push-state --as-needed and --pop-state instead of --as-needed and
14452         --no-as-needed if ld supports it.
14453         * configure: Regenerated.
14455         PR web/85578
14456         * doc/install.texi2html: Replace _002d with - and _002a with * in
14457         generated html files using sed.
14459 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14461         PR c++/85523
14462         * gcc-rich-location.c (blank_line_before_p): New function.
14463         (use_new_line): New function.
14464         (gcc_rich_location::add_fixit_insert_formatted): New function.
14465         * gcc-rich-location.h
14466         (gcc_rich_location::add_fixit_insert_formatted): New function.
14468 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14470         * selftest.c (assert_streq): Rename "expected" and "actual" to
14471         "val1" and "val2".  Extend NULL-handling to cover both inputs
14472         symmetrically, while still requiring both to be non-NULL for a pass.
14473         * selftest.h (assert_streq): Rename "expected" and "actual" to
14474         "val1" and "val2".
14475         (ASSERT_EQ): Likewise.
14476         (ASSERT_EQ_AT): Likewise.
14477         (ASSERT_KNOWN_EQ): Likewise.
14478         (ASSERT_KNOWN_EQ_AT): Likewise.
14479         (ASSERT_NE): Likewise.
14480         (ASSERT_MAYBE_NE): Likewise.
14481         (ASSERT_MAYBE_NE_AT): Likewise.
14482         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
14483         the assertion to pass.
14484         (ASSERT_STREQ_AT): Likewise.
14486 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
14488         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14489         interaction with -pie.
14491 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14493         * selftest.h: Fix alphabetization of per-source-file selftest
14494         declarations.
14496 2018-04-30  Jason Merrill  <jason@redhat.com>
14498         PR c++/61982 - dead stores to destroyed objects.
14499         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14500         of clobber.
14502 2018-04-30  Jason Merrill  <jason@redhat.com>
14504         * tree.c (build_clobber): New.
14505         * tree.h: Declare it.
14506         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14508 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14510         * diagnostic-show-locus.c (layout::layout): Update for
14511         location_get_source_line returning a char_span.
14512         (struct char_span): Move to input.h.
14513         (struct correction): Update for fields in char_span becoming
14514         private.
14515         (struct source_line): Update for location_get_source_line
14516         returning a char_span.
14517         (layout::print_line): Likewise.
14518         * edit-context.c (edited_file::print_content): Likewise.
14519         (edited_file::print_diff_hunk): Likewise.
14520         (edited_file::print_run_of_changed_lines): Likewise.
14521         (edited_file::get_num_lines): Likewise.
14522         (edited_line::edited_line): Likewise.
14523         * final.c (asm_show_source): Likewise.
14524         * input.c (location_get_source_line): Convert return type
14525         from const char * to char_span, losing the final "line_len"
14526         param.
14527         (dump_location_info): Update for the above.
14528         (get_substring_ranges_for_loc): Likewise.  Use a char_span
14529         when handling the literal within the line.
14530         (test_reading_source_line): Update for location_get_source_line
14531         returning a char_span.
14532         * input.h (class char_span): Move here from
14533         diagnostic-show-locus.c, converting from a struct to a class.
14534         Make data members private.
14535         (char_span::operator bool): New.
14536         (char_span::length): New.
14537         (char_span::get_buffer): New.
14538         (char_span::operator[]): New.
14539         (char_span::subspan): Make const.
14540         (char_span::xstrdup): New.
14541         (location_get_source_line): Convert return type from const char *
14542         to char_span, losing the final "line_size" param.
14544 2018-04-30  Jan Hubicka  <jh@suse.cz>
14546         * lto-wrapper.c (ltrans_priorities): New static var.
14547         (cmp_priority): New.
14548         (run_gcc): Read priorities and if doing parallel build order
14549         the Makefile by them.
14551 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14553         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14555 2018-04-30  Richard Biener  <rguenther@suse.de>
14557         * tree-cfg.c (verify_address): Remove base argument, add
14558         flag whether to check TREE_ADDRESSABLE and do that.
14559         (verify_expr): Remove.
14560         (verify_types_in_gimple_reference): Add pieces from verify_expr.
14561         (verify_gimple_assign_single): Likewise.
14562         (verify_gimple_switch): Likewise.
14563         (verify_expr_location_1): Dereference tp once.  Add (disabled)
14564         piece from verify_expr.
14565         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14567 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14569         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14571 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14573         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14574         (small_data_pattern): Likewise.
14575         (arc_rewrite_small_data): Likewise.
14576         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14577         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14578         (get_symbol_alignment): New function.
14579         (legitimate_small_data_address_p): Likewise.
14580         (legitimate_scaled_address): Update, call
14581         legitimate_small_data_address_p.
14582         (output_sdata): New static variable.
14583         (arc_print_operand): Update how we handle small data operands.
14584         (arc_print_operand_address): Likewise.
14585         (arc_legitimate_address_p): Update, use
14586         legitimate_small_data_address_p.
14587         (arc_rewrite_small_data_p): Remove.
14588         (arc_rewrite_small_data_1): Likewise.
14589         (arc_rewrite_small_data): Likewise.
14590         (small_data_pattern): Likewise.
14591         (compact_sda_memory_operand): Update to use
14592         legitimate_small_data_address_p and get_symbol_alignment.
14593         (prepare_move_operands): Don't rewite sdata pattern.
14594         (prepare_extend_operands): Remove.
14595         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14596         pattern.
14597         (zero_extendqisi2): Likewise.
14598         (zero_extendhisi2): Likewise.
14599         (extendqihi2): Likewise.
14600         (extendqisi2): Likewise.
14601         (extendhisi2): Likewise.
14602         (addsi3): Likewise.
14603         (subsi3): Likewise.
14604         (andsi3): Likewise.
14605         * config/arc/constraints.md (Usd): Change it to memory constraint.
14607 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14609         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14610         as source of std instructions.
14611         * config/arc/arc.md (movsi_insn): Update pattern predicate to
14612         allow 6-bit constants as source for store instructions.
14613         (movdi_insn): Update instruction pattern to allow 6-bit constants
14614         as source for store instructions.
14616 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
14618         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14620 2018-04-30  Nathan Sidwell  <nathan@acm.org>
14621             Sandra Loosemore <sandra@codesourcery.com>
14623         * dumpfile.c (dump_open): Allow '-' for stdout.
14624         * doc/invoke.texi (Developer Options): Document dump filename
14625         determination early.  Document stdin/stdout selection.
14627 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
14629         Microblaze Target: PIC data text relative
14631         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14632         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14633         Add declaration.
14634         * config/microblaze/microblaze.h (microblaze_constant_address_p):
14635         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14636         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14637         New addressing mode for data-text relative position indepenedent code.
14638         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14639         'ADDRESS_SYMBOLIC_TXT_REL'.
14640         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14641         (microblaze_legitimate_pic_operand): Exclude function calls from
14642         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14643         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14644         addresses cases.
14645         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14646         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14647         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14648         for 'address + offset'.
14649         (microblaze_expand_prologue): Add new function prologue call for
14650         'r20' assignation.
14651         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14652         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14653         table in case of TARGET_PIC_DATA_TEXT_REL.
14654         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14655         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14656         Add new macros 'UNSPEC_TEXT',
14657         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14658         + exclude function calls from 'UNSPEC_PLT' in case of data text
14659         relative mode.
14660         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14661         new target hook for generating address diff vector tables in case of
14662         flag_pic.
14663         * doc/tm.texi : Regenerate.
14664         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14665         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14666         of addr diff vector generation.
14667         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14668         target hook definition.
14669         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14670         Add default function for generate_pic_addr_diff_vec -> flag_pic.
14671         * doc/invoke.texi (Add new pic option): Add new microblaze pic
14672         option for data text relative.
14674 2018-04-30  Richard Biener  <rguenther@suse.de>
14676         * tree-chrec.h (evolution_function_is_constant_p): Remove
14677         redundant check.
14678         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14680 2018-04-30  Richard Biener  <rguenther@suse.de>
14682         PR bootstrap/85571
14683         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14685 2018-04-30  Richard Biener  <rguenther@suse.de>
14687         PR tree-optimization/28364
14688         PR tree-optimization/85275
14689         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14690         copying first exit test.
14692 2018-04-28  Mark Wielaard  <mark@klomp.org>
14694         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14695         dwarf_version >= 5.
14696         (dwarf_AT): Handle DW_AT_addr_base.
14697         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14699 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
14701         PR target/84431
14702         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14703         (*ashl<dwi>3_doubleword_mask_1): Ditto.
14704         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14705         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14707 2018-04-28  Richard Biener  <rguenther@suse.de>
14709         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14710         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14711         to reflect use.  Only add interesting stmts.
14713 2018-04-27  Martin Jambor  <mjambor@suse.cz>
14715         PR ipa/85549
14716         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14717         the jump function allows for passing through aggregate values.
14719 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
14721         * input.h (in_system_header_at): Convert from macro to inline
14722         function.
14723         (from_macro_expansion_at): Likewise.
14724         (from_macro_definition_at): Likewise.
14726 2018-04-27  Jeff Law  <law@redhat.com>
14728         * config.gcc: Mark tile* targets as deprecated/obsolete.
14730 2018-04-27  Richard Biener  <rguenther@suse.de>
14732         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14733         fix for ILP32.
14735 2018-04-27  Richard Biener  <rguenther@suse.de>
14737         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14739 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
14741         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14742         with Yd constraint. Set "preferred_for_speed" attribute from
14743         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14744         with Yd constraint.
14745         (*movdi_internal): Ditto.
14746         (movti_interunit splitters): Remove
14747         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14748         (movdi_interunit splitters): Ditto.
14749         * config/i386/constraints.md (Ye): Remove.
14750         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14752 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14754         PR target/85512
14755         * config/aarch64/constraints.md (Usg): Limit to 31.
14756         (Usj): Limit to 63.
14758 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
14760         PR tree-optimization/85529
14761         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14762         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14763         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14764         zero extension or masking of the MSB bit.
14765         (optimize_range_tests): Add FIRST_BB argument, pass it through
14766         to optimize_range_tests_var_bound.
14767         (maybe_optimize_range_tests, reassociate_bb): Adjust
14768         optimize_range_tests callers.
14770 2018-04-26  Richard Biener  <rguenther@suse.de>
14771             Jakub Jelinek  <jakub@redhat.com>
14773         * cgraph.h (symbol_table): Just declare debug method here.
14774         * symtab.c (symbol_table::debug): Define.
14776 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
14778         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14780 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
14782         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14783         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14784         (*movdi_internal): Substitute Yi and Yj constraint with x
14785         and Ym and Yn constraint with y constraint.  Update "isa"
14786         attribute and set "preferred_for_speed" attribute from
14787         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14788         (*movsi_internal): Ditto.
14789         (*movdf_internal): Ditto.
14790         (*movsf_internal): Ditto.
14791         (*zero_extendsidi2): Ditto.
14792         * config/i386/sse.md (vec_set<mode>_0): Ditto.
14793         (sse2_loadld): Ditto.
14794         (*vec_extract<ssevecmodelower>_0): Ditto.
14795         (*vec_extractv4si_0_zext_sse4): Ditto.
14796         (vec_concatv2di): Ditto.
14797         (*vec_dup<mode>): Ditto.
14798         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14799         * config/i386/constraints.md (Yi): Remove.
14800         (Yj): Remove.
14801         (Ym): Remove.
14802         (Yn): Remove.
14804 2018-04-26  Nathan Sidwell  <nathan@acm.org>
14806         * dumpfile.c (dump_open): New.
14807         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14808         (dump_finish): Detect stdio/stderr by value not name.
14810 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
14812         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14814 2018-04-26  Tom de Vries  <tom@codesourcery.com>
14816         PR target/84952
14817         * config/nvptx/nvptx.c (verify_neutering_jumps)
14818         (verify_neutering_labels): New function
14819         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14821 2018-04-26  Tom de Vries  <tom@codesourcery.com>
14823         PR target/84025
14824         * config/nvptx/nvptx.c (needs_neutering_p): New function.
14825         (nvptx_single): Use needs_neutering_p to skip over insns that do not
14826         need neutering.
14828 2018-04-26  Richard Biener <rguenther@suse.de>
14829             Tom de Vries  <tom@codesourcery.com>
14831         PR lto/85422
14832         * lto-streamer-out.c (output_function): Fixup loops if required to match
14833         discovery done in the reader.
14835 2018-04-26  Richard Biener  <rguenther@suse.de>
14837         PR tree-optimization/85116
14838         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14839         have a loop exit from the single latch predecessor.  Remove
14840         case of header with just condition.
14841         (ch_base::copy_headers): Exclude infinite loops from any
14842         processing.
14843         (pass_ch::execute): Record exits.
14845 2018-04-26  Richard Biener  <rguenther@suse.de>
14847         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14848         prologue cost vector and pass it to vect_get_load_cost.
14849         (vect_get_peeling_costs_all_drs): Likewise.
14850         (vect_peeling_hash_get_lowest_cost): Likewise.
14851         (vect_enhance_data_refs_alignment): Likewise.
14853 2018-04-26  Richard Biener  <rguenther@suse.de>
14855         PR middle-end/85450
14856         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14857         checking of integer<->pointer conversions.
14858         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14859         sign-/zero-extending pointer types.
14860         (expand_omp_for_static_chunk): Likewise.
14862 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
14863             Jean Lee  <xiaoyur347@gmail.com>
14865         * config/mips/mips.c (mips_asan_shadow_offset): New function.
14866         (TARGET_ASAN_SHADOW_OFFSET): Define.
14867         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14868         true for -fsanitize=address.
14870 2018-04-25  Mark Wielaard  <mark@klomp.org>
14872         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14873         shorter ones.
14875 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
14877         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14878         than "alu", remove explicit "memory" and "imm_disp" attributes.
14879         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14881         PR middle-end/85414
14882         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14883         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14884         gen_lowpart_no_emit.
14886 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
14888         PR target/85473
14889         * config/i386/i386.c (ix86_expand_builtin): Change memory
14890         operand to XI, extend p0 to Pmode.
14891         * config/i386/i386.md: Change unspec volatile and operand
14892         1 mode to XI, change operand 0 mode to P.
14894 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14896         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14897         GET_MODE_MASK before any checking.
14898         (nds32_can_use_bset_p): Likewise.
14899         (nds32_can_use_btgl_p): Likewise.
14901 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14903         * config/nds32/nds32-doubleword.md: New define_split pattern for
14904         illegal register number.
14906 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14908         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14910 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14912         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14914 2018-04-25  Richard Biener  <rguenther@suse.de>
14916         * lto-streamer.h (LTO_major_version): Bump to 8.
14918 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
14920         * BASE-VER: Set to 9.0.0.
14922 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
14924         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14925         in __abskf2 and __powikf2.
14927 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14929         PR target/85512
14930         * config/aarch64/constraints.md (Usg, Usj): New constraints.
14931         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14932         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14933         Use the above on operand 2.  Reindent.
14934         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14936 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
14938         PR target/85485
14939         * common/config/i386/i386-common.c (ix86_handle_option): Don't
14940         handle OPT_mcet.
14941         * config/i386/i386.opt (mcet): Removed.
14942         * doc/install.texi: Remove -mcet documentation.
14943         * doc/invoke.texi: Likewise.
14945 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
14947         PR target/85485
14948         * doc/install.texi: Remove -mcet from bootstrap-cet.
14950 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14952         PR target/85511
14953         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14954         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14955         if TARGET_64BIT.
14957         PR target/85503
14958         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14959         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14960         containing a CONST_VECTOR.
14962 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
14964         * doc/install.texi: Update newlib dependency for nvptx.
14966 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14968         PR target/85508
14969         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14970         instead of INTVAL when shifting x left.
14972 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
14974         PR tree-optimization/85478
14975         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14976         vect_grouped_store_supported for single element vectors.
14978 2018-04-24  Richard Biener  <rguenther@suse.de>
14980         PR target/85491
14981         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14982         load cost increase to the case of non-constant step.
14984 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14986         PR target/84828
14987         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14988         destination if any_malformed_asm.
14990 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
14992         PR middle-end/85496
14993         * expr.c (store_field): In the bitfield case, if the value comes from
14994         a function call and is returned in registers by means of a PARALLEL,
14995         do not change the mode of the temporary unless BLKmode and VOIDmode.
14997 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
14999         PR rtl-optimization/85423
15000         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
15001         dependencies to debug insns when the previous insn is non-debug.
15003 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
15005         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
15006         enums into a single definition.
15007         (fls): Fix predicates and printing.
15008         (seti): Likewise.
15010 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
15012         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
15013         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
15014         and short u6 immediate.
15015         (check_if_valid_sleep_operand): Remove.
15016         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
15018 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15020         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
15021         flag_always_save_lp condition.
15022         * config/nds32/nds32.opt (malways-save-lp): New option.
15024 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
15026         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
15027         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
15028         * config/nds32/nds32.h
15029         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
15030         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
15032 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
15034         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
15035         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
15037 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
15038             Chung-Ju Wu  <jasonwucj@gmail.com>
15040         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
15041         Declare.
15042         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
15043         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
15045 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15047         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
15049 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15051         * config/nds32/nds32-protos.h (nds32_data_alignment,
15052         nds32_local_alignment): Declare.
15053         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
15054         nds32_local_alignment): New functions.
15055         (TARGET_CONSTANT_ALIGNMENT): Define.
15056         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
15058 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15060         * config/nds32/nds32.c
15061         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
15062         (TARGET_MODES_TIEABLE_P): Likewise.
15064 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15066         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
15067         level Ofast and Og.
15069 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
15070             Chung-Ju Wu  <jasonwucj@gmail.com>
15072         * config/nds32/constants.md (unspec_volatile_element): Add enum values
15073         for unaligned access.
15074         * config/nds32/nds32-intrinsic.c: Implementation of expanding
15075         unaligned access.
15076         * config/nds32/nds32-intrinsic.md: Likewise.
15077         * config/nds32/nds32_intrinsic.h: Likewise.
15078         * config/nds32/nds32.h (nds32_builtins): Likewise.
15079         * config/nds32/nds32.opt (munaligned-access): New option.
15080         * config/nds32/nds32.c (nds32_asm_file_start): Display
15081         flag_unaligned_access status.
15083 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
15085         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
15086         -mno-relax is present.
15087         * config/riscv/linux.h (LINK_SPEC): Ditto.
15089 2018-04-20  Martin Sebor  <msebor@redhat.com>
15091         PR c/85365
15092         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
15093         for null pointers.
15094         (gimple_fold_builtin_stxcpy_chk): Same.
15095         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
15097 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
15099         PR target/85456
15100         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
15101         __powikf2 when long double is IEEE 128-bit.
15103 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
15105         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
15106         step to make sure stack always aligned.
15108 2018-04-20  Carl Love  <cel@us.ibm.com>
15110         PR target/83402
15111         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
15112         size check for arg0.
15114 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
15115             Tom de Vries  <tom@codesourcery.com>
15117         PR target/85445
15118         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
15119         Emit insns for calls too.
15120         (nvptx_find_par): Always look for worker-level predecessor insn.
15121         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
15122         calls.
15123         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
15124         (nvptx_process_pars): Propagate frames for calls.
15126 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
15128         PR target/85469
15129         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
15130         Removed.
15131         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
15132         (ix86_handle_option): Don't handle OPT_mibt.
15133         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
15134         __SHSTK__.
15135         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
15136         has_ibt and ibt.
15137         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
15138         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
15139         (ix86_target_macros): Define __CET__ with flag_cf_protection
15140         for -fcf-protection.
15141         * config/i386/i386.c (isa2_opts): Remove -mibt.
15142         * config/i386/i386.h (TARGET_IBT): Removed.
15143         (TARGET_IBT_P): Likewise.
15144         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
15145         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
15146         * config/i386/i386.opt (mcet): Update help message.
15147         (mshstk): Likewise.
15148         (mibt): Removed.
15149         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
15150         -mcet as an alias for -mshstk.
15152 2018-04-20  Richard Biener <rguenther@suse.de>
15154         PR middle-end/85475
15155         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
15156         complexity by forcing a single use of the multiply operand.
15158 2018-04-20  Martin Jambor  <mjambor@suse.cz>
15160         ipa/85449
15161         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
15162         recursion dependency to only apply to non-clones.
15164 2018-04-20  Martin Jambor  <mjambor@suse.cz>
15166         ipa/85447
15167         * ipa-cp.c (create_specialized_node): Check that clones of
15168         self-recursive edges exist during IPA-CP.
15170 2018-04-19  Toon Moene  <toon@moene.org>
15172         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15173         by -O3.
15175 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
15177         PR tree-optimization/85467
15178         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
15179         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
15180         VECTOR_CST element to type.
15182 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
15184         PR target/85397
15185         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15186         * config/i386/i386.md (builtin_setjmp_setup): Removed.
15187         (builtin_longjmp): Likewise.
15188         (save_stack_nonlocal): New pattern.
15189         (restore_stack_nonlocal): Likewise.
15191 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
15193         PR target/85404
15194         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15195         Replace ASM_OUTPUT_LABEL with fprintf.
15197 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
15199         PR target/85417
15200         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15201         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15202         * config/i386/i386-c.c (ix86_target_macros_internal): Also
15203         define __IBT__ and __SHSTK__ for -fcf-protection.
15204         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15205         TARGET_IBT.
15206         (ix86_trampoline_init): Likewise.
15207         (x86_output_mi_thunk): Likewise.
15208         (ix86_notrack_prefixed_insn_p): Likewise.
15209         (ix86_option_override_internal): Don't disallow -fcf-protection.
15210         * config/i386/i386.md (rdssp<mode>): Also enable for
15211         -fcf-protection.
15212         (incssp<mode>): Likewise.
15213         (nop_endbr): Likewise.
15214         * config/i386/i386.opt (mcet): Change help message to built-in
15215         functions only.
15216         (mibt): Likewise.
15217         (mshstk): Likewise.
15218         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15219         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
15220         enable CET built-in functions.
15222 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
15224         * common/config/i386/i386-common.c
15225         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15226         OPTION_MASK_ISA_MOVDIRI_UNSET,
15227         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15228         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15229         * config.gcc (movdirintrin.h): New header.
15230         * config/i386/cpuid.h (bit_MOVDIRI,
15231         bit_MOVDIR64B): New bits.
15232         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15233         and -mmvodir64b.
15234         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15235         (VOID, PVOID, PCVOID)): New function types.
15236         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15237         __builtin_ia32_directstoreu_u64,
15238         __builtin_ia32_movdir64b): New builtins.
15239         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15240         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15241         and -mmovdiri.
15242         (ix86_valid_target_attribute_inner_p): Ditto.
15243         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15244         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15245         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15246         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15247         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15248         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15249         (movdiri<mode>, movdir64b_<mode>): New.
15250         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15251         * config/i386/immintrin.h: Include movdirintrin.h.
15252         * config/i386/movdirintrin.h: New file.
15253         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15255 2018-04-19  Richard Biener  <rguenther@suse.de>
15257         PR middle-end/85455
15258         * cfg.c (clear_bb_flags): When loop state says we have
15259         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15261 2018-04-19  Richard Biener  <rguenther@suse.de>
15263         PR tree-optimization/84737
15264         * tree-vect-data-refs.c (vect_copy_ref_info): New function
15265         copying restrict info.
15266         (vect_setup_realignment): Use it.
15267         * tree-vectorizer.h (vect_copy_ref_info): Declare.
15268         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15269         the first DR to all generated stores.
15270         (vectorizable_load): Likewise for loads.
15272 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
15274         PR tree-optimization/85446
15275         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15276         the integral and pointer types to have the same precision.
15278         * doc/install.texi: Document --disable-cet being the default and
15279         --enable-cet=auto.
15281 2018-04-18  Martin Liska  <mliska@suse.cz>
15283         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15284         style.
15286 2018-04-18  Martin Liska  <mliska@suse.cz>
15288         Revert
15289         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
15291         PR ipa/83983
15292         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15293         arguments if they are comparable.
15295 2018-04-18  Martin Liska  <mliska@suse.cz>
15297         Revert
15298         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
15300         PR lto/84805
15301         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15302         incomplete types.
15304 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
15306         PR target/85388
15307         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15308         ENDBR after calling __morestack.
15310 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
15312         PR jit/85384
15313         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15314         by using gcc_base_ver to generate a gcc_driver_version, and use
15315         it when generating GCC_DRIVER_NAME.
15316         * configure: Regenerate.
15318 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
15320         PR target/81084
15321         * config.gcc: Obsolete powerpc*-*-*spe*.
15323 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15325         PR debug/84637
15326         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15327         (stabstr_D): Change type of unum from unsigned int to
15328         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
15329         type.
15331 2018-04-17  Jim Wilson  <jimw@sifive.com>
15333         PR 84856
15334         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15335         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15336         Set arg_pointer_offset after using pretend_args_size.
15338 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15340         PR rtl-optimization/85431
15341         * dse.c (record_store): Ignore zero width stores.
15343         PR sanitizer/85230
15344         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
15345         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15346         __builtin_stack_restore rather than after it.
15347         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15348         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15349         argument instead of virtual_dynamic_stack_rtx.
15351 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15353         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15354         New prototype.
15355         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15356         Add note to error message to explain internal mapping of overloaded
15357         built-in function name to non-overloaded built-in function name.
15358         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15359         function.
15361 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
15363         PR target/85424
15364         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15365         where the inputs overlap with the output.
15367 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15369         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15370         (=v, v) alternative and explicit "memory" attribute.
15371         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
15372         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15373         attributes.
15374         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15375         "sselog1" type instead of "sselog".
15376         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15377         "sselog".  Remove explicit "memory" attribute.
15378         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15379         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15380         attributes.
15381         (vec_extract_hi_v32hi): Merge all alternatives into one, use
15382         "sselog1" type instead of "sselog".  Remove explicit "memory"
15383         attribute.
15384         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
15385         use "sselog1" type instead of "sselog".  Remove explicit "memory"
15386         attribute.
15387         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
15388         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15389         attributes.
15390         (vec_extract_hi_v64qi): Merge all alternatives into one, use
15391         "sselog1" type instead of "sselog".  Remove explicit "memory"
15392         attribute.
15393         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
15394         use "sselog1" type instead of "sselog".  Remove explicit "memory"
15395         attribute.
15397         PR target/85430
15398         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
15400         PR middle-end/85414
15401         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
15402         on a SUBREG.
15404 2018-04-17  Martin Jambor  <mjambor@suse.cz>
15406         PR ipa/85421
15407         * ipa-cp.c (create_specialized_node): Call
15408         expand_all_artificial_thunks if necessary.
15410 2018-04-17  Martin Liska  <mliska@suse.cz>
15412         PR lto/85405
15413         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
15414         in message, remote space in between '_G' and '('.
15416 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15418         PR target/85281
15419         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
15420         avx512f_vmcmp<mode>3<round_saeonly_name>,
15421         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
15422         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
15423         avx512f_rndscale<mode><round_saeonly_name>,
15424         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
15425         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
15426         Use %<iptr>2 instead of %2 for -masm=intel.
15427         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
15428         avx512f_vcvttss2usi<round_saeonly_name>,
15429         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
15430         -masm=intel.
15431         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
15432         avx512f_vcvttsd2usi<round_saeonly_name>,
15433         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
15434         Use %q1 instead of %1 for -masm=intel.
15435         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
15436         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
15437         of %3 for -masm=intel.
15438         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
15439         -masm=intel.
15440         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
15441         -masm=intel.
15442         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
15443         -masm=intel.
15444         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
15445         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
15446         %g1.
15447         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
15448         -masm=intel.
15449         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15450         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15451         %g1 and one with %0 and %1.
15452         (avx512er_vmrcp28<mode><round_saeonly_name>,
15453         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15454         %1 for -masm=intel.
15455         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15456         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15457         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15458         of %0 and %{%4%} for -masm=intel.
15459         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15460         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15461         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15462         order of %0 and %{%5%}%{z%} for -masm=intel.
15464 2018-04-17  Jan Hubicka  <jh@suse.cz>
15466         PR lto/85405
15467         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15469 2018-04-17  Martin Liska  <mliska@suse.cz>
15471         PR ipa/85329
15472         * multiple_target.c (create_dispatcher_calls): Set apostrophes
15473         for target_clone error message.  Make default implementation
15474         clone to be a local declaration.
15475         (separate_attrs): Add new argument and check for an empty
15476         string.
15477         (expand_target_clones): Handle it.
15478         (ipa_target_clone): Make redirection just for target_clones
15479         functions.
15481 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
15482             Tom de Vries  <tom@codesourcery.com>
15484         PR middle-end/84955
15485         * omp-expand.c (expand_oacc_for): Add dummy false branch for
15486         tiled basic blocks without omp continue statements.
15488 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
15490         PR target/83660
15491         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15492         vec_extract expression as having side effects to make sure it gets
15493         a cleanup point.
15495 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
15497         PR target/85403
15498         * config/i386/i386.c (get_builtin_code_for_version): Check
15499         error_mark_node.
15501 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
15503         PR target/84331
15504         * config.gcc: Support "skylake".
15505         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15506         PROCESSOR_SKYLAKE.
15507         * config/i386/i386.c (m_SKYLAKE): Define.
15508         (processor_target_table): Add "skylake".
15509         (ix86_option_override_internal): Add "skylake".
15510         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15511         PROCESSOR_CANNONLAKE.
15512         (get_builtin_code_for_version): Fix priority for
15513         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15514         PROCESSOR_SKYLAKE-AVX512.
15515         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15516         (processor_type): Add PROCESSOR_SKYLAKE.
15518 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
15519             Jason Merrill  <jason@redhat.com>
15521         PR c++/85112
15522         * convert.c (convert_to_integer_1): Use direct recursion for
15523         enumeral types and types with a precision less than the number
15524         of bits in their mode.
15526 2018-04-16  Julia Koval  <julia.koval@intel.com>
15528         PR target/84413
15529         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15530         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15532 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
15534         PR target/85293
15535         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15536         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15537         and -mno-direct-move.
15539 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
15541         PR target/83402
15542         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15543         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15544         Ensure negative shifts result in {0}.
15546 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
15548         PR rtl-optimization/79916
15549         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15550         regs (if any) to define how to gnerate SD moves when LRA is in
15551         progress.
15553 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
15555         PR rtl-optimization/85393
15556         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15557         * except.c (expand_dw2_landing_pad_for_region): Make static.
15558         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15559         a label and unconditional jump to old_bb, rather than
15560         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15561         basic block.
15563         PR rtl-optimization/85376
15564         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15565         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15566         instead of a specific value.
15568 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
15569             Bin Cheng  <bin.cheng@arm.com>
15571         PR tree-optimization/82965
15572         PR tree-optimization/83991
15573         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15574         by_profile_only parameter.
15575         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15576         information if the loop was predicted to iterate too many times.
15577         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15579 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
15581         PR lto/71991
15582         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15583         always inline.
15585 2018-04-13  Martin Liska  <mliska@suse.cz>
15586             Jakub Jelinek  <jakub@redhat.com>
15588         PR middle-end/81657
15589         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15590         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15591         * builtins.c (expand_builtin_memory_copy_args): Use
15592         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15593         handle dest_addr == pc_rtx.
15595 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
15597         PR target/85291
15598         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15599         asked to not generate direct moves.
15600         (fix_trunc<mode>si2_stfiwx): Similar.
15601         (fix_trunc<mode>si2_internal): Similar.
15603 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15605         PR debug/83157
15606         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15607         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15608         lookup if dest in some wider mode is known to be const0_rtx and
15609         if so, record permanent equivalence for it to be ZERO_EXTEND of
15610         the narrower mode destination.
15612 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
15614         * lto-streamer-out.c (output_function): Revert 259346.
15615         * omp-expand.c (expand_oacc_for): Likewise.
15617 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
15619         PR rtl-optimization/85354
15620         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15621         * sel-sched.c (sel_global_init): ... here.
15623 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
15625         PR target/85238
15626         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15627         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15628         mode for PE-COFF targets.
15629         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15630         (i386_pe_asm_lto_end): Likewise.
15631         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15632         (TARGET_ASM_LTO_END): Likewise.
15633         * config/i386/winnt.c (saved_debug_info_level): New static variable.
15634         (i386_pe_asm_lto_start): New function.
15635         (i386_pe_asm_lto_end): Likewise.
15637 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
15638             Richard Biener  <rguenther@suse.de>
15640         PR middle-end/84955
15641         * lto-streamer-out.c (output_function): Fix CFG loop state before
15642         streaming out.
15643         * omp-expand.c (expand_oacc_for): Handle calls to internal
15644         functions like regular functions.
15646 2018-04-12  Richard Biener  <rguenther@suse.de>
15648         PR lto/85371
15649         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15650         for the early LTO debug to properly generate references to it
15651         during DIE emission.  Do not re-use that for the skeleton for
15652         split-dwarf.
15653         (dwarf2out_early_finish): Likewise.
15655 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15657         PR target/85328
15658         * config/i386/sse.md
15659         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15660         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15661         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15662         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15663         and output is a reg, avoid creating invalid lowpart subreg, but
15664         instead split into a 512-bit move.  Don't split if not AVX512VL,
15665         input is xmm16+ reg and output is a mem.
15666         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15667         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15668         xmm16+ reg and output is a mem.
15670 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15672         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15673         also for flag_dwarf2_cfi_asm.
15675 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15677         PR rtl-optimization/85342
15678         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15679         a bool scalar var inside of the loop instead.  Don't try to update
15680         recog_data.operand after failed apply_change_group.
15682 2018-04-12  Tom de Vries  <tom@codesourcery.com>
15684         PR target/85296
15685         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15686         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
15687         array with flexible array member as array without given dimension.
15688         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15689         argument for undefined param to true.
15691 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
15693         PR target/85321
15694         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15695         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15696         from PowerPC section.
15697         * config/rs6000/sysv4.opt (mcall-): Improve help text.
15698         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15699         help text that is too long.
15700         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15701         help text that is too long.
15702         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15703         help text that is too long.
15705 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
15707         * config/alpha/alpha.md (stack_probe_internal): Rename
15708         from "probe_stack".  Update all callers.
15710 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15712         PR rtl-optimization/84566
15713         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15714         sched_macro_fuse_insns.
15716 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15718         PR target/84301
15719         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15720         (compute_block_dependences): ... from here.
15722 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15724         PR tree-optimization/85331
15725         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15726         from int to HOST_WIDE_INT.
15728 2018-04-11  Martin Jambor  <mjambor@suse.cz>
15730         PR ipa/84149
15731         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15732         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15733         not the same as the source val.
15734         (cgraph_edge_brings_value_p): New parameter.
15735         (gather_edges_for_value): Pass destination value to
15736         cgraph_edge_brings_value_p.
15737         (perhaps_add_new_callers): Likewise.
15738         (get_info_about_necessary_edges): Likewise and exclude values brought
15739         only by self-recursive edges.
15740         (create_specialized_node): Redirect only clones of self-calling edges.
15741         (+self_recursive_pass_through_p): New function.
15742         (find_more_scalar_values_for_callers_subset): Use it.
15743         (find_aggregate_values_for_callers_subset): Likewise.
15744         (known_aggs_to_agg_replacement_list): Removed.
15745         (decide_whether_version_node): Re-calculate known constants for all
15746         remaining context clones.
15748 2018-04-11  Richard Biener  <rguenther@suse.de>
15750         PR lto/85339
15751         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15752         from early DWARF output.
15753         (dwarf2out_early_finish): Output line info unconditionally into
15754         early DWARF and add reference to it.
15756 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15758         PR target/85281
15759         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15760         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15761         other than V2DFmode using iptr mode attribute.
15762         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15764 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15766         PR rtl-optimization/84659
15767         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15769 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15771         PR debug/85302
15772         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15773         SIZEP is NULL.
15774         (output_loc_list): Pass address of a dummy size variable even in the
15775         locview handling loop.
15776         (index_location_lists): Add comment on why skip_loc_list_entry can't
15777         call size_of_locs.
15779 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
15781         PR target/85261
15782         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15783         into register.
15785 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
15787         PR target/85321
15788         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15789         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15790         and -mstring-compare-inline-limit.
15792 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15794         PR target/85287
15795         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15796         for stack clash protection in a register whenever we need it to be in
15797         a register.
15799 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15801         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15802         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15804 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15806         PR target/85321
15807         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15808         the help text.
15809         (mlong-double-): Ditto.
15810         * config/rs6000/sysv4.opt (msdata=): Ditto.
15811         (mtls-size=): Ditto.
15813 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15815         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15816         erroneous entries for
15817         "vector int vec_ldl (int, long int *)", and
15818         "vector unsigned int vec_ldl (int, unsigned long int *)".
15819         Add comments and entries for
15820         "vector bool char vec_ldl (int, bool char *)",
15821         "vector bool short vec_ldl (int, bool short *)",
15822         "vector bool int vec_ldl (int, bool int *)",
15823         "vector bool long long vec_ldl (int, bool long long *)",
15824         "vector pixel vec_ldl (int, pixel *)",
15825         "vector long long vec_ldl (int, long long *)",
15826         "vector unsigned long long vec_ldl (int, unsigned long long *)".
15827         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15828         type tree bool_long_long_type_node and correct definition of
15829         bool_V2DI_type_node to make reference to this new type tree.
15830         (rs6000_mangle_type): Replace erroneous reference to
15831         bool_long_type_node with bool_long_long_type_node.
15832         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15833         comments to emphasize sign distinctions for char and int types and
15834         replace RS6000_BTI_bool_long constant with
15835         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
15836         use of RS6000_BTI_pixel.
15837         (bool_long_type_node): Remove this macro definition.
15838         (bool_long_long_type_node): New macro definition
15840 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
15842         PR rtl-optimization/85300
15843         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15844         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15845         simplify_unary_operation fails.
15847 2018-04-10  Martin Liska  <mliska@suse.cz>
15849         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15850         cgraph_edge and ipa_ref.
15852 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
15854         PR target/85177
15855         PR target/85255
15856         * config/i386/sse.md
15857         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15858         computation of the VEC_MERGE selector from mask.
15859         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15860         Fix decoding of the VEC_MERGE selector into mask.
15862 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
15864         PR tree-optimization/85286
15865         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15867 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
15869         * final.c (final_1): Set insn_last_address as well as
15870         insn_current_address.
15872 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15874         PR target/85173
15875         * explow.c (emit_stack_probe): Call validize_mem on memory location
15876         before passing it to gen_probe_stack.  Create address operand and
15877         legitimize it for the probe_stack_address case.
15879 2018-04-09  Jan Hubicka  <jh@suse.cz>
15881         PR lto/85078
15882         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15883         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15884         * tree.c (free_lang_data_in_type): Fix handling of binfos;
15885         walk basetypes.
15886         (free_lang_data): Rebuild type inheritance graph.
15888 2018-04-09  Martin Sebor  <msebor@redhat.com>
15890         * invoke.texi (-finline-small-functions): Mention other optimization
15891         options.
15892         (-findirect-inlining, -fpartial-inlining): Same.
15893         (-finline-functions-called-once): Same.
15894         (-freorder-blocks-and-partition): Same.
15896 2018-04-09  Jan Hubicka  <jh@suse.cz>
15898         PR rtl/84058
15899         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15900         jumps; choose last target that matches the criteria (i.e.
15901         no partition changes for non-crossing jumps).
15902         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15903         support for redirecting crossing jumps to non-crossing.
15905 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
15907         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15908         also for naked functions.
15910 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
15912         * config/arc/arc.md (add_shift): New pattern.
15913         (add_shift2): Likewise.
15914         (sub_shift): Likewise.
15915         (sub_shift_cmp0_noout): Likewise.
15916         (compare_si_ashiftsi): Likewise.
15917         (xbfu_cmp0_noout): New combine pattern.
15918         (xbfu_cmp0"): Likewise.
15919         (movsi_set_cc_insn): Place the predicable variant first.
15920         (commutative_binary_cmp0_noout): Remove clobber.
15921         (commutative_binary_cmp0): New pattern.
15922         (noncommutative_binary_cmp0): Likewise.
15923         (noncommutative_binary_cmp0_noout): Likewise.
15924         (noncommutative_binary_comparison_result_used): Removed.
15925         (rsub_cmp0): New pattern.
15926         (rsub_cmp0_noout): Likewise.
15927         (extzvsi): Changed, keep only meaningful variants.
15928         (SQH, SEZ): New iterators.
15929         (SQH_postfix): New mode attribute.
15930         (SEZ_prefix): New code attribute.
15931         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15932         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15933         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15934         of numerical value.
15935         (noncommutative_operator): Check the availability of barrel
15936         shifter option.
15938 2018-04-09  Richard Biener  <rguenther@suse.de>
15940         PR tree-optimization/85284
15941         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15942         Only use the niter constraining form of simple_iv when the exit
15943         is always executed.
15945 2018-04-09  Tom de Vries  <tom@codesourcery.com>
15947         PR target/84041
15948         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15949         (define_expand "*memory_barrier"): New define_expand.
15950         (define_insn "memory_barrier"): New insn.
15952 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15954         PR rtl-optimization/80463
15955         PR rtl-optimization/83972
15956         PR rtl-optimization/83480
15958         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15959         correct producer for the insn.
15960         (tidy_control_flow): Fixup seqnos in case of debug insns.
15962 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15964         PR rtl-optimization/83913
15966         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15967         different sched-times when merging exprs.
15969 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15971         PR rtl-optimization/83962
15973         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15974         tidy_fallthru_edge and tidy_control_flow.
15976 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15978         PR rtl-optimization/83530
15980         * sel-sched.c (force_next_insn): New global variable.
15981         (remove_insn_for_debug): When force_next_insn is true, also leave only
15982         next insn in the ready list.
15983         (sel_sched_region): When the region wasn't scheduled, make another pass
15984         over it with force_next_insn set to 1.
15986 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
15988         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15989         into tm_file.
15990         * config/nds32/constants.md (unspec_volatile_element): Add enum values
15991         for interrupt control.
15992         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15993         functions for interrupt control.
15994         * config/nds32/nds32-intrinsic.md: Likewise.
15995         * config/nds32/nds32_intrinsic.h: Likewise.
15996         * config/nds32/nds32.h (nds32_builtins): Likewise.
15998 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
16000         * config/nds32/nds32.c (nds32_init_machine_status,
16001         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
16002         strict_aligned_p field.
16003         (nds32_expand_to_rtl_hook): New function.
16004         (TARGET_EXPAND_TO_RTL_HOOK): Define.
16005         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
16007 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16008             Chung-Ju Wu  <jasonwucj@gmail.com>
16010         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
16011         * config/nds32/nds32-n7.md: New file.
16012         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
16013         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
16014         pipeline.
16015         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
16016         * config/nds32/nds32.md (pipeline_model): Add n7.
16017         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
16018         * config/nds32/pipelines.md: Include n7 settings.
16020 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16021             Chung-Ju Wu  <jasonwucj@gmail.com>
16023         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
16024         * config/nds32/nds32-e8.md: New file.
16025         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
16026         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
16027         pipeline.
16028         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
16029         * config/nds32/nds32.md (pipeline_model): Add e8.
16030         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
16031         * config/nds32/pipelines.md: Include e8 settings.
16033 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16034             Chung-Ju Wu  <jasonwucj@gmail.com>
16036         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
16037         * config/nds32/nds32-n8.md: New file.
16038         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
16039         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
16040         pipeline.
16041         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
16042         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
16043         * config/nds32/nds32.md (pipeline_model): Add n8.
16044         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
16045         * config/nds32/pipelines.md: Include n8 settings.
16047 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16048             Chung-Ju Wu  <jasonwucj@gmail.com>
16050         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
16051         * config/nds32/nds32-n9-2r1w.md: New file.
16052         * config/nds32/nds32-n9-3r2w.md: New file.
16053         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
16054         nds32_register_ports): New or modify for cpu n9.
16055         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
16056         pipeline.
16057         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
16058         * config/nds32/nds32-utils.c: New file.
16059         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
16060         TARGET_MUL_SLOW): Define.
16061         * config/nds32/nds32.md (pipeline_model): New attribute.
16062         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
16063         New options that support cpu n9.
16064         * config/nds32/pipelines.md: Include n9 settings.
16065         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
16067 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
16069         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
16070         information if necessary.
16071         (output_cond_branch_compare_zero): Likewise.
16072         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
16073         (nds32_target_alignment): Refine for alignment.
16074         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
16075         (FUNCTION_BOUNDARY): Modify.
16076         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
16077         align case.
16078         * config/nds32/nds32.opt (malways-align, malign-functions): New.
16080 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
16082         * config/nds32/constants.md (unspec_volatile_element): Add values for
16083         TLB operation and data prefetch.
16084         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16085         functions for TLB operation and data prefetch.
16086         * config/nds32/nds32-intrinsic.md: Likewise.
16087         * config/nds32/nds32_intrinsic.h: Likewise.
16088         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
16089         (nds32_print_operand): Likewise.
16090         * config/nds32/nds32.h (nds32_builtins): Likewise.
16092 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
16093         Andrew Pinski <pinsika@gcc.gnu.org>
16095         PR middle-end/82976
16096         * match.pd: Use constant_boolean_node of correct type instead of
16097         boolean_true_node or boolean_false_node for simplifying
16098         pointer comparisons to zero.
16100 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
16102         PR tree-optimization/80021
16103         * tree.c (verify_type_variant): Make error call in verify_variant_match
16104         translatable and remove final full stop.
16106 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16108         * config/nds32/constants.md (unspec_volatile_element): Add
16109         UNSPEC_VOLATILE_EH_RETURN.
16110         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
16111         nds32_output_stack_pop): Support dwarf exception handling process.
16112         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
16113         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
16114         exception handling process.
16115         (nds32_compute_stack_frame): Likewise.
16116         (nds32_return_addr_rtx): Likewise.
16117         (nds32_initial_elimination_offset): Likewise.
16118         (nds32_expand_prologue): Likewise.
16119         (nds32_expand_epilogue): Likewise.
16120         (nds32_dynamic_chain_address): New function.
16121         * config/nds32/nds32.h (machine_function): Add fields for dwarf
16122         exception handling.
16123         (DYNAMIC_CHAIN_ADDRESS): Define.
16124         (EH_RETURN_DATA_REGNO): Define.
16125         (EH_RETURN_STACKADJ_RTX): Define.
16126         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
16127         patterns for dwarf exception handling.
16129 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16131         * config/nds32/nds32.h: Clean up obsolete macros.
16133 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16135         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16136         Add enum values for particular instructions.
16137         * config/nds32/nds32-intrinsic.c: Implementation of expanding
16138         particular intrinsic functions.
16139         * config/nds32/nds32-intrinsic.md: Likewise.
16140         * config/nds32/nds32_intrinsic.h: Likewise.
16141         * config/nds32/nds32.h (nds32_builtins): Likewise.
16142         * config/nds32/nds32.md (type): Add pbsad and pbsada.
16143         (btst, ave): New patterns for particular instructions.
16145 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16147         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16148         Add enum values for atomic load/store and memory sync.
16149         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
16150         and memory sync.
16151         * config/nds32/nds32-intrinsic.md: Likewise.
16152         * config/nds32/nds32_intrinsic.h: Likewise.
16153         * config/nds32/nds32.h (nds32_builtins): Likewise.
16155 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
16157         PR tree-optimization/85257
16158         * fold-const.c (native_encode_vector): If not all elts could fit
16159         and off is -1, return 0 rather than offset.
16160         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
16161         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
16162         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
16163         adjust buffer in native_interpret_expr call.
16165 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16167         * config/nds32/constants.md (unspec_volatile_element): Add cache
16168         control enum values.
16169         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16170         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
16171         * config/nds32/nds32.c (nds32_cctl_names): New.
16172         (nds32_print_operand): Handle cache control register names.
16173         * config/nds32/nds32.h (nds32_builtins): New enum values.
16174         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16175         macros.
16176         * config/nds32/nds32.md (type): Add mmu.
16177         * config/nds32/pipelines.md (simple_insn): Add mmu.
16179 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16181         * config/nds32/nds32.md (type): Remove call.
16182         * config/nds32/pipelines.md (simple_insn): Likewise.
16184 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16186         * config/nds32/constants.md (unspec_volatile_element): Add
16187         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16188         UNSPEC_VOLATILE_FMFCFG.
16189         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16190         description for fmfcfg and fmfcsr.
16191         (bdesc_1arg): Add fmtcsr.
16192         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16193         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16194         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16195         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16196         unspec_fmfcfg): New patterns.
16197         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16198         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16199         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16200         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16201         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16202         __nds32__fmfcfg): Define.
16204 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16206         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16207         intrinsic register names.
16208         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16209         intrinsic register enum values and macros.
16211 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16213         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16214         for load/store addressing form.
16215         (nds32_print_operand_address): Likewise.
16217 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
16219         PR target/85196
16220         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16221         based on LABEL_REF.  Remove useless assertion.
16222         (pic_address_needs_scratch): Fix formatting.
16223         (sparc_legitimize_pic_address): Minor tweaks.
16224         (sparc_delegitimize_address): Adjust assertion accordingly.
16225         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16226         into symbolic_operand.
16227         (movsi_high_pic_label_ref): Likewise.
16228         (movsi_lo_sum_pic_label_ref): Likewise.
16229         (movdi_pic_label_ref): Likewise.
16230         (movdi_high_pic_label_ref): Likewise.
16231         (movdi_lo_sum_pic_label_ref): Likewise.
16233 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
16235         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16236         custom LIB_SPEC setup.
16238 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
16239             Kito Cheng  <kito.cheng@gmail.com>
16241         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16242         * config/riscv/freebsd.h: New.
16244 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
16246         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16247         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16248         file.
16250 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
16251             Kito Cheng  <kito.cheng@gmail.com>
16253         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16254         nds32_output_call, nds32_symbol_binds_local_p): New functions.
16255         * config/nds32/nds32-protos.h (nds32_output_call,
16256         nds32_output_return): Declare.
16257         * config/nds32/nds32.md: Refine all the call and return patterns.
16259 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
16261         PR debug/85252
16262         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16263         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16265         PR rtl-optimization/84872
16266         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16267         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16268         EDGE_CROSSING edge.
16270 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
16272         * expr.c (copy_blkmode_to_reg): Revert 254862.
16273         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16275 2018-04-06  Richard Biener  <rguenther@suse.de>
16277         PR middle-end/85244
16278         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16279         after seeing a component reference with an adjacent field.  Treat
16280         refs to arrays at struct end of external decls similar to
16281         refs to unconstrained commons.
16283 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
16285         PR sanitizer/85213
16286         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16287         look through SAVE_EXPRs with non-side-effects argument.  Adjust
16288         recursive calls.
16289         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16290         save_p here.
16292 2018-04-06  Richard Biener  <rguenther@suse.de>
16294         PR middle-end/85180
16295         * alias.c (find_base_term): New wrapper around find_base_term
16296         unwinding CSELIB_VAL_PTR changes.
16297         (find_base_term): Do not restore CSELIB_VAL_PTR during the
16298         recursion.
16300 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
16302         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16303         instructions.
16304         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16305         constant definitions.
16306         ("nop"): lr 0,0 -> nopr r0
16307         ("nop_lr0", "nop_lr1"): New insn definitions.
16309 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
16311         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16312         NDS32_V3PUSH_AVAILABLE_P macro.
16314 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
16315             Chung-Ju Wu  <jasonwucj@gmail.com>
16317         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16318         (nds32*-*-*): Add float and fpu_config into supported_defaults.
16319         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16320         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16321         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16322         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16323         * config/nds32/constraints.md: New constraints and checking for hard
16324         float configuration.
16325         * config/nds32/iterators.md: New mode iterator and attribute for hard
16326         float configuration.
16327         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16328         patterns.
16329         * config/nds32/nds32-fpu.md: New file.
16330         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16331         deal with hard float code generation.
16332         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16333         ARCH_V3S.
16334         (abi_type, float_reg_number): New enum type.
16335         * config/nds32/nds32-predicates.c: New predicates for hard float.
16336         * config/nds32/nds32-protos.h: Declare functions for hard float.
16337         * config/nds32/nds32.c: Implementation for hard float configuration.
16338         * config/nds32/nds32.h: Definitions for hard float configuration.
16339         * config/nds32/nds32.md: Include hard float machine description and
16340         modify patterns for hard float configuration.
16341         * config/nds32/nds32.opt: New options for hard float configuration.
16342         * config/nds32/predicates.md: New predicates for hard float
16343         configuration.
16345 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
16347         * common/config/nds32/nds32-common.c
16348         (nds32_option_optimization_table): Enable -mreleax-hint by default.
16350 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
16352         PR middle-end/85195
16353         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16354         CONSTRUCTOR_ELT (ctor, ...)->value.
16356 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
16358         PR target/85193
16359         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16361 2018-04-05  Tom de Vries  <tom@codesourcery.com>
16363         PR target/85204
16364         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16365         cond jump.
16367 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
16368             Kito Cheng  <kito.cheng@gmail.com>
16370         * config/nds32/constraints.md (U33): Fine-tune checking condition.
16371         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16372         * config/nds32/nds32.h (nds32_16bit_address_type): Add
16373         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16375 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
16376             Kito Cheng  <kito.cheng@gmail.com>
16378         * config/nds32/constraints.md (Ufe): New memory constraint.
16379         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16380         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16381         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
16382         operands.
16383         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
16384         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
16386 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16388         * config/nds32/nds32.md: Use optimize_size in the condition for
16389         alu-shift instructions.
16391 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16393         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
16395 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16397         * config/nds32/nds32.md (negsi2): Refine pattern.
16399 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
16400             Chung-Ju Wu  <jasonwucj@gmail.com>
16402         * config/nds32/iterators.md (shift_rotate): New code iterator.
16403         (shift): New code attribute.
16404         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
16405         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
16406         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
16407         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
16408         bit-wise operations.
16409         (andsi3, *andsi3): Ditto.
16410         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
16411         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
16412         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
16413         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
16414         nds32_ior_operand, nds32_xor_operand): New predicates.
16416 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16418         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
16419         (addsi3, subsi3): ... this.
16421 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16423         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
16425 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16427         * config/nds32/nds32.md: Adjust indention.
16429 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
16431         * config/nds32/nds32.md (feature): New attribute.
16433 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16435         * config/nds32/nds32.md (subtype): New attribute.
16437 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
16439         PR target/85203
16440         * config/arm/arm-builtins.c (arm_expand_builtin): Change
16441         expansion to perform a bitwise AND of the argument followed by a
16442         boolean negation of the result.
16444 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
16446         PR rtl-optimization/84878
16447         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
16448         the basic block.  Assert the use reference is not artificial and that
16449         it has an associated insn.
16451 2018-04-04  Michael Matz  <matz@suse.de>
16453         * builtins.c (compute_objsize): Pass correct operand
16454         to array_at_struct_end_p.
16456 2018-04-04  Richard Biener  <rguenther@suse.de>
16458         PR lto/85176
16459         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16460         from contexts for DINFO_LEVEL_TERSE and below.
16462 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16464         * config/nds32/nds32-doubleword.md (move_<mode>): Require
16465         resiter_operand condition.
16466         * config/nds32/nds32.md (*move<mode>): Ditto.
16468 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16469             Monk Chiang  <sh.chiang04@gmail.com>
16471         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16473 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16475         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16477 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16478             Kito Cheng  <kito.cheng@gmail.com>
16480         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16481         nds32_cond_code_str, output_cond_branch,
16482         output_cond_branch_compare_zero, nds32_expand_cbranch,
16483         nds32_expand_cstore, nds32_expand_movcc,
16484         nds32_output_cbranchsi4_equality_zero,
16485         nds32_output_cbranchsi4_equality_reg,
16486         nds32_output_cbranchsi4_equality_reg_or_const_int,
16487         nds32_output_cbranchsi4_greater_less_zero: New functions.
16488         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16489         nds32_expand_cstore, nds32_expand_movcc,
16490         nds32_output_cbranchsi4_equality_zero,
16491         nds32_output_cbranchsi4_equality_reg,
16492         nds32_output_cbranchsi4_equality_reg_or_const_int,
16493         nds32_output_cbranchsi4_greater_less_zero): Declare.
16494         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16495         nds32_rimm11s_operand): New predicates.
16496         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16497         * config/nds32/nds32.md: Rewrite all the branch and conditional move
16498         patterns.
16500 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16502         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16503         * config/nds32/nds32.md: Ditto.
16504         * config/nds32/pipelines.md: Ditto.
16506 2018-04-04  Richard Biener  <rguenther@suse.de>
16508         PR tree-optimization/85168
16509         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16510         propagating abnormals.
16512 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16514         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16516 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16517             Kito Cheng  <kito.cheng@gmail.com>
16519         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16520         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16521         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16522         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16523         * config/nds32/nds32.md (sibcall_internal): New.
16524         (sibcall_register): Remove.
16525         (sibcall_immediate): Remove.
16526         (sibcall_value_internal): New.
16527         (sibcall_value_register): Remove.
16528         (sibcall_value_immediate): Remove.
16529         * config/nds32/predicates.md (nds32_general_register_operand): New.
16530         (nds32_call_address_operand): New.
16532 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
16534         PR rtl-optimization/85167
16535         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16536         bb_defs if *split_p, instead preinitialize it to NULL.
16538         PR tree-optimization/85156
16539         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16540         evaluating the argument multiple times.
16542 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
16544         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16545         than vector.
16546         (_mm_cvtpd_ps): Likewise.
16547         (_mm_cvttpd_epi32): Likewise.
16548         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16549         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16550         vector, pixel, and bool following altivec.h include.
16552 2018-04-03  Martin Sebor  <msebor@redhat.com>
16554         * doc/extend.texi (Common Function Attributes): Clarify.
16555         (const attribute): Likewise.
16556         (pure attribute): Likewise.
16558 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
16560         PR target/85169
16561         * config/i386/i386.c (ix86_expand_vector_set): Use
16562         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
16564 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
16566         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16567         instructions when changing rounding bits to preserve precision bits
16568         in the x87 control word.
16570 2018-04-03  Martin Liska  <mliska@suse.cz>
16572         PR tree-optimization/82491
16573         * rtl.h (strip_offset_and_add): Replace += suboffset with
16574         poly_uint64 () + suboffset.
16576 2018-03-29  Martin Liska  <mliska@suse.cz>
16577             Martin Jambor  <mjambor@suse.cz>
16579         PR ipa/84947
16580         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16581         param_type is not an integral or pointer type.
16583 2018-04-03  Richard Biener  <rguenther@suse.de>
16585         * sese.h (recompute_all_dominators): Remove.
16587 2018-04-02  Martin Sebor  <msebor@redhat.com>
16589         * doc/invoke.texi (-Wrestrict): Fix typos.
16591 2018-04-02  Jim Wilson  <jimw@sifive.com>
16593         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16594         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16595         (<optab>di3, <optab>si3_extend): Likewise.
16596         (<optab>si3_mask, <optab>si3_mask_1): New.
16597         (<optab>di3_mask, <optab>di3_mask_1): New.
16598         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16599         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16600         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16602 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
16604         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16605         example.
16607 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16609         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16610         (nds32_canonicalize_comparison): New function.
16612 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16613             Kito Cheng  <kito.cheng@gmail.com>
16614             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
16616         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16617         * config/nds32/constants.md (unspec_volatile_element): Add
16618         UNSPEC_VOLATILE_RELAX_GROUP.
16619         * config/nds32/nds32-relax-opt.c: New file.
16620         * config/nds32/nds32-predicates.c
16621         (nds32_symbol_load_store_p): New function.
16622         * config/nds32/nds32-protos.h
16623         (nds32_symbol_load_store_p): Declare function.
16624         (make_pass_nds32_relax_opt): Declare new rtl pass function.
16625         * config/nds32/nds32.c
16626         (nds32_register_pass): New function to register pass.
16627         (nds32_register_passes): New function to register passes.
16628         * config/nds32/nds32.md (relax_group): New pattern.
16629         * config/nds32/nds32.opt (mrelax-hint): New option.
16630         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16632 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
16634         * config/nds32/t-nds32: Modify files dependency.
16636 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16638         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16639         (PROFILE_HOOK): Define its implementation.
16641 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16643         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16644         type and 32-bit size.
16646 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
16648         PR middle-end/85090
16649         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16650         (V_128_256): New mode iterator.
16651         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16652         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16653         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16654         of V.
16655         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16656         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16658 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
16660         PR target/83315
16661         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16662         NaN inputs correctly.
16664 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
16666         PR target/80546
16667         * config/rs6000/vsx.md (??r): New mode attribute.
16668         (*vsx_mov<mode>_64bit): Use it.
16669         (*vsx_mov<mode>_32bit): Likewise.
16671 2018-03-30  Martin Sebor  <msebor@redhat.com>
16673         PR tree-optimization/84818
16674         * builtins.c (check_access): Use warning_n.
16676 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
16678         PR target/83822
16679         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16680         condition.
16681         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16682         condition.
16684 2018-03-30  Julia Koval  <julia.koval@intel.com>
16686         PR target/84413
16687         * x86-tune.def (movx, partial_reg_dependency): Enable for
16688         m_SKYLAKE_AVX512.
16690 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
16692         PR inline-asm/84985
16693         * lra-constraints.c (process_alt_operands): Move setting
16694         this_alternative_matches below.
16696 2018-03-29  Martin Liska  <mliska@suse.cz>
16698         PR lto/84995.
16699         * doc/invoke.texi: Document how LTO works with debug info.
16700         Describe auto-load support of binutils.  Mention 'x86-64'
16701         as valid option value of -march option.
16703 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
16705         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16707         PR c/85094
16708         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16709         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16710         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16711         checking.
16713 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
16715         PR target/84912
16716         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16717         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16718         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16719         for RS6000_BTM_POWERPC64.
16720         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16721         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16722         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16723         definition.
16724         (DIVDE): Use it.
16725         (DIVDEU): Likewise.
16727 2018-03-28  Carl Love  <cel@us.ibm.com>
16729         Revert
16730         2017-09-27  Carl Love  <cel@us.ibm.com>
16732         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16733         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16734         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16735         fctiw instruction.
16737 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
16739         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16740         instead of __vector bool.
16741         (_mm_max_pu8): Likewise.
16742         (_mm_min_pi16): Likewise.
16744 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
16746         PR target/84912
16747         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16748         (DIVWEUO): Likewise.
16749         (DIVDEO): Likewise.
16750         (DIVDEUO): Likewise.
16751         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16752         DIVWEUO and DIVDEUO.
16753         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16754         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16755         (div_extend): Likewise.
16756         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16757         builtin function.
16758         (__builtin_divweuo): Likewise.
16759         (__builtin_divdeo): Likewise.
16760         (__builtin_divdeuo): Likewise.
16762 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
16764         PR target/85095
16765         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16766         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16768         PR tree-optimization/82004
16769         * gimple-match-head.c (optimize_pow_to_exp): New function.
16770         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16771         Don't fold to exp if optimize_pow_to_exp is false.
16773 2018-03-28  Martin Liska  <mliska@suse.cz>
16775         PR other/84819
16776         * calls.c (initialize_argument_information): Fix trailing space.
16777         * common.opt: Fix typo and provide better explanation for
16778         -fsanitize-coverage option.
16779         * config/i386/i386.opt: Fix typo.
16781 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
16782             Martin Liska  <mliska@suse.cz>
16784         PR sanitizer/85081
16785         * gimplify.c (asan_poison_variable): Don't do the check for
16786         gimplify_omp_ctxp here.
16787         (gimplify_decl_expr): Do it here.
16788         (gimplify_target_expr): Likewise.
16790 2018-03-28  Martin Liska  <mliska@suse.cz>
16792         PR target/84988
16793         * config/i386/i386.c (ix86_function_arg_advance): Do not call
16794         chkp_type_bounds_count if MPX is not enabled.
16796 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
16798         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16800 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
16802         PR target/84914
16803         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16804         function to create the function decl for complex long double
16805         multiply and divide for -mabi=ieeelongdouble.
16806         (init_float128_ieee): Call it.
16808 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
16810         PR target/85044
16811         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16812         -fcf-protection=branch -mibt.
16813         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16815 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16817         PR target/81863
16818         * config/arm/arm.c (arm_valid_symbolic_address): Handle
16819         arm_word_relocations.
16821 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
16823         PR target/85056
16824         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16825         extern array declarations.
16827 2018-03-27  Richard Biener  <rguenther@suse.de>
16829         PR middle-end/84067
16830         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16831         explicit single_use checks.
16833 2018-03-27  Richard Biener  <rguenther@suse.de>
16835         PR tree-optimization/85082
16836         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16837         Valueize the VUSE.
16839 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16841         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16842         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16843         Turn on fasynchronous-unwind-tables and funwind-tables.
16845 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
16847         PR target/85073
16848         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16849         (*bmi_blsr_<mode>_ccz): Ditto.
16851 2018-03-26  Tom de Vries  <tom@codesourcery.com>
16853         PR tree-optimization/85063
16854         * omp-general.c (offloading_function_p): New function.  Factor out
16855         of ...
16856         * omp-offload.c (pass_omp_target_link::gate): ... here.
16857         * omp-general.h (offloading_function_p): Declare.
16858         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16859         with attribute omp declare target for offloading functions.
16861 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
16863         PR tree-optimization/84005
16864         * tree-data-ref.h (get_base_for_alignment): Declare.
16865         * tree-data-ref.c (get_base_for_alignment_1): New function.
16866         (get_base_for_alignment): Likewise.
16867         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16868         get_base_for_alignment to find a suitable base object, instead
16869         of always using drb->base_address.
16871 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
16873         PR inline-asm/85022
16874         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16875         known size by default.
16877 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
16879         PR inline-asm/85030
16880         * lra-constraints.c (process_alt_operands): Don't match BLKmode
16881         and non BLKmode operands.
16883 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16885         PR target/85026
16886         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16887         Clean up attributes.
16889 2018-03-23  Richard Biener  <rguenther@suse.de>
16891         PR debug/85020
16892         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16893         we are going to emit early debug for LTO.
16895 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
16897         PR inline-asm/85034
16898         * function.c (match_asm_constraints_1): Don't optimize if input
16899         doesn't satisfy general_operand predicate for output's mode.
16901         PR inline-asm/85022
16902         * alias.c (write_dependence_p): Don't require for x_canonicalized
16903         non-VOIDmode if x has VOIDmode.
16905         PR sanitizer/85029
16906         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16907         just don't try to optimize it rather than assert it never happens.
16909 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16911         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16912         macro expansions for definition of ST_INTERNAL_<mode> and
16913         LD_INTERNAL_<mode> builtins.
16914         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16915         Remove prototype.
16916         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16917         function.
16918         (altivec_expand_st_builtin): Likewise.
16919         (altivec_expand_builtin): Remove calls to deleted functions.
16920         (rs6000_address_for_altivec): Delete this function.
16921         * config/rs6000/vector.md: Remove expands for
16922         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16924 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
16926         PR target/84826
16927         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16928         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16929         re-computing once computed.
16930         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16931         (arm_init_machine_status): Initialize
16932         machine->static_chain_stack_bytes.
16934 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16936         PR target/84760
16937         * doc/extend.texi: Add four new prototypes for vec_ld.
16938         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16939         definitions for more logical presentation.
16940         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16941         entries for V1TI variants of __builtin_altivec_ld builtin.
16942         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16943         handling of V1TI variant of LVX icode pattern.
16944         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16945         (rs6000_gimple_fold_builtin): Likewise.
16946         (altivec_init_builtins): Add code to define
16947         __builtin_altivec_lvx_v1ti function.
16949 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
16951         PR inline-asm/84941
16952         * function.c (match_asm_constraints_1): Don't do the optimization
16953         if input isn't a REG, SUBREG, MEM or constant.
16955 2018-03-22  Tom de Vries  <tom@codesourcery.com>
16957         PR tree-optimization/84956
16958         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16959         bb_has_abnormal_pred.
16961 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
16963         PR sanitizer/85018
16964         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16965         DECL_INITIAL (decl) to decl at the end.
16966         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16967         adjust the comment.
16969 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
16971         * doc/extend.texi (__builtin_tgmath): Document when complex
16972         integer types are treated as _Complex _Float64.
16974 2018-03-21  Tom de Vries  <tom@codesourcery.com>
16976         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16978 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16980         PR tree-optimization/84960
16981         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16982         if it is ENTRY block, move them into single succ of ENTRY in that case.
16984 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
16986         PR tree-optimization/84811
16987         * poly-int.h (poly_span_traits): Remove the T3 parameter and
16988         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16989         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16990         (known_subrange_p): Update accordingly.  Cast each value involved
16991         in the size comparison, rather than casting the result of the
16992         subtraction.
16994 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16996         PR tree-optimization/84982
16997         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16998         by flipping the least significant bit rather than all bits from
16999         bitpos to bitpos + bitsize - 1.
17001 2018-03-21  Nathan Sidwell  <nathan@acm.org>
17003         * doc/extend.texi (Deprecated Features): Remove mention of
17004         long-deleted deprecations.
17006 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17008         PR jit/84288
17009         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
17010         * configure: Regenerate.
17012 2018-03-21  Tom de Vries  <tom@codesourcery.com>
17014         PR tree-optimization/83126
17015         * tree-parloops.c (num_phis): New function.
17016         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
17018 2018-03-21  Nathan Sidwell  <nathan@acm.org>
17020         * doc/extend.texi (Deprecated Features): Update deprecated flags,
17021         mention anon-struct/union members and trailing attributes.
17023 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
17025         PR tree-optimization/84969
17026         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
17027         builtin memset partitions if they set different rhs values.
17029 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
17031         PR rtl-optimization/84989
17032         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
17033         VEC_DUPLICATE with scalar result mode.
17035 2018-03-21  Martin Liska  <mliska@suse.cz>
17037         PR ipa/84963
17038         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
17039         not intended return statement.
17041 2018-03-21  Martin Liska  <mliska@suse.cz>
17043         PR target/84988
17044         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
17045         (chkp_find_bound_slots_1): Limit number of iterations.
17047 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
17049         PR target/84838
17050         * Minor grammar fixes for x86 options.
17052 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
17054         PR debug/84875
17055         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
17056         holding REG_CFA_RESTORE notes, instead turn them into a USE.
17058 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
17060         PR target/83789
17061         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
17062         (altivec_lvx_<mode>_1op): Likewise.
17063         (altivec_stvx_<mode>_2op): Likewise.
17064         (altivec_stvx_<mode>_1op): Likewise.
17065         (altivec_lvx_<VM2:mode>): New define_expand.
17066         (altivec_stvx_<VM2:mode>): Likewise.
17067         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
17068         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17069         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
17070         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17071         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
17072         (rs6000_gen_lvx): Likewise.
17073         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
17074         (altivec_expand_stv_builtin): Likewise.
17075         (altivec_expand_builtin): Likewise.
17076         * config/rs6000/vector.md: Likewise.
17078 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17080         PR target/82518
17081         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
17082         BYTES_BIG_ENDIAN.
17084 2018-03-20  Richard Biener  <rguenther@suse.de>
17086         PR target/84986
17087         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
17088         sign-conversions as zero, fall back to standard scalar_stmt
17089         cost for the rest.
17091 2018-03-20  Martin Liska  <mliska@suse.cz>
17093         PR ipa/84825
17094         * predict.c (rebuild_frequencies): Handle case when we have
17095         PROFILE_ABSENT, but flag_guess_branch_prob is false.
17097 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
17099         PR target/84990
17100         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
17101         flag_section_anchors.
17102         * varasm.c (use_blocks_for_decl_p): Remove hack for
17103         dw2_force_const_mem.
17105         PR target/84845
17106         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
17107         to ...
17108         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
17109         be created, use lowpart_subreg of operands[0] rather than operands[0]
17110         itself.
17111         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
17112         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
17113         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
17114         and n constraint instead of aarch64_shift_imm_di and Usd.
17115         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
17116         (*aarch64_<optab>_reg_minus<mode>3): ... this.
17118 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
17120         PR target/82989
17121         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
17122         to favor GPR over NEON registers.
17123         (<shift>di3_neon): Likewise.
17125 2018-03-20  Tom de Vries  <tom@codesourcery.com>
17127         PR target/84952
17128         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
17129         (nvptx_process_pars): Emit bar.sync asap and alap.
17131 2018-03-20  Tom de Vries  <tom@codesourcery.com>
17133         PR target/84954
17134         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
17135         seen_label if seen_label is already set.
17137 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
17139         PR target/84945
17140         * config/i386/i386.c (fold_builtin_cpu): For features above 31
17141         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
17142         Use 1U instead of 1.  Formatting fixes.
17144         PR c/84953
17145         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
17146         instead of TREE_TYPE (s1) for the return value.
17148 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
17150         PR tree-optimization/84946
17151         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
17152         bitsize + bitsize in poly_uint64 rather than poly_int64.
17154         PR sanitizer/78651
17155         * dwarf2asm.c: Include fold-const.c.
17156         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
17157         of decl rather than decl itself.
17159         PR rtl-optimization/84643
17160         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
17162 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
17164         PR sanitizer/78651
17165         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
17166         calling assemble_variable.
17168 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
17170         PR target/81647
17171         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17172         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17174 2018-03-19  Jim Wilson  <jimw@sifive.com>
17176         PR bootstrap/84856
17177         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
17178         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
17179         (riscv_first_stack_step): Likewise.
17180         (riscv_option_override): Use STACK_BOUNDARY instead of
17181         MIN_STACK_BOUNDARY.
17182         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17183         MIN_STACK_BOUNDARY.
17184         (BIGGEST_ALIGNMENT): Set to 128.
17185         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17186         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17187         STACK_BOUNDARY.
17189 2018-03-19  Richard Biener  <rguenther@suse.de>
17191         PR tree-optimization/84933
17192         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17193         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17195 2018-03-19  Richard Biener  <rguenther@suse.de>
17197         PR tree-optimization/84859
17198         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17199         (cond_if_else_store_replacement): Perform sinking operation on
17200         single-store BBs regardless of MAX_STORES_TO_SINK setting.
17201         Generalize what a BB with a single eligible store is.
17203 2018-03-19  Richard Biener  <rguenther@suse.de>
17205         PR tree-optimization/84929
17206         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17207         chrec_is_positive against non-chrec arg.
17209 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
17211         PR target/84711
17212         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17214 2018-03-18  Martin Liska  <mliska@suse.cz>
17216         PR rtl-optimization/84635
17217         * regrename.c (build_def_use): Use matches_mode only when
17218         matches >= 0.
17220 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
17222         PR tree-optimization/84913
17223         * tree-vect-loop.c (vectorizable_reduction): Don't try to
17224         vectorize chains of COND_EXPRs.
17226 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
17228         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17230 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
17232         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17234 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
17236         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17238 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
17239             Kito Cheng  <kito.cheng@gmail.com>
17241         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17242         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17243         (nds32_adjust_reg_alloc_order): New function.
17244         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17246 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
17248         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17249         nds32_print_operand, nds32_print_operand_address): Use
17250         HOST_WIDE_INT_PRINT_DEC instead.
17252 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
17254         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17256 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
17258         PR target/84902
17259         * config/i386/i386.c (initial_ix86_tune_features,
17260         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17261         unsigned long long.
17262         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17263         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17264         rather than 1u << ix86_tune.  Formatting fix.
17265         (ix86_option_override_internal): Change ix86_arch_mask from
17266         unsigned int to unsigned HOST_WIDE_INT, initialize to
17267         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17268         (ix86_function_specific_restore): Likewise.
17270 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
17272         PR target/84899
17273         * postreload.c (reload_combine_recognize_pattern): Perform
17274         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17275         truncate_int_for_mode the result for the destination's mode.
17277         PR c/84909
17278         * hsa-gen.c (mem_type_for_type): Fix comment typo.
17279         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17280         Likewise.
17281         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17282         Likewise.
17284 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
17286         PR target/84876
17287         * lra-assigns.c (lra_split_hard_reg_for): Don't use
17288         regno_allocno_class_array and sorted_pseudos.
17289         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17290         insns where regno is used.
17292 2018-03-16  Martin Liska  <mliska@suse.cz>
17294         PR ipa/84833
17295         * multiple_target.c (create_dispatcher_calls): Redirect
17296         reference in the symbol table.
17298 2018-03-16  Martin Liska  <mliska@suse.cz>
17300         PR ipa/84722
17301         * multiple_target.c (create_dispatcher_calls): Redirect also
17302         an alias.
17304 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
17306         PR c++/79937
17307         PR c++/82410
17308         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17309         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17310         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17312 2018-03-16  Julia Koval  <julia.koval@intel.com>
17314         * doc/invoke.texi (Skylake Server): Add CLWB.
17315         Cannonlake): Remove CLWB.
17317 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
17319         PR tree-optimization/84841
17320         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17321         1 << 3.
17322         (FLOAT_ONE_CONST_TYPE): Define.
17323         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17324         (sort_by_operand_rank): Put entries with higher constant_type last
17325         rather than first to match comments.
17327 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
17329         * config/nios2/nios2.md (movsi_internal): Fix thinko in
17330         split predicate.
17332 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
17334         PR c++/79085
17335         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17336         check and use address of target always.
17338 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
17340         PR target/84574
17341         * config/i386/i386.c (indirect_thunk_needed): Update comments.
17342         (indirect_thunk_bnd_needed): Likewise.
17343         (indirect_thunks_used): Likewise.
17344         (indirect_thunks_bnd_used): Likewise.
17345         (indirect_return_needed): New.
17346         (indirect_return_bnd_needed): Likewise.
17347         (output_indirect_thunk_function): Add a bool argument for
17348         function return.
17349         (output_indirect_thunk_function): Don't generate alias for
17350         function return thunk.
17351         (ix86_code_end): Call output_indirect_thunk_function to generate
17352         function return thunks.
17353         (ix86_output_function_return): Set indirect_return_bnd_needed
17354         and indirect_return_needed instead of indirect_thunk_bnd_needed
17355         and indirect_thunk_needed.
17357 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
17359         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17360         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17361         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17363 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
17364             Paul Hua <paul.hua.gm@gmail.com>
17366         PR c/84852
17367         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17369 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
17371         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17372         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17373         resp. SFmode cases.
17375 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
17377         PR target/84711
17378         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17379         instead of GET_MODE_SIZE when comparing Units.
17381 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
17383         PR target/68256
17384         * varasm.c (hash_section): Return an unchangeble hash value
17385         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
17386         Return !aarch64_can_use_per_function_literal_pools_p ().
17388 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
17390         PR target/84860
17391         * optabs.c (emit_conditional_move): Pass address of cmode's copy
17392         rather than address of cmode as last argument to prepare_cmp_insn.
17394 2018-03-15  Julia Koval  <julia.koval@intel.com>
17396         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17397         F_AVX512VNNI, F_AVX512BITALG): New.
17399 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
17401         PR target/83451
17402         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
17403         insn for floating-point loads and stores.
17405 2018-03-14  Carl Love  <cel@us.ibm.com>
17407         * config/rs6000/rs6000-c.c: Add macro definitions for
17408         ALTIVEC_BUILTIN_VEC_PERMXOR.
17409         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
17410         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
17411         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
17412         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
17413         UNSPEC_VPERMXOR.
17414         * config/doc/extend.texi: Add prototypes for vec_permxor.
17416 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
17418         PR c/84852
17419         * diagnostic-show-locus.c (class layout_point): Convert m_line
17420         from int to linenum_type.
17421         (line_span::comparator): Use linenum "compare" function when
17422         comparing line numbers.
17423         (test_line_span): New function.
17424         (layout_range::contains_point): Convert param "row" from int to
17425         linenum_type.
17426         (layout_range::intersects_line_p): Likewise.
17427         (layout::will_show_line_p): Likewise.
17428         (layout::print_source_line): Likewise.
17429         (layout::should_print_annotation_line_p): Likewise.
17430         (layout::print_annotation_line): Likewise.
17431         (layout::print_leading_fixits): Likewise.
17432         (layout::annotation_line_showed_range_p): Likewise.
17433         (struct line_corrections): Likewise for field m_row.
17434         (line_corrections::line_corrections): Likewise for param "row".
17435         (layout::print_trailing_fixits): Likewise.
17436         (layout::get_state_at_point): Likewise.
17437         (layout::get_x_bound_for_row): Likewise.
17438         (layout::print_line): Likewise.
17439         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
17440         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
17441         * input.c (selftest::test_linenum_comparisons): New function.
17442         (selftest::input_c_tests): Call it.
17443         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
17444         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
17445         * selftest.h (ASSERT_GT): New macro.
17446         (ASSERT_GT_AT): New macro.
17447         (ASSERT_LT): New macro.
17448         (ASSERT_LT_AT): New macro.
17450 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
17452         PR rtl-optimization/84780
17453         * combine.c (distribute_links): Don't make a link based on pc_rtx.
17455 2018-03-14  Martin Liska  <mliska@suse.cz>
17457         * tree.c (record_node_allocation_statistics): Use
17458         get_stats_node_kind.
17459         (get_stats_node_kind): New function extracted from
17460         record_node_allocation_statistics.
17461         (free_node): Use get_stats_node_kind.
17463 2018-03-14  Richard Biener  <rguenther@suse.de>
17465         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17466         that the value-set of ANTIC_IN doesn't grow.
17468         Revert
17469         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17470         member.
17471         (BB_VISITED_WITH_VISITED_SUCCS): New define.
17472         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17474 2018-03-14  Julia Koval  <julia.koval@intel.com>
17476         * config.gcc (icelake-client, icelake-server): New.
17477         (icelake): Remove.
17478         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17479         (initial_ix86_arch_features): Ditto.
17480         (PTA_SKYLAKE): Add SGX.
17481         (PTA_ICELAKE): Remove.
17482         (PTA_ICELAKE_CLIENT): New.
17483         (PTA_ICELAKE_SERVER): New.
17484         (ix86_option_override_internal): Split up icelake on icelake client and
17485         icelake server.
17486         (get_builtin_code_for_version): Ditto.
17487         (fold_builtin_cpu): Ditto.
17488         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17489         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17490         * config/i386/i386.h (processor_type): Ditto.
17491         * doc/invoke.texi: Ditto.
17493 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
17495         PR sanitizer/83392
17496         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17497         INTEGER_CST offset, add it together with bitpos / 8 and
17498         sign extend based on POINTER_SIZE.
17500         PR target/84844
17501         Revert
17502         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
17504         PR target/78090
17505         * config/i386/constraints.md (Yc): New register constraint.
17506         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17507         Use Yc constraint for alternative 2 of operand 0.  Remove
17508         preferred_for_speed attribute.
17510 2018-03-14  Richard Biener  <rguenther@suse.de>
17512         PR tree-optimization/84830
17513         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17514         with the old one to avoid oscillations.
17516 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
17518         PR target/83712
17519         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17520         pseudos.
17521         (assign_by_spills): Return a flag of reload assignment failure.
17522         Do not process the reload assignment failures.  Do not spill other
17523         reload pseudos if they has the same reg class.  Update n if
17524         necessary.
17525         (lra_assign): Add a return arg.  Set up from the result of
17526         assign_by_spills call.
17527         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17528         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17529         usage_insns if it is not NULL.
17530         (spill_hard_reg_in_range): New function.
17531         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17532         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17533         function prototypes.
17534         (lra_assign): Change prototype.
17535         * lra.c (lra): Add code to deal with fails by splitting hard reg
17536         live ranges.
17538 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
17540         * config/riscv/riscv.opt (mrelax): New option.
17541         * config/riscv/riscv.c (riscv_file_start): Emit ".option
17542         "norelax" when riscv_mrelax is disabled.
17543         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17545 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
17547         PR target/84743
17548         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17549         reassociation for int modes.
17551 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17553         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17554         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17555         for big-endian.
17556         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17557         * config/aarch64/aarch64-sve.md
17558         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17559         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17560         (*extend<mode><Vwide>2): Rename to...
17561         (aarch64_sve_extend<mode><Vwide>2): ...this.
17562         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17563         renaming the old pattern to...
17564         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
17565         unsigned packs.
17566         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17567         define_expand, renaming the old pattern to...
17568         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17569         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17570         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17571         account when deciding which SVE instruction the optab should use.
17572         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17574 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17576         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17577         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17578         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17579         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17580         (tlsdesc_small_<mode>): Turn a define_expand and use
17581         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
17582         (tlsdesc_small_advsimd_<mode>): ...this.
17583         (tlsdesc_small_sve_<mode>): New pattern.
17585 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17587         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17588         (UNSPEC_UMUL_HIGHPART): New constants.
17589         (MUL_HIGHPART): New int iteraor.
17590         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17591         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17592         define_expand.
17593         (*<su>mul<mode>3_highpart): New define_insn.
17595 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
17597         PR lto/84805
17598         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17599         incomplete types.
17601 2018-03-13  Martin Liska  <mliska@suse.cz>
17603         PR ipa/84658.
17604         * (sem_item_optimizer::sem_item_optimizer): Initialize new
17605         vector.
17606         (sem_item_optimizer::~sem_item_optimizer): Release it.
17607         (sem_item_optimizer::merge_classes): Register variable aliases.
17608         (sem_item_optimizer::fixup_pt_set): New function.
17609         (sem_item_optimizer::fixup_points_to_sets): Likewise.
17610         * ipa-icf.h: Declare new variables and functions.
17612 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
17614         PR middle-end/84834
17615         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17616         integer_pow2p@2 and test integer_pow2p in condition.
17617         (A < 0 ? C : 0): Similarly for @1.
17619         PR middle-end/84831
17620         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17621         characters starting at p contain '\0' character, don't look beyond
17622         that.
17624         PR target/84827
17625         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17626         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17628         PR target/84828
17629         * reg-stack.c (change_stack): Change update_end var from int to
17630         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17631         also call set_block_for_insn on the newly added insns and rescan.
17633         PR target/84786
17634         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17635         on the last operand.
17637         PR c++/84704
17638         * tree.c (stabilize_reference_1): Return save_expr (e) for
17639         STATEMENT_LIST even if it doesn't have side-effects.
17641 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
17643         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17645 2018-03-12  Renlin Li  <renlin.li@arm.com>
17647         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17648         aarch64_output_scalar_simd_mov_immediate.
17650 2018-03-12  Martin Sebor  <msebor@redhat.com>
17652         PR tree-optimization/83456
17653         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17654         for perfectly overlapping calls to memcpy.
17655         (gimple_fold_builtin_memory_chk): Same.
17656         (gimple_fold_builtin_strcpy): Handle no-warning.
17657         (gimple_fold_builtin_stxcpy_chk): Same.
17658         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17660 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
17662         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17663         parameter.  Use it for SFmode.
17664         (rs6000_function_arg_advance_1): Adjust.
17665         (rs6000_function_arg): Adjust.
17666         (rs6000_gimplify_va_arg): Pass false for that new parameter.
17668 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
17670         PR rtl-optimization/84169
17671         PR rtl-optimization/84780
17672         * combine.c (can_combine_p): Check for a 2-insn combination whether
17673         the destination register is used between the two insns, too.
17675 2018-03-12  Richard Biener  <rguenther@suse.de>
17677         PR tree-optimization/84803
17678         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17679         for refs DR analysis didn't process.
17681 2018-03-12  Richard Biener  <rguenther@suse.de>
17683         PR tree-optimization/84777
17684         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17685         force-vectorize loops ignore whether we are optimizing for size.
17687 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
17689         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17690         (TARGET_MD_ASM_ADJUST): Define.
17692 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
17693             Kito Cheng  <kito.cheng@gmail.com>
17694             Chung-Ju Wu  <jasonwucj@gmail.com>
17696         * config/nds32/nds32.c (nds32_compute_stack_frame,
17697         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17698         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17699         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17700         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17701         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17702         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17703         * config/nds32/nds32.md (prologue, epilogue): Use macro
17704         NDS32_V3PUSH_AVAILABLE_P to do checking.
17706 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
17708         PR debug/58150
17709         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17710         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17711         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
17712         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
17713         addition of most attributes on !orig_type_die or the attribute not
17714         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17716 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17717             Chung-Ju Wu  <jasonwucj@gmail.com>
17719         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17720         __NDS32_VH__ macro.
17721         * config/nds32/nds32.opt (mvh): New option.
17723 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17724             Chung-Ju Wu  <jasonwucj@gmail.com>
17726         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17727         function.
17728         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17729         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17730         definition.
17732 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17733             Chung-Ju Wu  <jasonwucj@gmail.com>
17735         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17736         function.
17737         * config/nds32/nds32-multiple.md (strlensi): New pattern.
17738         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17740 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
17741             Kito Cheng  <kito.cheng@gmail.com>
17742             Chung-Ju Wu  <jasonwucj@gmail.com>
17744         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17745         UNSPEC_FFMISM and UNSPEC_FLMISM.
17746         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17747         for ffb, ffmism and flmism.
17748         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17749         (unspec_ffmism): Ditto.
17750         (unspec_flmism): Ditto.
17751         (nds32_expand_builtin_impl): Check if string extension is available.
17752         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17753         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17755 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
17757         Reverting patch:
17758         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
17760         PR target/83712
17761         * lra-assigns.c (assign_by_spills): Return a flag of reload
17762         assignment failure.  Do not process the reload assignment
17763         failures.  Do not spill other reload pseudos if they has the same
17764         reg class.
17765         (lra_assign): Add a return arg.  Set up from the result of
17766         assign_by_spills call.
17767         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17768         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17769         usage_insns if it is not NULL.
17770         (spill_hard_reg_in_range): New function.
17771         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17772         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17773         function prototypes.
17774         (lra_assign): Change prototype.
17775         * lra.c (lra): Add code to deal with fails by splitting hard reg
17776         live ranges.
17778 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
17780         PR target/84807
17781         * config/i386/i386.opt: Replace Enforcment with Enforcement.
17783 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
17785         PR debug/84620
17786         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17787         (dw_val_node): Add val_symbolic_view.
17788         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17789         (symview_upper_bound): New.
17790         (new_line_info_table): Initialize symviews_since_reset.
17791         (dwarf2out_source_line): Count symviews_since_reset and set
17792         symview_upper_bound.
17793         (dw_val_equal_p): Handle symview.
17794         (add_AT_symview): New.
17795         (print_dw_val): Handle symview.
17796         (attr_checksum, attr_checksum_ordered): Likewise.
17797         (same_dw_val_p, size_of_die): Likewise.
17798         (value_format, output_die): Likewise.
17799         (add_high_low_attributes): Use add_AT_symview for entry_view.
17800         (dwarf2out_finish): Reset symview_upper_bound, clear
17801         zero_view_p.
17803 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
17805         PR target/83969
17806         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17807         Add strict argument and use it.
17808         (rs6000_split_multireg_move): Update for new strict argument.
17809         (mem_operand_gpr): Disallow all non-offsettable addresses.
17810         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17812 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
17814         PR target/84772
17815         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17816         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17817         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17819         PR c++/84767
17820         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17821         decl, use remap_type if we want to use the type.
17823 2018-03-09  Martin Sebor  <msebor@redhat.com>
17825         PR tree-optimization/84526
17826         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17827         Remove dead code.
17828         (builtin_access::generic_overlap): Be prepared to handle non-array
17829         base objects.
17831 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
17833         PR rtl-optimization/84682
17834         * lra-constraints.c (process_address_1): Check is_address flag
17835         for address constraints.
17836         (process_alt_operands): Likewise.
17837         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17838         preprocess_constraints.
17839         * recog.h (preprocess_constraints): Add oploc parameter.
17840         Adjust callers.
17841         * recog.c (preprocess_constraints): Test address_operand for
17842         CT_ADDRESS constraints.
17844 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
17846         PR target/83712
17847         * lra-assigns.c (assign_by_spills): Return a flag of reload
17848         assignment failure.  Do not process the reload assignment
17849         failures.  Do not spill other reload pseudos if they has the same
17850         reg class.
17851         (lra_assign): Add a return arg.  Set up from the result of
17852         assign_by_spills call.
17853         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17854         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17855         usage_insns if it is not NULL.
17856         (spill_hard_reg_in_range): New function.
17857         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17858         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17859         function prototypes.
17860         (lra_assign): Change prototype.
17861         * lra.c (lra): Add code to deal with fails by splitting hard reg
17862         live ranges.
17864 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17866         PR target/83193
17867         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17868         Accept complain bool parameter.  Only emit errors if it is true.
17869         (arm_parse_cpu_option_name): Likewise.
17870         (arm_target_thumb_only): Adjust callers of the above.
17871         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17872         prototype to take a default true bool parameter.
17873         (arm_parse_arch_option_name): Likewise.
17875 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
17876             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
17878         PR jit/64089
17879         PR jit/84288
17880         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17881         * configure: Regenerate.
17882         * configure.ac ("linker --version-script option"): New.
17883         ("linker soname option"): New.
17885 2018-03-09  Richard Biener  <rguenther@suse.de>
17887         PR tree-optimization/84775
17888         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17889         immediate uses of predicate stmts and mark them modified.
17891         Revert
17892         PR tree-optimization/84178
17893         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17894         to caller.
17895         (version_loop_for_if_conversion): Delay update_ssa call.
17896         (tree_if_conversion): Delay update_ssa until after predicate
17897         insertion.
17899 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
17901         PR target/84763
17902         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17903         when the function accesses prior frames.
17905 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
17907         PR debug/84456
17908         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17909         gen_llsym, otherwise call maybe_gen_llsym.
17911         PR inline-asm/84742
17912         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17913         has ',' character inside of it.
17915 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17917         PR target/84748
17918         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17919         as clobbering CC_REGNUM.
17921 2018-03-08  Richard Biener  <rguenther@suse.de>
17923         PR middle-end/84552
17924         * tree-scalar-evolution.c: Include tree-into-ssa.h.
17925         (follow_copies_to_constant): Do not follow SSA names registered
17926         for update.
17928 2018-03-08  Richard Biener  <rguenther@suse.de>
17930         PR tree-optimization/84178
17931         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17932         to caller.
17933         (version_loop_for_if_conversion): Delay update_ssa call.
17934         (tree_if_conversion): Delay update_ssa until after predicate
17935         insertion.
17937 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
17939         PR tree-optimization/84178
17940         * tree-if-conv.c (release_bb_predicate): Remove the
17941         the assertion that the stmts have NULL use_ops.
17942         Discard the statements, asserting that they haven't
17943         yet been added to a BB.
17945 2018-03-08  Richard Biener  <rguenther@suse.de>
17947         PR tree-optimization/84746
17948         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17949         (phi_translate): Pass in destination ANTIC_OUT set.
17950         (phi_translate_1): Likewise.  For a simplified result lookup
17951         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17952         (phi_translate_set): Adjust.
17953         (do_pre_regular_insertion): Likewise.
17954         (do_pre_partial_partial_insertion): Likewise.
17956 2018-03-08  Martin Liska  <mliska@suse.cz>
17958         PR gcov-profile/84735
17959         * doc/gcov.texi: Document usage of profile files.
17960         * gcov-io.h: Document changes in the format.
17962 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
17964         PR debug/84404
17965         PR debug/84408
17966         * dwarf2out.c (struct dw_line_info_table): Update comments for
17967         view == -1.
17968         (FORCE_RESET_NEXT_VIEW): New.
17969         (FORCE_RESETTING_VIEW_P): New.
17970         (RESETTING_VIEW_P): Check for -1 too.
17971         (ZERO_VIEW_P): Likewise.
17972         (new_line_info_table): Force-reset next view.
17973         (dwarf2out_begin_function): Likewise.
17974         (dwarf2out_source_line): Simplify zero_view_p initialization.
17975         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17976         view directly.  Omit view when omitting .loc at line 0.
17978 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
17980         PR tree-optimization/84740
17981         * tree-switch-conversion.c (process_switch): Call build_constructors
17982         only if info.phi_count is non-zero.
17984         PR tree-optimization/84739
17985         * tree-tailcall.c (find_tail_calls): Check call arguments against
17986         DECL_ARGUMENTS (current_function_decl) rather than
17987         DECL_ARGUMENTS (func) when checking for tail recursion.
17989 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
17991         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17992         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17993         Volker Reichelt's entry and add entries for people that perform
17994         GCC fuzzy testing and report numerous bugs.
17996 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
17998         PR target/82411
17999         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
18000         readonly data in sdata, if that is disabled.
18001         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
18002         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
18003         -mreadonly-in-sdata option.
18005 2018-03-07  Martin Sebor  <msebor@redhat.com>
18007         PR tree-optimization/84468
18008         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
18009         basic block when looking for nul assignment.
18011 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
18013         PR target/84277
18014         * except.h (output_function_exception_table): Adjust prototype.
18015         * except.c (output_function_exception_table): Remove FNNAME parameter
18016         and add SECTION parameter.  Ouput one part of the table at a time.
18017         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
18018         the first part of the exception table and emit unwind directives.
18019         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
18020         (i386_pe_seh_cold_init): Likewise.
18021         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
18022         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
18023         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
18024         (ix86_output_call_insn): Emit a nop in one more case for SEH.
18025         * config/i386/winnt.c: Include except.h.
18026         (struct seh_frame_state): Add reg_offset, after_prologue and
18027         in_cold_section fields.
18028         (i386_pe_seh_end_prologue): Set seh->after_prologue.
18029         (i386_pe_seh_cold_init): New function.
18030         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
18031         to seh->in_cold_section.
18032         (seh_emit_push): Record the offset of the push.
18033         (seh_emit_save): Record the offet of the save.
18034         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
18035         Test seh->after_prologue to disregard the epilogue.
18036         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
18037         (i386_pe_end_cold_function): New function.
18039 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
18041         PR fortran/84565
18042         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
18043         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
18045         PR c++/84704
18046         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
18047         on tmp_var.
18048         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
18049         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
18051         PR middle-end/84723
18052         * multiple_target.c: Include tree-inline.h and intl.h.
18053         (expand_target_clones): Diagnose and fail if node->definition and
18054         !tree_versionable_function_p (node->decl).
18056 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
18058         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
18059         sprint_ul.
18060         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
18061         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
18062         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
18064 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
18066         PR target/84710
18067         * combine.c (try_combine): Use reg_or_subregno instead of handling
18068         just paradoxical SUBREGs and REGs.
18070 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
18072         * config/arc/arc.c (arc_finalize_pic): Remove function.
18073         (arc_must_save_register): We use single base PIC register, remove
18074         checks to save/restore the PIC register.
18075         (arc_expand_prologue): Likewise.
18076         * config/arc/arc-protos.h (arc_set_default_type_attributes):
18077         Remove.
18078         (arc_verify_short): Likewise.
18079         (arc_attr_type): Likewise.
18080         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
18081         (walk_stores): Likewise.
18082         (arc_address_cost): Make it static.
18083         (arc_verify_short): Likewise.
18084         (branch_dest): Likewise.
18085         (arc_attr_type): Likewise.
18086         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
18087         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
18088         (arc_final_prescan_insn): Remove inserting the nops due to
18089         hardware hazards.  It is done in reorg step.
18090         (insn_length_variant_t): Remove.
18091         (insn_length_parameters_t): Likewise.
18092         (arc_insn_length_parameters): Likewise.
18093         (arc_get_insn_variants): Likewise.
18094         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
18096 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
18098         PR inline-asm/84683
18099         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
18100         assertion failure.
18102         PR tree-optimization/84687
18103         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
18104         on new_node->decl.
18105         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
18107 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18109         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
18110         Rename to ppc_speculation_barrier.
18111         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
18112         __builtin_ppc_speculation_barrier.
18114 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
18116         PR target/84700
18117         * combine.c (combine_simplify_rtx): Don't try to simplify if
18118         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
18119         are equal to x.
18121 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
18123         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
18124         to 32 bytes when compiling for POWER9.
18126 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
18128         PR target/84564
18129         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
18130         regparm >= 3 with no arg reg available also for calls with
18131         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
18133         PR target/84524
18134         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
18135         orig,vex.
18136         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
18138 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
18140         PR target/84264
18141         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
18143 2018-03-05  Richard Biener  <rguenther@suse.de>
18145         PR tree-optimization/84486
18146         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
18147         When inserting a __builtin_assume_aligned call set the LHS
18148         SSA name alignment info accordingly.
18150 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
18152         PR tree-optimization/84114
18153         * config/aarch64/aarch64.c (aarch64_reassociation_width)
18154         Avoid reassociation of FLOAT_MODE addition.
18156 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
18158         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
18159         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
18160         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
18161         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
18162         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
18163         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
18164         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
18165         and -mwbnoinvd.
18166         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
18167         __builtin_ia32_wbinvd): New builtins.
18168         (SPECIAL_ARGS2): New.
18169         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
18170         (SPECIAL_ARGS2): New.
18171         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18172         (ix86_valid_target_attribute_inner_p): Ditto.
18173         (ix86_init_mmx_sse_builtins): Add special_args2.
18174         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18175         TARGET_WBNOINVD_P): New.
18176         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18177         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
18178         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
18179         * config/i386/immintrin.h (_wbinvd): New intrinsic.
18180         * config/i386/pconfigintrin.h: New file.
18181         * config/i386/wbnoinvdintrin.h: Ditto.
18182         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18183         wbnoinvdintrin.h.
18184         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18186 2018-03-05  Richard Biener  <rguenther@suse.de>
18188         PR tree-optimization/84670
18189         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18190         member.
18191         (BB_VISITED_WITH_VISITED_SUCCS): New define.
18192         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18193         (compute_antic_aux): Only assert the number of values in ANTIC_IN
18194         doesn't grow if all successors (recursively) were visited at least
18195         once.
18197 2018-03-05  Richard Biener  <rguenther@suse.de>
18199         PR tree-optimization/84650
18200         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18201         if executed in the loop pipeline.
18203 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
18205         * doc/configfiles.texi (Configuration Files): Move info about
18206         conditionalizing $target-protos.h to...
18207         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
18208         differs from $target-protos.h.
18210 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
18211             Chung-Ju Wu  <jasonwucj@gmail.com>
18213         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18214         * config/nds32/nds32-multiple.md (setmemsi): Define.
18215         * config/nds32/nds32-memory-manipulation.c
18216         (nds32_gen_dup_4_byte_to_word_value): New.
18217         (emit_setmem_word_loop): New.
18218         (emit_setmem_byte_loop): New.
18219         (nds32_expand_setmem_loop): New.
18220         (nds32_expand_setmem_loop_v3m): New.
18221         (nds32_expand_setmem_unroll): New.
18222         (nds32_expand_setmem): New.
18224 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
18225             Chung-Ju Wu  <jasonwucj@gmail.com>
18227         * config/nds32/nds32-memory-manipulation.c
18228         (nds32_emit_load_store): New.
18229         (nds32_emit_post_inc_load_store): New.
18230         (nds32_emit_mem_move): New.
18231         (nds32_emit_mem_move_block): New.
18232         (nds32_expand_movmemsi_loop_unknown_size): New.
18233         (nds32_expand_movmemsi_loop_known_size): New.
18234         (nds32_expand_movmemsi_loop): New.
18235         (nds32_expand_movmemsi_unroll): New.
18236         (nds32_expand_movmemqi): Rename ...
18237         (nds32_expand_movmemsi): ... to this.
18238         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18239         (movmemsi): ... to this.
18240         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18241         (nds32_expand_movmemsi): ... to this.
18243 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
18244             Monk Chiang  <sh.chiang04@gmail.com>
18245             Chung-Ju Wu  <jasonwucj@gmail.com>
18247         * config/nds32/nds32-protos.h
18248         (nds32_expand_load_multiple): New arguments.
18249         (nds32_expand_store_multiple): Ditto.
18250         (nds32_valid_multiple_load_store): Rename ...
18251         (nds32_valid_multiple_load_store_p): ... to this.
18252         * config/nds32/nds32-memory-manipulation.c
18253         (nds32_expand_load_multiple): Refine implementation.
18254         (nds32_expand_store_multiple): Ditto.
18255         * config/nds32/nds32-multiple.md
18256         (load_multiple): Update nds32_expand_load_multiple interface.
18257         (store_multiple): Update nds32_expand_store_multiple interface.
18258         * config/nds32/nds32-predicates.c
18259         (nds32_valid_multiple_load_store): Rename ...
18260         (nds32_valid_multiple_load_store_p): ... to this and refine
18261         implementation.
18262         * config/nds32/predicates.md
18263         (nds32_load_multiple_and_update_address_operation): New predicate.
18264         (nds32_store_multiple_and_update_address_operation): New predicate.
18266 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
18267             Chung-Ju Wu  <jasonwucj@gmail.com>
18269         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18270         (combo): New attribute.
18271         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18273 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
18275         * config/nds32/nds32.opt: Change -mcmodel= default value.
18277 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
18278             Monk Chiang  <sh.chiang04@gmail.com>
18279             Chung-Ju Wu  <jasonwucj@gmail.com>
18281         * config/nds32/constants.md (unspec_element): New enum.
18282         * config/nds32/constraints.md (Umw): New constraint.
18283         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18284         * config/nds32/nds32-intrinsic.md: Likewise.
18285         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18286         (nds32_valid_smw_lwm_base_p): New.
18287         (nds32_output_smw_single_word): New.
18288         (nds32_output_lmw_single_word): New.
18289         (nds32_expand_unaligned_load): New.
18290         (nds32_expand_unaligned_store): New.
18291         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18292         (nds32_output_smw_single_word): Declare.
18293         (nds32_output_lmw_single_word): Declare.
18294         (nds32_expand_unaligned_load): Declare.
18295         (nds32_expand_unaligned_store): Declare.
18296         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18297         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18298         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18299         NDS32_BUILTIN_UASTORE_DW.
18300         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18301         predicate.
18303 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
18304             Kito Cheng  <kito.cheng@gmail.com>
18305             Chung-Ju Wu  <jasonwucj@gmail.com>
18307         * config/nds32/nds32-intrinsic.c
18308         (nds32_expand_builtin_null_ftype_reg): Delete.
18309         (nds32_expand_builtin_reg_ftype_imm): Ditto.
18310         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18311         (nds32_read_argument): New.
18312         (nds32_legitimize_target): Ditto.
18313         (nds32_legitimize_argument): Ditto.
18314         (nds32_check_constant_argument): Ditto.
18315         (nds32_expand_unop_builtin): Ditto.
18316         (nds32_expand_unopimm_builtin): Ditto.
18317         (nds32_expand_binop_builtin): Ditto.
18318         (nds32_builtin_decl_impl): Ditto.
18319         (builtin_description): Ditto.
18320         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18321         (nds32_init_builtins_impl): Ditto.
18322         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18323         (nds32_builtin_decl): New.
18324         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18325         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18327 2018-03-02  Jeff Law  <law@redhat.com>
18329         * reorg.c (stop_search_p): Handle DEBUG_INSN.
18330         (redundant_insn, fill_simple_delay_slots): Likewise.
18331         (fill_slots_from_thread): Likewise.
18332         * resource.c (mark_referenced_resources): Likewise.
18333         (mark_set_resources, find_dead_or_set_registers): Likewise.
18335 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18337         * substring-locations.h (format_warning_va): Formatting fix for
18338         ATTRIBUTE_GCC_DIAG.
18339         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18340         argument.
18341         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18342         * substring-locations.c: Include intl.h.
18343         (format_warning_va): Turned into small wrapper around
18344         format_warning_n_va, renamed to ...
18345         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18346         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18347         use ngettext.
18348         (format_warning_at_substring_n): New function.
18349         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18350         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
18351         format_warning_at_substring with just a shorter name instead of
18352         const function pointer.
18353         (fmtwarn_n): New function.
18354         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18355         appropriate, get rid of all the fmtstr temporaries, move conditionals
18356         with G_() wrapped string literals directly into fmtwarn arguments,
18357         cast dir.len to (int), formatting fixes.
18359 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
18361         * doc/invoke.texi: Remove "Cilk Plus" references.
18363 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18364             Richard Biener  <rguenther@suse.de>
18366         PR ipa/84628
18367         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18368         for error or warning attributes if CALL_FROM_THUNK_P is set.
18369         Formatting fixes.
18371 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18373         PR target/56540
18374         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18375         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18377         PR target/56540
18378         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18379         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18381         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
18382         instead of -1U in last predictors element's probability member.
18384 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
18386         PR ipa/83983
18387         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18388         arguments if they are comparable.
18390 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
18392         PR tree-optimization/84634
18393         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
18394         masks and masked_loop_p with a single loop_masks, making sure it's
18395         null for bb vectorization.
18397 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
18399         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
18400         (vect_analyze_data_ref_access): Use loop->safe_len rather than
18401         loop->force_vectorize to check whether there is no alias.
18403 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18405         PR target/84614
18406         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
18407         prototypes.
18408         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
18409         comments.
18410         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
18411         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
18412         instead of a loop around prev_real_insn.
18413         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
18414         prev_real_insn.
18416         PR inline-asm/84625
18417         * config/i386/i386.c (ix86_print_operand): Use conditional
18418         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
18419         zero vector.
18421 2018-03-02  Richard Biener  <rguenther@suse.de>
18423         PR tree-optimization/84427
18424         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
18425         (bitmap_set_subtract_values): Rewrite to handle multiple
18426         exprs per value.
18427         (clean): Likewise.
18428         (prune_clobbered_mems): Likewise.
18429         (phi_translate): Take edge instead of pred/phiblock.
18430         (phi_translate_1): Likewise.
18431         (phi_translate_set): Likewise.  Insert all translated
18432         exprs for a value into the set, keeping possibly multiple
18433         expressions per value.
18434         (compute_antic_aux): Adjust for phi_translate changes.
18435         When intersecting union the expressions and prune those
18436         not in the final value set, keeping possibly multiple
18437         expressions per value.  Do not use value-insertion
18438         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
18439         all expressions.  Add verification that the value-sets
18440         only shrink during iteration.
18441         (compute_partial_antic_aux): Adjust for the phi_translate changes.
18442         (do_pre_regular_insertion): Likewise.
18443         (do_pre_partial_partial_insertion): Likewise.
18445 2018-03-02  Richard Biener  <rguenther@suse.de>
18447         PR target/82005
18448         * config/darwin.c (saved_debug_info_level): New static global.
18449         (darwin_asm_lto_start): Disable debug info generation for LTO out.
18450         (darwin_asm_lto_end): Restore debug info generation settings.
18452 2018-03-01  Martin Liska  <mliska@suse.cz>
18454         PR sanitizer/82484
18455         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18456         volatile arguments.
18458 2018-03-01  Richard Biener  <rguenther@suse.de>
18460         PR debug/84645
18461         * dwarf2out.c (gen_variable_die): Properly handle late VLA
18462         type annotation with LTO when debug was disabled at compile-time.
18464 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
18466         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18467         XINT with INTVAL.
18468         (mips_final_postscan_insn): Likewise.
18470 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
18472         PR rtl-optimization/84528
18473         * alias.c (init_alias_target): Add commentary.
18474         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18475         a unique base value if the frame pointer is not eliminated
18476         to the stack pointer.
18478 2018-03-01  Tom de Vries  <tom@codesourcery.com>
18480         PR rtl-optimization/83327
18481         * lra-int.h (hard_regs_spilled_into): Declare.
18482         * lra.c (hard_regs_spilled_into): Define.
18483         (init_reg_info): Init hard_regs_spilled_into.
18484         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18485         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18486         (process_bb_lives): Handle hard_regs_spilled_into.
18487         (lra_create_live_ranges_1): Before doing liveness propagation, clear
18488         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18490 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
18492         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18493         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18494         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18495         * config/rs6000/aix72.h: New file.
18497 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
18499         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18500         instead of warning_at with conditional singular and plural messages
18501         where possible.
18503         PR target/52991
18504         * stor-layout.c (update_alignment_for_field): For
18505         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18506         && !DECL_PACKED (field), do the alignment update, just use
18507         only desired_align instead of MAX (type_align, desired_align)
18508         as the alignment.
18509         (place_field): Don't do known_align < desired_align handling
18510         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18511         is non-NULL, instead do it after rli->prev_field handling and
18512         only if not within a bitfield word.  For DECL_PACKED (field)
18513         use type_align of BITS_PER_UNIT.
18515 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
18517         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18518         superfluous parentheses and trailing spaces.
18520 2018-02-28  Richard Biener  <rguenther@suse.de>
18522         PR tree-optimization/84584
18523         * graphite-scop-detection.c (scop_detection::add_scop): Discard
18524         SCoPs with fake exit edge.
18526 2018-02-28  Martin Liska  <mliska@suse.cz>
18528         PR testsuite/84597
18529         * timevar.c (timer::print): Fix format to properly print 100%
18530         values.
18532 2018-02-28  Richard Biener  <rguenther@suse.de>
18534         PR middle-end/84607
18535         * genmatch.c (capture_info::walk_match): Do not mark
18536         captured expressions without operands as expr_p given
18537         they act more like predicates and should be subject to
18538         "lost tail" side-effect preserving.
18540 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
18542         PR rtl-optimization/81611
18543         * auto-inc-dec.c (attempt_change): Move dead note from
18544         mem_insn if it's the next use of regno
18545         (find_address): Take address use of reg holding
18546         non-incremented value.  Add parm to limit search to the named
18547         reg only.
18548         (merge_in_block): Attempt to use a mem insn that is the next
18549         use of the original regno.
18551 2018-02-27  Martin Sebor  <msebor@redhat.com>
18553         PR c++/83871
18554         * doc/invoke.texi (-Wmissing-attributes): New option.
18555         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18557 2018-02-27  Martin Sebor  <msebor@redhat.com>
18559         PR translation/84207
18560         * diagnostic-core.h (warning_n, error_n, inform_n): Change
18561         n argument to unsigned HOST_WIDE_INT.
18562         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18563         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
18564         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18565         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18567 2018-02-27  Richard Biener  <rguenther@suse.de>
18569         PR tree-optimization/84512
18570         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18571         Do not use the estimate returned from record_stmt_cost for
18572         the scalar iteration cost but sum properly using add_stmt_cost.
18574 2018-02-27  Richard Biener  <rguenther@suse.de>
18576         PR tree-optimization/84466
18577         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18578         Adjust last change to less strictly validate use operands.
18580 2018-02-27  Martin Liska  <mliska@suse.cz>
18582         PR gcov-profile/84548
18583         * gcov.c (process_file): Allow partial overlap and consider it
18584         also as group functions.
18585         (output_lines): Properly calculate range of lines for a group.
18587 2018-02-27  Martin Liska  <mliska@suse.cz>
18589         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18590         'ggc' suffixes.  Change first column width.
18591         (timer::print): Fix formatting of the column.
18593 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
18595         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18596         preserve inline entry blocks for the sake of debug inline
18597         entry point markers alone.
18598         (remove_unused_locals): Suggest in comments a better place to
18599         force the preservation of inline entry blocks that are
18600         otherwise unused, but do not preserve them.
18602 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18604         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18606 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18608         PR target/84039
18609         * config/i386/constraints.md (Bs): Replace
18610         ix86_indirect_branch_register with
18611         TARGET_INDIRECT_BRANCH_REGISTER.
18612         (Bw): Likewise.
18613         * config/i386/i386.md (indirect_jump): Likewise.
18614         (tablejump): Likewise.
18615         (*sibcall_memory): Likewise.
18616         (*sibcall_value_memory): Likewise.
18617         Peepholes of indirect call and jump via memory: Likewise.
18618         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18619         (*sibcall_value_GOT_32): Likewise.
18620         * config/i386/predicates.md (indirect_branch_operand): Likewise.
18621         (GOT_memory_operand): Likewise.
18622         (call_insn_operand): Likewise.
18623         (sibcall_insn_operand): Likewise.
18624         (GOT32_symbol_operand): Likewise.
18625         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18627 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
18629         PR rtl-optimization/83496
18630         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18631         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
18632         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18633         redundant insn, if any.
18634         (relax_delay_slots): Likewise.
18635         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18637 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
18639         PR tree-optimization/83965
18640         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18641         that grouped statements are part of a reduction chain.  Return
18642         true if the statement is not marked as a reduction itself but
18643         is part of a group.
18644         (vect_recog_dot_prod_pattern): Don't check whether the statement
18645         is part of a group here.
18646         (vect_recog_sad_pattern): Likewise.
18647         (vect_recog_widen_sum_pattern): Likewise.
18649 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
18651         PR debug/84545
18652         * final.c (rest_of_clean_state): Also look for calls inside sequences.
18654 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18656         PR target/84530
18657         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18658         the bool argument.
18659         (ix86_output_indirect_function_return): New prototype.
18660         (ix86_split_simple_return_pop_internal): Likewise.
18661         * config/i386/i386.c (indirect_return_via_cx): New.
18662         (indirect_return_via_cx_bnd): Likewise.
18663         (indirect_thunk_name): Handle return va CX_REG.
18664         (output_indirect_thunk_function): Create alias for
18665         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18666         (ix86_output_indirect_jmp): Remove the bool argument.
18667         (ix86_output_indirect_function_return): New function.
18668         (ix86_split_simple_return_pop_internal): Likewise.
18669         * config/i386/i386.md (*indirect_jump): Don't pass false
18670         to ix86_output_indirect_jmp.
18671         (*tablejump_1): Likewise.
18672         (simple_return_pop_internal): Change it to define_insn_and_split.
18673         Call ix86_split_simple_return_pop_internal to split it for
18674         -mfunction-return=.
18675         (simple_return_indirect_internal): Call
18676         ix86_output_indirect_function_return instead of
18677         ix86_output_indirect_jmp.
18679 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
18681         PR bootstrap/84405
18682         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18683         memset and value initialization afterwards.
18685 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
18687         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18689 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18691         PR target/84521
18692         * common/config/aarch64/aarch64-common.c
18693         (aarch_option_optimization_table[]): Switch
18694         off fomit-frame-pointer
18696 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
18697             Chung-Ju Wu  <jasonwucj@gmail.com>
18699         * config/nds32/nds32-multiple.md (load_multiple): Disallow
18700         volatile memory.
18701         (store_multiple): Ditto.
18703 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
18705         * config.gcc: Add --with-cpu support for nds32 target.
18706         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18707         * config/nds32/nds32.opt: Add -mcpu= option.
18709 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
18711         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18712         isel=yes): Warn for these deprecated options.
18714 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
18716         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18717         ISA_2_5_MASKS_EMBEDDED.
18719 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
18721         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18722         p->max as pointers rather than using iterative_hash_expr.
18724 2018-02-23  Carl Love  <cel@us.ibm.com>
18726         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18727         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18728         BU_P8V_OVERLOAD_2.
18729         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18730         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18731         P8V_BUILTIN_VEC_VUNSIGNED2.
18733 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
18735         PR target/81572
18736         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18737         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18738         LRA_UNKNOWN_ALT.
18739         * lra-constraints.c (curr_insn_transform): Set up
18740         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
18741         LRA_UNKNOWN_ALT.
18742         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18743         * lra-eliminations.c (spill_pseudos): Ditto.
18744         (process_insn_for_elimination): Ditto.
18745         * lra-lives.c (reg_early_clobber_p): Use the new macros.
18746         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18747         LRA_NON_CLOBBERED_ALT.
18749 2018-02-22  Martin Sebor  <msebor@redhat.com>
18751         PR tree-optimization/84480
18752         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18753         to maybe_diag_stxncpy_trunc.  Call it.
18754         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18755         from gimple_fold_builtin_strcpy.  Print inlining stack.
18756         (handle_builtin_stxncpy): Print inlining stack.
18757         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18759 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
18761         PR target/84176
18762         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18763         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18764         and -fcheck-pointer-bounds are used together.
18765         (indirect_thunk_prefix): New enum.
18766         (indirect_thunk_need_prefix): New function.
18767         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
18768         "_nt" instead of "_bnd" for NOTRACK prefix.
18769         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18770         (output_indirect_thunk_function): Likewise.
18771         (): Likewise.
18772         (ix86_code_end): Update output_indirect_thunk_function calls.
18773         (ix86_output_indirect_branch_via_reg): Replace
18774         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18775         (ix86_output_indirect_branch_via_push): Likewise.
18776         (ix86_output_function_return): Likewise.
18777         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18778         incompatible with -fcf-protection=branch and
18779         -fcheck-pointer-bounds.
18781 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
18783         PR target/83335
18784         * config/aarch64/aarch64.c (aarch64_print_address_internal):
18785         Change gcc_assert call to output_operand_lossage.
18787 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
18789         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18791 2018-02-22  DJ Delorie  <dj@redhat.com>
18792             Sebastian Perta  <sebastian.perta@renesas.com>
18793             Oleg Endo  <olegendo@gcc.gnu.org>
18795         * config/rx/rx.c (rx_rtx_costs): New function.
18796         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18798 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18800         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18802 2018-02-22  Martin Liska  <mliska@suse.cz>
18804         PR driver/83193
18805         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18806         Add "native" as a possible value.
18808 2018-02-22  Martin Liska  <mliska@suse.cz>
18810         PR driver/83193
18811         * config/i386/i386.c (ix86_option_override_internal):
18812         Add "native" as a possible value for -march and -mtune.
18814 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
18816         PR target/84502
18817         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18818         to all type variants.
18820         PR tree-optimization/84503
18821         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18822         width as info->bitpos + info->bitsize - start.
18823         (merged_store_group::merge_overlapping): Simplify width computation.
18824         (check_no_overlap): New function.
18825         (imm_store_chain_info::try_coalesce_bswap): Compute expected
18826         start + width and last_order of the group, fail if check_no_overlap
18827         fails.
18828         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18829         to group if check_no_overlap fails.
18831 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
18833         * config/rs6000/altivec.md: Delete contraint arguments to
18834         define_expand, define_split, and define_peephole2, and in
18835         define_insn_and_split if always unused.
18836         * config/rs6000/darwin.md: Ditto.
18837         * config/rs6000/dfp.md: Ditto.
18838         * config/rs6000/rs6000.md: Ditto.
18839         * config/rs6000/sync.md: Ditto.
18840         * config/rs6000/vector.md: Ditto.
18841         * config/rs6000/vsx.md: Ditto.
18843 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
18845         * config/rs6000/altivec.md: Write output control strings as braced
18846         blocks instead of double-quoted strings.
18847         * config/rs6000/darwin.md: Ditto.
18848         * config/rs6000/rs6000.md: Ditto.
18849         * config/rs6000/vector.md: Ditto.
18850         * config/rs6000/vsx.md: Ditto.
18852 2018-02-21  Jason Merrill  <jason@redhat.com>
18854         PR c++/84314 - ICE with templates and fastcall attribute.
18855         * attribs.c (build_type_attribute_qual_variant): Remove assert.
18857 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
18859         * ipa-cp.c (determine_versionability): Fix comment typos.
18861 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
18863         PR c/84229
18864         * ipa-cp.c (determine_versionability): Do not version functions caling
18865         va_arg_pack.
18867 2018-02-21  Martin Liska  <mliska@suse.cz>
18869         PR driver/83193
18870         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18871         Add "native" as a possible value.
18872         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
18873         the macro when native cpu detection is available.
18875 2018-02-21  Martin Liska  <mliska@suse.cz>
18877         PR driver/83193
18878         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18879         Add "native" as a possible value.
18880         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18881         when native cpu detection is available.
18883 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
18884             Martin Sebor  <msebor@redhat.com>
18886         PR tree-optimization/84478
18887         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18888         false.
18889         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18890         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18891         support which is conservatively correct, for 2 only stay conservative
18892         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
18893         argument to the 2 argument get_range_strlen, adjust 6 arg
18894         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18895         false.
18896         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18897         (gimple_fold_builtin_strlen): Pass true as last argument to
18898         get_range_strlen.
18900 2018-02-20  Martin Sebor  <msebor@redhat.com>
18902         PR middle-end/84095
18903         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18904         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
18905         (builtin_memref::builtin_memref): Factor out parts into
18906         set_base_and_offset and call it.
18908 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
18910         PR middle-end/84406
18911         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18912         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18913         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
18914         search at the associated MODE_INT.
18916 2018-02-20  Jeff Law  <law@redhat.com>
18918         PR middle-end/82123
18919         PR tree-optimization/81592
18920         PR middle-end/79257
18921         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18922         for range data rather than using global data.
18923         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18924         range data rather than using global data.
18925         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18926         pass it to children as needed.
18927         (struct directive::fmtresult): Similarly.
18928         (struct directive::set_width): Similarly.
18929         (struct directive::set_precision): Similarly.
18930         (format_integer, format_directive, parse_directive): Similarly.
18931         (format_none): Accept unnamed vr_values parameter.
18932         (format_percent, format_floating, format_character): Similarly.
18933         (format_string, format_plain): Similarly.
18934         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18935         the EVRP range analyzer for range data rather than using global data.
18936         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18937         gimple-ssa-evrp-analyze.h
18938         (class sprintf_dom_walker): Add after_dom_children member function.
18939         Add evrp_range_analyzer member.
18940         (sprintf_dom_walker::before_dom_children): Call into the EVRP
18941         range analyzer as needed.
18942         (sprintf_dom_walker::after_dom_children): New member function.
18943         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18944         if not optimizing.
18945         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18946         (evrp_range_analyzer::pop_to_marker): Likewise.
18948 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
18950         PR tree-optimization/84419
18951         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18952         with the required type if its current type is compatible but
18953         different.
18955 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
18957         PR middle-end/82004
18958         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18959         after vectorization.
18961 2018-02-20  Martin Liska  <mliska@suse.cz>
18963         PR driver/83193
18964         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18965         possible values if we don't have a hint.
18967 2018-02-20  Martin Liska  <mliska@suse.cz>
18969         PR c/84310
18970         PR target/79747
18971         * final.c (shorten_branches): Build align_tab array with one
18972         more element.
18973         * opts.c (finish_options): Add alignment option limit check.
18974         (MAX_CODE_ALIGN): Likewise.
18975         (MAX_CODE_ALIGN_VALUE): Likewise.
18976         * doc/invoke.texi: Document maximum allowed option value for
18977         all -falign-* options.
18979 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
18981         PR target/84146
18982         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18983         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18984         * var-tracking.c (emit_note_insn_var_location): Remove all references
18985         to NOTE_INSN_CALL_ARG_LOCATION.
18986         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18987         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18988         Use copy_rtx_if_shared.
18989         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18990         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18991         (dwarf2out_var_location): Remove handling of
18992         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18993         on call_insn.
18994         * final.c (final_scan_insn): Remove all references to
18995         NOTE_INSN_CALL_ARG_LOCATION.
18996         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
18997         before dumping final insns.
18998         * except.c (emit_note_eh_region_end): Remove all references to
18999         NOTE_INSN_CALL_ARG_LOCATION.
19000         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
19001         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
19002         * config/arc/arc.c (hwloop_optimize): Likewise.
19003         * config/arm/arm.c (create_fix_barrier): Likewise.
19004         * config/s390/s390.c (s390_chunkify_start): Likewise.
19005         * config/sh/sh.c (find_barrier): Likewise.
19006         * config/i386/i386.c (rest_of_insert_endbranch,
19007         ix86_seh_fixup_eh_fallthru): Likewise.
19008         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
19009         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
19010         * config/frv/frv.c (frv_function_prologue): Likewise.
19011         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
19012         reg note.
19013         (note_outside_basic_block_p): Remove all references to
19014         NOTE_INSN_CALL_ARG_LOCATION.
19015         * gengtype.c (adjust_field_rtx_def): Likewise.
19016         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
19017         Likewise.
19018         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
19019         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
19021         PR c++/84444
19022         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
19023         is ADDR_EXPR.
19025         PR tree-optimization/84452
19026         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
19027         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
19028         is NULL.
19030 2018-02-19  Martin Liska  <mliska@suse.cz>
19032         PR sanitizer/82183
19033         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
19035 2018-02-19  Martin Liska  <mliska@suse.cz>
19036             Richard Sandiford  <richard.sandiford@linaro.org>
19038         PR tree-optimization/82491
19039         * gimple-fold.c (get_base_constructor): Make earlier bail out
19040         to prevent ubsan.
19042 2018-02-19  Carl Love  <cel@us.ibm.com>
19044         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
19045         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
19046         BU_P8V_OVERLOAD_1.
19047         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
19048         P8V_BUILTIN_VEC_NEG.
19050 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
19052         * config/rl78/rl78.md (movdf): New define expand.
19054 2018-02-19  Martin Liska  <mliska@suse.cz>
19056         PR other/80589
19057         * doc/invoke.texi: Fix typo.
19058         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
19060 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
19062         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
19063         handle rs6000_single_float and rs6000_double_float specially for
19064         e500 family CPUs.
19066 2018-02-16  Jeff Law  <law@redhat.com>
19068         * config/rx/rx.c (add_pop_cfi_notes): New function.;
19069         (pop_regs): Use it.
19071 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
19073         PR ipa/84425
19074         * ipa-inline.c (inline_small_functions): Fix a typo.
19076 2018-02-16  Nathan Sidwell  <nathan@acm.org>
19078         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
19080 2018-02-16  Carl Love  <cel@us.ibm.com>
19082         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
19083         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
19084         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
19085         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
19086         expansion to P8V_BUILTIN_VEC_FLOAT2.
19088 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
19090         PR rtl-optimization/70023
19091         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
19092         src_regno into account.
19094 2018-02-16  Carl Love  <cel@us.ibm.com>
19096         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
19097         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
19098         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
19099         * config/rs6000/rs6000.c: Remove case statements for
19100         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
19101         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
19102         and P9V_BUILTIN_VEC_VINSERT4B.
19103         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
19104         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
19105         * config/rs6000/vsx.md:
19106         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
19107         vec_insert4b.
19109 2018-02-16  Carl Love  <cel@us.ibm.com>
19111         * config/rs6000/altivec.h: Add builtin names vec_extract4b
19112         vec_insert4b.
19113         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
19114         definitions.
19115         * config/rs6000/rs6000-c.c: Add the definitions for
19116         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
19117         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
19118         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
19119         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
19120         definition for insert4b and define insn *insert3b_internal.
19121         * doc/extend.texi: Add documentation for vec_extract4b.
19123 2018-02-16  Nathan Sidwell  <nathan@acm.org>
19125         * doc/extend.texi (Backwards Compatibility): Mention friend
19126         injection.  Note for-scope is deprecated.
19127         * doc/invoke.texi (-ffriend-injection): Deprecate.
19129 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
19131         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
19132         that moved to I2, also allow destinations that are a paradoxical
19133         subreg (instead of a normal reg).
19135 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
19137         PR target/83831
19138         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
19139         to QImode.
19141 2018-02-16  Richard Biener  <rguenther@suse.de>
19143         PR tree-optimization/84037
19144         PR tree-optimization/84016
19145         PR target/82862
19146         * config/i386/i386.c (ix86_builtin_vectorization_cost):
19147         Adjust vec_construct for the fact we need additional higher latency
19148         128bit inserts for AVX256 and AVX512 vector builds.
19149         (ix86_add_stmt_cost): Scale vector construction cost for
19150         elementwise loads.
19152 2018-02-16  Richard Biener  <rguenther@suse.de>
19154         PR tree-optimization/84417
19155         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
19156         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
19157         (non_rewritable_lvalue_p): Likewise, use poly-ints.
19159 2018-02-16  Martin Liska  <mliska@suse.cz>
19161         PR sanitizer/84307
19162         * internal-fn.def (ASAN_CHECK): Set proper flags.
19163         (ASAN_MARK): Likewise.
19165 2018-02-16  Julia Koval  <julia.koval@intel.com>
19167         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
19168         from PTA_CANNONLAKE.
19170 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
19172         PR target/84272
19173         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19174         Use ++iter rather than iter++ for std::list iterators.
19175         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
19176         defer deleting them until all nodes in the forest are processed.  Do
19177         free even leaf nodes.  Change to_process into auto_vec.
19179         PR bootstrap/84405
19180         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19181         * vec.h (vec_default_construct): Use memset instead of placement new
19182         if BROKEN_VALUE_INITIALIZATION is defined.
19183         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19184         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19185         is defined.
19187         PR rtl-optimization/83723
19188         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19189         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
19190         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
19191         recursive calls.
19192         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19193         callers.
19194         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19196 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
19198         PR rtl-optimization/81443
19199         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19200         from inner REGs to paradoxical SUBREGs.
19202 2018-02-16  Richard Biener  <rguenther@suse.de>
19204         PR tree-optimization/84399
19205         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19206         For operands we can analyze at their definition make sure we can
19207         analyze them at each use as well.
19209 2018-02-16  Richard Biener  <rguenther@suse.de>
19211         PR tree-optimization/84190
19212         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19213         volatile accesses if the decl isn't volatile.
19215 2018-02-15  Jason Merrill  <jason@redhat.com>
19217         PR c++/84314 - ICE with templates and fastcall attribute.
19218         * attribs.c (build_type_attribute_qual_variant): Don't clobber
19219         TYPE_CANONICAL on an existing type.
19221 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
19223         PR tree-optimization/84383
19224         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19225         dstoff nor call operand_equal_p if dstbase is NULL.
19227         PR tree-optimization/84334
19228         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19229         also a CONSTANT_CLASS_P, punt.
19231 2018-02-14  Jim Wilson  <jimw@sifive.com>
19233         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19234         first SMALL_OPERAND check.  New local min_second_step.  Move assert
19235         to where locals are set.  Add TARGET_RVC support.
19236         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19238 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
19240         * doc/invoke.texi: Correct -Wformat-overflow code sample.
19242 2018-02-14  Martin Sebor  <msebor@redhat.com>
19244         PR tree-optimization/83698
19245         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19246         arrays constrain the offset range to their bounds.
19247         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19248         (builtin_access::overlap): Avoid setting the size of overlap if it's
19249         already been set.
19250         (maybe_diag_overlap): Also consider arrays when deciding what values
19251         of offsets to include in diagnostics.
19253 2018-02-14  Martin Sebor  <msebor@redhat.com>
19255         PR c/84108
19256         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19257         that correspond to the kind of a declaration.
19259 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
19261         PR target/83984
19262         * config/pa/pa.md: Load address of PIC label using the linkage table
19263         if the label is nonlocal.
19265 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19267         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19268         warning message if user requests -maltivec=be.
19269         * doc/invoke.texi: Document deprecation of -maltivec=be.
19271 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
19273         PR target/84220
19274         * config/rs6000/rs6000-c.c: Update definitions for
19275         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19276         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19278 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
19280         PR target/84239
19281         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19282         add _get_ssp intrinsics. Remove argument from
19283         __builtin_ia32_rdssp[d|q].
19284         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19285         * config/i386/i386-builtin.def: Remove argument from
19286         __builtin_ia32_rdssp[d|q].
19287         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19288         ix86_expand_special_args_builtin for _rdssp[d|q].
19289         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19290         Clear register before usage.
19291         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19292         Add documentation for new _get_ssp and _inc_ssp intrinsics.
19294 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
19296         PR tree-optimization/84357
19297         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19298         operand 1 of an ARRAY_REF too.
19300 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
19302         PR target/83831
19303         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19304         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19305         declarations.
19306         (set_of_reg): New struct.
19307         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19308         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19309         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19310         functions.
19311         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19312         Split into bitclr, bitset, bitinvert patterns if appropriate.
19313         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19314         use rx_fuse_in_memory_bitop.
19315         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19316         to named insn, correct maximum insn length.
19318 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
19320         PR target/79242
19321         * machmode.def: Define a complex mode for PARTIAL_INT.
19322         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19323         MODE_PARTIAL_INT.
19324         * doc/rtl.texi: Document CSPImode.
19325         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19326         handling.
19327         (msp430_hard_regno_nregs_with_padding): Likewise.
19329 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
19331         PR target/84279
19332         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19334 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
19336         PR rtl-optimization/84169
19337         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
19338         we generated a parallel as new i3 and we split that to new i2 and i3
19339         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
19340         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
19341         those to i2, not i1.  Partially rewrite this scan code.
19343 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
19345         PR c/82210
19346         * stor-layout.c (place_field): For variable length fields, adjust
19347         offset_align afterwards not just based on the field's alignment,
19348         but also on the size.
19350         PR middle-end/84309
19351         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19352         of exps and logs in the use_exp2 case.
19354 2018-02-13  Jeff Law  <law@redhat.com>
19356         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19357         entry for "vector".
19359         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19360         ARGS as unused.
19362 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
19364         PR debug/84342
19365         PR debug/84319
19366         * common.opt (gas-loc-support, gas-locview-support): New.
19367         (ginline-points, ginternal-reset-location-views): New.
19368         * doc/invoke.texi: Document them.  Use @itemx where intended.
19369         (gvariable-location-views): Adjust.
19370         * target.def (reset_location_view): New.
19371         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19372         (TARGET_RESET_LOCATION_VIEW): New.
19373         * doc/tm.texi: Rebuilt.
19374         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19375         (dwarf2out_default_as_locview_support): New.
19376         (output_asm_line_debug_info): Use option variables.
19377         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19378         (output_loc_list): Likewise.
19379         (add_high_low_attributes): Check option variables.
19380         Don't output entry view attribute in strict mode.
19381         (gen_inlined_subroutine_die): Check option variables.
19382         (dwarf2out_inline_entry): Likewise.
19383         (init_sections_and_labels): Likewise.
19384         (dwarf2out_early_finish): Likewise.
19385         (maybe_reset_location_view): New, from...
19386         (dwarf2out_var_location): ... here.  Call it.
19387         * debug.h (dwarf2out_default_as_loc_support): Declare.
19388         (dwarf2out_default_as_locview_support): Declare.
19389         * hooks.c (hook_int_rtx_insn_0): New.
19390         * hooks.h (hook_int_rtx_insn_0): Declare.
19391         * toplev.c (process_options): Take -gas-loc-support and
19392         -gas-locview-support from dwarf2out.  Enable
19393         -gvariable-location-views by default only with locview
19394         assembler support.  Enable -ginternal-reset-location-views by
19395         default only if the target defines the corresponding hook.
19396         Enable -ginline-points by default if location views are
19397         enabled; force it disabled if statement frontiers are
19398         disabled.
19399         * tree-inline.c (expand_call_inline): Check option variables.
19400         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
19402 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
19404         PR tree-optimization/84321
19405         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
19406         handling.  Also check whether the anti-range contains any values
19407         that satisfy the mask; switch to a VR_RANGE if not.
19409 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
19411         PR sanitizer/84340
19412         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
19414 2018-02-13  Martin Jambor  <mjambor@suse.cz>
19416         PR c++/83990
19417         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
19418         of call statements, also set location of a load to a temporary.
19420 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
19422         * config/rl78/rl78.c (add_vector_labels): New function.
19423         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
19424         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
19425         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
19426         which checks that no arguments are passed.
19427         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
19428         * doc/extend.texi: Documentation for the new attribute.
19430 2018-02-13  Andreas Schwab  <schwab@suse.de>
19432         * config/riscv/linux.h (CPP_SPEC): Define.
19434 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
19436         PR target/84335
19437         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
19438         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
19439         OPTION_MASK_ISA_AES as first argument to def_builtin_const
19440         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
19441         instead of OPTION_MASK_ISA_PCLMUL as first argument to
19442         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
19443         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
19444         temporarily for AES and PCLMUL builtins.
19446         PR tree-optimization/84339
19447         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
19448         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19449         Formatting fixes.
19451         PR middle-end/84309
19452         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19453         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19454         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19455         inline function.
19456         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19457         inline function.
19458         * omp-simd-clone.h: New file.
19459         * omp-simd-clone.c: Include omp-simd-clone.h.
19460         (expand_simd_clones): No longer static.
19461         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19462         cgraph.h and omp-simd-clone.h.
19463         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19464         (vect_recog_widen_shift_pattern): Formatting fix.
19465         (vect_pattern_recog_1): Don't check optab for calls.
19467         PR target/84336
19468         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19469         operands[2] into a REG before using gen_lowpart on it.
19471 2018-02-12  Jeff Law  <law@redhat.com>
19473         PR target/83760
19474         * config/sh/sh.c (find_barrier): Consider a sibling call
19475         a barrier as well.
19477         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19478         successfully back substituting a reg.
19480 2018-02-12  Richard Biener  <rguenther@suse.de>
19482         PR tree-optimization/84037
19483         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19484         parameter, move visited init to caller.
19485         (vect_slp_analyze_operations): Separate cost from validity
19486         check, initialize visited once for all instances.
19487         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19488         for all instances.
19489         * tree-vect-stmts.c (vect_model_simple_cost): Make early
19490         out an assert.
19491         (vect_model_promotion_demotion_cost): Likewise.
19492         (vectorizable_bswap): Guard cost modeling with !slp_node
19493         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19494         SLP stmts.
19495         (vectorizable_call): Likewise.
19496         (vectorizable_conversion): Likewise.
19497         (vectorizable_assignment): Likewise.
19498         (vectorizable_shift): Likewise.
19499         (vectorizable_operation): Likewise.
19500         (vectorizable_store): Likewise.
19501         (vectorizable_load): Likewise.
19502         (vectorizable_condition): Likewise.
19503         (vectorizable_comparison): Likewise.
19505 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
19507         PR sanitizer/84307
19508         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19509         (ASAN_MARK): Fix fnspec to account for return value, change pointer
19510         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19512 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
19514         PR middle-end/83665
19515         * params.def (inline-min-speedup): Increase from 8 to 15.
19516         (max-inline-insns-auto): Decrease from 40 to 30.
19517         * ipa-split.c (consider_split): Add some buffer for function to
19518         be considered inlining candidate.
19519         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19520         default values.
19522 2018-02-12  Richard Biener  <rguenther@suse.de>
19524         PR tree-optimization/84037
19525         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19526         matched stmts if we cannot swap the non-matched ones.
19528 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
19530         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19531         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19532         _mm_maskz_scalef_round_ss): New intrinsics.
19533         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19534         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19535         __builtin_ia32_scalefss_round): Remove.
19536         (__builtin_ia32_scalefsd_mask_round,
19537         __builtin_ia32_scalefss_mask_round): New intrinsics.
19538         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19539         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19540         ((match_operand:VF_128 2 "<round_nimm_predicate>"
19541         "<round_constraint>")): Changed to ...
19542         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19543         "<round_scalar_constraint>")): ... this.
19544         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19545         %0, %1, %2<round_op3>}"): Changed to ...
19546         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19547         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19548         %2<round_scalar_mask_op3>}"): ... this.
19549         * config/i386/subst.md (round_scalar_nimm_predicate): New.
19551 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
19553         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19554         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19555         (_mm_maskz_sqrt_round_ss): New intrinsics.
19556         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19557         (__builtin_ia32_sqrtsd_mask_round)
19558         (__builtin_ia32_sqrtss_mask_round): New builtins.
19559         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19560         (__builtin_ia32_sqrtss_round): Remove.
19561         (__builtin_ia32_sqrtsd_mask_round)
19562         (__builtin_ia32_sqrtss_mask_round): New builtins.
19563         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19564         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19565         ((match_operand:VF_128 1 "vector_operand"
19566         "xBm,<round_constraint>")): Changed to ...
19567         ((match_operand:VF_128 1 "vector_operand"
19568         "xBm,<round_scalar_constraint>")): ... this.
19569         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19570         %0, %2, %<iptr>1<round_op3>}): Changed to ...
19571         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19572         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19573         %<iptr>1<round_scalar_mask_op3>}): ... this.
19574         ((set_attr "prefix" "<round_prefix>")): Changed to ...
19575         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19577 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
19579         PR target/84266
19580         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19581         Cast vec_cmpeq result to correct type.
19582         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19583         Cast vec_cmpgt result to correct type.
19585 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
19587         * final.c (final_scan_insn_1): Renamed from...
19588         (final_scan_insn): ... this.  New wrapper, to recover
19589         seen from the outermost call in recursive ones.
19590         * config/sparc/sparc.c (output_return): Drop seen from call.
19591         (output_sibcall): Likewise.
19592         * config/visium/visium.c (output_branch): Likewise.
19594 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
19596         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19597         function label.
19599 2018-02-10  Alan Modra  <amodra@gmail.com>
19601         PR target/84300
19602         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19603         Specify LR as an input.
19605 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
19607         PR sanitizer/83987
19608         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19609         remove_member_access_dummy_vars): New functions.
19610         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19611         lower_omp_1, execute_lower_omp): Use them.
19613         PR rtl-optimization/84308
19614         * shrink-wrap.c (spread_components): Release todo vector.
19616 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
19618         PR rtl-optimization/57193
19619         * ira-color.c (struct allocno_color_data): Add member
19620         conflict_allocno_hard_prefs.
19621         (update_conflict_allocno_hard_prefs): New.
19622         (bucket_allocno_compare_func): Add a preference based on
19623         conflict_allocno_hard_prefs.
19624         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19625         (color_allocnos): Remove a dead code.  Initiate
19626         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
19628 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
19630         PR target/84226
19631         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19632         constraint from =wa to wa.  Avoid a subreg on the output operand,
19633         instead use a pseudo and subreg it in a move.
19634         (p9_xxbrd_<mode>): Changed to ...
19635         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19636         (p9_xxbrd_v2df): New expander.
19637         (p9_xxbrw_<mode>): Changed to ...
19638         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19639         (p9_xxbrw_v4sf): New expander.
19641 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
19643         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19645 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
19647         PR target/83926
19648         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19649         multiply in 32-bit mode.
19650         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19651         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19652         mode.
19654 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
19656         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19657         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19658         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19659         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19661 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
19663         PR lto/84213
19664         * dwarf2out.c (is_trivial_indirect_ref): New function.
19665         (dwarf2out_late_global_decl): Do not generate a location
19666         attribute for variables that have a non-trivial DECL_VALUE_EXPR
19667         and that are not defined in the current unit.
19669 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
19671         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19672         instead of a libcall for UNORDERED.
19674 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
19676         PR target/82641
19677         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19678         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19680 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19682         PR target/PR84295
19683         * config/s390/s390.c (s390_set_current_function): Invoke
19684         s390_indirect_branch_settings also if fndecl didn't change.
19686 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
19688         * config/rs6000/rs6000.md (blockage): Set length to zero.
19690 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
19692         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19694 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
19696         PR sanitizer/84285
19697         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19698         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19699         -static-lib*san.
19701         PR debug/84252
19702         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19703         PARALLEL incoming that failed vt_get_decl_and_offset check.
19705         PR middle-end/84237
19706         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19707         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19708         TREE_READONLY bit.
19709         (get_variable_section): For decls in named .bss* sections pass true as
19710         second argument to bss_initializer_p.
19712 2018-02-09  Marek Polacek  <polacek@redhat.com>
19713             Jakub Jelinek  <jakub@redhat.com>
19715         PR c++/83659
19716         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19717         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
19718         Sync some changes from cxx_fold_indirect_ref.
19720 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
19722         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19723         markers.
19724         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19725         (BLOCK_INLINE_ENTRY_LABEL): New.
19726         (dwarf2out_var_location): Disregard inline entry markers.
19727         (inline_entry_data): New struct.
19728         (inline_entry_data_hasher): New hashtable type.
19729         (inline_entry_data_hasher::hash): New.
19730         (inline_entry_data_hasher::equal): New.
19731         (inline_entry_data_table): New variable.
19732         (add_high_low_attributes): Add DW_AT_entry_pc and
19733         DW_AT_GNU_entry_view attributes if a pending entry is found
19734         in inline_entry_data_table.  Add old entry_pc attribute only
19735         if debug nonbinding markers are disabled.
19736         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19737         markers are enabled.
19738         (block_within_block_p, dwarf2out_inline_entry): New.
19739         (dwarf2out_finish): Check that no entries remained in
19740         inline_entry_data_table.
19741         * final.c (reemit_insn_block_notes): Handle inline entry notes.
19742         (final_scan_insn, notice_source_line): Likewise.
19743         (rest_of_clean_state): Skip inline entry markers.
19744         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19745         markers.
19746         * gimple.c (gimple_build_debug_inline_entry): New.
19747         * gimple.h (enum gimple_debug_subcode): Add
19748         GIMPLE_DEBUG_INLINE_ENTRY.
19749         (gimple_build_debug_inline_entry): Declare.
19750         (gimple_debug_inline_entry_p): New.
19751         (gimple_debug_nonbind_marker_p): Adjust.
19752         * insn-notes.def (INLINE_ENTRY): New.
19753         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19754         inline entry marker notes.
19755         (print_insn): Likewise.
19756         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19757         (INSN_DEBUG_MARKER_KIND): Likewise.
19758         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19759         * tree-inline.c (expand_call_inline): Build and insert
19760         debug_inline_entry stmt.
19761         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19762         inline entry blocks early, if nonbind markers are enabled.
19763         (dump_scope_block): Dump fragment info.
19764         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19765         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19766         (gimple_build_debug_inline_entry): New.
19767         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19768         Enable/disable inline entry points too.
19769         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19770         (DEBUG_INSN): Describe inline entry markers.
19772         * common.opt (gvariable-location-views): New.
19773         (gvariable-location-views=incompat5): New.
19774         * config.in: Rebuilt.
19775         * configure: Rebuilt.
19776         * configure.ac: Test assembler for view support.
19777         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19778         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19779         * dwarf2out.c (var_loc_view): New typedef.
19780         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19781         (dwarf2out_locviews_in_attribute): New.
19782         (dwarf2out_locviews_in_loclist): New.
19783         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19784         (enum dw_line_info_opcode): Add LI_adv_address.
19785         (struct dw_line_info_table): Add view.
19786         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19787         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19788         (zero_view_p): New variable.
19789         (ZERO_VIEW_P): New macro.
19790         (output_asm_line_debug_info): New.
19791         (struct var_loc_node): Add view.
19792         (add_AT_view_list, AT_loc_list): New.
19793         (add_var_loc_to_decl): Add view param.  Test it against last.
19794         (new_loc_list): Add view params.  Record them.
19795         (AT_loc_list_ptr): Handle loc and view lists.
19796         (view_list_to_loc_list_val_node): New.
19797         (print_dw_val): Handle dw_val_class_view_list.
19798         (size_of_die): Likewise.
19799         (value_format): Likewise.
19800         (loc_list_has_views): New.
19801         (gen_llsym): Set vl_symbol too.
19802         (maybe_gen_llsym, skip_loc_list_entry): New.
19803         (dwarf2out_maybe_output_loclist_view_pair): New.
19804         (output_loc_list): Output view list or entries too.
19805         (output_view_list_offset): New.
19806         (output_die): Handle dw_val_class_view_list.
19807         (output_dwarf_version): New.
19808         (output_compilation_unit_header): Use it.
19809         (output_skeleton_debug_sections): Likewise.
19810         (output_rnglists, output_line_info): Likewise.
19811         (output_pubnames, output_aranges): Update version comments.
19812         (output_one_line_info_table): Output view numbers in asm comments.
19813         (dw_loc_list): Determine current endview, pass it to new_loc_list.
19814         Call maybe_gen_llsym.
19815         (loc_list_from_tree_1): Adjust.
19816         (add_AT_location_description): Create view list attribute if
19817         needed, check it's absent otherwise.
19818         (convert_cfa_to_fb_loc_list): Adjust.
19819         (maybe_emit_file): Call output_asm_line_debug_info for test.
19820         (dwarf2out_var_location): Reset views as needed.  Precompute
19821         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
19822         attribute.  Set view.
19823         (new_line_info_table): Reset next view.
19824         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19825         (dwarf2out_source_line): Likewise.  Output view resets and labels to
19826         the assembler, or select appropriate line info opcodes.
19827         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19828         (optimize_string_length): Catch it.  Adjust.
19829         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
19830         dw_val_class_view_list, and remove it if no longer needed.
19831         (hash_loc_list): Hash view numbers.
19832         (loc_list_hasher::equal): Compare them.
19833         (optimize_location_lists): Check whether a view list symbol is
19834         needed, and whether the locview attribute is present, and
19835         whether they match.  Remove the locview attribute if no longer
19836         needed.
19837         (index_location_lists): Call skip_loc_list_entry for test.
19838         (dwarf2out_finish): Call output_asm_line_debug_info for test.
19839         Use output_dwarf_version.
19840         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19841         (struct dw_val_node): Add val_view_list.
19842         * final.c (SEEN_NEXT_VIEW): New.
19843         (set_next_view_needed): New.
19844         (clear_next_view_needed): New.
19845         (maybe_output_next_view): New.
19846         (final_start_function): Rename to...
19847         (final_start_function_1): ... this.  Take pointer to FIRST,
19848         add SEEN parameter.  Emit param bindings in the initial view.
19849         (final_start_function): Reintroduce SEEN-less interface.
19850         (final): Rename to...
19851         (final_1): ... this.  Take SEEN parameter.  Output final pending
19852         next view at the end.
19853         (final): Reintroduce seen-less interface.
19854         (final_scan_insn): Output pending next view before switching
19855         sections or ending a block.  Mark the next view as needed when
19856         outputting variable locations.  Notify debug backend of section
19857         changes, and of location view changes.
19858         (rest_of_handle_final): Adjust.
19859         * toplev.c (process_options): Autodetect value for debug variable
19860         location views option.  Warn on incompat5 without -gdwarf-5.
19861         * doc/invoke.texi (gvariable-location-views): New.
19862         (gvariable-location-views=incompat5): New.
19863         (gno-variable-location-views): New.
19865 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
19867         PR tree-optimization/84136
19868         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19869         that the result of find_edge is non-NULL.
19871 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
19873         PR target/83008
19874         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19875         storing integer register in SImode.  Fix cost of 256 and 512
19876         byte aligned SSE register store.
19878 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
19880         * config/i386/i386.c (ix86_multiplication_cost): Fix
19881         multiplication cost for TARGET_AVX512DQ.
19883 2018-02-08  Marek Polacek  <polacek@redhat.com>
19885         PR tree-optimization/84238
19886         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19887         get_range_strlen.
19889 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19891         PR tree-optimization/84265
19892         * tree-vect-stmts.c (vectorizable_store): Don't treat
19893         VMAT_CONTIGUOUS accesses as grouped.
19894         (vectorizable_load): Likewise.
19896 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19898         PR tree-optimization/81635
19899         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19900         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19901         (test_round_for_mask): New functions.
19902         (wide_int_cc_tests): Call test_round_for_mask.
19903         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19904         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19905         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19906         range returned by get_range_info.
19908 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
19910         PR ipa/81360
19911         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19912         * symtab.c: Include builtins.h
19913         (symtab_node::output_to_lto_symbol_table_p): Move here
19914         from lto-streamer-out.c:output_symbol_p.
19915         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19916         (output_symbol_p): Move all logic to symtab.c
19917         (produce_symtab): Update.
19919 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19921         * config/s390/s390-opts.h (enum indirect_branch): Define.
19922         * config/s390/s390-protos.h (s390_return_addr_from_memory)
19923         (s390_indirect_branch_via_thunk)
19924         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19925         (enum s390_indirect_branch_type): Define.
19926         * config/s390/s390.c (struct s390_frame_layout, struct
19927         machine_function): Remove.
19928         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19929         (indirect_branch_table_label_no, indirect_branch_table_name):
19930         Define variables.
19931         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19932         (enum s390_indirect_branch_option): Define.
19933         (s390_return_addr_from_memory): New function.
19934         (s390_handle_string_attribute): New function.
19935         (s390_attribute_table): Add new attribute handler.
19936         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19937         (s390_indirect_branch_via_thunk): New function.
19938         (s390_indirect_branch_via_inline_thunk): New function.
19939         (s390_function_ok_for_sibcall): When jumping via thunk disallow
19940         sibling call optimization for non z10 compiles.
19941         (s390_emit_call): Force indirect branch target to be a single
19942         register.  Add r1 clobber for non-z10 compiles.
19943         (s390_emit_epilogue): Emit return jump via return_use expander.
19944         (s390_reorg): Handle JUMP_INSNs as execute targets.
19945         (s390_option_override_internal): Perform validity checks for the
19946         new command line options.
19947         (s390_indirect_branch_attrvalue): New function.
19948         (s390_indirect_branch_settings): New function.
19949         (s390_set_current_function): Invoke s390_indirect_branch_settings.
19950         (s390_output_indirect_thunk_function):  New function.
19951         (s390_code_end): Implement target hook.
19952         (s390_case_values_threshold): Implement target hook.
19953         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19954         macros.
19955         * config/s390/s390.h (struct s390_frame_layout)
19956         (struct machine_function): Move here from s390.c.
19957         (TARGET_INDIRECT_BRANCH_NOBP_RET)
19958         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19959         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19960         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19961         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19962         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19963         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19964         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19965         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19966         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19967         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19968         (mnemonic attribute): Add values which aren't recognized
19969         automatically.
19970         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19971         pattern for branch conversion.  Fix mnemonic attribute.
19972         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19973         indirect branch via thunk if requested.
19974         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19975         ("*indirect_jump"): Disable for branch conversion using out of
19976         line thunks.
19977         ("indirect_jump_via_thunk<mode>_z10")
19978         ("indirect_jump_via_thunk<mode>")
19979         ("indirect_jump_via_inlinethunk<mode>_z10")
19980         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19981         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19982         ("casesi_jump_via_inlinethunk<mode>_z10")
19983         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19984         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19985         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19986         ("*indirect2_jump"): Disable for branch conversion.
19987         ("casesi_jump"): Turn into expander and expand patterns for branch
19988         conversion.
19989         ("return_use"): New expander.
19990         ("*return"): Emit return via thunk and rename it to ...
19991         ("*return<mode>"): ... this one.
19992         * config/s390/s390.opt: Add new options and and enum for the
19993         option values.
19995 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19997         * lra-constraints.c (match_reload): Unconditionally use
19998         gen_lowpart_SUBREG, rather than selecting between that
19999         and equivalent gen_rtx_SUBREG code.
20001 2018-02-08  Richard Biener  <rguenther@suse.de>
20003         PR tree-optimization/84233
20004         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
20005         changed flag instead of boguously re-using phi_inserted.
20007 2018-02-08  Martin Jambor  <mjambor@suse.cz>
20009         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
20010         static local variables.
20012 2018-02-08  Richard Biener  <rguenther@suse.de>
20014         PR tree-optimization/84278
20015         * tree-vect-stmts.c (vectorizable_store): When looking for
20016         smaller vector types to perform grouped strided loads/stores
20017         make sure the mode is supported by the target.
20018         (vectorizable_load): Likewise.
20020 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
20022         * config/aarch64/aarch64.c (aarch64_components_for_bb):
20023         Increase LDP/STP opportunities by adding adjacent callee-saves.
20025 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
20027         PR rtl-optimization/84068
20028         PR rtl-optimization/83459
20029         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
20031 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
20033         PR tree-optimization/84224
20034         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
20035         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
20036         non-zero arguments.
20038 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
20040         PR target/84113
20041         * config/rs6000/altivec.md (*restore_world): Remove LR use.
20042         * config/rs6000/predicates.md (restore_world_operation): Adjust op
20043         count, remove one USE.
20045 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
20047         * doc/install.texi (Configuration): Document the
20048         --with-long-double-format={ibm,ieee} PowerPC configuration
20049         options.
20051         PR target/84154
20052         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
20053         Convert from define_expand to be define_insn_and_split.  Rework
20054         float/double/_Float128 conversions to QI/HI/SImode to work with
20055         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
20056         conversions to QI/HImode types did a store and then a load to
20057         truncate the value.  For conversions to VSX registers, don't split
20058         the insn, instead emit the code directly.  Use the code iterator
20059         any_fix to combine signed and unsigned conversions.
20060         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
20061         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
20062         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20063         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
20064         (fix_<mode>di2_hw): Likewise.
20065         (fixuns_<mode>di2_hw): Likewise.
20066         (fix_<mode>si2_hw): Likewise.
20067         (fixuns_<mode>si2_hw): Likewise.
20068         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
20069         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20070         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
20071         fix<uns>_trunc<SFDF:mode>si2_p8.
20072         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
20073         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
20074         (fix<uns>_<mode>_mem): Likewise.
20075         (fctiw<u>z_<mode>_mem): Likewise.
20076         (fix<uns>_<mode>_mem): Likewise.
20077         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
20078         the register allocator from doing a direct move to the GPRs to do
20079         a store, and instead use the ISA 3.0 store byte/half-word from
20080         vector register instruction.  For IEEE 128-bit floating point,
20081         also optimize stores of 32-bit ints.
20082         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
20084 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
20086         * genextract.c (push_pathstr_operand): New function to support
20087         [a-zA-Z].
20088         (walk_rtx): Call push_pathstr_operand.
20089         (print_path): Support [a-zA-Z].
20091 2018-02-07  Richard Biener  <rguenther@suse.de>
20093         PR tree-optimization/84037
20094         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
20095         (cse_and_gimplify_to_preheader): Declare.
20096         (vect_get_place_in_interleaving_chain): Likewise.
20097         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20098         ivexpr_map.
20099         (_loop_vec_info::~_loop_vec_info): Delete it.
20100         (cse_and_gimplify_to_preheader): New function.
20101         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
20102         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
20103         (vectorizable_load): Likewise.  For grouped stores always base
20104         the IV on the first element.
20105         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
20106         condition before gimplifying.
20108 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
20110         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
20111         *DIV_EXPR and *MOD_EXPR.
20113 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
20115         PR target/84248
20116         * config/i386/i386.c (ix86_option_override_internal): Mask out
20117         the CF_SET bit when checking -fcf-protection.
20119 2018-02-07  Tom de Vries  <tom@codesourcery.com>
20121         PR libgomp/84217
20122         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
20123         enough.
20125 2018-02-07  Richard Biener  <rguenther@suse.de>
20127         PR tree-optimization/84204
20128         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
20129         this place.
20131         PR tree-optimization/84205
20132         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
20133         special-case isl_ast_op_zdiv_r.
20135         PR tree-optimization/84223
20136         * graphite-scop-detection.c (gather_bbs::before_dom_children):
20137         Only add conditions from within the region.
20138         (gather_bbs::after_dom_children): Adjust.
20140 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
20142         PR target/84209
20143         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
20144         * config/avr/avr.md: Only post-reload split REG-REG moves if
20145         either register is GENERAL_REG_P.
20147 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
20149         PR tree-optimization/84235
20150         * tree-ssa-scopedtables.c
20151         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
20152         if the subtraction is performed in floating point type where NaNs are
20153         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
20154         build 1.  Formatting fix.
20156 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
20158         PR target/84146
20159         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
20160         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
20161         and skip it regardless of bb boundaries.  Use CALL_P macro,
20162         don't test INSN_P (insn) together with CALL_P or JUMP_P check
20163         unnecessarily, formatting fix.
20165 2018-02-06  Michael Collison  <michael.collison@arm.com>
20167         * config/arm/thumb2.md:
20168         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
20169         (*thumb_mov_notscc): Ditto.
20171 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
20173         PR target/84154
20174         * config/rs6000/rs6000.md (su code attribute): Use "u" for
20175         unsigned_fix, not "s".
20177 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20179         * configure.ac (gcc_fn_eh_frame_ro): New function.
20180         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20181         correct .eh_frame permissions.
20182         * configure: Regenerate.
20184 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
20186         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20187         irrelevant options.
20189 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20191         * config/rs6000/rs6000.c (rs6000_option_override_internal):
20192         Display warning message for -mno-speculate-indirect-jumps.
20194 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
20196         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20197         Undocumented.
20198         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20200 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
20202         PR tree-optimization/84225
20203         * tree-eh.c (find_trapping_overflow): Only call
20204         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20206 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
20208         PR target/84145
20209         * config/i386/i386.c: Reimplement the check of possible options
20210         -mibt/-mshstk conbination. Change error messages.
20211         * doc/invoke.texi: Fix a typo: remove extra '='.
20213 2018-02-06  Marek Polacek  <polacek@redhat.com>
20215         PR tree-optimization/84228
20216         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20218 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
20220         PR target/82641
20221         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20222         emitted arch directives.
20223         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20224         __ARM_FEATURE_COPROC before changing architectures.
20226 2018-02-06  Richard Biener  <rguenther@suse.de>
20228         * config/i386/i386.c (print_reg): Fix typo.
20229         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20231 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
20233         * configure: Regenerate.
20235 2018-02-05  Martin Sebor  <msebor@redhat.com>
20237         PR tree-optimization/83369
20238         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20239         inlining context.
20241 2018-02-05  Martin Liska  <mliska@suse.cz>
20243         * doc/invoke.texi: Cherry-pick upstream r323995.
20245 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
20247         * ira.c (ira_init_register_move_cost): Adjust comment.
20249 2018-02-05  Martin Liska  <mliska@suse.cz>
20251         PR gcov-profile/84137
20252         * doc/gcov.texi: Fix typo in documentation.
20254 2018-02-05  Martin Liska  <mliska@suse.cz>
20256         PR gcov-profile/83879
20257         * doc/gcov.texi: Document necessity of --dynamic-list-data when
20258         using dlopen functionality.
20260 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
20262         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20263         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20264         _mm_maskz_range_ss, _mm_mask_range_round_ss,
20265         _mm_maskz_range_round_ss): New intrinsics.
20266         (__builtin_ia32_rangesd128_round)
20267         (__builtin_ia32_rangess128_round): Remove.
20268         (__builtin_ia32_rangesd128_mask_round,
20269         __builtin_ia32_rangess128_mask_round): New builtins.
20270         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20271         __builtin_ia32_rangess128_round): Remove.
20272         (__builtin_ia32_rangesd128_mask_round,
20273         __builtin_ia32_rangess128_mask_round): New builtins.
20274         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20275         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20276         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20277         "<round_saeonly_constraint>")): Changed to ...
20278         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20279         "<round_saeonly_scalar_constraint>")): ... this.
20280         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20281         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20282         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20283         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20284         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20286 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
20288         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20289         options.
20290         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20291         Remove all values except native, 8540 and 8548.
20293 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
20295         * config/i386/i386.c (ix86_output_function_return): Pass
20296         INVALID_REGNUM, instead of -1, as invalid register number to
20297         indirect_thunk_name and output_indirect_thunk.
20299 2018-02-02  Julia Koval  <julia.koval@intel.com>
20301         * config.gcc: Add -march=icelake.
20302         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20303         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20304         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20305         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20306         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20307         (processor_target_table): Add icelake.
20308         (ix86_option_override_internal): Handle new PTAs.
20309         (get_builtin_code_for_version): Handle icelake.
20310         (M_INTEL_COREI7_ICELAKE): New.
20311         (fold_builtin_cpu): Handle icelake.
20312         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20313         * doc/invoke.texi: Add -march=icelake.
20315 2018-02-02  Julia Koval  <julia.koval@intel.com>
20317         * config/i386/i386.c (ix86_option_override_internal): Change flags type
20318         to wide_int_bitmask.
20319         * wide-int-bitmask.h: New.
20321 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
20323         PR target/84066
20324         * config/i386/i386.md: Replace Pmode with word_mode in
20325         builtin_setjmp_setup and builtin_longjmp to support x32.
20327 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
20329         PR target/56010
20330         PR target/83743
20331         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20332         #include "opts.h".
20333         (rs6000_supported_cpu_names): New static variable.
20334         (linux_cpu_translation_table): Likewise.
20335         (elf_platform) <cpu>: Define new static variable and use it.
20336         Translate kernel AT_PLATFORM name to canonical name if needed.
20337         Error if platform name is unknown.
20339 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
20341         PR target/84089
20342         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20344 2018-02-01  Jeff Law  <law@redhat.com>
20346         PR target/84128
20347         * config/i386/i386.c (release_scratch_register_on_entry): Add new
20348         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
20349         the scratch if RELEASE_VIA_POP is false.
20350         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20351         If we have to save a temporary register, decrement SIZE appropriately.
20352         Pass new arguments to release_scratch_register_on_entry.
20353         (ix86_adjust_stack_and_probe): Likewise.
20354         (ix86_emit_probe_stack_range): Pass new arguments to
20355         release_scratch_register_on_entry.
20357 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
20359         PR rtl-optimization/84157
20360         * combine.c (change_zero_ext): Use REG_P predicate in
20361         front of HARD_REGISTER_P predicate.
20363 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
20365         * config/avr/avr.c (avr_option_override): Move disabling of
20366         -fdelete-null-pointer-checks to...
20367         * common/config/avr/avr-common.c (avr_option_optimization_table):
20368         ...here.
20370 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20372         PR tree-optimization/81635
20373         * tree-data-ref.c (split_constant_offset_1): For types that
20374         wrap on overflow, try to use range info to prove that wrapping
20375         cannot occur.
20377 2018-02-01  Renlin Li  <renlin.li@arm.com>
20379         PR target/83370
20380         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20381         TAILCALL_ADDR_REGS.
20382         (aarch64_register_move_cost): Likewise.
20383         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
20384         TAILCALL_ADDR_REGS.
20385         (REG_CLASS_NAMES): Likewise.
20386         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
20387         TAILCALL_ADDR_REGS. Remove IP registers.
20388         * config/aarch64/aarch64.md (Ucs): Update register constraint.
20390 2018-02-01  Richard Biener  <rguenther@suse.de>
20392         * domwalk.h (dom_walker::dom_walker): Add additional constructor
20393         for specifying RPO order and allow NULL for that.
20394         * domwalk.c (dom_walker::dom_walker): Likewise.
20395         (dom_walker::walk): Handle NULL RPO order.
20396         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
20397         in RPO order.
20398         (rewrite_update_dom_walker): Likewise.
20399         (mark_def_dom_walker): Likewise.
20401 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20403         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
20404         (aarch64_maybe_expand_sve_subreg_move): Declare.
20405         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
20406         * config/aarch64/predicates.md (aarch64_any_register_operand): New
20407         predicate.
20408         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
20409         that are semantically a reverse operation.
20410         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
20411         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
20412         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
20413         functions.
20414         (aarch64_can_change_mode_class): For big-endian, forbid changes
20415         between two SVE modes if they have different element sizes.
20417 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20419         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
20420         the TImode handling for big-endian targets.
20422 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20424         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
20425         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
20426         not just bytes.
20427         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
20428         Remove BSWAP handing for big-endian targets and use the form of
20429         LD1RQ appropariate for the mode.
20431 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20433         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
20434         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
20435         duplicated element.
20437 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20439         PR tearget/83845
20440         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
20441         check for operands that need to go through aarch64_sve_reload_be.
20443 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
20445         PR tree-optimization/81661
20446         PR tree-optimization/84117
20447         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
20448         * tree-eh.c: Include gimplify.h.
20449         (find_trapping_overflow, replace_trapping_overflow,
20450         rewrite_to_non_trapping_overflow): New functions.
20451         * tree-vect-loop.c: Include tree-eh.h.
20452         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20453         * tree-data-ref.c: Include tree-eh.h.
20454         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20456 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
20458         PR rtl-optimization/84123
20459         * combine.c (change_zero_ext): Check if hard register satisfies
20460         can_change_dest_mode before calling gen_lowpart_SUBREG.
20462 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
20464         PR target/82444
20465         * ira.c (ira_init_register_move_cost): Remove assert.
20467 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20469         PR rtl-optimization/84071
20470         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20471         * doc/tm.texi: Regenerate.
20473 2018-01-31  Richard Biener  <rguenther@suse.de>
20475         PR tree-optimization/84132
20476         * tree-data-ref.c (analyze_miv_subscript): Properly
20477         check whether evolution_function_is_affine_multivariate_p
20478         before calling gcd_of_steps_may_divide_p.
20480 2018-01-31  Julia Koval  <julia.koval@intel.com>
20482         PR target/83618
20483         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20484         * config/i386/i386.md (rdpid_rex64) New.
20485         (rdpid): Make 32bit only.
20487 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
20489         PR lto/84105
20490         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20491         an IDENTIFIER_NODE for FUNCTION_TYPE's.
20493 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20495         Revert
20496         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
20498         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20500 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20502         PR rtl-optimization/84071
20503         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20504         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20506 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
20508         * config/arc/arc.c (arc_handle_aux_attribute): New function.
20509         (arc_attribute_table): Add 'aux' attribute.
20510         (arc_in_small_data_p): Consider aux like variables.
20511         (arc_is_aux_reg_p): New function.
20512         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20513         (arc_get_aux_arg): New function.
20514         (prepare_move_operands): Handle aux-register access.
20515         (arc_handle_aux_attribute): New function.
20516         * doc/extend.texi (ARC Variable attributes): Add subsection.
20518 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
20520         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20521         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20522         (arc_attribute_table): Add 'uncached' attribute.
20523         (arc_print_operand): Print '.di' flag for uncached memory
20524         accesses.
20525         (arc_in_small_data_p): Do not consider for small data the uncached
20526         types.
20527         (arc_is_uncached_mem_p): New function.
20528         * config/arc/predicates.md (compact_store_memory_operand): Check
20529         for uncached memory accesses.
20530         (nonvol_nonimm_operand): Likewise.
20531         * doc/extend.texi (ARC Type Attribute): New subsection.
20533 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
20535         PR c/84100
20536         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20537         falign-loops=): Add Optimization flag.
20539 2018-01-30  Jeff Law  <law@redhat.com>
20541         PR target/84064
20542         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20543         INT_REGISTERS_SAVED.  Check it prior to calling
20544         get_scratch_register_on_entry.
20545         (ix86_adjust_stack_and_probe): Similarly.
20546         (ix86_emit_probe_stack_range): Similarly.
20547         (ix86_expand_prologue): Corresponding changes.
20549 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20551         PR target/40411
20552         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20553         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20555 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
20557         PR target/84112
20558         * lra-constraints.c (curr_insn_transform): Process AND in the
20559         address.
20561 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
20563         PR rtl-optimization/83986
20564         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20565         dependence against last_pending_memory_flush in addition to
20566         pending_jump_insns.
20568 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
20570         PR tree-optimization/81611
20571         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20572         copies.
20574 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20576         PR target/83758
20577         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20578         a reg rtx.
20580 2018-01-30  Richard Biener  <rguenther@suse.de>
20581             Jakub Jelinek  <jakub@redhat.com>
20583         PR tree-optimization/84111
20584         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20585         inner loops added during recursion, as they don't have up-to-date
20586         SSA form.
20588 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
20590         PR ipa/81360
20591         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20592         (can_inline_edge_by_limits_p): ... here.
20593         (can_early_inline_edge_p, check_callers,
20594         update_caller_keys, update_callee_keys, recursive_inlining,
20595         add_new_edges_to_heap, speculation_useful_p,
20596         inline_small_functions,
20597         inline_small_functions, flatten_function,
20598         inline_to_all_callers_1): Update.
20600 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
20602         * profile-count.c (profile_count::combine_with_ipa_count): Handle
20603         zeros correctly.
20605 2018-01-30  Richard Biener  <rguenther@suse.de>
20607         PR tree-optimization/83008
20608         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20609         invariant and constant vector uses in stmts when they need
20610         more than one stmt.
20612 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20614         PR bootstrap/84017
20615         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20616         * configure: Regenerate.
20618 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20620         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20621         pattern.
20622         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20623         Use gen_rtx_REG rather than gen_lowpart.
20625 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20627         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20628         rather than 0 when creating partial subregs.
20630 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20632         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20633         of usage.
20635 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20637         PR target/81550
20638         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20639         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20640         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20641         flags.  This restores the settings used before the 2017-07-24.
20642         Turning off pre increment/decrement/modify allows IVOPTS to
20643         optimize DF/SF loops where the index is an int.
20645 2018-01-29  Richard Biener  <rguenther@suse.de>
20646             Kelvin Nilsen  <kelvin@gcc.gnu.org>
20648         PR bootstrap/80867
20649         * tree-vect-stmts.c (vectorizable_call): Don't call
20650         targetm.vectorize_builtin_md_vectorized_function if callee is
20651         NULL.
20653 2018-01-22  Carl Love  <cel@us.ibm.com>
20655         * doc/extend.tex: Fix typo in second arg in
20656         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20658 2018-01-29  Richard Biener  <rguenther@suse.de>
20660         PR tree-optimization/84086
20661         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20662         (flush_ssaname_freelist): When SSA names were released reset
20663         the SCEV hash table.
20665 2018-01-29  Richard Biener  <rguenther@suse.de>
20667         PR tree-optimization/84057
20668         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20669         removed paths when removing edges.
20671 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
20673         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20674         -mfunction-return=@var{choice}.
20676 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
20678         PR diagnostic/84034
20679         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20680         Handle CR like TAB.
20681         (layout::print_source_line): Likewise.
20682         (test_get_line_width_without_trailing_whitespace): Add test cases.
20684 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
20686         PR middle-end/84040
20687         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20688         debug insns.
20690 2018-01-26  Jim Wilson  <jimw@sifive.com>
20692         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20694         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20695         specified.
20697 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20699         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20700         and CMP + SUB-immediate -> SUBS.
20702 2018-01-26  Martin Sebor  <msebor@redhat.com>
20704         PR tree-optimization/83896
20705         * tree-ssa-strlen.c (get_string_len): Rename...
20706         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
20707         Avoid assuming length is constant.
20708         (handle_char_store): Use HOST_WIDE_INT for string length.
20710 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
20712         PR target/81763
20713         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20714         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20716 2018-01-26  Richard Biener  <rguenther@suse.de>
20718         PR rtl-optimization/84003
20719         * dse.c (record_store): Only record redundant stores when
20720         the earlier store aliases at least all accesses the later one does.
20722 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
20724         PR rtl-optimization/83985
20725         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20726         REG_CFA_RESTORE insns.
20727         (delete_unmarked_insns): Don't ignore separate shrink wrapping
20728         REG_CFA_RESTORE insns here.
20730         PR c/83989
20731         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20732         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20734 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20736         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20737         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20738         (arc_init): Likewise.
20739         (arc_override_options): Likewise.
20740         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20741         value.
20742         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20743         support.
20744         * config/arc/arc.h (TARGET_DBNZ): Define.
20745         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20746         properly set the tune attribute.
20747         (dbnz): Use TARGET_DBNZ guard.
20748         * config/arc/arc.opt (mtune): Add core3 option.
20750 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20752         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20753         recognize new pic like addresses.
20754         (arc_delegitimize_address): Clean up.
20756 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20758         * config/arc/arc-arches.def: Option mrf16 valid for all
20759         architectures.
20760         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20761         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20762         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20763         * config/arc/arc-tables.opt: Regenerate.
20764         * config/arc/arc.c (arc_conditional_register_usage): Handle
20765         reduced register file case.
20766         (arc_file_start): Set must have build attributes.
20767         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20768         mrf16 option value.
20769         * config/arc/arc.opt (mrf16): Add new option.
20770         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20771         * config/arc/genmultilib.awk: Handle new mrf16 option.
20772         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20773         * config/arc/t-multilib: Regenerate.
20774         * doc/invoke.texi (ARC Options): Document mrf16 option.
20776 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20778         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20779         * config/arc/arc.c (arc_handle_secure_attribute): New function.
20780         (arc_attribute_table): Add 'secure_call' attribute.
20781         (arc_print_operand): Print secure call operand.
20782         (arc_function_ok_for_sibcall): Don't optimize tail calls when
20783         secure.
20784         (arc_is_secure_call_p): New function.  * config/arc/arc.md
20785         (call_i): Add support for sjli instruction.
20786         (call_value_i): Likewise.
20787         * config/arc/constraints.md (Csc): New constraint.
20789 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20790             John Eric Martin  <John.Martin@emmicro-us.com>
20792         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20793         * config/arc/arc.c (_arc_jli_section): New struct.
20794         (arc_jli_section): New type.
20795         (rc_jli_sections): New static variable.
20796         (arc_handle_jli_attribute): New function.
20797         (arc_attribute_table): Add jli_always and jli_fixed attribute.
20798         (arc_file_end): New function.
20799         (TARGET_ASM_FILE_END): Define.
20800         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20801         (arc_add_jli_section): New function.
20802         (jli_call_scan): Likewise.
20803         (arc_reorg): Call jli_call_scan.
20804         (arc_output_addsi): Remove 'S' from printing asm operand.
20805         (arc_is_jli_call_p): New function.
20806         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20807         operand.
20808         (movhi_insn): Likewise.
20809         (movsi_insn): Likewise.
20810         (movsi_set_cc_insn): Likewise.
20811         (loadqi_update): Likewise.
20812         (load_zeroextendqisi_update): Likewise.
20813         (load_signextendqisi_update): Likewise.
20814         (loadhi_update): Likewise.
20815         (load_zeroextendhisi_update): Likewise.
20816         (load_signextendhisi_update): Likewise.
20817         (loadsi_update): Likewise.
20818         (loadsf_update): Likewise.
20819         (movsicc_insn): Likewise.
20820         (bset_insn): Likewise.
20821         (bxor_insn): Likewise.
20822         (bclr_insn): Likewise.
20823         (bmsk_insn): Likewise.
20824         (bicsi3_insn): Likewise.
20825         (cmpsi_cc_c_insn): Likewise.
20826         (movsi_ne): Likewise.
20827         (movsi_cond_exec): Likewise.
20828         (clrsbsi2): Likewise.
20829         (norm_f): Likewise.
20830         (normw): Likewise.
20831         (swap): Likewise.
20832         (divaw): Likewise.
20833         (flag): Likewise.
20834         (sr): Likewise.
20835         (kflag): Likewise.
20836         (ffs): Likewise.
20837         (ffs_f): Likewise.
20838         (fls): Likewise.
20839         (call_i): Remove 'S' asm letter, add jli instruction.
20840         (call_value_i): Likewise.
20841         * config/arc/arc.op (mjli-always): New option.
20842         * config/arc/constraints.md (Cji): New constraint.
20843         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20844         operand.
20845         (subsf3_fpx): Likewise.
20846         (mulsf3_fpx): Likewise.
20847         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20848         asm operand.
20849         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20850         function attrbutes.
20851         * doc/invoke.texi (ARC): Document mjli-always option.
20853 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
20855         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20856         avoid addition with 0 and use incw and decw where possible.
20858 2018-01-26  Richard Biener  <rguenther@suse.de>
20860         PR tree-optimization/81082
20861         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20862         association if it requires casting to unsigned.
20863         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20864         from fold_plusminus_mult_expr to catch important cases late when
20865         range info is available.
20867 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20869         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20870         * configure.ac (hidden_linkonce): New test.
20871         * configure: Regenerate.
20872         * config.in: Regenerate.
20874 2018-01-26  Julia Koval  <julia.koval@intel.com>
20876         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20877         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20878         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20879         _mm_mask_bitshuffle_epi64_mask): Fix type.
20880         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20881         USI_FTYPE_V4DI_V4DI_USI): Remove.
20882         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20883         __builtin_ia32_vpshufbitqmb256_mask,
20884         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20885         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20886         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20888 2018-01-26  Alan Modra  <amodra@gmail.com>
20890         PR target/84033
20891         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20892         UNSPEC_VBPERMQ.  Sort other unspecs.
20894 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
20896         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20898 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
20900         PR middle-end/83055
20901         * predict.c (drop_profile): Do not push/pop cfun; update also
20902         node->count.
20903         (handle_missing_profiles): Fix logic looking for zero profiles.
20905 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
20907         PR middle-end/83977
20908         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20909         on functions with #pragma omp declare simd or functions with simd
20910         attribute.
20911         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20912         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20913         Remove trailing \n from warning_at calls.
20915 2018-01-25  Tom de Vries  <tom@codesourcery.com>
20917         PR target/84028
20918         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20919         for neutered workers.
20921 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
20923         PR target/68467
20924         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20925         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20927 2018-01-24  Jeff Law  <law@redhat.com>
20929         PR target/83994
20930         * i386.c (get_probe_interval): Move to earlier point.
20931         (ix86_compute_frame_layout): If -fstack-clash-protection and
20932         the frame is larger than the probe interval, then use pushes
20933         to save registers rather than reg->mem moves.
20934         (ix86_expand_prologue): Remove conditional for int_registers_saved
20935         assertion.
20937 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
20939         PR target/84014
20940         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20941         min/max for never referenced object.
20943 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
20945         PR middle-end/83977
20946         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20947         here.
20948         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20949         attributes from DECL_ATTRIBUTES (decl) without affecting
20950         DECL_ATTRIBUTES (current_function_decl).
20951         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20952         functions with non-NULL DECL_ABSTRACT_ORIGIN.
20954 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
20956         PR tree-optimization/83979
20957         * fold-const.c (fold_comparison): Use constant_boolean_node
20958         instead of boolean_{true,false}_node.
20960 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
20962         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20963         with zero counts.
20965 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20967         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20968         Simplify the clause that sets the length attribute.
20969         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20970         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20971         clause that sets the length attribute.
20972         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20974 2018-01-24  Tom de Vries  <tom@codesourcery.com>
20976         PR target/83589
20977         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20978         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20979         Add strict parameter.
20980         (prevent_branch_around_nothing): Insert dummy insn between branch to
20981         label and label with no ptx insn inbetween.
20982         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20984 2018-01-24  Tom de Vries  <tom@codesourcery.com>
20986         PR target/81352
20987         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20988         for neutered threads in warp.
20989         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20991 2018-01-24  Richard Biener  <rguenther@suse.de>
20993         PR tree-optimization/83176
20994         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20995         operands.
20997 2018-01-24  Richard Biener  <rguenther@suse.de>
20999         PR tree-optimization/82819
21000         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
21001         code generating pluses that are no-ops in the target precision.
21003 2018-01-24  Richard Biener  <rguenther@suse.de>
21005         PR middle-end/84000
21006         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
21008 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
21010         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
21011         to merge probabilities.
21012         * predict.c (probably_never_executed): Also mark as cold functions
21013         with global 0 profile and guessed local profile.
21014         * profile-count.c (profile_probability::combine_with_count): New
21015         member function.
21016         * profile-count.h (profile_probability::operator*,
21017         profile_probability::operator*=, profile_probability::operator/,
21018         profile_probability::operator/=): Reduce precision to adjusted
21019         and set value to guessed on contradictory divisions.
21020         (profile_probability::combine_with_freq): Remove.
21021         (profile_probability::combine_wiht_count): Declare.
21022         (profile_count::force_nonzero):: Set to adjusted.
21023         (profile_count::probability_in):: Set quality to adjusted.
21024         * tree-ssa-tail-merge.c (replace_block_by): Use
21025         combine_with_count.
21027 2018-01-23  Andrew Waterman  <andrew@sifive.com>
21028             Jim Wilson  <jimw@sifive.com>
21030         * config/riscv/riscv.c (riscv_stack_boundary): New.
21031         (riscv_option_override): Set riscv_stack_boundary.  Handle
21032         riscv_preferred_stack_boundary_arg.
21033         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
21034         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
21035         (STACK_BOUNDARY): Set to riscv_stack_boundary.
21036         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
21037         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
21038         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
21040 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
21042         PR target/83905
21043         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
21044         of struct ix86_frame.
21045         (ix86_expand_epilogue): Likewise.  Add a local variable for
21046         the reg_save_offset field in struct ix86_frame.
21048 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
21050         PR tree-optimization/82604
21051         * tree-loop-distribution.c (enum partition_kind): New enum item
21052         PKIND_PARTIAL_MEMSET.
21053         (partition_builtin_p): Support above new enum item.
21054         (generate_code_for_partition): Ditto.
21055         (compute_access_range): Differentiate cases that equality can be
21056         proven at all loops, the innermost loops or no loops.
21057         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
21058         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
21059         (finalize_partitions, distribute_loop): Don't fuse partition of
21060         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
21061         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
21062         parloop is enabled.
21064 2018-01-23  Martin Liska  <mliska@suse.cz>
21066         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
21067         order to ignore the predictor.
21068         (PRED_POLYMORPHIC_CALL): Likewise.
21069         (PRED_RECURSIVE_CALL): Likewise.
21071 2018-01-23  Martin Liska  <mliska@suse.cz>
21073         * tree-profile.c (tree_profiling): Print function header to
21074         aware reader which function we are working on.
21075         * value-prof.c (gimple_find_values_to_profile): Do not print
21076         not interesting value histograms.
21078 2018-01-23  Martin Liska  <mliska@suse.cz>
21080         * profile-count.h (enum profile_quality): Add
21081         profile_uninitialized as the first value. Do not number values
21082         as they are zero based.
21083         (profile_count::verify): Update sanity check.
21084         (profile_probability::verify): Likewise.
21086 2018-01-23  Nathan Sidwell  <nathan@acm.org>
21088         * doc/invoke.texi (ffor-scope): Deprecate.
21090 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
21092         PR tree-optimization/83510
21093         * domwalk.c (set_all_edges_as_executable): New function.
21094         (dom_walker::dom_walker): Convert bool param
21095         "skip_unreachable_blocks" to enum reachability.  Move setup of
21096         edge flags to set_all_edges_as_executable and only do it when
21097         reachability is REACHABLE_BLOCKS.
21098         * domwalk.h (enum dom_walker::reachability): New enum.
21099         (dom_walker::dom_walker): Convert bool param
21100         "skip_unreachable_blocks" to enum reachability.
21101         (set_all_edges_as_executable): New decl.
21102         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
21103         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
21104         "reachability".
21105         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
21106         but converting true to REACHABLE_BLOCKS.
21107         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
21108         * tree-vrp.c
21109         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
21110         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
21111         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
21112         REACHABLE_BLOCKS.
21113         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
21114         if check_all_array_refs will be called.
21116 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
21118         * tree.c (selftest::test_location_wrappers): Add more test
21119         coverage.
21121 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
21123         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
21124         (selftest::test_bit_in_range): Likewise.
21126 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
21128         PR testsuite/83888
21129         * doc/sourcebuild.texi (vect_float): Say that the selector
21130         only describes the situation when -funsafe-math-optimizations is on.
21131         (vect_float_strict): Document.
21133 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
21135         PR tree-optimization/83965
21136         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
21137         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
21138         instead of checking only for a reduction.
21139         (vect_recog_widen_sum_pattern): Likewise.
21141 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
21143         * predict.c (probably_never_executed): Only use precise profile info.
21144         (compute_function_frequency): Skip after inlining hack since we now
21145         have quality checking.
21147 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
21149         * profile-count.h (profile_probability::very_unlikely,
21150         profile_probability::unlikely, profile_probability::even): Set
21151         precision to guessed.
21153 2018-01-23  Richard Biener  <rguenther@suse.de>
21155         PR tree-optimization/83963
21156         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
21157         Properly terminate dominator walk when crossing the exit edge not
21158         when visiting its source block.
21160 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
21162         PR c++/83918
21163         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
21164         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
21166 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
21168         PR tree-optimization/83957
21169         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
21170         semicolon after for body surrounded by braces.
21172         PR tree-optimization/83081
21173         * profile-count.h (profile_probability::split): New method.
21174         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21175         Use profile_probability::split.
21176         (do_compare_rtx_and_jump): Fix adjustment of probabilities
21177         when splitting a single conditional jump into 2.
21179 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
21181         PR tree-optimization/69452
21182         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21183         decl.
21185 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21187         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21188         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21189         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21191 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21193         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21194         declaration.
21195         * config/rl78/rl78.md (movdi): New define_expand.
21196         * config/rl78/rl78.c (rl78_split_movdi): New function.
21198 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
21200         PR target/83862
21201         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21202         no longer used.
21203         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21204         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21205         128-bit to produce an UNSPEC move to get the double word with the
21206         signbit and then a shift directly to do signbit.
21207         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21208         implementation with a new version that just does either a direct
21209         move or a regular move.  Move memory interface to separate insns.
21210         Move insns so they are next to the expander.
21211         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21212         with signbit move.  Split big and little endian case.
21213         (signbit<mode>2_dm_mem_le): Likewise.
21214         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21215         (signbit<mode>2_dm2): Likewise.
21217 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21219         * config/rl78/rl78.md (anddi3): New define_expand.
21221 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21223         * config/rl78/rl78.md (umindi3): New define_expand.
21225 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21227         * config/rl78/rl78.md (smindi3): New define_expand.
21229 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21231         * config/rl78/rl78.md (smaxdi3): New define_expand.
21233 2018-01-22  Carl Love  <cel@us.ibm.com>
21235         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21236         LVX_V1TI): Add macro expansion.
21237         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21238         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21239         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21240         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21241         Change check to determine if the instruction is a byte reversing
21242         entry.  Fix typo in comment.
21243         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21244         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21245         Add def_builtin calls for new builtins.
21246         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21247         Add define_insn expansion.
21249 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21251         * config/rl78/rl78.md (umaxdi3): New define_expand.
21253 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21255         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21256         for non-QImode registers.
21258 2018-01-22  Richard Biener  <rguenther@suse.de>
21260         PR tree-optimization/83963
21261         * graphite-scop-detection.c (scop_detection::get_sese): Delay
21262         including the loop exit block.
21263         (scop_detection::merge_sese): Likewise.
21264         (scop_detection::add_scop): Do it here instead.
21266 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21268         * doc/sourcebuild.texi (arm_softfloat): Document.
21270 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
21272         PR gcc/77734
21273         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21274         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21275         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21277 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21278             David Edelsohn  <dje.gcc@gmail.com>
21280         PR target/83946
21281         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21282         Change "crset eq" to "crset 2".
21283         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21284         (*call_indirect_aix<mode>_nospec): Likewise.
21285         (*call_value_indirect_aix<mode>_nospec): Likewise.
21286         (*call_indirect_elfv2<mode>_nospec): Likewise.
21287         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21288         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21289         change assembly output from . to $.
21290         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21291         (indirect_jump<mode>_nospec): Change assembly output from . to $.
21292         (*tablejump<mode>_internal1_nospec): Likewise.
21294 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
21296         PR target/80870
21297         * config/sh/sh_optimize_sett_clrt.cc:
21298         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21300 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
21302         PR tree-optimization/83940
21303         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21304         offset_dt to vect_constant_def rather than vect_unknown_def_type.
21305         (vect_check_load_store_mask): Add a mask_dt_out parameter and
21306         use it to pass back the definition type.
21307         (vect_check_store_rhs): Likewise rhs_dt_out.
21308         (vect_build_gather_load_calls): Add a mask_dt argument and use
21309         it instead of a call to vect_is_simple_use.
21310         (vectorizable_store): Update calls to vect_check_load_store_mask
21311         and vect_check_store_rhs.  Use the dt returned by the latter instead
21312         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
21313         instead of calls to vect_is_simple_use.  Pass the scalar rather
21314         than the vector operand to vect_is_simple_use when handling
21315         second and subsequent copies of an rhs value.
21316         (vectorizable_load): Update calls to vect_check_load_store_mask
21317         and vect_build_gather_load_calls.  Use the cached mask_dt and
21318         gs_info.offset_dt instead of calls to vect_is_simple_use.
21320 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
21322         PR middle-end/83945
21323         * tree-emutls.c: Include gimplify.h.
21324         (lower_emutls_2): New function.
21325         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21326         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21327         it before further processing.
21329         PR target/83930
21330         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21331         UINTVAL (trueop1) instead of INTVAL (op1).
21333 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
21335         PR debug/81570
21336         PR debug/83728
21337         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21338         INCOMING_FRAME_SP_OFFSET if not defined.
21339         (scan_trace): Add ENTRY argument.  If true and
21340         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21341         emit a note to adjust the CFA offset.
21342         (create_cfi_notes): Adjust scan_trace callers.
21343         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21344         INCOMING_FRAME_SP_OFFSET in the CIE.
21345         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21346         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21347         Likewise.
21348         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21349         * doc/tm.texi: Regenerated.
21351 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21353         PR rtl-optimization/83147
21354         * lra-constraints.c (remove_inheritance_pseudos): Use
21355         lra_substitute_pseudo_within_insn.
21357 2018-01-19  Tom de Vries  <tom@codesourcery.com>
21358             Cesar Philippidis  <cesar@codesourcery.com>
21360         PR target/83920
21361         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21363 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
21365         PR target/83790
21366         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21367         spaces for function labels.
21369 2018-01-19  Martin Liska  <mliska@suse.cz>
21371         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21372         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21373         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21374         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21375         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21376         (PRED_CONST_RETURN): Change from 69 to 65.
21377         (PRED_NULL_RETURN): Change from 91 to 71.
21378         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21379         (PRED_LOOP_GUARD): Change from 66 to 73.
21381 2018-01-19  Martin Liska  <mliska@suse.cz>
21383         * predict.c (predict_insn_def): Add new assert.
21384         (struct branch_predictor): Change type to signed integer.
21385         (test_prediction_value_range): Amend test to cover
21386         PROB_UNINITIALIZED.
21387         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
21388         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
21389         (PRED_LOOP_ITERATIONS_MAX): Likewise.
21390         (PRED_LOOP_IV_COMPARE): Likewise.
21391         * predict.h (PROB_UNINITIALIZED): Define new constant.
21393 2018-01-19  Martin Liska  <mliska@suse.cz>
21395         * predict.c (dump_prediction): Add new format for
21396         analyze_brprob.py script which is enabled with -details
21397         suboption.
21398         * profile-count.h (precise_p): New function.
21400 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
21402         PR tree-optimization/83922
21403         * tree-vect-loop.c (vect_verify_full_masking): Return false if
21404         there are no statements that need masking.
21405         (vect_active_double_reduction_p): New function.
21406         (vect_analyze_loop_operations): Use it when handling phis that
21407         are not in the loop header.
21409 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
21411         PR tree-optimization/83914
21412         * tree-vect-loop.c (vectorizable_induction): Don't convert
21413         init_expr or apply the peeling adjustment for inductions
21414         that are nested within the vectorized loop.
21416 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21418         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
21419         instead of NEG.
21421 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
21423         PR sanitizer/81715
21424         PR testsuite/83882
21425         * function.h (gimplify_parameters): Add gimple_seq * argument.
21426         * function.c: Include gimple.h and options.h.
21427         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
21428         for the added local temporaries if needed.
21429         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
21430         if there are any parameter cleanups, wrap whole body into a
21431         try/finally with the cleanups.
21433 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
21435         PR target/82964
21436         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
21437         Use GET_MODE_CLASS for scalar floating point.
21439 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
21441         PR ipa/82256
21442         patch by PaX Team
21443         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
21444         Fix call of call_cgraph_insertion_hooks.
21446 2018-01-18  Martin Sebor  <msebor@redhat.com>
21448         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21450 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
21452         PR ipa/83619
21453         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21454         frequencies.
21456 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
21458         PR other/70268
21459         * common.opt: (-ffile-prefix-map): New option.
21460         * opts.c (common_handle_option): Defer it.
21461         * opts-global.c (handle_common_deferred_options): Handle it.
21462         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21463         * file-prefix-map.h: New file.
21464         (remap_debug_filename, add_debug_prefix_map): ...here.
21465         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21466         * final.c (debug_prefix_map, add_debug_prefix_map
21467         remap_debug_filename): Move to...
21468         * file-prefix-map.c: New file.
21469         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21470         generalize, get rid of alloca(), use strrchr() instead of strchr().
21471         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21472         Implement in terms of add_prefix_map().
21473         (remap_macro_filename, remap_debug_filename): Implement in term of
21474         remap_filename().
21475         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21476         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21477         * dbxout.c: Include file-prefix-map.h.
21478         * varasm.c: Likewise.
21479         * vmsdbgout.c: Likewise.
21480         * xcoffout.c: Likewise.
21481         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21482         * doc/cppopts.texi (-fmacro-prefix-map): Document.
21483         * doc/invoke.texi (-ffile-prefix-map): Document.
21484         (-fdebug-prefix-map): Update description.
21486 2018-01-18  Martin Liska  <mliska@suse.cz>
21488         * config/i386/i386.c (indirect_thunk_name): Document that also
21489         lfence is emitted.
21490         (output_indirect_thunk): Document why both instructions
21491         (pause and lfence) are generated.
21493 2018-01-18  Richard Biener  <rguenther@suse.de>
21495         PR tree-optimization/83887
21496         * graphite-scop-detection.c
21497         (scop_detection::get_nearest_dom_with_single_entry): Remove.
21498         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21499         (scop_detection::merge_sese): Re-implement with a flood-fill
21500         algorithm that properly finds a SESE region if it exists.
21502 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
21504         PR c/61240
21505         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21506         pointer_diff optimizations use view_convert instead of convert.
21508 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21510         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21511         Generate different code for -mno-speculate-indirect-jumps.
21512         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21513         (*call_indirect_aix<mode>): Disable for
21514         -mno-speculate-indirect-jumps.
21515         (*call_indirect_aix<mode>_nospec): New define_insn.
21516         (*call_value_indirect_aix<mode>): Disable for
21517         -mno-speculate-indirect-jumps.
21518         (*call_value_indirect_aix<mode>_nospec): New define_insn.
21519         (*sibcall_nonlocal_sysv<mode>): Generate different code for
21520         -mno-speculate-indirect-jumps.
21521         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21523 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21525         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21526         long double type, set the flags for noting the default long double
21527         type, even if we don't pass or return a long double type.
21529 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
21531         PR ipa/83051
21532         * ipa-inline.c (flatten_function): Do not overwrite final inlining
21533         failure.
21535 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
21537         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21538         support for merge[hl].
21539         (fold_mergehl_helper): New helper function.
21540         (tree-vector-builder.h): New #include for tree_vector_builder usage.
21541         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21542         (altivec_vmrglw_direct): Add xxmrglw insn.
21544 2018-01-17  Andrew Waterman  <andrew@sifive.com>
21546         * config/riscv/riscv.c (riscv_conditional_register_usage): If
21547         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21549 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
21551         PR lto/83121
21552         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21553         call the lto_location_cache before reading the
21554         DECL_SOURCE_LOCATION of the types.
21556 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
21557             Richard Sandiford  <richard.sandiford@linaro.org>
21559         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21560         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21561         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21562         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21563         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21564         Add declaration.
21565         * config/aarch64/constraints.md (aarch64_movti_operand):
21566         Limit immediates.
21567         * config/aarch64/predicates.md (Uti): Add new constraint.
21569 2018-01-17  Carl Love  <cel@us.ibm.com>
21571         * config/rs6000/vsx.md (define_expand xl_len_r,
21572         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21573         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21574         lxvll.
21575         (define_expand, define_insn): Move the shift left from  the
21576         define_insn to the define_expand for lxvl and stxvl instructions.
21577         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21578         and XL_LEN_R definitions to PURE.
21580 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
21582         * config/i386/i386.c (indirect_thunk_name): Declare regno
21583         as unsigned int.  Compare regno with INVALID_REGNUM.
21584         (output_indirect_thunk): Ditto.
21585         (output_indirect_thunk_function): Ditto.
21586         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
21587         in the call to output_indirect_thunk_function.
21589 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
21591         PR middle-end/83884
21592         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21593         rather than the size of inner_type to determine the stack slot size
21594         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21596 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
21598         PR target/83546
21599         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21600         to PTA_SILVERMONT.
21602 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
21604         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21605         endian Linux systems to optionally enable multilibs for selecting
21606         the long double type if the user configured an explicit type.
21607         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21608         have no long double multilibs if not defined.
21609         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21610         warn if the user used -mabi={ieee,ibm}longdouble and we built
21611         multilibs for long double.
21612         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21613         appropriate multilib option.
21614         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21615         multilib options.
21616         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21617         for building long double multilibs.
21618         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21620 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
21622         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21623         copies.
21625         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21626         64 bits.
21627         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21628         128 bits.
21630         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21631         variables.
21633         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21634         return value.
21636 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
21638         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21639         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21641 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21643         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21644         different rtl trees depending on TARGET_64BIT.
21645         (rs6000_gen_lvx): Likewise.
21647 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
21649         * config/visium/visium.md (nop): Tweak comment.
21650         (hazard_nop): Likewise.
21652 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21654         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21655         -mspeculate-indirect-jumps.
21656         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21657         for -mno-speculate-indirect-jumps.
21658         (*call_indirect_elfv2<mode>_nospec): New define_insn.
21659         (*call_value_indirect_elfv2<mode>): Disable for
21660         -mno-speculate-indirect-jumps.
21661         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21662         (indirect_jump): Emit different RTL for
21663         -mno-speculate-indirect-jumps.
21664         (*indirect_jump<mode>): Disable for
21665         -mno-speculate-indirect-jumps.
21666         (*indirect_jump<mode>_nospec): New define_insn.
21667         (tablejump): Emit different RTL for
21668         -mno-speculate-indirect-jumps.
21669         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21670         (tablejumpsi_nospec): New define_expand.
21671         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21672         (tablejumpdi_nospec): New define_expand.
21673         (*tablejump<mode>_internal1): Disable for
21674         -mno-speculate-indirect-jumps.
21675         (*tablejump<mode>_internal1_nospec): New define_insn.
21676         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21677         option.
21679 2018-01-16  Artyom Skrobov tyomitch@gmail.com
21681         * caller-save.c (insert_save): Drop unnecessary parameter.  All
21682         callers updated.
21684 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21685             Richard Biener  <rguenth@suse.de>
21687         PR libgomp/83590
21688         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21689         return early, inline manually is_gimple_sizepos.  Make sure if we
21690         call gimplify_expr we don't end up with a gimple constant.
21691         * tree.c (variably_modified_type_p): Don't return true for
21692         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
21693         * gimplify.h (is_gimple_sizepos): Remove.
21695 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21697         PR tree-optimization/83857
21698         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21699         vectorizable_live_operation for pure SLP statements.
21700         (vectorizable_live_operation): Handle PHIs.
21702 2018-01-16  Richard Biener  <rguenther@suse.de>
21704         PR tree-optimization/83867
21705         * tree-vect-stmts.c (vect_transform_stmt): Precompute
21706         nested_in_vect_loop_p since the scalar stmt may get invalidated.
21708 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21710         PR c/83844
21711         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21712         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21713         If off is not INTEGER_CST, issue a may not be aligned warning
21714         rather than isn't aligned.  Use isn%'t rather than isn't.
21715         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21716         into MULT_EXPR.
21717         <case MULT_EXPR>: Improve the case when bottom and one of the
21718         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21719         operand, in that case check if the other operand is multiple of
21720         bottom divided by the INTEGER_CST operand.
21722 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21724         PR target/83858
21725         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21726         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21727         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21728         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21729         * config/pa/pa.c (pa_function_arg_advance): Likewise.
21730         (pa_function_arg, pa_arg_partial_bytes): Likewise.
21731         (pa_function_arg_size): New function.
21733 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21735         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21736         in a separate statement.
21738 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21740         PR tree-optimization/83847
21741         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21742         group gathers and scatters.
21744 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21746         PR rtl-optimization/86620
21747         * params.def (max-sched-ready-insns): Bump minimum value to 1.
21749         PR rtl-optimization/83213
21750         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21751         to last if both are JUMP_INSNs.
21753         PR tree-optimization/83843
21754         * gimple-ssa-store-merging.c
21755         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21756         store_immediate_info for bswap/nop orig_stores.
21758 2018-01-15  Andrew Waterman  <andrew@sifive.com>
21760         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21761         !TARGET_MUL.
21762         <UDIV>: Increase cost if !TARGET_DIV.
21764 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
21766         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21767         (define_attr "cr_logical_3op"): New.
21768         (cceq_ior_compare): Adjust.
21769         (cceq_ior_compare_complement): Adjust.
21770         (*cceq_rev_compare): Adjust.
21771         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21772         (is_cracked_insn): Adjust.
21773         (insn_must_be_first_in_group): Adjust.
21774         * config/rs6000/40x.md: Adjust.
21775         * config/rs6000/440.md: Adjust.
21776         * config/rs6000/476.md: Adjust.
21777         * config/rs6000/601.md: Adjust.
21778         * config/rs6000/603.md: Adjust.
21779         * config/rs6000/6xx.md: Adjust.
21780         * config/rs6000/7450.md: Adjust.
21781         * config/rs6000/7xx.md: Adjust.
21782         * config/rs6000/8540.md: Adjust.
21783         * config/rs6000/cell.md: Adjust.
21784         * config/rs6000/e300c2c3.md: Adjust.
21785         * config/rs6000/e500mc.md: Adjust.
21786         * config/rs6000/e500mc64.md: Adjust.
21787         * config/rs6000/e5500.md: Adjust.
21788         * config/rs6000/e6500.md: Adjust.
21789         * config/rs6000/mpc.md: Adjust.
21790         * config/rs6000/power4.md: Adjust.
21791         * config/rs6000/power5.md: Adjust.
21792         * config/rs6000/power6.md: Adjust.
21793         * config/rs6000/power7.md: Adjust.
21794         * config/rs6000/power8.md: Adjust.
21795         * config/rs6000/power9.md: Adjust.
21796         * config/rs6000/rs64.md: Adjust.
21797         * config/rs6000/titan.md: Adjust.
21799 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21801         * config/i386/predicates.md (indirect_branch_operand): Rewrite
21802         ix86_indirect_branch_register logic.
21804 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21806         * config/i386/constraints.md (Bs): Update
21807         ix86_indirect_branch_register check.  Don't check
21808         ix86_indirect_branch_register with GOT_memory_operand.
21809         (Bw): Likewise.
21810         * config/i386/predicates.md (GOT_memory_operand): Don't check
21811         ix86_indirect_branch_register here.
21812         (GOT32_symbol_operand): Likewise.
21814 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21816         * config/i386/predicates.md (constant_call_address_operand):
21817         Rewrite ix86_indirect_branch_register logic.
21818         (sibcall_insn_operand): Likewise.
21820 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21822         * config/i386/constraints.md (Bs): Replace
21823         ix86_indirect_branch_thunk_register with
21824         ix86_indirect_branch_register.
21825         (Bw): Likewise.
21826         * config/i386/i386.md (indirect_jump): Likewise.
21827         (tablejump): Likewise.
21828         (*sibcall_memory): Likewise.
21829         (*sibcall_value_memory): Likewise.
21830         Peepholes of indirect call and jump via memory: Likewise.
21831         * config/i386/i386.opt: Likewise.
21832         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21833         (GOT_memory_operand): Likewise.
21834         (call_insn_operand): Likewise.
21835         (sibcall_insn_operand): Likewise.
21836         (GOT32_symbol_operand): Likewise.
21838 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
21840         PR middle-end/83837
21841         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21842         type rather than type addr's type points to.
21843         (expand_omp_atomic_mutex): Likewise.
21844         (expand_omp_atomic): Likewise.
21846 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21848         PR target/83839
21849         * config/i386/i386.c (output_indirect_thunk_function): Use
21850         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21851         for  __x86_return_thunk.
21853 2018-01-15  Richard Biener  <rguenther@suse.de>
21855         PR middle-end/83850
21856         * expmed.c (extract_bit_field_1): Fix typo.
21858 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21860         PR target/83687
21861         * config/arm/iterators.md (VF): New mode iterator.
21862         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21863         Remove integer-related logic from pattern.
21864         (neon_vabd<mode>_3): Likewise.
21866 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
21868         PR middle-end/82694
21869         * common.opt (fstrict-overflow): No longer an alias.
21870         (fwrapv-pointer): New option.
21871         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21872         also for pointer types based on flag_wrapv_pointer.
21873         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21874         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21875         opts->x_flag_wrapv got set.
21876         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21877         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21878         POINTER_TYPE_OVERFLOW_UNDEFINED.
21879         * match.pd: Likewise in address comparison pattern.
21880         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21882 2018-01-15  Richard Biener  <rguenther@suse.de>
21884         PR lto/83804
21885         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21886         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
21887         Reset type names to their identifier if their TYPE_DECL doesn't
21888         have linkage (and thus is used for ODR and devirt).
21889         (save_debug_info_for_decl): Remove.
21890         (save_debug_info_for_type): Likewise.
21891         (add_tree_to_fld_list): Adjust.
21892         * tree-pretty-print.c (dump_generic_node): Make dumping of
21893         type names more robust.
21895 2018-01-15  Richard Biener  <rguenther@suse.de>
21897         * BASE-VER: Bump to 8.0.1.
21899 2018-01-14  Martin Sebor  <msebor@redhat.com>
21901         PR other/83508
21902         * builtins.c (check_access): Avoid warning when the no-warning bit
21903         is set.
21905 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
21907         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21908         * ira-color (allocno_hard_regs_compare): Likewise.
21910 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
21912         PR target/83013
21913         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21914         Use .pushsection/.popsection.
21916 2018-01-14  Martin Sebor  <msebor@redhat.com>
21918         PR c++/81327
21919         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21921 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
21923         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21924         entry from extra_headers.
21925         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21926         extra_headers, make the list bitwise identical to the i?86-*-* one.
21928 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21930         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21931         -mcmodel=large with -mindirect-branch=thunk,
21932         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21933         -mfunction-return=thunk-extern.
21934         * doc/invoke.texi: Document -mcmodel=large is incompatible with
21935         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21936         -mfunction-return=thunk and -mfunction-return=thunk-extern.
21938 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21940         * config/i386/i386.c (print_reg): Print the name of the full
21941         integer register without '%'.
21942         (ix86_print_operand): Handle 'V'.
21943         * doc/extend.texi: Document 'V' modifier.
21945 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21947         * config/i386/constraints.md (Bs): Disallow memory operand for
21948         -mindirect-branch-register.
21949         (Bw): Likewise.
21950         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21951         (GOT_memory_operand): Likewise.
21952         (call_insn_operand): Likewise.
21953         (sibcall_insn_operand): Likewise.
21954         (GOT32_symbol_operand): Likewise.
21955         * config/i386/i386.md (indirect_jump): Call convert_memory_address
21956         for -mindirect-branch-register.
21957         (tablejump): Likewise.
21958         (*sibcall_memory): Likewise.
21959         (*sibcall_value_memory): Likewise.
21960         Disallow peepholes of indirect call and jump via memory for
21961         -mindirect-branch-register.
21962         (*call_pop): Replace m with Bw.
21963         (*call_value_pop): Likewise.
21964         (*sibcall_pop_memory): Replace m with Bs.
21965         * config/i386/i386.opt (mindirect-branch-register): New option.
21966         * doc/invoke.texi: Document -mindirect-branch-register option.
21968 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21970         * config/i386/i386-protos.h (ix86_output_function_return): New.
21971         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21972         set function_return_type.
21973         (indirect_thunk_name): Add ret_p to indicate thunk for function
21974         return.
21975         (output_indirect_thunk_function): Pass false to
21976         indirect_thunk_name.
21977         (ix86_output_indirect_branch_via_reg): Likewise.
21978         (ix86_output_indirect_branch_via_push): Likewise.
21979         (output_indirect_thunk_function): Create alias for function
21980         return thunk if regno < 0.
21981         (ix86_output_function_return): New function.
21982         (ix86_handle_fndecl_attribute): Handle function_return.
21983         (ix86_attribute_table): Add function_return.
21984         * config/i386/i386.h (machine_function): Add
21985         function_return_type.
21986         * config/i386/i386.md (simple_return_internal): Use
21987         ix86_output_function_return.
21988         (simple_return_internal_long): Likewise.
21989         * config/i386/i386.opt (mfunction-return=): New option.
21990         (indirect_branch): Mention -mfunction-return=.
21991         * doc/extend.texi: Document function_return function attribute.
21992         * doc/invoke.texi: Document -mfunction-return= option.
21994 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21996         * config/i386/i386-opts.h (indirect_branch): New.
21997         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21998         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21999         with local indirect jump when converting indirect call and jump.
22000         (ix86_set_indirect_branch_type): New.
22001         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
22002         (indirectlabelno): New.
22003         (indirect_thunk_needed): Likewise.
22004         (indirect_thunk_bnd_needed): Likewise.
22005         (indirect_thunks_used): Likewise.
22006         (indirect_thunks_bnd_used): Likewise.
22007         (INDIRECT_LABEL): Likewise.
22008         (indirect_thunk_name): Likewise.
22009         (output_indirect_thunk): Likewise.
22010         (output_indirect_thunk_function): Likewise.
22011         (ix86_output_indirect_branch_via_reg): Likewise.
22012         (ix86_output_indirect_branch_via_push): Likewise.
22013         (ix86_output_indirect_branch): Likewise.
22014         (ix86_output_indirect_jmp): Likewise.
22015         (ix86_code_end): Call output_indirect_thunk_function if needed.
22016         (ix86_output_call_insn): Call ix86_output_indirect_branch if
22017         needed.
22018         (ix86_handle_fndecl_attribute): Handle indirect_branch.
22019         (ix86_attribute_table): Add indirect_branch.
22020         * config/i386/i386.h (machine_function): Add indirect_branch_type
22021         and has_local_indirect_jump.
22022         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
22023         to true.
22024         (tablejump): Likewise.
22025         (*indirect_jump): Use ix86_output_indirect_jmp.
22026         (*tablejump_1): Likewise.
22027         (simple_return_indirect_internal): Likewise.
22028         * config/i386/i386.opt (mindirect-branch=): New option.
22029         (indirect_branch): New.
22030         (keep): Likewise.
22031         (thunk): Likewise.
22032         (thunk-inline): Likewise.
22033         (thunk-extern): Likewise.
22034         * doc/extend.texi: Document indirect_branch function attribute.
22035         * doc/invoke.texi: Document -mindirect-branch= option.
22037 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
22039         PR ipa/83051
22040         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
22042 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
22044         * ipa-inline.c (want_inline_small_function_p): Return false if
22045         inlining has already failed with CIF_FINAL_ERROR.
22046         (update_caller_keys): Call want_inline_small_function_p before
22047         can_inline_edge_p.
22048         (update_callee_keys): Likewise.
22050 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22052         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
22053         New function.
22054         (rs6000_quadword_masked_address_p): Likewise.
22055         (quad_aligned_load_p): Likewise.
22056         (quad_aligned_store_p): Likewise.
22057         (const_load_sequence_p): Add comment to describe the outer-most loop.
22058         (mimic_memory_attributes_and_flags): New function.
22059         (rs6000_gen_stvx): Likewise.
22060         (replace_swapped_aligned_store): Likewise.
22061         (rs6000_gen_lvx): Likewise.
22062         (replace_swapped_aligned_load): Likewise.
22063         (replace_swapped_load_constant): Capitalize argument name in
22064         comment describing this function.
22065         (rs6000_analyze_swaps): Add a third pass to search for vector loads
22066         and stores that access quad-word aligned addresses and replace
22067         with stvx or lvx instructions when appropriate.
22068         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
22069         New function prototype.
22070         (rs6000_quadword_masked_address_p): Likewise.
22071         (rs6000_gen_lvx): Likewise.
22072         (rs6000_gen_stvx): Likewise.
22073         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
22074         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
22075         when memory address is aligned.
22076         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
22077         this split to select lvx instruction when memory address is aligned.
22078         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
22079         instruction when memory address is aligned.
22080         (*vsx_le_perm_load_v16qi): Likewise.
22081         (four unnamed splitters): Modify to select the stvx instruction
22082         when memory is aligned.
22084 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
22086         * predict.c (determine_unlikely_bbs): Handle correctly BBs
22087         which appears in the queue multiple times.
22089 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22090             Alan Hayward  <alan.hayward@arm.com>
22091             David Sherwood  <david.sherwood@arm.com>
22093         * tree-vectorizer.h (vec_lower_bound): New structure.
22094         (_loop_vec_info): Add check_nonzero and lower_bounds.
22095         (LOOP_VINFO_CHECK_NONZERO): New macro.
22096         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
22097         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
22098         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
22099         fields.  Make seg_len the distance travelled, not including the
22100         access size.
22101         (dr_direction_indicator): Declare.
22102         (dr_zero_step_indicator): Likewise.
22103         (dr_known_forward_stride_p): Likewise.
22104         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
22105         tree-ssanames.h.
22106         (runtime_alias_check_p): Allow runtime alias checks with
22107         variable strides.
22108         (operator ==): Compare access_size and align.
22109         (prune_runtime_alias_test_list): Rework for new distinction between
22110         the access_size and seg_len.
22111         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
22112         segment lengths.
22113         (get_segment_min_max): New function.
22114         (create_intersect_range_checks): Use it.
22115         (dr_step_indicator): New function.
22116         (dr_direction_indicator): Likewise.
22117         (dr_zero_step_indicator): Likewise.
22118         (dr_known_forward_stride_p): Likewise.
22119         * tree-loop-distribution.c (data_ref_segment_size): Return
22120         DR_STEP * (niters - 1).
22121         (compute_alias_check_pairs): Update call to the dr_with_seg_len
22122         constructor.
22123         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
22124         (vect_preserves_scalar_order_p): New function, split out from...
22125         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
22126         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
22127         (vect_vfa_access_size): New function.
22128         (vect_vfa_align): Likewise.
22129         (vect_compile_time_alias): Take access_size_a and access_b arguments.
22130         (dump_lower_bound): New function.
22131         (vect_check_lower_bound): Likewise.
22132         (vect_small_gap_p): Likewise.
22133         (vectorizable_with_step_bound_p): Likewise.
22134         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
22135         depencies if the vectorization factor is 1.  Convert the checks
22136         for nonzero steps into checks on the bounds of DR_STEP.  Try using
22137         a bunds check for variable steps if the minimum required step is
22138         relatively small. Update calls to the dr_with_seg_len
22139         constructor and to vect_compile_time_alias.
22140         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
22141         function.
22142         (vect_loop_versioning): Call it.
22143         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
22144         when retrying.
22145         (vect_estimate_min_profitable_iters): Account for any bounds checks.
22147 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22148             Alan Hayward  <alan.hayward@arm.com>
22149             David Sherwood  <david.sherwood@arm.com>
22151         * doc/sourcebuild.texi (vect_scatter_store): Document.
22152         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
22153         optabs.
22154         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
22155         Document.
22156         * genopinit.c (main): Add supports_vec_scatter_store and
22157         supports_vec_scatter_store_cached to target_optabs.
22158         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
22159         IFN_MASK_SCATTER_STORE.
22160         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
22161         functions.
22162         * internal-fn.h (internal_store_fn_p): Declare.
22163         (internal_fn_stored_value_index): Likewise.
22164         * internal-fn.c (scatter_store_direct): New macro.
22165         (expand_scatter_store_optab_fn): New function.
22166         (direct_scatter_store_optab_supported_p): New macro.
22167         (internal_store_fn_p): New function.
22168         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
22169         IFN_MASK_SCATTER_STORE.
22170         (internal_fn_mask_index): Likewise.
22171         (internal_fn_stored_value_index): New function.
22172         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22173         for scatter stores.
22174         * optabs-query.h (supports_vec_scatter_store_p): Declare.
22175         * optabs-query.c (supports_vec_scatter_store_p): New function.
22176         * tree-vectorizer.h (vect_get_store_rhs): Declare.
22177         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
22178         true for scatter stores.
22179         (vect_gather_scatter_fn_p): Handle scatter stores too.
22180         (vect_check_gather_scatter): Consider using scatter stores if
22181         supports_vec_scatter_store_p.
22182         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22183         scatter stores too.
22184         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22185         internal_fn_stored_value_index.
22186         (check_load_store_masking): Handle scatter stores too.
22187         (vect_get_store_rhs): Make public.
22188         (vectorizable_call): Use internal_store_fn_p.
22189         (vectorizable_store): Handle scatter store internal functions.
22190         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22191         when deciding whether the end of the group has been reached.
22192         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22193         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22194         (mask_scatter_store<mode>): New insns.
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         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22201         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22202         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22203         function.
22204         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22205         Use vect_truncate_gather_scatter_offset if we can't treat the
22206         operation as a normal gather load or scatter store.
22207         (get_group_load_store_type): Take the gather_scatter_info
22208         as argument.  Try using a gather load or scatter store for
22209         single-element groups.
22210         (get_load_store_type): Update calls to get_group_load_store_type
22211         and vect_use_strided_gather_scatters_p.
22213 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22214             Alan Hayward  <alan.hayward@arm.com>
22215             David Sherwood  <david.sherwood@arm.com>
22217         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22218         optional tree argument.
22219         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22220         null target hooks.
22221         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22222         but continue to use the current value as a fallback.
22223         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22224         to compare the updates.
22225         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22226         (get_load_store_type): Use it when handling a strided access.
22227         (vect_get_strided_load_store_ops): New function.
22228         (vect_get_data_ptr_increment): Likewise.
22229         (vectorizable_load): Handle strided gather loads.  Always pass
22230         a step to vect_create_data_ref_ptr and bump_vector_ptr.
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         * doc/md.texi (gather_load@var{m}): Document.
22237         (mask_gather_load@var{m}): Likewise.
22238         * genopinit.c (main): Add supports_vec_gather_load and
22239         supports_vec_gather_load_cached to target_optabs.
22240         * optabs-tree.c (init_tree_optimization_optabs): Use
22241         ggc_cleared_alloc to allocate target_optabs.
22242         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22243         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22244         functions.
22245         * internal-fn.h (internal_load_fn_p): Declare.
22246         (internal_gather_scatter_fn_p): Likewise.
22247         (internal_fn_mask_index): Likewise.
22248         (internal_gather_scatter_fn_supported_p): Likewise.
22249         * internal-fn.c (gather_load_direct): New macro.
22250         (expand_gather_load_optab_fn): New function.
22251         (direct_gather_load_optab_supported_p): New macro.
22252         (direct_internal_fn_optab): New function.
22253         (internal_load_fn_p): Likewise.
22254         (internal_gather_scatter_fn_p): Likewise.
22255         (internal_fn_mask_index): Likewise.
22256         (internal_gather_scatter_fn_supported_p): Likewise.
22257         * optabs-query.c (supports_at_least_one_mode_p): New function.
22258         (supports_vec_gather_load_p): Likewise.
22259         * optabs-query.h (supports_vec_gather_load_p): Declare.
22260         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22261         and memory_type field.
22262         (NUM_PATTERNS): Bump to 15.
22263         * tree-vect-data-refs.c: Include internal-fn.h.
22264         (vect_gather_scatter_fn_p): New function.
22265         (vect_describe_gather_scatter_call): Likewise.
22266         (vect_check_gather_scatter): Try using internal functions for
22267         gather loads.  Recognize existing calls to a gather load function.
22268         (vect_analyze_data_refs): Consider using gather loads if
22269         supports_vec_gather_load_p.
22270         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22271         (vect_get_gather_scatter_offset_type): Likewise.
22272         (vect_convert_mask_for_vectype): Likewise.
22273         (vect_add_conversion_to_patterm): Likewise.
22274         (vect_try_gather_scatter_pattern): Likewise.
22275         (vect_recog_gather_scatter_pattern): New pattern recognizer.
22276         (vect_vect_recog_func_ptrs): Add it.
22277         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22278         internal_fn_mask_index and internal_gather_scatter_fn_p.
22279         (check_load_store_masking): Take the gather_scatter_info as an
22280         argument and handle gather loads.
22281         (vect_get_gather_scatter_ops): New function.
22282         (vectorizable_call): Check internal_load_fn_p.
22283         (vectorizable_load): Likewise.  Handle gather load internal
22284         functions.
22285         (vectorizable_store): Update call to check_load_store_masking.
22286         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22287         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22288         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22289         (aarch64_gather_scale_operand_d): New predicates.
22290         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22291         (mask_gather_load<mode>): New insns.
22293 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22294             Alan Hayward  <alan.hayward@arm.com>
22295             David Sherwood  <david.sherwood@arm.com>
22297         * optabs.def (fold_left_plus_optab): New optab.
22298         * doc/md.texi (fold_left_plus_@var{m}): Document.
22299         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22300         * internal-fn.c (fold_left_direct): Define.
22301         (expand_fold_left_optab_fn): Likewise.
22302         (direct_fold_left_optab_supported_p): Likewise.
22303         * fold-const-call.c (fold_const_fold_left): New function.
22304         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22305         * tree-parloops.c (valid_reduction_p): New function.
22306         (gather_scalar_reductions): Use it.
22307         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22308         (vect_finish_replace_stmt): Declare.
22309         * tree-vect-loop.c (fold_left_reduction_fn): New function.
22310         (needs_fold_left_reduction_p): New function, split out from...
22311         (vect_is_simple_reduction): ...here.  Accept reductions that
22312         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22313         (vect_force_simple_reduction): Also store the reduction type in
22314         the assignment's STMT_VINFO_REDUC_TYPE.
22315         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22316         (merge_with_identity): New function.
22317         (vect_expand_fold_left): Likewise.
22318         (vectorize_fold_left_reduction): Likewise.
22319         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
22320         scalar phi in place for it.  Check for target support and reject
22321         cases that would reassociate the operation.  Defer the transform
22322         phase to vectorize_fold_left_reduction.
22323         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22324         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22325         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22327 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22329         * tree-if-conv.c (predicate_mem_writes): Remove redundant
22330         call to ifc_temp_var.
22332 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22333             Alan Hayward  <alan.hayward@arm.com>
22334             David Sherwood  <david.sherwood@arm.com>
22336         * target.def (legitimize_address_displacement): Take the original
22337         offset as a poly_int.
22338         * targhooks.h (default_legitimize_address_displacement): Update
22339         accordingly.
22340         * targhooks.c (default_legitimize_address_displacement): Likewise.
22341         * doc/tm.texi: Regenerate.
22342         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22343         as an argument, moving assert of ad->disp == ad->disp_term to...
22344         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
22345         Try calling targetm.legitimize_address_displacement before expanding
22346         the address rather than afterwards, and adjust for the new interface.
22347         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22348         Match the new hook interface.  Handle SVE addresses.
22349         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22350         new hook interface.
22352 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22354         * Makefile.in (OBJS): Add early-remat.o.
22355         * target.def (select_early_remat_modes): New hook.
22356         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22357         * doc/tm.texi: Regenerate.
22358         * targhooks.h (default_select_early_remat_modes): Declare.
22359         * targhooks.c (default_select_early_remat_modes): New function.
22360         * timevar.def (TV_EARLY_REMAT): New timevar.
22361         * passes.def (pass_early_remat): New pass.
22362         * tree-pass.h (make_pass_early_remat): Declare.
22363         * early-remat.c: New file.
22364         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22365         function.
22366         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22368 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22369             Alan Hayward  <alan.hayward@arm.com>
22370             David Sherwood  <david.sherwood@arm.com>
22372         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22373         vfm1 with a bound_epilog parameter.
22374         (vect_do_peeling): Update calls accordingly, and move the prologue
22375         call earlier in the function.  Treat the base bound_epilog as 0 for
22376         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
22377         this base when peeling for gaps.
22378         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22379         with fully-masked loops.
22380         (vect_estimate_min_profitable_iters): Handle the single peeled
22381         iteration in that case.
22383 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22384             Alan Hayward  <alan.hayward@arm.com>
22385             David Sherwood  <david.sherwood@arm.com>
22387         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
22388         single-element interleaving even if the size is not a power of 2.
22389         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
22390         accesses for single-element interleaving if the group size is
22391         not a power of 2.
22393 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22394             Alan Hayward  <alan.hayward@arm.com>
22395             David Sherwood  <david.sherwood@arm.com>
22397         * doc/md.texi (fold_extract_last_@var{m}): Document.
22398         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
22399         * optabs.def (fold_extract_last_optab): New optab.
22400         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
22401         * internal-fn.c (fold_extract_direct): New macro.
22402         (expand_fold_extract_optab_fn): Likewise.
22403         (direct_fold_extract_optab_supported_p): Likewise.
22404         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
22405         * tree-vect-loop.c (vect_model_reduction_cost): Handle
22406         EXTRACT_LAST_REDUCTION.
22407         (get_initial_def_for_reduction): Do not create an initial vector
22408         for EXTRACT_LAST_REDUCTION reductions.
22409         (vectorizable_reduction): Leave the scalar phi in place for
22410         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
22411         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
22412         epilogue code for EXTRACT_LAST_REDUCTION and defer the
22413         transform phase to vectorizable_condition.
22414         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
22415         split out from...
22416         (vect_finish_stmt_generation): ...here.
22417         (vect_finish_replace_stmt): New function.
22418         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
22419         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
22420         pattern.
22421         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
22423 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22424             Alan Hayward  <alan.hayward@arm.com>
22425             David Sherwood  <david.sherwood@arm.com>
22427         * doc/md.texi (extract_last_@var{m}): Document.
22428         * optabs.def (extract_last_optab): New optab.
22429         * internal-fn.def (EXTRACT_LAST): New internal function.
22430         * internal-fn.c (cond_unary_direct): New macro.
22431         (expand_cond_unary_optab_fn): Likewise.
22432         (direct_cond_unary_optab_supported_p): Likewise.
22433         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
22434         loops using EXTRACT_LAST.
22435         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
22436         (extract_last_<mode>): ...this optab.
22437         (vec_extract<mode><Vel>): Update accordingly.
22439 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22440             Alan Hayward  <alan.hayward@arm.com>
22441             David Sherwood  <david.sherwood@arm.com>
22443         * target.def (empty_mask_is_expensive): New hook.
22444         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
22445         * doc/tm.texi: Regenerate.
22446         * targhooks.h (default_empty_mask_is_expensive): Declare.
22447         * targhooks.c (default_empty_mask_is_expensive): New function.
22448         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22449         if the target says that empty masks are expensive.
22450         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22451         New function.
22452         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22454 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22455             Alan Hayward  <alan.hayward@arm.com>
22456             David Sherwood  <david.sherwood@arm.com>
22458         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22459         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22460         (vect_use_loop_mask_for_alignment_p): New function.
22461         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22462         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22463         niters_skip argument.  Make sure that the first niters_skip elements
22464         of the first iteration are inactive.
22465         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22466         Update call to vect_set_loop_masks_directly.
22467         (get_misalign_in_elems): New function, split out from...
22468         (vect_gen_prolog_loop_niters): ...here.
22469         (vect_update_init_of_dr): Take a code argument that specifies whether
22470         the adjustment should be added or subtracted.
22471         (vect_update_init_of_drs): Likewise.
22472         (vect_prepare_for_masked_peels): New function.
22473         (vect_do_peeling): Skip prologue peeling if we're using a mask
22474         instead.  Update call to vect_update_inits_of_drs.
22475         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22476         mask_skip_niters.
22477         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22478         alignment.  Do not include the number of peeled iterations in
22479         the minimum threshold in that case.
22480         (vectorizable_induction): Adjust the start value down by
22481         LOOP_VINFO_MASK_SKIP_NITERS iterations.
22482         (vect_transform_loop): Call vect_prepare_for_masked_peels.
22483         Take the number of skipped iterations into account when calculating
22484         the loop bounds.
22485         * tree-vect-stmts.c (vect_gen_while_not): New function.
22487 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22488             Alan Hayward  <alan.hayward@arm.com>
22489             David Sherwood  <david.sherwood@arm.com>
22491         * doc/sourcebuild.texi (vect_fully_masked): Document.
22492         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22493         default value to 0.
22494         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22495         split out from...
22496         (vect_analyze_loop_2): ...here. Don't check the vectorization
22497         factor against the number of loop iterations if the loop is
22498         fully-masked.
22500 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22501             Alan Hayward  <alan.hayward@arm.com>
22502             David Sherwood  <david.sherwood@arm.com>
22504         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22505         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22506         (dump_groups): Update accordingly.
22507         (iv_use::mem_type): New member variable.
22508         (address_p): New function.
22509         (record_use): Add a mem_type argument and initialize the new
22510         mem_type field.
22511         (record_group_use): Add a mem_type argument.  Use address_p.
22512         Remove obsolete null checks of base_object.  Update call to record_use.
22513         (find_interesting_uses_op): Update call to record_group_use.
22514         (find_interesting_uses_cond): Likewise.
22515         (find_interesting_uses_address): Likewise.
22516         (get_mem_type_for_internal_fn): New function.
22517         (find_address_like_use): Likewise.
22518         (find_interesting_uses_stmt): Try find_address_like_use before
22519         calling find_interesting_uses_op.
22520         (addr_offset_valid_p): Use the iv mem_type field as the type
22521         of the addressed memory.
22522         (add_autoinc_candidates): Likewise.
22523         (get_address_cost): Likewise.
22524         (split_small_address_groups_p): Use address_p.
22525         (split_address_groups): Likewise.
22526         (add_iv_candidate_for_use): Likewise.
22527         (autoinc_possible_for_pair): Likewise.
22528         (rewrite_groups): Likewise.
22529         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22530         (determine_group_iv_cost): Update after split of USE_ADDRESS.
22531         (get_alias_ptr_type_for_ptr_address): New function.
22532         (rewrite_use_address): Rewrite address uses in calls that were
22533         identified by find_address_like_use.
22535 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22536             Alan Hayward  <alan.hayward@arm.com>
22537             David Sherwood  <david.sherwood@arm.com>
22539         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22540         TARGET_MEM_REFs.
22541         * gimple-expr.h (is_gimple_addressable: Likewise.
22542         * gimple-expr.c (is_gimple_address): Likewise.
22543         * internal-fn.c (expand_call_mem_ref): New function.
22544         (expand_mask_load_optab_fn): Use it.
22545         (expand_mask_store_optab_fn): Likewise.
22547 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22548             Alan Hayward  <alan.hayward@arm.com>
22549             David Sherwood  <david.sherwood@arm.com>
22551         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22552         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22553         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22554         (cond_umax@var{mode}): Document.
22555         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22556         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22557         (cond_umin_optab, cond_umax_optab): New optabs.
22558         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22559         (COND_IOR, COND_XOR): New internal functions.
22560         * internal-fn.h (get_conditional_internal_fn): Declare.
22561         * internal-fn.c (cond_binary_direct): New macro.
22562         (expand_cond_binary_optab_fn): Likewise.
22563         (direct_cond_binary_optab_supported_p): Likewise.
22564         (get_conditional_internal_fn): New function.
22565         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22566         Cope with reduction statements that are vectorized as calls rather
22567         than assignments.
22568         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22569         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22570         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22571         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22572         (UNSPEC_COND_EOR): New unspecs.
22573         (optab): Add mappings for them.
22574         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22575         (sve_int_op, sve_fp_op): New int attributes.
22577 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22578             Alan Hayward  <alan.hayward@arm.com>
22579             David Sherwood  <david.sherwood@arm.com>
22581         * optabs.def (while_ult_optab): New optab.
22582         * doc/md.texi (while_ult@var{m}@var{n}): Document.
22583         * internal-fn.def (WHILE_ULT): New internal function.
22584         * internal-fn.h (direct_internal_fn_supported_p): New override
22585         that takes two types as argument.
22586         * internal-fn.c (while_direct): New macro.
22587         (expand_while_optab_fn): New function.
22588         (convert_optab_supported_p): Likewise.
22589         (direct_while_optab_supported_p): New macro.
22590         * wide-int.h (wi::udiv_ceil): New function.
22591         * tree-vectorizer.h (rgroup_masks): New structure.
22592         (vec_loop_masks): New typedef.
22593         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22594         and fully_masked_p.
22595         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22596         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22597         (vect_max_vf): New function.
22598         (slpeel_make_loop_iterate_ntimes): Delete.
22599         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22600         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22601         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22602         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22603         internal-fn.h, stor-layout.h and optabs-query.h.
22604         (vect_set_loop_mask): New function.
22605         (add_preheader_seq): Likewise.
22606         (add_header_seq): Likewise.
22607         (interleave_supported_p): Likewise.
22608         (vect_maybe_permute_loop_masks): Likewise.
22609         (vect_set_loop_masks_directly): Likewise.
22610         (vect_set_loop_condition_masked): Likewise.
22611         (vect_set_loop_condition_unmasked): New function, split out from
22612         slpeel_make_loop_iterate_ntimes.
22613         (slpeel_make_loop_iterate_ntimes): Rename to..
22614         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
22615         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22616         (vect_do_peeling): Update call accordingly.
22617         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22618         loops.
22619         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22620         mask_compare_type, can_fully_mask_p and fully_masked_p.
22621         (release_vec_loop_masks): New function.
22622         (_loop_vec_info): Use it to free the loop masks.
22623         (can_produce_all_loop_masks_p): New function.
22624         (vect_get_max_nscalars_per_iter): Likewise.
22625         (vect_verify_full_masking): Likewise.
22626         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22627         retries, and free the mask rgroups before retrying.  Check loop-wide
22628         reasons for disallowing fully-masked loops.  Make the final decision
22629         about whether use a fully-masked loop or not.
22630         (vect_estimate_min_profitable_iters): Do not assume that peeling
22631         for the number of iterations will be needed for fully-masked loops.
22632         (vectorizable_reduction): Disable fully-masked loops.
22633         (vectorizable_live_operation): Likewise.
22634         (vect_halve_mask_nunits): New function.
22635         (vect_double_mask_nunits): Likewise.
22636         (vect_record_loop_mask): Likewise.
22637         (vect_get_loop_mask): Likewise.
22638         (vect_transform_loop): Handle the case in which the final loop
22639         iteration might handle a partial vector.  Call vect_set_loop_condition
22640         instead of slpeel_make_loop_iterate_ntimes.
22641         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22642         (check_load_store_masking): New function.
22643         (prepare_load_store_mask): Likewise.
22644         (vectorizable_store): Handle fully-masked loops.
22645         (vectorizable_load): Likewise.
22646         (supportable_widening_operation): Use vect_halve_mask_nunits for
22647         booleans.
22648         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22649         (vect_gen_while): New function.
22650         * config/aarch64/aarch64.md (umax<mode>3): New expander.
22651         (aarch64_uqdec<mode>): New insn.
22653 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22654             Alan Hayward  <alan.hayward@arm.com>
22655             David Sherwood  <david.sherwood@arm.com>
22657         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22658         (reduc_xor_scal_optab): New optabs.
22659         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22660         (reduc_xor_scal_@var{m}): Document.
22661         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22662         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22663         internal functions.
22664         * fold-const-call.c (fold_const_call): Handle them.
22665         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22666         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22667         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22668         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22669         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22670         (UNSPEC_XORV): New unspecs.
22671         (optab): Add entries for them.
22672         (BITWISEV): New int iterator.
22673         (bit_reduc_op): New int attributes.
22675 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22676             Alan Hayward  <alan.hayward@arm.com>
22677             David Sherwood  <david.sherwood@arm.com>
22679         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22680         * internal-fn.def (VEC_SHL_INSERT): New internal function.
22681         * optabs.def (vec_shl_insert_optab): New optab.
22682         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22683         (duplicate_and_interleave): Likewise.
22684         * tree-vect-loop.c: Include internal-fn.h.
22685         (neutral_op_for_slp_reduction): New function, split out from
22686         get_initial_defs_for_reduction.
22687         (get_initial_def_for_reduction): Handle option 2 for variable-length
22688         vectors by loading the neutral value into a vector and then shifting
22689         the initial value into element 0.
22690         (get_initial_defs_for_reduction): Replace the code argument with
22691         the neutral value calculated by neutral_op_for_slp_reduction.
22692         Use gimple_build_vector for constant-length vectors.
22693         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22694         but the first group_size elements have a neutral value.
22695         Use duplicate_and_interleave otherwise.
22696         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22697         Update call to get_initial_defs_for_reduction.  Handle SLP
22698         reductions for variable-length vectors by creating one vector
22699         result for each scalar result, with the elements associated
22700         with other scalar results stubbed out with the neutral value.
22701         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22702         Require IFN_VEC_SHL_INSERT for double reductions on
22703         variable-length vectors, or SLP reductions that have
22704         a neutral value.  Require can_duplicate_and_interleave_p
22705         support for variable-length unchained SLP reductions if there
22706         is no neutral value, such as for MIN/MAX reductions.  Also require
22707         the number of vector elements to be a multiple of the number of
22708         SLP statements when doing variable-length unchained SLP reductions.
22709         Update call to vect_create_epilog_for_reduction.
22710         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22711         and remove initial values.
22712         (duplicate_and_interleave): Make public.
22713         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22714         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22716 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22717             Alan Hayward  <alan.hayward@arm.com>
22718             David Sherwood  <david.sherwood@arm.com>
22720         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22721         (can_duplicate_and_interleave_p): New function.
22722         (vect_get_and_check_slp_defs): Take the vector of statements
22723         rather than just the current one.  Remove excess parentheses.
22724         Restriction rejectinon of vect_constant_def and vect_external_def
22725         for variable-length vectors to boolean types, or types for which
22726         can_duplicate_and_interleave_p is false.
22727         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22728         (duplicate_and_interleave): New function.
22729         (vect_get_constant_vectors): Use gimple_build_vector for
22730         constant-length vectors and suitable variable-length constant
22731         vectors.  Use duplicate_and_interleave for other variable-length
22732         vectors.  Don't defer the update when inserting new statements.
22734 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22735             Alan Hayward  <alan.hayward@arm.com>
22736             David Sherwood  <david.sherwood@arm.com>
22738         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22739         min_profitable_iters doesn't go negative.
22741 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22742             Alan Hayward  <alan.hayward@arm.com>
22743             David Sherwood  <david.sherwood@arm.com>
22745         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22746         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22747         * optabs.def (vec_mask_load_lanes_optab): New optab.
22748         (vec_mask_store_lanes_optab): Likewise.
22749         * internal-fn.def (MASK_LOAD_LANES): New internal function.
22750         (MASK_STORE_LANES): Likewise.
22751         * internal-fn.c (mask_load_lanes_direct): New macro.
22752         (mask_store_lanes_direct): Likewise.
22753         (expand_mask_load_optab_fn): Handle masked operations.
22754         (expand_mask_load_lanes_optab_fn): New macro.
22755         (expand_mask_store_optab_fn): Handle masked operations.
22756         (expand_mask_store_lanes_optab_fn): New macro.
22757         (direct_mask_load_lanes_optab_supported_p): Likewise.
22758         (direct_mask_store_lanes_optab_supported_p): Likewise.
22759         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22760         parameter.
22761         (vect_load_lanes_supported): Likewise.
22762         * tree-vect-data-refs.c (strip_conversion): New function.
22763         (can_group_stmts_p): Likewise.
22764         (vect_analyze_data_ref_accesses): Use it instead of checking
22765         for a pair of assignments.
22766         (vect_store_lanes_supported): Take a masked_p parameter.
22767         (vect_load_lanes_supported): Likewise.
22768         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22769         vect_store_lanes_supported and vect_load_lanes_supported.
22770         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22771         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22772         parameter.  Don't allow gaps for masked accesses.
22773         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
22774         and vect_load_lanes_supported.
22775         (get_load_store_type): Take a masked_p parameter and update
22776         call to get_group_load_store_type.
22777         (vectorizable_store): Update call to get_load_store_type.
22778         Handle IFN_MASK_STORE_LANES.
22779         (vectorizable_load): Update call to get_load_store_type.
22780         Handle IFN_MASK_LOAD_LANES.
22782 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22783             Alan Hayward  <alan.hayward@arm.com>
22784             David Sherwood  <david.sherwood@arm.com>
22786         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22787         modes for SVE.
22788         * config/aarch64/aarch64-protos.h
22789         (aarch64_sve_struct_memory_operand_p): Declare.
22790         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22791         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22792         (VPRED, vpred): Handle SVE structure modes.
22793         * config/aarch64/constraints.md (Utx): New constraint.
22794         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22795         (aarch64_sve_struct_nonimmediate_operand): New predicates.
22796         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22797         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22798         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22799         structure modes.  Split into pieces after RA.
22800         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22801         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22802         New patterns.
22803         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22804         SVE structure modes.
22805         (aarch64_classify_address): Likewise.
22806         (sizetochar): Move earlier in file.
22807         (aarch64_print_operand): Handle SVE register lists.
22808         (aarch64_array_mode): New function.
22809         (aarch64_sve_struct_memory_operand_p): Likewise.
22810         (TARGET_ARRAY_MODE): Redefine.
22812 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22813             Alan Hayward  <alan.hayward@arm.com>
22814             David Sherwood  <david.sherwood@arm.com>
22816         * target.def (array_mode): New target hook.
22817         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22818         * doc/tm.texi: Regenerate.
22819         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22820         * hooks.c (hook_optmode_mode_uhwi_none): New function.
22821         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22822         targetm.array_mode.
22823         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
22824         type sizes.
22826 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22827             Alan Hayward  <alan.hayward@arm.com>
22828             David Sherwood  <david.sherwood@arm.com>
22830         * fold-const.c (fold_binary_loc): Check the argument types
22831         rather than the result type when testing for a vector operation.
22833 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22835         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22836         * doc/tm.texi: Regenerate.
22838 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22839             Alan Hayward  <alan.hayward@arm.com>
22840             David Sherwood  <david.sherwood@arm.com>
22842         * doc/invoke.texi (-msve-vector-bits=): Document new option.
22843         (sve): Document new AArch64 extension.
22844         * doc/md.texi (w): Extend the description of the AArch64
22845         constraint to include SVE vectors.
22846         (Upl, Upa): Document new AArch64 predicate constraints.
22847         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22848         enum.
22849         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22850         (msve-vector-bits=): New option.
22851         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22852         SVE when these are disabled.
22853         (sve): New extension.
22854         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22855         modes.  Adjust their number of units based on aarch64_sve_vg.
22856         (MAX_BITSIZE_MODE_ANY_MODE): Define.
22857         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22858         aarch64_addr_query_type.
22859         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22860         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22861         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22862         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22863         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22864         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22865         (aarch64_simd_imm_zero_p): Delete.
22866         (aarch64_check_zero_based_sve_index_immediate): Declare.
22867         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22868         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22869         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22870         (aarch64_sve_float_mul_immediate_p): Likewise.
22871         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22872         rather than an rtx.
22873         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22874         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22875         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22876         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22877         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22878         (aarch64_regmode_natural_size): Likewise.
22879         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22880         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22881         left one place.
22882         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22883         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22884         for VG and the SVE predicate registers.
22885         (V_ALIASES): Add a "z"-prefixed alias.
22886         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22887         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22888         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22889         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22890         (REG_CLASS_NAMES): Add entries for them.
22891         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
22892         and the predicate registers.
22893         (aarch64_sve_vg): Declare.
22894         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22895         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22896         (REGMODE_NATURAL_SIZE): Define.
22897         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22898         SVE macros.
22899         * config/aarch64/aarch64.c: Include cfgrtl.h.
22900         (simd_immediate_info): Add a constructor for series vectors,
22901         and an associated step field.
22902         (aarch64_sve_vg): New variable.
22903         (aarch64_dbx_register_number): Handle VG and the predicate registers.
22904         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22905         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22906         (VEC_ANY_DATA, VEC_STRUCT): New constants.
22907         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22908         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22909         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22910         (aarch64_get_mask_mode): New functions.
22911         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22912         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22913         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
22914         predicate modes and predicate registers.  Explicitly restrict
22915         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
22916         to store a vector mode if it is recognized by
22917         aarch64_classify_vector_mode.
22918         (aarch64_regmode_natural_size): New function.
22919         (aarch64_hard_regno_caller_save_mode): Return the original mode
22920         for predicates.
22921         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22922         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22923         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22924         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22925         functions.
22926         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
22927         does not overlap dest if the function is frame-related.  Handle
22928         SVE constants.
22929         (aarch64_split_add_offset): New function.
22930         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22931         them aarch64_add_offset.
22932         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22933         and update call to aarch64_sub_sp.
22934         (aarch64_add_cfa_expression): New function.
22935         (aarch64_expand_prologue): Pass extra temporary registers to the
22936         functions above.  Handle the case in which we need to emit new
22937         DW_CFA_expressions for registers that were originally saved
22938         relative to the stack pointer, but now have to be expressed
22939         relative to the frame pointer.
22940         (aarch64_output_mi_thunk): Pass extra temporary registers to the
22941         functions above.
22942         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
22943         IP0 and IP1 values for SVE frames.
22944         (aarch64_expand_vec_series): New function.
22945         (aarch64_expand_sve_widened_duplicate): Likewise.
22946         (aarch64_expand_sve_const_vector): Likewise.
22947         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22948         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
22949         into the register, rather than emitting a SET directly.
22950         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22951         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22952         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22953         (offset_9bit_signed_scaled_p): New functions.
22954         (aarch64_replicate_bitmask_imm): New function.
22955         (aarch64_bitmask_imm): Use it.
22956         (aarch64_cannot_force_const_mem): Reject expressions involving
22957         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
22958         (aarch64_classify_index): Handle SVE indices, by requiring
22959         a plain register index with a scale that matches the element size.
22960         (aarch64_classify_address): Handle SVE addresses.  Assert that
22961         the mode of the address is VOIDmode or an integer mode.
22962         Update call to aarch64_classify_symbol.
22963         (aarch64_classify_symbolic_expression): Update call to
22964         aarch64_classify_symbol.
22965         (aarch64_const_vec_all_in_range_p): New function.
22966         (aarch64_print_vector_float_operand): Likewise.
22967         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
22968         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
22969         and the FP immediates 1.0 and 0.5.
22970         (aarch64_print_address_internal): Handle SVE addresses.
22971         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22972         (aarch64_regno_regclass): Handle predicate registers.
22973         (aarch64_secondary_reload): Handle big-endian reloads of SVE
22974         data modes.
22975         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22976         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22977         (aarch64_convert_sve_vector_bits): New function.
22978         (aarch64_override_options): Use it to handle -msve-vector-bits=.
22979         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22980         rather than an rtx.
22981         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22982         Handle SVE vector and predicate modes.  Accept VL-based constants
22983         that need only one temporary register, and VL offsets that require
22984         no temporary registers.
22985         (aarch64_conditional_register_usage): Mark the predicate registers
22986         as fixed if SVE isn't available.
22987         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22988         Return true for SVE vector and predicate modes.
22989         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22990         rather than an unsigned int.  Handle SVE modes.
22991         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
22992         SVE modes.
22993         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22994         if SVE is enabled.
22995         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22996         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22997         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22998         (aarch64_sve_float_mul_immediate_p): New functions.
22999         (aarch64_sve_valid_immediate): New function.
23000         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
23001         Explicitly reject structure modes.  Check for INDEX constants.
23002         Handle PTRUE and PFALSE constants.
23003         (aarch64_check_zero_based_sve_index_immediate): New function.
23004         (aarch64_simd_imm_zero_p): Delete.
23005         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
23006         vector modes.  Accept constants in the range of CNT[BHWD].
23007         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
23008         ask for an Advanced SIMD mode.
23009         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
23010         (aarch64_simd_vector_alignment): Handle SVE predicates.
23011         (aarch64_vectorize_preferred_vector_alignment): New function.
23012         (aarch64_simd_vector_alignment_reachable): Use it instead of
23013         the vector size.
23014         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
23015         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
23016         functions.
23017         (MAX_VECT_LEN): Delete.
23018         (expand_vec_perm_d): Add a vec_flags field.
23019         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
23020         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
23021         (aarch64_evpc_ext): Don't apply a big-endian lane correction
23022         for SVE modes.
23023         (aarch64_evpc_rev): Rename to...
23024         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
23025         (aarch64_evpc_rev_global): New function.
23026         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
23027         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
23028         MAX_VECT_LEN.
23029         (aarch64_evpc_sve_tbl): New function.
23030         (aarch64_expand_vec_perm_const_1): Update after rename of
23031         aarch64_evpc_rev.  Handle SVE permutes too, trying
23032         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
23033         than aarch64_evpc_tbl.
23034         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
23035         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
23036         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
23037         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
23038         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
23039         (aarch64_expand_sve_vcond): New functions.
23040         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
23041         of aarch64_vector_mode_p.
23042         (aarch64_dwarf_poly_indeterminate_value): New function.
23043         (aarch64_compute_pressure_classes): Likewise.
23044         (aarch64_can_change_mode_class): Likewise.
23045         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
23046         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
23047         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
23048         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
23049         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
23050         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
23051         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
23052         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
23053         constraints.
23054         (Dn, Dl, Dr): Accept const as well as const_vector.
23055         (Dz): Likewise.  Compare against CONST0_RTX.
23056         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
23057         of "vector" where appropriate.
23058         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
23059         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
23060         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
23061         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
23062         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
23063         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
23064         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
23065         (v_int_equiv): Extend to SVE modes.
23066         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
23067         mode attributes.
23068         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
23069         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
23070         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
23071         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
23072         (SVE_COND_FP_CMP): New int iterators.
23073         (perm_hilo): Handle the new unpack unspecs.
23074         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
23075         attributes.
23076         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
23077         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
23078         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
23079         (aarch64_equality_operator, aarch64_constant_vector_operand)
23080         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
23081         (aarch64_sve_nonimmediate_operand): Likewise.
23082         (aarch64_sve_general_operand): Likewise.
23083         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
23084         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
23085         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
23086         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
23087         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
23088         (aarch64_sve_float_arith_immediate): Likewise.
23089         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
23090         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
23091         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
23092         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
23093         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
23094         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
23095         (aarch64_sve_float_arith_operand): Likewise.
23096         (aarch64_sve_float_arith_with_sub_operand): Likewise.
23097         (aarch64_sve_float_mul_operand): Likewise.
23098         (aarch64_sve_vec_perm_operand): Likewise.
23099         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
23100         (aarch64_mov_operand): Accept const_poly_int and const_vector.
23101         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
23102         as well as const_vector.
23103         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
23104         in file.  Use CONST0_RTX and CONSTM1_RTX.
23105         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
23106         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
23107         Use aarch64_simd_imm_zero.
23108         * config/aarch64/aarch64-sve.md: New file.
23109         * config/aarch64/aarch64.md: Include it.
23110         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
23111         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
23112         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
23113         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
23114         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
23115         (sve): New attribute.
23116         (enabled): Disable instructions with the sve attribute unless
23117         TARGET_SVE.
23118         (movqi, movhi): Pass CONST_POLY_INT operaneds through
23119         aarch64_expand_mov_immediate.
23120         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
23121         CNT[BHSD] immediates.
23122         (movti): Split CONST_POLY_INT moves into two halves.
23123         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
23124         Split additions that need a temporary here if the destination
23125         is the stack pointer.
23126         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
23127         (*add<mode>3_poly_1): New instruction.
23128         (set_clobber_cc): New expander.
23130 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23132         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
23133         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
23134         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
23135         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
23136         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
23137         Change innermode from fixed_mode_size to machine_mode.
23138         (simplify_subreg): Update call accordingly.  Handle a constant-sized
23139         subreg of a variable-length CONST_VECTOR.
23141 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23142             Alan Hayward  <alan.hayward@arm.com>
23143             David Sherwood  <david.sherwood@arm.com>
23145         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
23146         (add_offset_to_base): New function, split out from...
23147         (create_mem_ref): ...here.  When handling a scale other than 1,
23148         check first whether the address is valid without the offset.
23149         Add it into the base if so, leaving the index and scale as-is.
23151 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
23153         PR c++/83778
23154         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
23155         fold_for_warn before checking if arg2 is INTEGER_CST.
23157 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
23159         * config/rs6000/predicates.md (load_multiple_operation): Delete.
23160         (store_multiple_operation): Delete.
23161         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
23162         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
23163         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
23164         guarded by TARGET_STRING.
23165         (rs6000_output_load_multiple): Delete.
23166         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
23167         OPTION_MASK_STRING / TARGET_STRING handling.
23168         (print_operand) <'N', 'O'>: Add comment that these are unused now.
23169         (const rs6000_opt_masks) <"string">: Change mask to 0.
23170         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
23171         (MASK_STRING): Delete.
23172         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23173         parts.  Simplify.
23174         (load_multiple): Delete.
23175         (*ldmsi8): Delete.
23176         (*ldmsi7): Delete.
23177         (*ldmsi6): Delete.
23178         (*ldmsi5): Delete.
23179         (*ldmsi4): Delete.
23180         (*ldmsi3): Delete.
23181         (store_multiple): Delete.
23182         (*stmsi8): Delete.
23183         (*stmsi7): Delete.
23184         (*stmsi6): Delete.
23185         (*stmsi5): Delete.
23186         (*stmsi4): Delete.
23187         (*stmsi3): Delete.
23188         (movmemsi_8reg): Delete.
23189         (corresponding unnamed define_insn): Delete.
23190         (movmemsi_6reg): Delete.
23191         (corresponding unnamed define_insn): Delete.
23192         (movmemsi_4reg): Delete.
23193         (corresponding unnamed define_insn): Delete.
23194         (movmemsi_2reg): Delete.
23195         (corresponding unnamed define_insn): Delete.
23196         (movmemsi_1reg): Delete.
23197         (corresponding unnamed define_insn): Delete.
23198         * config/rs6000/rs6000.opt (mno-string): New.
23199         (mstring): Replace by deprecation warning stub.
23200         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23202 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
23204         * regrename.c (regrename_do_replace): If replacing the same
23205         reg multiple times, try to reuse last created gen_raw_REG.
23207         PR debug/81155
23208         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23209         main to workaround a bug in GDB.
23211 2018-01-12  Tom de Vries  <tom@codesourcery.com>
23213         PR target/83737
23214         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23216 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
23218         PR rtl-optimization/80481
23219         * ira-color.c (get_cap_member): New function.
23220         (allocnos_conflict_by_live_ranges_p): Use it.
23221         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23222         (setup_slot_coalesced_allocno_live_ranges): Ditto.
23224 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
23226         PR target/83628
23227         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23228         (*saddl_se_1): Ditto.
23229         (*ssubsi_1): Ditto.
23230         (*ssubl_se_1): Ditto.
23232 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
23234         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23235         rather than wi::to_widest for DR_INITs.
23236         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23237         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23238         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23239         INTEGER_CSTs.
23240         (vect_analyze_group_access_1): Note that here.
23242 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
23244         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23245         polynomial type sizes.
23247 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
23249         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23250         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23251         (gimple_add_tmp_var): Likewise.
23253 2018-01-12  Martin Liska  <mliska@suse.cz>
23255         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23256         (gimple_alloc_sizes): Likewise.
23257         (dump_gimple_statistics): Use PRIu64 in printf format.
23258         * gimple.h: Change uint64_t to int.
23260 2018-01-12  Martin Liska  <mliska@suse.cz>
23262         * tree-core.h: Use uint64_t instead of int.
23263         * tree.c (tree_node_counts): Likewise.
23264         (tree_node_sizes): Likewise.
23265         (dump_tree_statistics): Use PRIu64 in printf format.
23267 2018-01-12  Martin Liska  <mliska@suse.cz>
23269         * Makefile.in: As qsort_chk is implemented in vec.c, add
23270         vec.o to linkage of gencfn-macros.
23271         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23272         passing the info to record_node_allocation_statistics.
23273         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23274         and pass the info.
23275         * ggc-common.c (struct ggc_usage): Add operator== and use
23276         it in operator< and compare function.
23277         * mem-stats.h (struct mem_usage): Likewise.
23278         * vec.c (struct vec_usage): Remove operator< and compare
23279         function. Can be simply inherited.
23281 2018-01-12  Martin Jambor  <mjambor@suse.cz>
23283         PR target/81616
23284         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23285         * tree-ssa-math-opts.c: Include domwalk.h.
23286         (convert_mult_to_fma_1): New function.
23287         (fma_transformation_info): New type.
23288         (fma_deferring_state): Likewise.
23289         (cancel_fma_deferring): New function.
23290         (result_of_phi): Likewise.
23291         (last_fma_candidate_feeds_initial_phi): Likewise.
23292         (convert_mult_to_fma): Added deferring logic, split actual
23293         transformation to convert_mult_to_fma_1.
23294         (math_opts_dom_walker): New type.
23295         (math_opts_dom_walker::after_dom_children): New method, body moved
23296         here from pass_optimize_widening_mul::execute, added deferring logic
23297         bits.
23298         (pass_optimize_widening_mul::execute): Moved most of code to
23299         math_opts_dom_walker::after_dom_children.
23300         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23301         * config/i386/i386.c (ix86_option_override_internal): Added
23302         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23304 2018-01-12  Richard Biener  <rguenther@suse.de>
23306         PR debug/83157
23307         * dwarf2out.c (gen_variable_die): Do not reset old_die for
23308         inline instance vars.
23310 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
23312         PR target/81819
23313         * config/rx/rx.c (rx_is_restricted_memory_address):
23314         Handle SUBREG case.
23316 2018-01-12  Richard Biener  <rguenther@suse.de>
23318         PR tree-optimization/80846
23319         * target.def (split_reduction): New target hook.
23320         * targhooks.c (default_split_reduction): New function.
23321         * targhooks.h (default_split_reduction): Declare.
23322         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23323         target requests first reduce vectors by combining low and high
23324         parts.
23325         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23326         (get_vectype_for_scalar_type_and_size): Export.
23327         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23328         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23329         * doc/tm.texi: Regenerate.
23330         * config/i386/i386.c (ix86_split_reduction): Implement
23331         TARGET_VECTORIZE_SPLIT_REDUCTION.
23333 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23335         PR target/83368
23336         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23337         in PIC mode except for TARGET_VXWORKS_RTP.
23338         * config/sparc/sparc.c: Include cfgrtl.h.
23339         (TARGET_INIT_PIC_REG): Define.
23340         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23341         (sparc_pic_register_p): New predicate.
23342         (sparc_legitimate_address_p): Use it.
23343         (sparc_legitimize_pic_address): Likewise.
23344         (sparc_delegitimize_address): Likewise.
23345         (sparc_mode_dependent_address_p): Likewise.
23346         (gen_load_pcrel_sym): Remove 4th parameter.
23347         (load_got_register): Adjust call to above.  Remove obsolete stuff.
23348         (sparc_expand_prologue): Do not call load_got_register here.
23349         (sparc_flat_expand_prologue): Likewise.
23350         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23351         (sparc_use_pseudo_pic_reg): New function.
23352         (sparc_init_pic_reg): Likewise.
23353         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23354         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23356 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
23358         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23359         Add item for branch_cost.
23361 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23363         PR rtl-optimization/83565
23364         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23365         not extend the result to a larger mode for rotate operations.
23366         (num_sign_bit_copies1): Likewise.
23368 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23370         PR target/40411
23371         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23372         -symbolic.
23373         Use values-Xc.o for -pedantic.
23374         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23376 2018-01-12  Martin Liska  <mliska@suse.cz>
23378         PR ipa/83054
23379         * ipa-devirt.c (final_warning_record::grow_type_warnings):
23380         New function.
23381         (possible_polymorphic_call_targets): Use it.
23382         (ipa_devirt): Likewise.
23384 2018-01-12  Martin Liska  <mliska@suse.cz>
23386         * profile-count.h (enum profile_quality): Use 0 as invalid
23387         enum value of profile_quality.
23389 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23391         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
23392         -mext-string options.
23394 2018-01-12  Richard Biener  <rguenther@suse.de>
23396         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
23397         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
23398         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
23399         Likewise.
23400         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
23402 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
23404         * configure.ac (--with-long-double-format): Add support for the
23405         configuration option to change the default long double format on
23406         PowerPC systems.
23407         * config.gcc (powerpc*-linux*-*): Likewise.
23408         * configure: Regenerate.
23409         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
23410         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
23411         used without modification.
23413 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23415         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
23416         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
23417         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
23418         MISC_BUILTIN_SPEC_BARRIER.
23419         (rs6000_init_builtins): Likewise.
23420         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
23421         enum value.
23422         (speculation_barrier): New define_insn.
23423         * doc/extend.texi: Document __builtin_speculation_barrier.
23425 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
23427         PR target/83203
23428         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
23429         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
23430         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
23431         iterators.
23432         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
23433         integral modes instead of "ss" and "sd".
23434         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
23435         vectors with 32-bit and 64-bit elements.
23436         (vecdupssescalarmodesuffix): New mode attribute.
23437         (vec_dup<mode>): Use it.
23439 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
23441         PR target/83330
23442         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
23443         frame if argument is passed on stack.
23445 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
23447         PR target/82682
23448         * ree.c (combine_reaching_defs): Optimize also
23449         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23450         reg2=any_extend(exp); reg1=reg2;, formatting fix.
23452 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
23454         PR middle-end/83189
23455         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23457 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
23459         PR middle-end/83718
23460         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23461         after they are computed.
23463 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
23465         PR tree-optimization/83695
23466         * gimple-loop-linterchange.cc
23467         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23468         reset cached scev information after interchange.
23469         (pass_linterchange::execute): Remove call to scev_reset_htab.
23471 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23473         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23474         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23475         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23476         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23477         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23478         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23479         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23480         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23481         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23482         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23483         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23484         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23485         (V_lane_reg): Likewise.
23486         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23487         New define_expand.
23488         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23489         (vfmal_lane_low<mode>_intrinsic,
23490         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23491         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23492         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23493         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23494         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23495         vfmsl_lane_high<mode>_intrinsic): New define_insns.
23497 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23499         * config/arm/arm-cpus.in (fp16fml): New feature.
23500         (ALL_SIMD): Add fp16fml.
23501         (armv8.2-a): Add fp16fml as an option.
23502         (armv8.3-a): Likewise.
23503         (armv8.4-a): Add fp16fml as part of fp16.
23504         * config/arm/arm.h (TARGET_FP16FML): Define.
23505         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23506         when appropriate.
23507         * config/arm/arm-modes.def (V2HF): Define.
23508         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23509         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23510         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23511         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23512         vfmsl_low, vfmsl_high): New set of builtins.
23513         * config/arm/iterators.md (PLUSMINUS): New code iterator.
23514         (vfml_op): New code attribute.
23515         (VFMLHALVES): New int iterator.
23516         (VFML, VFMLSEL): New mode attributes.
23517         (V_reg): Define mapping for V2HF.
23518         (V_hi, V_lo): New mode attributes.
23519         (VF_constraint): Likewise.
23520         (vfml_half, vfml_half_selector): New int attributes.
23521         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23522         define_expand.
23523         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23524         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23525         New define_insn.
23526         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23527         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23528         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23529         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
23530         documentation.
23531         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23532         Document new effective target and option set.
23534 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23536         * config/arm/arm-cpus.in (armv8_4): New feature.
23537         (ARMv8_4a): New fgroup.
23538         (armv8.4-a): New arch.
23539         * config/arm/arm-tables.opt: Regenerate.
23540         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23541         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23542         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23543         Add matching rules for -march=armv8.4-a and extensions.
23544         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23546 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
23548         PR target/81821
23549         * config/rx/rx.md (BW): New mode attribute.
23550         (sync_lock_test_and_setsi): Add mode suffix to insn output.
23552 2018-01-11  Richard Biener  <rguenther@suse.de>
23554         PR tree-optimization/83435
23555         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23556         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23557         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23559 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23560             Alan Hayward  <alan.hayward@arm.com>
23561             David Sherwood  <david.sherwood@arm.com>
23563         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23564         field.
23565         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
23566         (aarch64_print_address_internal): Use it to check for a zero offset.
23568 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23569             Alan Hayward  <alan.hayward@arm.com>
23570             David Sherwood  <david.sherwood@arm.com>
23572         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23573         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23574         Return a poly_int64 rather than a HOST_WIDE_INT.
23575         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23576         rather than a HOST_WIDE_INT.
23577         * config/aarch64/aarch64.h (aarch64_frame): Protect with
23578         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
23579         hard_fp_offset, frame_size, initial_adjust, callee_offset and
23580         final_offset from HOST_WIDE_INT to poly_int64.
23581         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23582         to_constant when getting the number of units in an Advanced SIMD
23583         mode.
23584         (aarch64_builtin_vectorized_function): Check for a constant number
23585         of units.
23586         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23587         GET_MODE_SIZE.
23588         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23589         attribute instead of GET_MODE_NUNITS.
23590         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23591         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23592         GET_MODE_SIZE for fixed-size registers.
23593         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23594         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23595         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23596         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23597         (aarch64_print_operand, aarch64_print_address_internal)
23598         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23599         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23600         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23601         Handle polynomial GET_MODE_SIZE.
23602         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
23603         wider than SImode without modification.
23604         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23605         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23606         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23607         passing and returning SVE modes.
23608         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23609         rather than GEN_INT.
23610         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23611         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23612         (aarch64_allocate_and_probe_stack_space): Likewise.
23613         (aarch64_layout_frame): Cope with polynomial offsets.
23614         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23615         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
23616         polynomial offsets.
23617         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23618         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23619         poly_int64 rather than a HOST_WIDE_INT.
23620         (aarch64_get_separate_components, aarch64_process_components)
23621         (aarch64_expand_prologue, aarch64_expand_epilogue)
23622         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23623         (aarch64_anchor_offset): New function, split out from...
23624         (aarch64_legitimize_address): ...here.
23625         (aarch64_builtin_vectorization_cost): Handle polynomial
23626         TYPE_VECTOR_SUBPARTS.
23627         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23628         GET_MODE_NUNITS.
23629         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23630         number of elements from the PARALLEL rather than the mode.
23631         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23632         rather than GET_MODE_BITSIZE.
23633         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23634         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23635         (aarch64_expand_vec_perm_const_1): Handle polynomial
23636         d->perm.length () and d->perm elements.
23637         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
23638         Apply to_constant to d->perm elements.
23639         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23640         polynomial CONST_VECTOR_NUNITS.
23641         (aarch64_move_pointer): Take amount as a poly_int64 rather
23642         than an int.
23643         (aarch64_progress_pointer): Avoid temporary variable.
23644         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23645         the mode attribute instead of GET_MODE.
23647 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23648             Alan Hayward  <alan.hayward@arm.com>
23649             David Sherwood  <david.sherwood@arm.com>
23651         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23652         x exists before using it.
23653         (aarch64_add_constant_internal): Rename to...
23654         (aarch64_add_offset_1): ...this.  Replace regnum with separate
23655         src and dest rtxes.  Handle the case in which they're different,
23656         including when the offset is zero.  Replace scratchreg with an rtx.
23657         Use 2 additions if there is no spare register into which we can
23658         move a 16-bit constant.
23659         (aarch64_add_constant): Delete.
23660         (aarch64_add_offset): Replace reg with separate src and dest
23661         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
23662         Use aarch64_add_offset_1.
23663         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23664         an rtx rather than an int.  Take the delta as a poly_int64
23665         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
23666         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23667         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23668         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23669         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23670         and aarch64_add_sp.
23671         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23672         aarch64_add_constant.
23674 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23676         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23677         Use scalar_float_mode.
23679 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23681         * config/aarch64/aarch64-simd.md
23682         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23683         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23684         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23685         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23686         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23687         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23688         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23689         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23690         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23691         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23693 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23695         PR target/83514
23696         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23697         targ_options->x_arm_arch_string is non NULL.
23699 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
23701         * config/aarch64/aarch64.h
23702         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
23704 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
23706         PR target/82096
23707         * expmed.c (emit_store_flag_force): Swap if const op0
23708         and change VOIDmode to mode of op0.
23710 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23712         PR rtl-optimization/83761
23713         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23714         than bytes to mode_for_size.
23716 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23718         PR middle-end/83189
23719         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23720         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23721         profile.
23723 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23725         PR middle-end/83575
23726         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23727         when in layout mode.
23728         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23729         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23730         partition fixup.
23732 2018-01-10  Michael Collison  <michael.collison@arm.com>
23734         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23735         * config/aarch64/aarch64-option-extension.def: Add
23736         AARCH64_OPT_EXTENSION of 'fp16fml'.
23737         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23738         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23739         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23740         * config/aarch64/constraints.md (Ui7): New constraint.
23741         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23742         (VFMLA_SEL_W): Ditto.
23743         (f16quad): Ditto.
23744         (f16mac1): Ditto.
23745         (VFMLA16_LOW): New int iterator.
23746         (VFMLA16_HIGH): Ditto.
23747         (UNSPEC_FMLAL): New unspec.
23748         (UNSPEC_FMLSL): Ditto.
23749         (UNSPEC_FMLAL2): Ditto.
23750         (UNSPEC_FMLSL2): Ditto.
23751         (f16mac): New code attribute.
23752         * config/aarch64/aarch64-simd-builtins.def
23753         (aarch64_fmlal_lowv2sf): Ditto.
23754         (aarch64_fmlsl_lowv2sf): Ditto.
23755         (aarch64_fmlalq_lowv4sf): Ditto.
23756         (aarch64_fmlslq_lowv4sf): Ditto.
23757         (aarch64_fmlal_highv2sf): Ditto.
23758         (aarch64_fmlsl_highv2sf): Ditto.
23759         (aarch64_fmlalq_highv4sf): Ditto.
23760         (aarch64_fmlslq_highv4sf): Ditto.
23761         (aarch64_fmlal_lane_lowv2sf): Ditto.
23762         (aarch64_fmlsl_lane_lowv2sf): Ditto.
23763         (aarch64_fmlal_laneq_lowv2sf): Ditto.
23764         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23765         (aarch64_fmlalq_lane_lowv4sf): Ditto.
23766         (aarch64_fmlsl_lane_lowv4sf): Ditto.
23767         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23768         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23769         (aarch64_fmlal_lane_highv2sf): Ditto.
23770         (aarch64_fmlsl_lane_highv2sf): Ditto.
23771         (aarch64_fmlal_laneq_highv2sf): Ditto.
23772         (aarch64_fmlsl_laneq_highv2sf): Ditto.
23773         (aarch64_fmlalq_lane_highv4sf): Ditto.
23774         (aarch64_fmlsl_lane_highv4sf): Ditto.
23775         (aarch64_fmlalq_laneq_highv4sf): Ditto.
23776         (aarch64_fmlsl_laneq_highv4sf): Ditto.
23777         * config/aarch64/aarch64-simd.md:
23778         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23779         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23780         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23781         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23782         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23783         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23784         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23785         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23786         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23787         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23788         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23789         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23790         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23791         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23792         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23793         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23794         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23795         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23796         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23797         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23798         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23799         (vfmlsl_low_u32): Ditto.
23800         (vfmlalq_low_u32): Ditto.
23801         (vfmlslq_low_u32): Ditto.
23802         (vfmlal_high_u32): Ditto.
23803         (vfmlsl_high_u32): Ditto.
23804         (vfmlalq_high_u32): Ditto.
23805         (vfmlslq_high_u32): Ditto.
23806         (vfmlal_lane_low_u32): Ditto.
23807         (vfmlsl_lane_low_u32): Ditto.
23808         (vfmlal_laneq_low_u32): Ditto.
23809         (vfmlsl_laneq_low_u32): Ditto.
23810         (vfmlalq_lane_low_u32): Ditto.
23811         (vfmlslq_lane_low_u32): Ditto.
23812         (vfmlalq_laneq_low_u32): Ditto.
23813         (vfmlslq_laneq_low_u32): Ditto.
23814         (vfmlal_lane_high_u32): Ditto.
23815         (vfmlsl_lane_high_u32): Ditto.
23816         (vfmlal_laneq_high_u32): Ditto.
23817         (vfmlsl_laneq_high_u32): Ditto.
23818         (vfmlalq_lane_high_u32): Ditto.
23819         (vfmlslq_lane_high_u32): Ditto.
23820         (vfmlalq_laneq_high_u32): Ditto.
23821         (vfmlslq_laneq_high_u32): Ditto.
23822         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23823         (AARCH64_FL_FOR_ARCH8_4): New.
23824         (AARCH64_ISA_F16FML): New ISA flag.
23825         (TARGET_F16FML): New feature flag for fp16fml.
23826         (doc/invoke.texi): Document new fp16fml option.
23828 2018-01-10  Michael Collison  <michael.collison@arm.com>
23830         * config/aarch64/aarch64-builtins.c:
23831         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23832         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23833         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23834         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23835         (AARCH64_ISA_SHA3): New ISA flag.
23836         (TARGET_SHA3): New feature flag for sha3.
23837         * config/aarch64/iterators.md (sha512_op): New int attribute.
23838         (CRYPTO_SHA512): New int iterator.
23839         (UNSPEC_SHA512H): New unspec.
23840         (UNSPEC_SHA512H2): Ditto.
23841         (UNSPEC_SHA512SU0): Ditto.
23842         (UNSPEC_SHA512SU1): Ditto.
23843         * config/aarch64/aarch64-simd-builtins.def
23844         (aarch64_crypto_sha512hqv2di): New builtin.
23845         (aarch64_crypto_sha512h2qv2di): Ditto.
23846         (aarch64_crypto_sha512su0qv2di): Ditto.
23847         (aarch64_crypto_sha512su1qv2di): Ditto.
23848         (aarch64_eor3qv8hi): Ditto.
23849         (aarch64_rax1qv2di): Ditto.
23850         (aarch64_xarqv2di): Ditto.
23851         (aarch64_bcaxqv8hi): Ditto.
23852         * config/aarch64/aarch64-simd.md:
23853         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23854         (aarch64_crypto_sha512su0qv2di): Ditto.
23855         (aarch64_crypto_sha512su1qv2di): Ditto.
23856         (aarch64_eor3qv8hi): Ditto.
23857         (aarch64_rax1qv2di): Ditto.
23858         (aarch64_xarqv2di): Ditto.
23859         (aarch64_bcaxqv8hi): Ditto.
23860         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23861         (vsha512h2q_u64): Ditto.
23862         (vsha512su0q_u64): Ditto.
23863         (vsha512su1q_u64): Ditto.
23864         (veor3q_u16): Ditto.
23865         (vrax1q_u64): Ditto.
23866         (vxarq_u64): Ditto.
23867         (vbcaxq_u16): Ditto.
23868         * config/arm/types.md (crypto_sha512): New type attribute.
23869         (crypto_sha3): Ditto.
23870         (doc/invoke.texi): Document new sha3 option.
23872 2018-01-10  Michael Collison  <michael.collison@arm.com>
23874         * config/aarch64/aarch64-builtins.c:
23875         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23876         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23877         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23878         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23879         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23880         (AARCH64_ISA_SM4): New ISA flag.
23881         (TARGET_SM4): New feature flag for sm4.
23882         * config/aarch64/aarch64-simd-builtins.def
23883         (aarch64_sm3ss1qv4si): Ditto.
23884         (aarch64_sm3tt1aq4si): Ditto.
23885         (aarch64_sm3tt1bq4si): Ditto.
23886         (aarch64_sm3tt2aq4si): Ditto.
23887         (aarch64_sm3tt2bq4si): Ditto.
23888         (aarch64_sm3partw1qv4si): Ditto.
23889         (aarch64_sm3partw2qv4si): Ditto.
23890         (aarch64_sm4eqv4si): Ditto.
23891         (aarch64_sm4ekeyqv4si): Ditto.
23892         * config/aarch64/aarch64-simd.md:
23893         (aarch64_sm3ss1qv4si): Ditto.
23894         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23895         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23896         (aarch64_sm4eqv4si): Ditto.
23897         (aarch64_sm4ekeyqv4si): Ditto.
23898         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23899         (sm3part_op): Ditto.
23900         (CRYPTO_SM3TT): Ditto.
23901         (CRYPTO_SM3PART): Ditto.
23902         (UNSPEC_SM3SS1): New unspec.
23903         (UNSPEC_SM3TT1A): Ditto.
23904         (UNSPEC_SM3TT1B): Ditto.
23905         (UNSPEC_SM3TT2A): Ditto.
23906         (UNSPEC_SM3TT2B): Ditto.
23907         (UNSPEC_SM3PARTW1): Ditto.
23908         (UNSPEC_SM3PARTW2): Ditto.
23909         (UNSPEC_SM4E): Ditto.
23910         (UNSPEC_SM4EKEY): Ditto.
23911         * config/aarch64/constraints.md (Ui2): New constraint.
23912         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23913         * config/arm/types.md (crypto_sm3): New type attribute.
23914         (crypto_sm4): Ditto.
23915         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23916         (vsm3tt1aq_u32): Ditto.
23917         (vsm3tt1bq_u32): Ditto.
23918         (vsm3tt2aq_u32): Ditto.
23919         (vsm3tt2bq_u32): Ditto.
23920         (vsm3partw1q_u32): Ditto.
23921         (vsm3partw2q_u32): Ditto.
23922         (vsm4eq_u32): Ditto.
23923         (vsm4ekeyq_u32): Ditto.
23924         (doc/invoke.texi): Document new sm4 option.
23926 2018-01-10  Michael Collison  <michael.collison@arm.com>
23928         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23929         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23930         (AARCH64_FL_FOR_ARCH8_4): New.
23931         (AARCH64_FL_V8_4): New flag.
23932         (doc/invoke.texi): Document new armv8.4-a option.
23934 2018-01-10  Michael Collison  <michael.collison@arm.com>
23936         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23937         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23938         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23939         * config/aarch64/aarch64-option-extension.def: Add
23940         AARCH64_OPT_EXTENSION of 'sha2'.
23941         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23942         (crypto): Disable sha2 and aes if crypto disabled.
23943         (crypto): Enable aes and sha2 if enabled.
23944         (simd): Disable sha2 and aes if simd disabled.
23945         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23946         New flags.
23947         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23948         (TARGET_SHA2): New feature flag for sha2.
23949         (TARGET_AES): New feature flag for aes.
23950         * config/aarch64/aarch64-simd.md:
23951         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23952         conditional on TARGET_AES.
23953         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23954         (aarch64_crypto_sha1hsi): Make pattern conditional
23955         on TARGET_SHA2.
23956         (aarch64_crypto_sha1hv4si): Ditto.
23957         (aarch64_be_crypto_sha1hv4si): Ditto.
23958         (aarch64_crypto_sha1su1v4si): Ditto.
23959         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23960         (aarch64_crypto_sha1su0v4si): Ditto.
23961         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23962         (aarch64_crypto_sha256su0v4si): Ditto.
23963         (aarch64_crypto_sha256su1v4si): Ditto.
23964         (doc/invoke.texi): Document new aes and sha2 options.
23966 2018-01-10  Martin Sebor  <msebor@redhat.com>
23968         PR tree-optimization/83781
23969         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23970         as string arrays.
23972 2018-01-11  Martin Sebor  <msebor@gmail.com>
23973             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23975         PR tree-optimization/83501
23976         PR tree-optimization/81703
23978         * tree-ssa-strlen.c (get_string_cst): Rename...
23979         (get_string_len): ...to this.  Handle global constants.
23980         (handle_char_store): Adjust.
23982 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
23983             Jim Wilson  <jimw@sifive.com>
23985         * config/riscv/riscv-protos.h (riscv_output_return): New.
23986         * config/riscv/riscv.c (struct machine_function): New naked_p field.
23987         (riscv_attribute_table, riscv_output_return),
23988         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23989         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23990         (riscv_compute_frame_info): Only compute frame->mask if not a naked
23991         function.
23992         (riscv_expand_prologue): Add early return for naked function.
23993         (riscv_expand_epilogue): Likewise.
23994         (riscv_function_ok_for_sibcall): Return false for naked function.
23995         (riscv_set_current_function): New.
23996         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23997         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23998         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23999         * doc/extend.texi (RISC-V Function Attributes): New.
24001 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
24003         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
24004         check for 128-bit long double before checking TCmode.
24005         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
24006         128-bit long doubles before checking TFmode or TCmode.
24007         (FLOAT128_IBM_P): Likewise.
24009 2018-01-10  Martin Sebor  <msebor@redhat.com>
24011         PR tree-optimization/83671
24012         * builtins.c (c_strlen): Unconditionally return zero for the empty
24013         string.
24014         Use -Warray-bounds for warnings.
24015         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
24016         for non-constant array indices with COMPONENT_REF, arrays of
24017         arrays, and pointers to arrays.
24018         (gimple_fold_builtin_strlen): Determine and set length range for
24019         non-constant character arrays.
24021 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
24023         PR middle-end/81897
24024         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
24025         empty blocks.
24027 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
24029         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
24031 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
24033         PR target/83399
24034         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
24035         VECTOR_MEM_ALTIVEC_OR_VSX_P.
24036         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
24037         indexed_or_indirect_operand predicate.
24038         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
24039         (*vsx_le_perm_load_v8hi): Likewise.
24040         (*vsx_le_perm_load_v16qi): Likewise.
24041         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
24042         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
24043         (*vsx_le_perm_store_v8hi): Likewise.
24044         (*vsx_le_perm_store_v16qi): Likewise.
24045         (eight unnamed splitters): Likewise.
24047 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
24049         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
24050         * config/rs6000/emmintrin.h: Likewise.
24051         * config/rs6000/mmintrin.h: Likewise.
24052         * config/rs6000/xmmintrin.h: Likewise.
24054 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
24056         PR c++/43486
24057         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
24058         "public_flag".
24059         * tree.c (tree_nop_conversion): Return true for location wrapper
24060         nodes.
24061         (maybe_wrap_with_location): New function.
24062         (selftest::check_strip_nops): New function.
24063         (selftest::test_location_wrappers): New function.
24064         (selftest::tree_c_tests): Call it.
24065         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
24066         (maybe_wrap_with_location): New decl.
24067         (EXPR_LOCATION_WRAPPER_P): New macro.
24068         (location_wrapper_p): New inline function.
24069         (tree_strip_any_location_wrapper): New inline function.
24071 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
24073         PR target/83735
24074         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
24075         stack_realign_offset for the largest alignment of stack slot
24076         actually used.
24077         (ix86_find_max_used_stack_alignment): New function.
24078         (ix86_finalize_stack_frame_flags): Use it.  Set
24079         max_used_stack_alignment if we don't realign stack.
24080         * config/i386/i386.h (machine_function): Add
24081         max_used_stack_alignment.
24083 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
24085         * config/arm/arm.opt (-mbranch-cost): New option.
24086         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
24087         account.
24089 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
24091         PR target/83629
24092         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
24093         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
24095 2018-01-10  Richard Biener  <rguenther@suse.de>
24097         PR debug/83765
24098         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
24099         early out so it also covers the case where we have a non-NULL
24100         origin.
24102 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
24104         PR tree-optimization/83753
24105         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
24106         for non-strided grouped accesses if the number of elements is 1.
24108 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
24110         PR target/81616
24111         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
24112         * i386.h (TARGET_USE_GATHER): Define.
24113         * x86-tune.def (X86_TUNE_USE_GATHER): New.
24115 2018-01-10  Martin Liska  <mliska@suse.cz>
24117         PR bootstrap/82831
24118         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
24119         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
24120         partitioning.
24121         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
24122         CLEANUP_NO_PARTITIONING is not set.
24124 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
24126         * doc/rtl.texi: Remove documentation of (const ...) wrappers
24127         for vectors, as a partial revert of r254296.
24128         * rtl.h (const_vec_p): Delete.
24129         (const_vec_duplicate_p): Don't test for vector CONSTs.
24130         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
24131         * expmed.c (make_tree): Likewise.
24133         Revert:
24134         * common.md (E, F): Use CONSTANT_P instead of checking for
24135         CONST_VECTOR.
24136         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
24137         checking for CONST_VECTOR.
24139 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
24141         PR middle-end/83575
24142         * predict.c (force_edge_cold): Handle in more sane way edges
24143         with no prediction.
24145 2018-01-09  Carl Love  <cel@us.ibm.com>
24147         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
24148         V4SI, V4SF types.
24149         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
24150         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
24151         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
24152         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
24153         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
24154         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
24155         * config/rs6000/rs6000-protos.h: Add extern defition for
24156         rs6000_generate_float2_double_code.
24157         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
24158         function.
24159         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
24160         (float2_v2df): Add define_expand.
24162 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
24164         PR target/83628
24165         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
24166         op_mode in the force_to_mode call.
24168 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
24170         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
24171         instead of checking each element individually.
24172         (aarch64_evpc_uzp): Likewise.
24173         (aarch64_evpc_zip): Likewise.
24174         (aarch64_evpc_ext): Likewise.
24175         (aarch64_evpc_rev): Likewise.
24176         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24177         instead of checking each element individually.  Return true without
24178         generating rtl if
24179         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
24180         whether all selected elements come from the same input, instead of
24181         checking each element individually.  Remove calls to gen_rtx_REG,
24182         start_sequence and end_sequence and instead assert that no rtl is
24183         generated.
24185 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
24187         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24188         order of HIGH and CONST checks.
24190 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
24192         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24193         if the destination isn't an SSA_NAME.
24195 2018-01-09  Richard Biener  <rguenther@suse.de>
24197         PR tree-optimization/83668
24198         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24199         move prologue...
24200         (canonicalize_loop_form): ... here, renamed from ...
24201         (canonicalize_loop_closed_ssa_form): ... this and amended to
24202         swap successor edges for loop exit blocks to make us use
24203         the RPO order we need for initial schedule generation.
24205 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
24207         PR tree-optimization/64811
24208         * match.pd: When optimizing comparisons with Inf, avoid
24209         introducing or losing exceptions from comparisons with NaN.
24211 2018-01-09  Martin Liska  <mliska@suse.cz>
24213         PR sanitizer/82517
24214         * asan.c (shadow_mem_size): Add gcc_assert.
24216 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
24218         Don't save registers in main().
24220         PR target/83738
24221         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24222         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24223         * config/avr/avr.c (avr_set_current_function): Don't error if
24224         naked, OS_task or OS_main are specified at the same time.
24225         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24226         OS_main.
24227         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24228         attribute.
24229         * common/config/avr/avr-common.c (avr_option_optimization_table):
24230         Switch on -mmain-is-OS_task for optimizing compilations.
24232 2018-01-09  Richard Biener  <rguenther@suse.de>
24234         PR tree-optimization/83572
24235         * graphite.c: Include cfganal.h.
24236         (graphite_transform_loops): Connect infinite loops to exit
24237         and remove fake edges at the end.
24239 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
24241         * ipa-inline.c (edge_badness): Revert accidental checkin.
24243 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
24245         PR ipa/80763
24246         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24247         symbols; not inline clones.
24249 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
24251         PR target/83507
24252         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24253         hard registers.  Formatting fixes.
24255         PR preprocessor/83722
24256         * gcc.c (try_generate_repro): Pass
24257         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24258         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24259         do_report_bug.
24261 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
24262             Kito Cheng  <kito.cheng@gmail.com>
24264         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24265         (riscv_leaf_function_p): Delete.
24266         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24268 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24270         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24271         function.
24272         (do_ifelse): New function.
24273         (do_isel): New function.
24274         (do_sub3): New function.
24275         (do_add3): New function.
24276         (do_load_mask_compare): New function.
24277         (do_overlap_load_compare): New function.
24278         (expand_compare_loop): New function.
24279         (expand_block_compare): Call expand_compare_loop() when appropriate.
24280         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24281         option description.
24282         (-mblock-compare-inline-loop-limit): New option.
24284 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24286         PR target/83677
24287         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24288         Reverse order of second and third operands in first alternative.
24289         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24290         of first and second elements in UNSPEC_VPERMR vector.
24291         (altivec_expand_vec_perm_le): Likewise.
24293 2018-01-08  Jeff Law  <law@redhat.com>
24295         PR rtl-optimizatin/81308
24296         * tree-switch-conversion.c (cfg_altered): New file scoped static.
24297         (process_switch): If group_case_labels makes a change, then set
24298         cfg_altered.
24299         (pass_convert_switch::execute): If a switch is converted, then
24300         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
24302         PR rtl-optimization/81308
24303         * recog.c (split_all_insns): Conditionally cleanup the CFG after
24304         splitting insns.
24306 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
24308         PR target/83663 - Revert r255946
24309         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24310         generation for cases where splatting a value is not useful.
24311         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24312         across a vec_duplicate and a paradoxical subreg forming a vector
24313         mode to a vec_concat.
24315 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24317         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24318         -march=armv8.3-a variants.
24319         * config/arm/t-multilib: Likewise.
24320         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
24322 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24324         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24325         to generate rtl.
24326         (cceq_ior_compare_complement): Give it a name so I can use it, and
24327         change boolean_or_operator predicate to boolean_operator so it can
24328         be used to generate a crand.
24329         (eqne): New code iterator.
24330         (bd/bd_neg): New code_attrs.
24331         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24332         a single define_insn.
24333         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24334         decrement (bdnzt/bdnzf/bdzt/bdzf).
24335         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24336         with the new names of the branch decrement patterns, and added the
24337         names of the branch decrement conditional patterns.
24339 2018-01-08  Richard Biener  <rguenther@suse.de>
24341         PR tree-optimization/83563
24342         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24343         cache.
24345 2018-01-08  Richard Biener  <rguenther@suse.de>
24347         PR middle-end/83713
24348         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24350 2018-01-08  Richard Biener  <rguenther@suse.de>
24352         PR tree-optimization/83685
24353         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24354         references to abnormals.
24356 2018-01-08  Richard Biener  <rguenther@suse.de>
24358         PR lto/83719
24359         * dwarf2out.c (output_indirect_strings): Handle empty
24360         skeleton_debug_str_hash.
24361         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24363 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
24365         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24366         (emit_store_direct): Likewise.
24367         (arc_trampoline_adjust_address): Likewise.
24368         (arc_asm_trampoline_template): New function.
24369         (arc_initialize_trampoline): Use asm_trampoline_template.
24370         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24371         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24372         * config/arc/arc.md (flush_icache): Delete pattern.
24374 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
24376         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24377         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24378         munaligned-access.
24380 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24382         PR target/83681
24383         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
24384         by not USED_FOR_TARGET.
24385         (make_pass_resolve_sw_modes): Likewise.
24387 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24389         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
24390         USED_FOR_TARGET.
24392 2018-01-08  Richard Biener  <rguenther@suse.de>
24394         PR middle-end/83580
24395         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
24397 2018-01-08  Richard Biener  <rguenther@suse.de>
24399         PR middle-end/83517
24400         * match.pd ((t * 2) / 2) -> t): Add missing :c.
24402 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
24404         PR middle-end/81897
24405         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
24406         basic blocks with a small number of successors.
24407         (convert_control_dep_chain_into_preds): Improve handling of
24408         forwarder blocks.
24409         (dump_predicates): Split apart into...
24410         (dump_pred_chain): ...here...
24411         (dump_pred_info): ...and here.
24412         (can_one_predicate_be_invalidated_p): Add debugging printfs.
24413         (can_chain_union_be_invalidated_p): Improve check for invalidation
24414         of paths.
24415         (uninit_uses_cannot_happen): Avoid unnecessary if
24416         convert_control_dep_chain_into_preds yielded nothing.
24418 2018-01-06  Martin Sebor  <msebor@redhat.com>
24420         PR tree-optimization/83640
24421         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
24422         subtracting negative offset from size.
24423         (builtin_access::overlap): Adjust offset bounds of the access to fall
24424         within the size of the object if possible.
24426 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
24428         PR rtl-optimization/83699
24429         * expmed.c (extract_bit_field_1): Restrict the vector usage of
24430         extract_bit_field_as_subreg to cases in which the extracted
24431         value is also a vector.
24433         * lra-constraints.c (process_alt_operands): Test for the equivalence
24434         substitutions when detecting a possible reload cycle.
24436 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
24438         PR debug/83480
24439         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
24440         by default if flag_selective_schedling{,2}.  Formatting fixes.
24442         PR rtl-optimization/83682
24443         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
24444         if it has non-VECTOR_MODE element mode.
24445         (vec_duplicate_p): Likewise.
24447         PR middle-end/83694
24448         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24449         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24451 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
24453         PR target/83604
24454         * config/i386/i386-builtin.def
24455         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24456         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24457         Require also OPTION_MASK_ISA_AVX512F in addition to
24458         OPTION_MASK_ISA_GFNI.
24459         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24460         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24461         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24462         to OPTION_MASK_ISA_GFNI.
24463         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24464         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24465         OPTION_MASK_ISA_AVX512BW.
24466         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24467         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24468         addition to OPTION_MASK_ISA_GFNI.
24469         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24470         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24471         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24472         to OPTION_MASK_ISA_GFNI.
24473         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24474         a requirement for all ISAs rather than any of them with a few
24475         exceptions.
24476         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24477         processing.
24478         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24479         bitmasks to be enabled with 3 exceptions, instead of requiring any
24480         enabled ISA with lots of exceptions.
24481         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24482         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24483         Change avx512bw in isa attribute to avx512f.
24484         * config/i386/sgxintrin.h: Add license boilerplate.
24485         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
24486         to __AVX512F__ and __AVX512VL to __AVX512VL__.
24487         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24488         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24489         defined.
24490         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24491         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24492         temporarily sse2 rather than sse if not enabled already.
24494         PR target/83604
24495         * config/i386/sse.md (VI248_VLBW): Rename to ...
24496         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
24497         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24498         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24499         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24500         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24501         mode iterator instead of VI248_VLBW.
24503 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
24505         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24506         (record_modified): Skip clobbers; add debug output.
24507         (param_change_prob): Use sreal frequencies.
24509 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
24511         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24512         punt for user-aligned variables.
24514 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
24516         * tree-chrec.c (chrec_contains_symbols): Return true for
24517         POLY_INT_CST.
24519 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
24521         PR target/82439
24522         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24523         of (x|y) == x for BICS pattern.
24525 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
24527         PR tree-optimization/83605
24528         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24529         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24530         can throw.
24532 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24534         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24535         * config/epiphany/rtems.h: New file.
24537 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24538             Uros Bizjak  <ubizjak@gmail.com>
24540         PR target/83554
24541         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24542         QIreg_operand instead of register_operand predicate.
24543         * config/i386/i386.c (ix86_rop_should_change_byte_p,
24544         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24545         comments instead of -fmitigate[-_]rop.
24547 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24549         PR bootstrap/81926
24550         * cgraphunit.c (symbol_table::compile): Switch to text_section
24551         before calling assembly_start debug hook.
24552         * run-rtl-passes.c (run_rtl_passes): Likewise.
24553         Include output.h.
24555 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24557         * tree-vrp.c (extract_range_from_binary_expr_1): Check
24558         range_int_cst_p rather than !symbolic_range_p before calling
24559         extract_range_from_multiplicative_op_1.
24561 2018-01-04  Jeff Law  <law@redhat.com>
24563         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24564         redundant test in assertion.
24566 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24568         * doc/rtl.texi: Document machine_mode wrapper classes.
24570 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24572         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24573         using tree_to_uhwi.
24575 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24577         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24578         the VEC_PERM_EXPR fold to fail.
24580 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24582         PR debug/83585
24583         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24584         to switched_sections.
24586 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24588         PR target/83680
24589         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24590         test for d.testing.
24592 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
24594         PR target/83387
24595         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24596         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24598 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24600         PR debug/83666
24601         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24602         is BLKmode and bitpos not zero or mode change is needed.
24604 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24606         PR target/83675
24607         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24608         TARGET_VIS2.
24610 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
24612         PR target/83628
24613         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24614         instead of MULT rtx.  Update all corresponding splitters.
24615         (*saddl_se): Ditto.
24616         (*ssub<modesuffix>): Ditto.
24617         (*ssubl_se): Ditto.
24618         (*cmp_sadd_di): Update split patterns.
24619         (*cmp_sadd_si): Ditto.
24620         (*cmp_sadd_sidi): Ditto.
24621         (*cmp_ssub_di): Ditto.
24622         (*cmp_ssub_si): Ditto.
24623         (*cmp_ssub_sidi): Ditto.
24624         * config/alpha/predicates.md (const23_operand): New predicate.
24625         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24626         Look for ASHIFT, not MULT inner operand.
24627         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24629 2018-01-04  Martin Liska  <mliska@suse.cz>
24631         PR gcov-profile/83669
24632         * gcov.c (output_intermediate_file): Add version to intermediate
24633         gcov file.
24634         * doc/gcov.texi: Document new field 'version' in intermediate
24635         file format. Fix location of '-k' option of gcov command.
24637 2018-01-04  Martin Liska  <mliska@suse.cz>
24639         PR ipa/82352
24640         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24642 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24644         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24646 2018-01-03  Martin Sebor  <msebor@redhat.com>
24648         PR tree-optimization/83655
24649         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24650         checking calls with invalid arguments.
24652 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24654         * tree-vect-stmts.c (vect_get_store_rhs): New function.
24655         (vectorizable_mask_load_store): Delete.
24656         (vectorizable_call): Return false for masked loads and stores.
24657         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
24658         instead of gimple_assign_rhs1.
24659         (vectorizable_load): Handle IFN_MASK_LOAD.
24660         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24662 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24664         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24665         split out from..,
24666         (vectorizable_mask_load_store): ...here.
24667         (vectorizable_load): ...and here.
24669 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24671         * tree-vect-stmts.c (vect_build_all_ones_mask)
24672         (vect_build_zero_merge_argument): New functions, split out from...
24673         (vectorizable_load): ...here.
24675 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24677         * tree-vect-stmts.c (vect_check_store_rhs): New function,
24678         split out from...
24679         (vectorizable_mask_load_store): ...here.
24680         (vectorizable_store): ...and here.
24682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24684         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24685         split out from...
24686         (vectorizable_mask_load_store): ...here.
24688 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24690         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24691         (vect_model_store_cost): Take a vec_load_store_type instead of a
24692         vect_def_type.
24693         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24694         (vect_model_store_cost): Take a vec_load_store_type instead of a
24695         vect_def_type.
24696         (vectorizable_mask_load_store): Update accordingly.
24697         (vectorizable_store): Likewise.
24698         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24700 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24702         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24703         IFN_MASK_LOAD calls here rather than...
24704         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24706 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24707             Alan Hayward  <alan.hayward@arm.com>
24708             David Sherwood  <david.sherwood@arm.com>
24710         * expmed.c (extract_bit_field_1): For vector extracts,
24711         fall back to extract_bit_field_as_subreg if vec_extract
24712         isn't available.
24714 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24715             Alan Hayward  <alan.hayward@arm.com>
24716             David Sherwood  <david.sherwood@arm.com>
24718         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24719         they are variable or constant sized.
24720         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24721         slots for constant-sized data.
24723 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24724             Alan Hayward  <alan.hayward@arm.com>
24725             David Sherwood  <david.sherwood@arm.com>
24727         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24728         handling COND_EXPRs with boolean comparisons, try to find a better
24729         basis for the mask type than the boolean itself.
24731 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24733         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24734         is calculated and how it can be overridden.
24735         * genmodes.c (max_bitsize_mode_any_mode): New variable.
24736         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24737         if defined.
24738         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24739         if nonzero.
24741 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24742             Alan Hayward  <alan.hayward@arm.com>
24743             David Sherwood  <david.sherwood@arm.com>
24745         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24746         Remove the mode argument.
24747         (aarch64_simd_valid_immediate): Remove the mode and inverse
24748         arguments.
24749         * config/aarch64/iterators.md (bitsize): New iterator.
24750         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24751         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24752         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24753         aarch64_simd_valid_immediate.
24754         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24755         (aarch64_reg_or_bic_imm): Likewise.
24756         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24757         with an insn_type enum and msl with a modifier_type enum.
24758         Replace element_width with a scalar_mode.  Change the shift
24759         to unsigned int.  Add constructors for scalar_float_mode and
24760         scalar_int_mode elements.
24761         (aarch64_vect_float_const_representable_p): Delete.
24762         (aarch64_can_const_movi_rtx_p)
24763         (aarch64_simd_scalar_immediate_valid_for_move)
24764         (aarch64_simd_make_constant): Update call to
24765         aarch64_simd_valid_immediate.
24766         (aarch64_advsimd_valid_immediate_hs): New function.
24767         (aarch64_advsimd_valid_immediate): Likewise.
24768         (aarch64_simd_valid_immediate): Remove mode and inverse
24769         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
24770         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24771         and aarch64_float_const_representable_p on the result.
24772         (aarch64_output_simd_mov_immediate): Remove mode argument.
24773         Update call to aarch64_simd_valid_immediate and use of
24774         simd_immediate_info.
24775         (aarch64_output_scalar_simd_mov_immediate): Update call
24776         accordingly.
24778 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24779             Alan Hayward  <alan.hayward@arm.com>
24780             David Sherwood  <david.sherwood@arm.com>
24782         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24783         (mode_nunits): Likewise CONST_MODE_NUNITS.
24784         * machmode.def (ADJUST_NUNITS): Document.
24785         * genmodes.c (mode_data::need_nunits_adj): New field.
24786         (blank_mode): Update accordingly.
24787         (adj_nunits): New variable.
24788         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24789         parameter.
24790         (emit_mode_size_inline): Set need_bytesize_adj for all modes
24791         listed in adj_nunits.
24792         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24793         listed in adj_nunits.  Don't emit case statements for such modes.
24794         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24795         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
24796         nothing if adj_nunits is nonnull.
24797         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24798         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24799         (emit_mode_fbit): Update use of print_maybe_const_decl.
24800         (emit_move_size): Likewise.  Treat the array as non-const
24801         if adj_nunits.
24802         (emit_mode_adjustments): Handle adj_nunits.
24804 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24806         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24807         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24808         (VECTOR_MODES): Use it.
24809         (make_vector_modes): Take the prefix as an argument.
24811 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24812             Alan Hayward  <alan.hayward@arm.com>
24813             David Sherwood  <david.sherwood@arm.com>
24815         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24816         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24817         for MODE_VECTOR_BOOL.
24818         * machmode.def (VECTOR_BOOL_MODE): Document.
24819         * genmodes.c (VECTOR_BOOL_MODE): New macro.
24820         (make_vector_bool_mode): New function.
24821         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24822         MODE_VECTOR_BOOL.
24823         * lto-streamer-in.c (lto_input_mode_table): Likewise.
24824         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24825         Likewise.
24826         * stor-layout.c (int_mode_for_mode): Likewise.
24827         * tree.c (build_vector_type_for_mode): Likewise.
24828         * varasm.c (output_constant_pool_2): Likewise.
24829         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24830         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
24831         for MODE_VECTOR_BOOL.
24832         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24833         of mode class checks.
24834         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24835         instead of a list of mode class checks.
24836         (expand_vector_scalar_condition): Likewise.
24837         (type_for_widest_vector_mode): Handle BImode as an inner mode.
24839 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24840             Alan Hayward  <alan.hayward@arm.com>
24841             David Sherwood  <david.sherwood@arm.com>
24843         * machmode.h (mode_size): Change from unsigned short to
24844         poly_uint16_pod.
24845         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24846         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24847         or if measurement_type is not polynomial.
24848         (fixed_size_mode::includes_p): Check for constant-sized modes.
24849         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24850         return a poly_uint16 rather than an unsigned short.
24851         (emit_mode_size): Change the type of mode_size from unsigned short
24852         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
24853         (emit_mode_adjustments): Cope with polynomial vector sizes.
24854         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24855         for GET_MODE_SIZE.
24856         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24857         for GET_MODE_SIZE.
24858         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24859         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24860         * caller-save.c (setup_save_areas): Likewise.
24861         (replace_reg_with_saved_mem): Likewise.
24862         * calls.c (emit_library_call_value_1): Likewise.
24863         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24864         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24865         (gen_lowpart_for_combine): Likewise.
24866         * convert.c (convert_to_integer_1): Likewise.
24867         * cse.c (equiv_constant, cse_insn): Likewise.
24868         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24869         (cselib_subst_to_values): Likewise.
24870         * dce.c (word_dce_process_block): Likewise.
24871         * df-problems.c (df_word_lr_mark_ref): Likewise.
24872         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24873         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24874         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24875         (rtl_for_decl_location): Likewise.
24876         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24877         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24878         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24879         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24880         (expand_expr_real_1): Likewise.
24881         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24882         (pad_below): Likewise.
24883         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24884         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24885         * ira.c (get_subreg_tracking_sizes): Likewise.
24886         * ira-build.c (ira_create_allocno_objects): Likewise.
24887         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24888         (ira_sort_regnos_for_alter_reg): Likewise.
24889         * ira-costs.c (record_operand_costs): Likewise.
24890         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24891         (resolve_simple_move): Likewise.
24892         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24893         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24894         (lra_constraints): Likewise.
24895         (CONST_POOL_OK_P): Reject variable-sized modes.
24896         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24897         (add_pseudo_to_slot, lra_spill): Likewise.
24898         * omp-low.c (omp_clause_aligned_alignment): Likewise.
24899         * optabs-query.c (get_best_extraction_insn): Likewise.
24900         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24901         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24902         (expand_mult_highpart, valid_multiword_target_p): Likewise.
24903         * recog.c (offsettable_address_addr_space_p): Likewise.
24904         * regcprop.c (maybe_mode_change): Likewise.
24905         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24906         * regrename.c (build_def_use): Likewise.
24907         * regstat.c (dump_reg_info): Likewise.
24908         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24909         (find_reloads, find_reloads_subreg_address): Likewise.
24910         * reload1.c (eliminate_regs_1): Likewise.
24911         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24912         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24913         (simplify_binary_operation_1, simplify_subreg): Likewise.
24914         * targhooks.c (default_function_arg_padding): Likewise.
24915         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24916         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24917         (verify_gimple_assign_ternary): Likewise.
24918         * tree-inline.c (estimate_move_cost): Likewise.
24919         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24920         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24921         (get_address_cost_ainc): Likewise.
24922         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24923         (vect_supportable_dr_alignment): Likewise.
24924         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24925         (vectorizable_reduction): Likewise.
24926         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24927         (vectorizable_operation, vectorizable_load): Likewise.
24928         * tree.c (build_same_sized_truth_vector_type): Likewise.
24929         * valtrack.c (cleanup_auto_inc_dec): Likewise.
24930         * var-tracking.c (emit_note_insn_var_location): Likewise.
24931         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24932         (ADDR_VEC_ALIGN): Likewise.
24934 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24935             Alan Hayward  <alan.hayward@arm.com>
24936             David Sherwood  <david.sherwood@arm.com>
24938         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24939         unsigned short.
24940         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24941         or if measurement_type is polynomial.
24942         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24943         * combine.c (make_extraction): Likewise.
24944         * dse.c (find_shift_sequence): Likewise.
24945         * dwarf2out.c (mem_loc_descriptor): Likewise.
24946         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24947         (extract_bit_field, extract_low_bits): Likewise.
24948         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24949         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24950         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24951         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24952         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24953         * reload.c (find_reloads): Likewise.
24954         * reload1.c (alter_reg): Likewise.
24955         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24956         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24957         * tree-if-conv.c (predicate_mem_writes): Likewise.
24958         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24959         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24960         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24961         * valtrack.c (dead_debug_insert_temp): Likewise.
24962         * varasm.c (mergeable_constant_section): Likewise.
24963         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24965 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24966             Alan Hayward  <alan.hayward@arm.com>
24967             David Sherwood  <david.sherwood@arm.com>
24969         * expr.c (expand_assignment): Cope with polynomial mode sizes
24970         when assigning to a CONCAT.
24972 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24973             Alan Hayward  <alan.hayward@arm.com>
24974             David Sherwood  <david.sherwood@arm.com>
24976         * machmode.h (mode_precision): Change from unsigned short to
24977         poly_uint16_pod.
24978         (mode_to_precision): Return a poly_uint16 rather than an unsigned
24979         short.
24980         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24981         or if measurement_type is not polynomial.
24982         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
24983         in which the mode is already known to be a scalar_int_mode.
24984         * genmodes.c (emit_mode_precision): Change the type of mode_precision
24985         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
24986         initializer.
24987         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24988         for GET_MODE_PRECISION.
24989         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24990         for GET_MODE_PRECISION.
24991         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24992         as polynomial.
24993         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24994         (expand_field_assignment, make_extraction): Likewise.
24995         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24996         (get_last_value): Likewise.
24997         * convert.c (convert_to_integer_1): Likewise.
24998         * cse.c (cse_insn): Likewise.
24999         * expr.c (expand_expr_real_1): Likewise.
25000         * lra-constraints.c (simplify_operand_subreg): Likewise.
25001         * optabs-query.c (can_atomic_load_p): Likewise.
25002         * optabs.c (expand_atomic_load): Likewise.
25003         (expand_atomic_store): Likewise.
25004         * ree.c (combine_reaching_defs): Likewise.
25005         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
25006         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
25007         * tree.h (type_has_mode_precision_p): Likewise.
25008         * ubsan.c (instrument_si_overflow): Likewise.
25010 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25011             Alan Hayward  <alan.hayward@arm.com>
25012             David Sherwood  <david.sherwood@arm.com>
25014         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
25015         polynomial numbers of units.
25016         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
25017         (valid_vector_subparts_p): New function.
25018         (build_vector_type): Remove temporary shim and take the number
25019         of units as a poly_uint64 rather than an int.
25020         (build_opaque_vector_type): Take the number of units as a
25021         poly_uint64 rather than an int.
25022         * tree.c (build_vector_from_ctor): Handle polynomial
25023         TYPE_VECTOR_SUBPARTS.
25024         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
25025         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
25026         (build_vector_from_val): If the number of units is variable,
25027         use build_vec_duplicate_cst for constant operands and
25028         VEC_DUPLICATE_EXPR otherwise.
25029         (make_vector_type): Remove temporary is_constant ().
25030         (build_vector_type, build_opaque_vector_type): Take the number of
25031         units as a poly_uint64 rather than an int.
25032         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
25033         VECTOR_CST_NELTS.
25034         * cfgexpand.c (expand_debug_expr): Likewise.
25035         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
25036         (store_constructor, expand_expr_real_1): Likewise.
25037         (const_scalar_mask_from_tree): Likewise.
25038         * fold-const-call.c (fold_const_reduction): Likewise.
25039         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
25040         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
25041         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
25042         (fold_relational_const): Likewise.
25043         (native_interpret_vector): Likewise.  Change the size from an
25044         int to an unsigned int.
25045         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
25046         TYPE_VECTOR_SUBPARTS.
25047         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
25048         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
25049         duplicating a non-constant operand into a variable-length vector.
25050         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
25051         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
25052         * ipa-icf.c (sem_variable::equals): Likewise.
25053         * match.pd: Likewise.
25054         * omp-simd-clone.c (simd_clone_subparts): Likewise.
25055         * print-tree.c (print_node): Likewise.
25056         * stor-layout.c (layout_type): Likewise.
25057         * targhooks.c (default_builtin_vectorization_cost): Likewise.
25058         * tree-cfg.c (verify_gimple_comparison): Likewise.
25059         (verify_gimple_assign_binary): Likewise.
25060         (verify_gimple_assign_ternary): Likewise.
25061         (verify_gimple_assign_single): Likewise.
25062         * tree-pretty-print.c (dump_generic_node): Likewise.
25063         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25064         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
25065         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
25066         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
25067         (vect_shift_permute_load_chain): Likewise.
25068         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
25069         (expand_vector_condition, optimize_vector_constructor): Likewise.
25070         (lower_vec_perm, get_compute_type): Likewise.
25071         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25072         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
25073         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
25074         (vect_recog_mask_conversion_pattern): Likewise.
25075         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
25076         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
25077         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25078         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
25079         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
25080         (vectorizable_shift, vectorizable_operation, vectorizable_store)
25081         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
25082         (supportable_widening_operation): Likewise.
25083         (supportable_narrowing_operation): Likewise.
25084         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
25085         Likewise.
25086         * varasm.c (output_constant): Likewise.
25088 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25089             Alan Hayward  <alan.hayward@arm.com>
25090             David Sherwood  <david.sherwood@arm.com>
25092         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
25093         so that both the length == 3 and length != 3 cases set up their
25094         own permute vectors.  Add comments explaining why we know the
25095         number of elements is constant.
25096         (vect_permute_load_chain): Likewise.
25098 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25099             Alan Hayward  <alan.hayward@arm.com>
25100             David Sherwood  <david.sherwood@arm.com>
25102         * machmode.h (mode_nunits): Change from unsigned char to
25103         poly_uint16_pod.
25104         (ONLY_FIXED_SIZE_MODES): New macro.
25105         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
25106         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
25107         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
25108         New typedefs.
25109         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
25110         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
25111         or if measurement_type is not polynomial.
25112         * genmodes.c (ZERO_COEFFS): New macro.
25113         (emit_mode_nunits_inline): Make mode_nunits_inline return a
25114         poly_uint16.
25115         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
25116         Use ZERO_COEFFS when emitting initializers.
25117         * data-streamer.h (bp_pack_poly_value): New function.
25118         (bp_unpack_poly_value): Likewise.
25119         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25120         for GET_MODE_NUNITS.
25121         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25122         for GET_MODE_NUNITS.
25123         * tree.c (make_vector_type): Remove temporary shim and make
25124         the real function take the number of units as a poly_uint64
25125         rather than an int.
25126         (build_vector_type_for_mode): Handle polynomial nunits.
25127         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
25128         * emit-rtl.c (const_vec_series_p_1): Likewise.
25129         (gen_rtx_CONST_VECTOR): Likewise.
25130         * fold-const.c (test_vec_duplicate_folding): Likewise.
25131         * genrecog.c (validate_pattern): Likewise.
25132         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
25133         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25134         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
25135         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
25136         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
25137         * rtlanal.c (subreg_get_info): Likewise.
25138         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25139         (vect_grouped_load_supported): Likewise.
25140         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
25141         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
25142         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
25143         (simplify_const_unary_operation, simplify_binary_operation_1)
25144         (simplify_const_binary_operation, simplify_ternary_operation)
25145         (test_vector_ops_duplicate, test_vector_ops): Likewise.
25146         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
25147         instead of CONST_VECTOR_NUNITS.
25148         * varasm.c (output_constant_pool_2): Likewise.
25149         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
25150         explicit-encoded elements in the XVEC for variable-length vectors.
25152 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25154         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
25156 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25157             Alan Hayward  <alan.hayward@arm.com>
25158             David Sherwood  <david.sherwood@arm.com>
25160         * coretypes.h (fixed_size_mode): Declare.
25161         (fixed_size_mode_pod): New typedef.
25162         * builtins.h (target_builtins::x_apply_args_mode)
25163         (target_builtins::x_apply_result_mode): Change type to
25164         fixed_size_mode_pod.
25165         * builtins.c (apply_args_size, apply_result_size, result_vector)
25166         (expand_builtin_apply_args_1, expand_builtin_apply)
25167         (expand_builtin_return): Update accordingly.
25169 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25171         * cse.c (hash_rtx_cb): Hash only the encoded elements.
25172         * cselib.c (cselib_hash_rtx): Likewise.
25173         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25174         CONST_VECTOR encoding.
25176 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25177             Jeff Law  <law@redhat.com>
25179         PR target/83641
25180         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25181         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25182         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25183         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25185         PR target/83641
25186         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25187         explicitly probe *sp in a noreturn function if there were any callee
25188         register saves or frame pointer is needed.
25190 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25192         PR debug/83621
25193         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25194         BLKmode for ternary, binary or unary expressions.
25196         PR debug/83645
25197         * var-tracking.c (delete_vta_debug_insn): New inline function.
25198         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25199         insns from get_insns () to NULL instead of each bb separately.
25200         Use delete_vta_debug_insn.  No longer static.
25201         (vt_debug_insns_local, variable_tracking_main_1): Adjust
25202         delete_vta_debug_insns callers.
25203         * rtl.h (delete_vta_debug_insns): Declare.
25204         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25205         instead of variable_tracking_main.
25207 2018-01-03  Martin Sebor  <msebor@redhat.com>
25209         PR tree-optimization/83603
25210         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25211         arguments past the endof the argument list in functions declared
25212         without a prototype.
25213         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25214         Avoid checking when arguments are null.
25216 2018-01-03  Martin Sebor  <msebor@redhat.com>
25218         PR c/83559
25219         * doc/extend.texi (attribute const): Fix a typo.
25220         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25221         issuing -Wsuggest-attribute for void functions.
25223 2018-01-03  Martin Sebor  <msebor@redhat.com>
25225         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25226         offset_int::from instead of wide_int::to_shwi.
25227         (maybe_diag_overlap): Remove assertion.
25228         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25229         * gimple-ssa-sprintf.c (format_directive): Same.
25230         (parse_directive): Same.
25231         (sprintf_dom_walker::compute_format_length): Same.
25232         (try_substitute_return_value): Same.
25234 2018-01-03  Jeff Law  <law@redhat.com>
25236         PR middle-end/83654
25237         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25238         non-constant residual for zero at runtime and avoid probing in
25239         that case.  Reorganize code for trailing problem to mirror handling
25240         of the residual.
25242 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25244         PR tree-optimization/83501
25245         * tree-ssa-strlen.c (get_string_cst): New.
25246         (handle_char_store): Call get_string_cst.
25248 2018-01-03  Martin Liska  <mliska@suse.cz>
25250         PR tree-optimization/83593
25251         * tree-ssa-strlen.c: Include tree-cfg.h.
25252         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25253         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25254         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25255         to false.
25256         (strlen_dom_walker::before_dom_children): Call
25257         gimple_purge_dead_eh_edges. Dump tranformation with details
25258         dump flags.
25259         (strlen_dom_walker::before_dom_children): Update call by adding
25260         new argument cleanup_eh.
25261         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25263 2018-01-03  Martin Liska  <mliska@suse.cz>
25265         PR ipa/83549
25266         * cif-code.def (VARIADIC_THUNK): New enum value.
25267         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25268         thunks.
25270 2018-01-03  Jan Beulich  <jbeulich@suse.com>
25272         * sse.md (mov<mode>_internal): Tighten condition for when to use
25273         vmovdqu<ssescalarsize> for TI and OI modes.
25275 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25277         Update copyright years.
25279 2018-01-03  Martin Liska  <mliska@suse.cz>
25281         PR ipa/83594
25282         * ipa-visibility.c (function_and_variable_visibility): Skip
25283         functions with noipa attribure.
25285 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25287         * gcc.c (process_command): Update copyright notice dates.
25288         * gcov-dump.c (print_version): Ditto.
25289         * gcov.c (print_version): Ditto.
25290         * gcov-tool.c (print_version): Ditto.
25291         * gengtype.c (create_file): Ditto.
25292         * doc/cpp.texi: Bump @copying's copyright year.
25293         * doc/cppinternals.texi: Ditto.
25294         * doc/gcc.texi: Ditto.
25295         * doc/gccint.texi: Ditto.
25296         * doc/gcov.texi: Ditto.
25297         * doc/install.texi: Ditto.
25298         * doc/invoke.texi: Ditto.
25300 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25302         * vector-builder.h (vector_builder::m_full_nelts): Change from
25303         unsigned int to poly_uint64.
25304         (vector_builder::full_nelts): Update prototype accordingly.
25305         (vector_builder::new_vector): Likewise.
25306         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25307         (vector_builder::operator ==): Likewise.
25308         (vector_builder::finalize): Likewise.
25309         * int-vector-builder.h (int_vector_builder::int_vector_builder):
25310         Take the number of elements as a poly_uint64 rather than an
25311         unsigned int.
25312         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25313         from unsigned int to poly_uint64.
25314         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25315         (vec_perm_indices::new_vector): Likewise.
25316         (vec_perm_indices::length): Likewise.
25317         (vec_perm_indices::nelts_per_input): Likewise.
25318         (vec_perm_indices::input_nelts): Likewise.
25319         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25320         number of elements per input as a poly_uint64 rather than an
25321         unsigned int.  Use the original encoding for variable-length
25322         vectors, rather than clamping each individual element.
25323         For the second and subsequent elements in each pattern,
25324         clamp the step and base before clamping their sum.
25325         (vec_perm_indices::series_p): Handle polynomial element counts.
25326         (vec_perm_indices::all_in_range_p): Likewise.
25327         (vec_perm_indices_to_tree): Likewise.
25328         (vec_perm_indices_to_rtx): Likewise.
25329         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25330         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25331         (tree_vector_builder::new_binary_operation): Handle polynomial
25332         element counts.  Return false if we need to know the number
25333         of elements at compile time.
25334         * fold-const.c (fold_vec_perm): Punt if the number of elements
25335         isn't known at compile time.
25337 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25339         * vec-perm-indices.h (vec_perm_builder): Change element type
25340         from HOST_WIDE_INT to poly_int64.
25341         (vec_perm_indices::element_type): Update accordingly.
25342         (vec_perm_indices::clamp): Handle polynomial element_types.
25343         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25344         (vec_perm_indices::all_in_range_p): Likewise.
25345         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25346         than shwi trees.
25347         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25348         polynomial vec_perm_indices element types.
25349         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25350         * fold-const.c (fold_vec_perm): Likewise.
25351         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25352         * tree-vect-generic.c (lower_vec_perm): Likewise.
25353         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25354         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25355         element type to HOST_WIDE_INT.
25357 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25358             Alan Hayward  <alan.hayward@arm.com>
25359             David Sherwood  <david.sherwood@arm.com>
25361         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25362         rather than an int.  Use plus_constant.
25363         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25364         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
25366 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25367             Alan Hayward  <alan.hayward@arm.com>
25368             David Sherwood  <david.sherwood@arm.com>
25370         * calls.c (emit_call_1, expand_call): Change struct_value_size from
25371         a HOST_WIDE_INT to a poly_int64.
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         * calls.c (load_register_parameters): Cope with polynomial
25378         mode sizes.  Require a constant size for BLKmode parameters
25379         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
25380         forces a parameter to be padded at the lsb end in order to
25381         fill a complete number of words, require the parameter size
25382         to be ordered wrt UNITS_PER_WORD.
25384 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25385             Alan Hayward  <alan.hayward@arm.com>
25386             David Sherwood  <david.sherwood@arm.com>
25388         * reload1.c (spill_stack_slot_width): Change element type
25389         from unsigned int to poly_uint64_pod.
25390         (alter_reg): Treat mode sizes as polynomial.
25392 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25393             Alan Hayward  <alan.hayward@arm.com>
25394             David Sherwood  <david.sherwood@arm.com>
25396         * reload.c (complex_word_subreg_p): New function.
25397         (reload_inner_reg_of_subreg, push_reload): Use it.
25399 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25400             Alan Hayward  <alan.hayward@arm.com>
25401             David Sherwood  <david.sherwood@arm.com>
25403         * lra-constraints.c (process_alt_operands): Reject matched
25404         operands whose sizes aren't ordered.
25405         (match_reload): Refer to this check here.
25407 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25408             Alan Hayward  <alan.hayward@arm.com>
25409             David Sherwood  <david.sherwood@arm.com>
25411         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
25412         that the mode size is in the set {1, 2, 4, 8, 16}.
25414 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25415             Alan Hayward  <alan.hayward@arm.com>
25416             David Sherwood  <david.sherwood@arm.com>
25418         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
25419         Use plus_constant instead of gen_rtx_PLUS.
25421 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25422             Alan Hayward  <alan.hayward@arm.com>
25423             David Sherwood  <david.sherwood@arm.com>
25425         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
25426         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
25427         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
25428         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
25429         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
25430         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25431         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
25432         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
25433         * config/i386/i386.c (ix86_push_rounding): ...this new function.
25434         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
25435         a poly_int64.
25436         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
25437         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
25438         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
25439         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
25440         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
25441         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
25442         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
25443         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
25444         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25445         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
25446         function.
25447         * expr.c (emit_move_resolve_push): Treat the input and result
25448         of PUSH_ROUNDING as a poly_int64.
25449         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25450         (emit_push_insn): Likewise.
25451         * lra-eliminations.c (mark_not_eliminable): Likewise.
25452         * recog.c (push_operand): Likewise.
25453         * reload1.c (elimination_effects): Likewise.
25454         * rtlanal.c (nonzero_bits1): Likewise.
25455         * calls.c (store_one_arg): Likewise.  Require the padding to be
25456         known at compile time.
25458 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25459             Alan Hayward  <alan.hayward@arm.com>
25460             David Sherwood  <david.sherwood@arm.com>
25462         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25463         Use plus_constant instead of gen_rtx_PLUS.
25465 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25466             Alan Hayward  <alan.hayward@arm.com>
25467             David Sherwood  <david.sherwood@arm.com>
25469         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25470         rather than an int.
25472 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25473             Alan Hayward  <alan.hayward@arm.com>
25474             David Sherwood  <david.sherwood@arm.com>
25476         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25477         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25478         via stack temporaries.  Treat the mode size as polynomial too.
25480 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25481             Alan Hayward  <alan.hayward@arm.com>
25482             David Sherwood  <david.sherwood@arm.com>
25484         * expr.c (expand_expr_real_2): When handling conversions involving
25485         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25486         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
25487         as a poly_uint64 too.
25489 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25490             Alan Hayward  <alan.hayward@arm.com>
25491             David Sherwood  <david.sherwood@arm.com>
25493         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25495 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25496             Alan Hayward  <alan.hayward@arm.com>
25497             David Sherwood  <david.sherwood@arm.com>
25499         * combine.c (can_change_dest_mode): Handle polynomial
25500         REGMODE_NATURAL_SIZE.
25501         * expmed.c (store_bit_field_1): Likewise.
25502         * expr.c (store_constructor): Likewise.
25503         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25504         and polynomial REGMODE_NATURAL_SIZE.
25505         (gen_lowpart_common): Likewise.
25506         * reginfo.c (record_subregs_of_mode): Likewise.
25507         * rtlanal.c (read_modify_subreg_p): Likewise.
25509 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25510             Alan Hayward  <alan.hayward@arm.com>
25511             David Sherwood  <david.sherwood@arm.com>
25513         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25514         numbers of elements.
25516 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25517             Alan Hayward  <alan.hayward@arm.com>
25518             David Sherwood  <david.sherwood@arm.com>
25520         * match.pd: Cope with polynomial numbers of vector elements.
25522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25523             Alan Hayward  <alan.hayward@arm.com>
25524             David Sherwood  <david.sherwood@arm.com>
25526         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25527         in a POINTER_PLUS_EXPR.
25529 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25530             Alan Hayward  <alan.hayward@arm.com>
25531             David Sherwood  <david.sherwood@arm.com>
25533         * omp-simd-clone.c (simd_clone_subparts): New function.
25534         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25535         (ipa_simd_modify_function_body): Likewise.
25537 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25538             Alan Hayward  <alan.hayward@arm.com>
25539             David Sherwood  <david.sherwood@arm.com>
25541         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25542         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25543         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25544         (expand_vector_condition, vector_element): Likewise.
25545         (subparts_gt): New function.
25546         (get_compute_type): Use subparts_gt.
25547         (count_type_subparts): Delete.
25548         (expand_vector_operations_1): Use subparts_gt instead of
25549         count_type_subparts.
25551 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25552             Alan Hayward  <alan.hayward@arm.com>
25553             David Sherwood  <david.sherwood@arm.com>
25555         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25556         (vect_compile_time_alias): ...this new function.  Do the calculation
25557         on poly_ints rather than trees.
25558         (vect_prune_runtime_alias_test_list): Update call accordingly.
25560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25561             Alan Hayward  <alan.hayward@arm.com>
25562             David Sherwood  <david.sherwood@arm.com>
25564         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25565         numbers of units.
25566         (vect_schedule_slp_instance): Likewise.
25568 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25569             Alan Hayward  <alan.hayward@arm.com>
25570             David Sherwood  <david.sherwood@arm.com>
25572         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25573         constant and extern definitions for variable-length vectors.
25574         (vect_get_constant_vectors): Note that the number of units
25575         is known to be constant.
25577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25578             Alan Hayward  <alan.hayward@arm.com>
25579             David Sherwood  <david.sherwood@arm.com>
25581         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25582         of units as polynomial.  Choose between WIDE and NARROW based
25583         on multiple_p.
25585 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25586             Alan Hayward  <alan.hayward@arm.com>
25587             David Sherwood  <david.sherwood@arm.com>
25589         * tree-vect-stmts.c (simd_clone_subparts): New function.
25590         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25592 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25593             Alan Hayward  <alan.hayward@arm.com>
25594             David Sherwood  <david.sherwood@arm.com>
25596         * tree-vect-stmts.c (vectorizable_call): Treat the number of
25597         vectors as polynomial.  Use build_index_vector for
25598         IFN_GOMP_SIMD_LANE.
25600 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25601             Alan Hayward  <alan.hayward@arm.com>
25602             David Sherwood  <david.sherwood@arm.com>
25604         * tree-vect-stmts.c (get_load_store_type): Treat the number of
25605         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25606         for variable-length vectors.
25607         (vectorizable_mask_load_store): Treat the number of units as
25608         polynomial, asserting that it is constant if the condition has
25609         already been enforced.
25610         (vectorizable_store, vectorizable_load): Likewise.
25612 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25613             Alan Hayward  <alan.hayward@arm.com>
25614             David Sherwood  <david.sherwood@arm.com>
25616         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25617         of units as polynomial.  Punt if we can't tell at compile time
25618         which vector contains the final result.
25620 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25621             Alan Hayward  <alan.hayward@arm.com>
25622             David Sherwood  <david.sherwood@arm.com>
25624         * tree-vect-loop.c (vectorizable_induction): Treat the number
25625         of units as polynomial.  Punt on SLP inductions.  Use an integer
25626         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
25627         cast of such a series for variable-length floating-point
25628         reductions.
25630 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25631             Alan Hayward  <alan.hayward@arm.com>
25632             David Sherwood  <david.sherwood@arm.com>
25634         * tree.h (build_index_vector): Declare.
25635         * tree.c (build_index_vector): New function.
25636         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25637         of units as polynomial, forcibly converting it to a constant if
25638         vectorizable_reduction has already enforced the condition.
25639         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
25640         to create a {1,2,3,...} vector.
25641         (vectorizable_reduction): Treat the number of units as polynomial.
25642         Choose vectype_in based on the largest scalar element size rather
25643         than the smallest number of units.  Enforce the restrictions
25644         relied on above.
25646 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25647             Alan Hayward  <alan.hayward@arm.com>
25648             David Sherwood  <david.sherwood@arm.com>
25650         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25651         number of units as polynomial.
25653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25654             Alan Hayward  <alan.hayward@arm.com>
25655             David Sherwood  <david.sherwood@arm.com>
25657         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25658         * target.def (autovectorize_vector_sizes): Return the vector sizes
25659         by pointer, using vector_sizes rather than a bitmask.
25660         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25661         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25662         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25663         Likewise.
25664         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25665         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25666         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25667         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25668         * omp-general.c (omp_max_vf): Likewise.
25669         * omp-low.c (omp_clause_aligned_alignment): Likewise.
25670         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25671         * tree-vect-loop.c (vect_analyze_loop): Likewise.
25672         * tree-vect-slp.c (vect_slp_bb): Likewise.
25673         * doc/tm.texi: Regenerate.
25674         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25675         to a poly_uint64.
25676         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25677         the vector size as a poly_uint64 rather than an unsigned int.
25678         (current_vector_size): Change from an unsigned int to a poly_uint64.
25679         (get_vectype_for_scalar_type): Update accordingly.
25680         * tree.h (build_truth_vector_type): Take the size and number of
25681         units as a poly_uint64 rather than an unsigned int.
25682         (build_vector_type): Add a temporary overload that takes
25683         the number of units as a poly_uint64 rather than an unsigned int.
25684         * tree.c (make_vector_type): Likewise.
25685         (build_truth_vector_type): Take the number of units as a poly_uint64
25686         rather than an unsigned int.
25688 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25689             Alan Hayward  <alan.hayward@arm.com>
25690             David Sherwood  <david.sherwood@arm.com>
25692         * target.def (get_mask_mode): Take the number of units and length
25693         as poly_uint64s rather than unsigned ints.
25694         * targhooks.h (default_get_mask_mode): Update accordingly.
25695         * targhooks.c (default_get_mask_mode): Likewise.
25696         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25697         * doc/tm.texi: Regenerate.
25699 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25700             Alan Hayward  <alan.hayward@arm.com>
25701             David Sherwood  <david.sherwood@arm.com>
25703         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25704         * omp-general.c (omp_max_vf): Likewise.
25705         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25706         (expand_omp_simd): Handle polynomial safelen.
25707         * omp-low.c (omplow_simd_context): Add a default constructor.
25708         (omplow_simd_context::max_vf): Change from int to poly_uint64.
25709         (lower_rec_simd_input_clauses): Update accordingly.
25710         (lower_rec_input_clauses): Likewise.
25712 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25713             Alan Hayward  <alan.hayward@arm.com>
25714             David Sherwood  <david.sherwood@arm.com>
25716         * tree-vectorizer.h (vect_nunits_for_cost): New function.
25717         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25718         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25719         (vect_analyze_slp_cost): Likewise.
25720         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25721         (vect_model_load_cost): Likewise.
25723 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25724             Alan Hayward  <alan.hayward@arm.com>
25725             David Sherwood  <david.sherwood@arm.com>
25727         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25728         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25729         from an unsigned int * to a poly_uint64_pod *.
25730         (calculate_unrolling_factor): New function.
25731         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
25733 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25734             Alan Hayward  <alan.hayward@arm.com>
25735             David Sherwood  <david.sherwood@arm.com>
25737         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25738         from an unsigned int to a poly_uint64.
25739         (_loop_vec_info::slp_unrolling_factor): Likewise.
25740         (_loop_vec_info::vectorization_factor): Change from an int
25741         to a poly_uint64.
25742         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25743         (vect_get_num_vectors): New function.
25744         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25745         (vect_get_num_copies): Use vect_get_num_vectors.
25746         (vect_analyze_data_ref_dependences): Change max_vf from an int *
25747         to an unsigned int *.
25748         (vect_analyze_data_refs): Change min_vf from an int * to a
25749         poly_uint64 *.
25750         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25751         than an unsigned HOST_WIDE_INT.
25752         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25753         (vect_analyze_data_ref_dependence): Change max_vf from an int *
25754         to an unsigned int *.
25755         (vect_analyze_data_ref_dependences): Likewise.
25756         (vect_compute_data_ref_alignment): Handle polynomial vf.
25757         (vect_enhance_data_refs_alignment): Likewise.
25758         (vect_prune_runtime_alias_test_list): Likewise.
25759         (vect_shift_permute_load_chain): Likewise.
25760         (vect_supportable_dr_alignment): Likewise.
25761         (dependence_distance_ge_vf): Take the vectorization factor as a
25762         poly_uint64 rather than an unsigned HOST_WIDE_INT.
25763         (vect_analyze_data_refs): Change min_vf from an int * to a
25764         poly_uint64 *.
25765         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25766         vfm1 as a poly_uint64 rather than an int.  Make the same change
25767         for the returned bound_scalar.
25768         (vect_gen_vector_loop_niters): Handle polynomial vf.
25769         (vect_do_peeling): Likewise.  Update call to
25770         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25771         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25772         be constant.
25773         * tree-vect-loop.c (vect_determine_vectorization_factor)
25774         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25775         (vect_get_known_peeling_cost): Likewise.
25776         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25777         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25778         (vect_transform_loop): Likewise.  Use the lowest possible VF when
25779         updating the upper bounds of the loop.
25780         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
25781         rather than an int.
25782         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25783         polynomial unroll factors.
25784         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25785         (vect_make_slp_decision): Likewise.
25786         (vect_supported_load_permutation_p): Likewise, and polynomial
25787         vf too.
25788         (vect_analyze_slp_cost): Handle polynomial vf.
25789         (vect_slp_analyze_node_operations): Likewise.
25790         (vect_slp_analyze_bb_1): Likewise.
25791         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25792         than an unsigned HOST_WIDE_INT.
25793         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25794         (vectorizable_load): Handle polynomial vf.
25795         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25796         a poly_uint64.
25797         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25799 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25800             Alan Hayward  <alan.hayward@arm.com>
25801             David Sherwood  <david.sherwood@arm.com>
25803         * match.pd: Handle bit operations involving three constants
25804         and try to fold one pair.
25806 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25808         * tree-vect-loop-manip.c: Include gimple-fold.h.
25809         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25810         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
25811         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25812         Add a path that uses a step of VF instead of 1, but disable it
25813         for now.
25814         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25815         and niters_no_overflow parameters.  Update calls to
25816         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25817         Create a new SSA name if the latter choses to use a ste other
25818         than zero, and return it via niters_vector_mult_vf_var.
25819         * tree-vect-loop.c (vect_transform_loop): Update calls to
25820         vect_do_peeling, vect_gen_vector_loop_niters and
25821         slpeel_make_loop_iterate_ntimes.
25822         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25823         (vect_gen_vector_loop_niters): Update declarations after above changes.
25825 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
25827         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25828         128-bit round to integer instructions.
25829         (ceil<mode>2): Likewise.
25830         (btrunc<mode>2): Likewise.
25831         (round<mode>2): Likewise.
25833 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25835         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25836         unaligned VSX load/store on P8/P9.
25837         (expand_block_clear): Allow the use of unaligned VSX
25838         load/store on P8/P9.
25840 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25842         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25843         New function.
25844         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25845         swap associated with both a load and a store.
25847 2018-01-02  Andrew Waterman  <andrew@sifive.com>
25849         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25850         * config/riscv/riscv.md (clear_cache): Use it.
25852 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
25854         * web.c: Remove out-of-date comment.
25856 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25858         * expr.c (fixup_args_size_notes): Check that any existing
25859         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25860         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25861         (emit_single_push_insn): ...here.
25863 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25865         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25866         (const_vector_encoded_nelts): New function.
25867         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25868         (const_vector_int_elt, const_vector_elt): Declare.
25869         * emit-rtl.c (const_vector_int_elt_1): New function.
25870         (const_vector_elt): Likewise.
25871         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25872         of CONST_VECTOR_ELT.
25874 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25876         * expr.c: Include rtx-vector-builder.h.
25877         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25878         directly on the tree encoding.
25879         (const_vector_from_tree): Likewise.
25880         * optabs.c: Include rtx-vector-builder.h.
25881         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25882         sequence of "u" values.
25883         * vec-perm-indices.c: Include rtx-vector-builder.h.
25884         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25885         directly on the vec_perm_indices encoding.
25887 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25889         * doc/rtl.texi (const_vector): Describe new encoding scheme.
25890         * Makefile.in (OBJS): Add rtx-vector-builder.o.
25891         * rtx-vector-builder.h: New file.
25892         * rtx-vector-builder.c: Likewise.
25893         * rtl.h (rtx_def::u2): Add a const_vector field.
25894         (CONST_VECTOR_NPATTERNS): New macro.
25895         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25896         (CONST_VECTOR_DUPLICATE_P): Likewise.
25897         (CONST_VECTOR_STEPPED_P): Likewise.
25898         (CONST_VECTOR_ENCODED_ELT): Likewise.
25899         (const_vec_duplicate_p): Check for a duplicated vector encoding.
25900         (unwrap_const_vec_duplicate): Likewise.
25901         (const_vec_series_p): Check for a non-duplicated vector encoding.
25902         Say that the function only returns true for integer vectors.
25903         * emit-rtl.c: Include rtx-vector-builder.h.
25904         (gen_const_vec_duplicate_1): Delete.
25905         (gen_const_vector): Call gen_const_vec_duplicate instead of
25906         gen_const_vec_duplicate_1.
25907         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25908         (gen_const_vec_duplicate): Use rtx_vector_builder.
25909         (gen_const_vec_series): Likewise.
25910         (gen_rtx_CONST_VECTOR): Likewise.
25911         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25912         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25913         Build a new vector rather than modifying a CONST_VECTOR in-place.
25914         (handle_special_swappables): Update call accordingly.
25915         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25916         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25917         Build a new vector rather than modifying a CONST_VECTOR in-place.
25918         (handle_special_swappables): Update call accordingly.
25920 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25922         * simplify-rtx.c (simplify_const_binary_operation): Use
25923         CONST_VECTOR_ELT instead of XVECEXP.
25925 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25927         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25928         the selector elements to be different from the data elements
25929         if the selector is a VECTOR_CST.
25930         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25931         ssizetype for the selector.
25933 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25935         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25936         before testing each element individually.
25937         * tree-vect-generic.c (lower_vec_perm): Likewise.
25939 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25941         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25942         * selftest-run-tests.c (selftest::run_tests): Call it.
25943         * vector-builder.h (vector_builder::operator ==): New function.
25944         (vector_builder::operator !=): Likewise.
25945         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25946         (vec_perm_indices::all_from_input_p): New function.
25947         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25948         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25949         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25950         instead of reading the VECTOR_CST directly.  Detect whether both
25951         vector inputs are the same before constructing the vec_perm_indices,
25952         and update the number of inputs argument accordingly.  Use the
25953         utility functions added above.  Only construct sel2 if we need to.
25955 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25957         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25958         the broadcast of the low byte.
25959         (expand_mult_highpart): Use an explicit encoding for the permutes.
25960         * optabs-query.c (can_mult_highpart_p): Likewise.
25961         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25962         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25963         (vectorizable_bswap): Likewise.
25964         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25965         explicit encoding for the power-of-2 permutes.
25966         (vect_permute_store_chain): Likewise.
25967         (vect_grouped_load_supported): Likewise.
25968         (vect_permute_load_chain): Likewise.
25970 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25972         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25973         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25974         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25975         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25976         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25977         (vect_gen_perm_mask_any): Likewise.
25979 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25981         * int-vector-builder.h: New file.
25982         * vec-perm-indices.h: Include int-vector-builder.h.
25983         (vec_perm_indices): Redefine as an int_vector_builder.
25984         (auto_vec_perm_indices): Delete.
25985         (vec_perm_builder): Redefine as a stand-alone class.
25986         (vec_perm_indices::vec_perm_indices): New function.
25987         (vec_perm_indices::clamp): Likewise.
25988         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25989         (vec_perm_indices::new_vector): New function.
25990         (vec_perm_indices::new_expanded_vector): Update for new
25991         vec_perm_indices class.
25992         (vec_perm_indices::rotate_inputs): New function.
25993         (vec_perm_indices::all_in_range_p): Operate directly on the
25994         encoded form, without computing elided elements.
25995         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25996         encoding.  Update for new vec_perm_indices class.
25997         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25998         the given vec_perm_builder.
25999         (expand_vec_perm_var): Update vec_perm_builder constructor.
26000         (expand_mult_highpart): Use vec_perm_builder instead of
26001         auto_vec_perm_indices.
26002         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
26003         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
26004         or double series encoding as appropriate.
26005         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
26006         vec_perm_indices instead of auto_vec_perm_indices.
26007         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26008         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26009         (vect_permute_store_chain): Likewise.
26010         (vect_grouped_load_supported): Likewise.
26011         (vect_permute_load_chain): Likewise.
26012         (vect_shift_permute_load_chain): Likewise.
26013         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26014         (vect_transform_slp_perm_load): Likewise.
26015         (vect_schedule_slp_instance): Likewise.
26016         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26017         (vectorizable_mask_load_store): Likewise.
26018         (vectorizable_bswap): Likewise.
26019         (vectorizable_store): Likewise.
26020         (vectorizable_load): Likewise.
26021         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
26022         vec_perm_indices instead of auto_vec_perm_indices.  Use
26023         tree_to_vec_perm_builder to read the vector from a tree.
26024         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
26025         vec_perm_builder instead of a vec_perm_indices.
26026         (have_whole_vector_shift): Use vec_perm_builder and
26027         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
26028         truncation to calc_vec_perm_mask_for_shift.
26029         (vect_create_epilog_for_reduction): Likewise.
26030         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
26031         from auto_vec_perm_indices to vec_perm_indices.
26032         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26033         instead of changing individual elements.
26034         (aarch64_vectorize_vec_perm_const): Use new_vector to install
26035         the vector in d.perm.
26036         * config/arm/arm.c (expand_vec_perm_d::perm): Change
26037         from auto_vec_perm_indices to vec_perm_indices.
26038         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26039         instead of changing individual elements.
26040         (arm_vectorize_vec_perm_const): Use new_vector to install
26041         the vector in d.perm.
26042         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
26043         Update vec_perm_builder constructor.
26044         (rs6000_expand_interleave): Likewise.
26045         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
26046         (rs6000_expand_interleave): Likewise.
26048 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26050         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
26051         to qimode could truncate the indices.
26052         * optabs.c (expand_vec_perm_var): Likewise.
26054 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26056         * Makefile.in (OBJS): Add vec-perm-indices.o.
26057         * vec-perm-indices.h: New file.
26058         * vec-perm-indices.c: Likewise.
26059         * target.h (vec_perm_indices): Replace with a forward class
26060         declaration.
26061         (auto_vec_perm_indices): Move to vec-perm-indices.h.
26062         * optabs.h: Include vec-perm-indices.h.
26063         (expand_vec_perm): Delete.
26064         (selector_fits_mode_p, expand_vec_perm_var): Declare.
26065         (expand_vec_perm_const): Declare.
26066         * target.def (vec_perm_const_ok): Replace with...
26067         (vec_perm_const): ...this new hook.
26068         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
26069         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
26070         * doc/tm.texi: Regenerate.
26071         * optabs.def (vec_perm_const): Delete.
26072         * doc/md.texi (vec_perm_const): Likewise.
26073         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
26074         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
26075         expand_vec_perm for constant permutation vectors.  Assert that
26076         the mode of variable permutation vectors is the integer equivalent
26077         of the mode that is being permuted.
26078         * optabs-query.h (selector_fits_mode_p): Declare.
26079         * optabs-query.c: Include vec-perm-indices.h.
26080         (selector_fits_mode_p): New function.
26081         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
26082         is defined, instead of checking whether the vec_perm_const_optab
26083         exists.  Use targetm.vectorize.vec_perm_const instead of
26084         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
26085         fit in the vector mode before using a variable permute.
26086         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
26087         vec_perm_indices instead of an rtx.
26088         (expand_vec_perm): Replace with...
26089         (expand_vec_perm_const): ...this new function.  Take the selector
26090         as a vec_perm_indices rather than an rtx.  Also take the mode of
26091         the selector.  Update call to shift_amt_for_vec_perm_mask.
26092         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
26093         Use vec_perm_indices::new_expanded_vector to expand the original
26094         selector into bytes.  Check whether the indices fit in the vector
26095         mode before using a variable permute.
26096         (expand_vec_perm_var): Make global.
26097         (expand_mult_highpart): Use expand_vec_perm_const.
26098         * fold-const.c: Includes vec-perm-indices.h.
26099         * tree-ssa-forwprop.c: Likewise.
26100         * tree-vect-data-refs.c: Likewise.
26101         * tree-vect-generic.c: Likewise.
26102         * tree-vect-loop.c: Likewise.
26103         * tree-vect-slp.c: Likewise.
26104         * tree-vect-stmts.c: Likewise.
26105         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
26106         Delete.
26107         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
26108         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
26109         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
26110         (aarch64_vectorize_vec_perm_const): ...this new function.
26111         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26112         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26113         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
26114         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
26115         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26116         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26117         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
26118         into...
26119         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
26120         check for NEON modes.
26121         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
26122         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
26123         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
26124         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
26125         into...
26126         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
26127         the old VEC_PERM_CONST conditions.
26128         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
26129         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
26130         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
26131         (ia64_vectorize_vec_perm_const_ok): Merge into...
26132         (ia64_vectorize_vec_perm_const): ...this new function.
26133         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
26134         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
26135         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
26136         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
26137         * config/mips/mips.c (mips_expand_vec_perm_const)
26138         (mips_vectorize_vec_perm_const_ok): Merge into...
26139         (mips_vectorize_vec_perm_const): ...this new function.
26140         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
26141         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
26142         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
26143         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
26144         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
26145         (rs6000_expand_vec_perm_const): Delete.
26146         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
26147         Delete.
26148         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26149         (altivec_expand_vec_perm_const_le): Take each operand individually.
26150         Operate on constant selectors rather than rtxes.
26151         (altivec_expand_vec_perm_const): Likewise.  Update call to
26152         altivec_expand_vec_perm_const_le.
26153         (rs6000_expand_vec_perm_const): Delete.
26154         (rs6000_vectorize_vec_perm_const_ok): Delete.
26155         (rs6000_vectorize_vec_perm_const): New function.
26156         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26157         an element count and rtx array.
26158         (rs6000_expand_extract_even): Update call accordingly.
26159         (rs6000_expand_interleave): Likewise.
26160         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
26161         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
26162         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
26163         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
26164         (rs6000_expand_vec_perm_const): Delete.
26165         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26166         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26167         (altivec_expand_vec_perm_const_le): Take each operand individually.
26168         Operate on constant selectors rather than rtxes.
26169         (altivec_expand_vec_perm_const): Likewise.  Update call to
26170         altivec_expand_vec_perm_const_le.
26171         (rs6000_expand_vec_perm_const): Delete.
26172         (rs6000_vectorize_vec_perm_const_ok): Delete.
26173         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
26174         reference to the SPE evmerge intructions.
26175         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26176         an element count and rtx array.
26177         (rs6000_expand_extract_even): Update call accordingly.
26178         (rs6000_expand_interleave): Likewise.
26179         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
26180         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26181         new function.
26182         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26184 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26186         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26187         vector mode and that that mode matches the mode of the data
26188         being permuted.
26189         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26190         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
26191         directly using expand_vec_perm_1 when forcing selectors into
26192         registers.
26193         (expand_vec_perm_var): New function, split out from expand_vec_perm.
26195 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26197         * optabs-query.h (can_vec_perm_p): Delete.
26198         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26199         * optabs-query.c (can_vec_perm_p): Split into...
26200         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26201         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26202         particular selector is valid.
26203         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26204         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26205         (vect_grouped_load_supported): Likewise.
26206         (vect_shift_permute_load_chain): Likewise.
26207         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26208         (vect_transform_slp_perm_load): Likewise.
26209         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26210         (vectorizable_bswap): Likewise.
26211         (vect_gen_perm_mask_checked): Likewise.
26212         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
26213         implementations of variable permutation vectors into account
26214         when deciding which selector to use.
26215         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26216         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26217         with a false third argument.
26218         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26219         to test whether the constant selector is valid and can_vec_perm_var_p
26220         to test whether a variable selector is valid.
26222 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26224         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26225         * optabs-query.c (can_vec_perm_p): Likewise.
26226         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26227         instead of vec_perm_indices.
26228         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26229         (vect_gen_perm_mask_checked): Likewise,
26230         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26231         (vect_gen_perm_mask_checked): Likewise,
26233 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26235         * optabs-query.h (qimode_for_vec_perm): Declare.
26236         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26237         (qimode_for_vec_perm): ...this new function.
26238         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26240 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26242         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26243         does not have a conditional at the top.
26245 2018-01-02  Richard Biener  <rguenther@suse.de>
26247         * ipa-inline.c (big_speedup_p): Fix expression.
26249 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
26251         PR target/81616
26252         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26253         for generic 4->6.
26255 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
26257         PR target/81616
26258         Generic tuning.
26259         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26260         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26261         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26262         cond_taken_branch_cost 3->4.
26264 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
26266         PR tree-optimization/83581
26267         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26268         TODO_cleanup_cfg if any changes have been made.
26270         PR middle-end/83608
26271         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26272         convert_modes if target mode has the right side, but different mode
26273         class.
26275         PR middle-end/83609
26276         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26277         last argument when extracting from CONCAT.  If either from_real or
26278         from_imag is NULL, use expansion through memory.  If result is not
26279         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26280         the parts directly to inner mode, if even that fails, use expansion
26281         through memory.
26283         PR middle-end/83623
26284         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26285         check for bswap in mode rather than HImode and use that in expand_unop
26286         too.
26288 Copyright (C) 2018 Free Software Foundation, Inc.
26290 Copying and distribution of this file, with or without modification,
26291 are permitted in any medium without royalty provided the copyright
26292 notice and this notice are preserved.