b94b4c0ecc129accf71ed0b2d2fc7afa16c84430
[platform/upstream/gcc.git] / gcc / ChangeLog
1 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2
3         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
4         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
5
6 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7
8         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
9         (PTRDIFF_TYPE): Likewise.
10
11 2018-09-21  Olivier Hainque  <hainque@adacore.com>
12
13         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
14         triplet, similar to support for VxWorks7.
15         * config/vxworks-dummy.h: Provide a default definition
16         of TARGET_VXWORKS64 to 0.
17
18 2018-09-21  Olivier Hainque  <hainque@adacore.com>
19
20         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
21         * config/vxworks-dummy.h: here.
22
23 2018-09-21  Olivier Hainque  <hainque@adacore.com>
24
25         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
26
27 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
28             Bo Zhou  <zbo.zhou@hisilicon.com>
29
30         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
31         * config/aarch64/aarch64-tune.md: Regenerated.
32         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
33         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
34         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
35
36 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
37             Julian Brown  <julian@codesourcery.com>
38
39         * builtins.c (get_builtin_sync_mem): Handle address spaces.
40
41 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
42
43         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
44         if the call takes a static chain.
45
46 2018-09-21  Martin Liska  <mliska@suse.cz>
47
48         * auto-profile.c (autofdo_source_profile::read): Do not
49         set sum_all.
50         (read_profile): Do not add working sets.
51         (read_autofdo_file): Remove sum_all.
52         (afdo_callsite_hot_enough_for_early_inline): Remove const
53         qualifier.
54         * coverage.c (struct counts_entry): Remove gcov_summary.
55         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
56         do not support GCOV_TAG_PROGRAM_SUMMARY.
57         (get_coverage_counts): Remove summary and expected
58         arguments.
59         * coverage.h (get_coverage_counts): Likewise.
60         * doc/gcov-dump.texi: Remove -w option.
61         * gcov-dump.c (dump_working_sets): Remove.
62         (main): Do not support '-w' option.
63         (print_usage): Likewise.
64         (tag_summary): Likewise.
65         * gcov-io.c (gcov_write_summary): Do not dump
66         histogram.
67         (gcov_read_summary): Likewise.
68         (gcov_histo_index): Remove.
69         (gcov_histogram_merge): Likewise.
70         (compute_working_sets): Likewise.
71         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
72         it not obsolete.
73         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
74         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
75         (GCOV_HISTOGRAM_SIZE): Remove.
76         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
77         (struct gcov_summary): Simplify rapidly just
78         to runs and sum_max fields.
79         (gcov_histo_index): Remove.
80         (NUM_GCOV_WORKING_SETS): Likewise.
81         (compute_working_sets): Likewise.
82         * gcov-tool.c (print_overlap_usage_message): Remove
83         trailing empty line.
84         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
85         (output_lines): Remove program related line.
86         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
87         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
88         histogram.
89         (input_profile_summary): Do not read it.
90         (merge_profile_summaries): And do not merge it.
91         (input_symtab): Do not call removed function.
92         * modulo-sched.c (sms_schedule): Do not print sum_max.
93         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
94         removed when histogram method was invented.
95         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
96         mode.
97         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
98         GCOV coding style.
99         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
100         and dump selected value.
101         * profile.c (add_working_set): Remove.
102         (get_working_sets): Likewise.
103         (find_working_set): Likewise.
104         (get_exec_counts): Do not work with working sets.
105         (read_profile_edge_counts): Do not inform as sum_max is removed.
106         (compute_branch_probabilities): Likewise.
107         (compute_value_histograms): Remove argument for call of
108         get_coverage_counts.
109         * profile.h: Do not make gcov_summary const.
110
111 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
112
113         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
114
115 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
116
117         PR tree-optimization/86990
118         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
119         Check that the entire merged store group is made of constants only for
120         overlapping stores.
121
122 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
123
124         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
125         (VTABLE_VERIFICATION_SPEC): Likewise.
126         (SANITIZER_EARLY_SPEC): Likewise.
127         (SANITIZER_SPEC): Likewise.
128         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
129         * doc/invoke.texi (Link Options): Document -r.
130
131 2018-09-20  Richard Biener <rguenther@suse.de>
132
133         PR middle-end/87054
134         * gimplify.c (gimplify_expr): Retain alignment of
135         addressable lvalue in dereference.
136
137 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
138
139         PR bootstrap/87013
140         * configure.ac: Check for .loc is_stmt support.
141         * configure, config.in: Rebuilt.
142         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
143         if not supported.
144
145 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
146
147         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
148         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
149         -misel=no.
150
151 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
152
153         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
154         VECTOR_OTHER.
155         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
156         case VECTOR_OTHER.
157
158 2018-09-20  Marek Polacek  <polacek@redhat.com>
159
160         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
161
162 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
163
164         PR tree-optimization/87288
165         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
166         into account when determining PEELING_FOR_NITERS.
167
168 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
169
170         PR tree-optimization/86877
171         * tree-vect-loop.c (vect_analyze_loop_2): Call
172         vect_verify_datarefs_alignment.
173
174 2018-09-19  Marek Polacek  <polacek@redhat.com>
175
176         * doc/invoke.texi: Document -Wclass-conversion.
177
178 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
179
180         * config/pa/pa.c (pa_adjust_priority): Delete.
181         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
182
183         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
184         (atomic_storehi): Likewise.
185         (atomic_storesi): Likewise.
186         (atomic_loaddi): Restore compare and swap exchange loop code.
187
188 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
189
190         PR rtl-optimization/86902
191         * combine.c (try_combine): When changing the CC mode used, don't change
192         an unrelated mode in other_insn to that new CC mode.
193
194 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
195
196         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
197         with %T in place of calls to dump_generic_expr.
198         (prune_runtime_alias_test_list): Likewise.
199         (create_runtime_alias_checks): Likewise.
200         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
201         (vect_analyze_data_ref_dependence): Likewise.
202         (vect_slp_analyze_data_ref_dependence): Likewise.
203         (vect_record_base_alignment): Likewise.  Use %G in place of call
204         to dump_gimple_stmt.
205         (vect_compute_data_ref_alignment): Likewise.
206         (verify_data_ref_alignment): Likewise.
207         (vect_find_same_alignment_drs): Likewise.
208         (vect_analyze_group_access_1): Likewise.
209         (vect_analyze_data_ref_accesses): Likewise.
210         (dependence_distance_ge_vf): Likewise.
211         (dump_lower_bound): Likewise.
212         (vect_prune_runtime_alias_test_list): Likewise.
213         (vect_find_stmt_data_reference): Likewise.
214         (vect_analyze_data_refs): Likewise.
215         (vect_create_addr_base_for_vector_ref): Likewise.
216         (vect_create_data_ref_ptr): Likewise.
217         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
218         (vect_can_advance_ivs_p): Likewise.
219         (vect_update_ivs_after_vectorizer): Likewise.
220         (vect_gen_prolog_loop_niters): Likewise.
221         (vect_prepare_for_masked_peels): Likewise.
222         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
223         (vect_determine_vectorization_factor): Likewise.
224         (vect_is_simple_iv_evolution): Likewise.
225         (vect_analyze_scalar_cycles_1): Likewise.
226         (vect_analyze_loop_operations): Likewise.
227         (report_vect_op): Likewise.
228         (vect_is_slp_reduction): Likewise.
229         (check_reduction_path): Likewise.
230         (vect_is_simple_reduction): Likewise.
231         (vect_create_epilog_for_reduction): Likewise.
232         (vect_finalize_reduction:): Likewise.
233         (vectorizable_induction): Likewise.
234         (vect_transform_loop_stmt): Likewise.
235         (vect_transform_loop): Likewise.
236         (optimize_mask_stores): Likewise.
237         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
238         (vect_split_statement): Likewise.
239         (vect_recog_over_widening_pattern): Likewise.
240         (vect_recog_average_pattern): Likewise.
241         (vect_determine_min_output_precision_1): Likewise.
242         (vect_determine_precisions_from_range): Likewise.
243         (vect_determine_precisions_from_users): Likewise.
244         (vect_mark_pattern_stmts): Likewise.
245         (vect_pattern_recog_1): Likewise.
246         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
247         (vect_record_max_nunits): Likewise.
248         (vect_build_slp_tree_1): Likewise.
249         (vect_build_slp_tree_2): Likewise.
250         (vect_print_slp_tree): Likewise.
251         (vect_analyze_slp_instance): Likewise.
252         (vect_detect_hybrid_slp_stmts): Likewise.
253         (vect_detect_hybrid_slp_1): Likewise.
254         (vect_slp_analyze_operations): Likewise.
255         (vect_slp_analyze_bb_1): Likewise.
256         (vect_transform_slp_perm_load): Likewise.
257         (vect_schedule_slp_instance): Likewise.
258         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
259         (vect_mark_stmts_to_be_vectorized): Likewise.
260         (vect_init_vector_1): Likewise.
261         (vect_get_vec_def_for_operand): Likewise.
262         (vect_finish_stmt_generation_1): Likewise.
263         (vect_check_load_store_mask): Likewise.
264         (vectorizable_call): Likewise.
265         (vectorizable_conversion): Likewise.
266         (vectorizable_operation): Likewise.
267         (vectorizable_load): Likewise.
268         (vect_analyze_stmt): Likewise.
269         (vect_is_simple_use): Likewise.
270         (vect_get_vector_types_for_stmt): Likewise.
271         (vect_get_mask_type_for_stmt): Likewise.
272         * tree-vectorizer.c (increase_alignment): Likewise.
273
274 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
275
276         * doc/rtl.texi: Adjust vec_select description.
277         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
278         non-constant selectors.
279
280 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
281
282         * config/aarch64/aarch64-protos.h
283         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
284         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
285         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
286         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
287         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
288         (AARCH64_FL_PROFILE): Move index so flags are ordered.
289         (AARCH64_ISA_RCPC8_4): New flag.
290         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
291         to aarch64_offset_9bit_signed_unscaled_p.
292         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
293         and use stlur.
294         * config/aarch64/constraints.md (Ust): New constraint.
295         * config/aarch64/predicates.md.
296         (aarch64_9bit_offset_memory_operand): New predicate.
297         (aarch64_rcpc_memory_operand): New predicate.
298
299 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
300
301         PR rtl-optimization/87361
302         * rtlanal.c (nonzero_bits1): Revert accidental change.
303
304 2018-09-19  Richard Biener  <rguenther@suse.de>
305
306         PR tree-optimization/87349
307         PR tree-optimization/87342
308         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
309
310 2018-09-18  Marek Polacek  <polacek@redhat.com>
311
312         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
313         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
314
315 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
316
317         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
318
319 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
320
321         PR rtl-optimization/86882
322         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
323
324 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
325
326         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
327         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
328
329 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
330
331         PR other/87353
332         * doc/invoke.texi (Link Options): Fix formatting and grammar.
333
334 2018-09-18  Richard Biener  <rguenther@suse.de>
335
336         PR middle-end/63155
337         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
338         (compute_samebase_partition_bases): Likewise.
339         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
340         (gimple_can_coalesce_p): Simplify.
341
342 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
343
344         Handle a library implementation of ffs calling __builtin_ffs.
345         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
346         (mmix_init_libfuncs): New function: make __builtin_ffs expand
347         to __ffsdi2.
348
349 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
350
351         * diagnostic-show-locus.c (class layout_range): Add field
352         "m_original_idx".
353         (layout_range::layout_range): Add "original_idx" param and use it
354         to initialize new field.
355         (make_range): Use 0 for original_idx.
356         (layout::layout): Pass in index to calls to
357         maybe_add_location_range.
358         (layout::maybe_add_location_range): Add param "original_idx" and
359         pass it on to layout_range.
360         (layout::print_any_labels): Pass on range->m_original_idx to
361         get_text call.
362         (gcc_rich_location::add_location_if_nearby): Use 0 for
363         original_idx.
364         * gcc-rich-location.h (text_range_label::get_text): Update for new
365         param.
366         (range_label_for_type_mismatch::get_text): Likewise.
367
368 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
369
370         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
371
372 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
373
374         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
375         format_string_diagnostic_t.
376         (fmtwarn_n): Likewise.
377         * substring-locations.c
378         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
379         (format_warning_n_va): Convert to...
380         (format_string_diagnostic_t::emit_warning_n_va): ...this.
381         (format_warning_va): Convert to...
382         (format_string_diagnostic_t::emit_warning_va): ...this.
383         (format_warning_at_substring): Convert to...
384         (format_string_diagnostic_t::emit_warning): ...this.
385         (format_warning_at_substring_n): Convert to...
386         (format_string_diagnostic_t::emit_warning_n): ...this.
387         * substring-locations.h (class format_string_diagnostic_t): New
388         class.
389         (format_warning_va): Convert to
390         format_string_diagnostic_t::emit_warning_va.
391         (format_warning_n_va): Convert to
392         format_string_diagnostic_t::emit_warning_n_va.
393         (format_warning_at_substring): Convert to
394         format_string_diagnostic_t::emit_warning.
395         (format_warning_at_substring_n): Convert to
396         format_string_diagnostic_t::emit_warning_n.
397
398 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
399             Bernd Schmidt <bernds_cb1@t-online.de>
400
401         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
402         SImode args.
403
404 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
405
406         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
407         operand 0 predicate to nonimmediate operand.
408         (rint<mode>2_frndint): Remove insn pattern.
409         (rint<mode>2): Change operand 1 predicate to general_operand.
410         Extend operand 1 to XFmode and generate rintxf2 insn.
411         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
412         Do not use X87MODEF mode macro.
413         (frndintxf2_<rounding>_i387): Rename from
414         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
415         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
416         to XFmode and generate significandxf3 insn.
417
418 2018-09-17  Richard Biener  <rguenther@suse.de>
419
420         PR tree-optimization/87328
421         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
422         visiting unexecutable backedges when not iterating.
423         (do_rpo_vn): Mark all edges not executable even when not
424         iterating.
425
426 2018-09-17  Martin Jambor  <mjambor@suse.cz>
427
428         PR c/63886
429         * doc/invoke.texi (Warning Options): Likewise.
430
431 2018-09-17  Richard Biener  <rguenther@suse.de>
432
433         PR tree-optimization/87301
434         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
435         clean EH info from leftover copy assignments.
436
437 2018-09-17  Martin Liska  <mliska@suse.cz>
438
439         PR gcov-profile/85871
440         * gcov.c (output_intermediate_file): Fix out of bounds
441         access.
442
443 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
444
445         * config/arc/arc.c: Object attributes for core4 not reflected
446         correctly.
447         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
448         core3).
449
450 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
451
452         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
453
454 2018-09-17  Martin Liska  <mliska@suse.cz>
455
456         * doc/gcov.texi: Document new option --use-hotness-colors.
457         * gcov.c (struct source_info): Declare new field.
458         (source_info::source_info): Set default for maximum_count.
459         (print_usage): Add new -q option.
460         (process_args): Process it.
461         (accumulate_line_info): Save src->maximum_count.
462         (output_line_beginning): Make color line number if
463         flag_use_hotness_colors is set.
464         (output_line_details): Pass default argument value.
465         (output_lines): Pass src->maximum_count.
466
467 2018-09-17  Martin Liska  <mliska@suse.cz>
468
469         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
470         Use processor_names table.
471         * config/i386/i386.c (ix86_default_align): Use
472         processor_cost_table for alignment values.
473         (ix86_option_override_internal): Use processor_names.
474         (ix86_function_specific_print): Likewise.
475         * config/i386/i386.h (struct processor_costs):
476         Add alignment values.
477         (struct ptt): Remove and replace with const char *.
478         * config/i386/x86-tune-costs.h (struct processor_costs):
479         Declare default alignments for all costs.
480
481 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
482
483         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
484         symbolics or VR_VARYING ranges for ABS_EXPR.
485         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
486         when range will wrap.
487
488 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
489
490         PR middle-end/86864
491         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
492         before and after a JUMP_TABLE_DATA.
493
494 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
495
496         PR middle-end/87188
497         * dojump.c (do_compare_and_jump): Canonicalize function pointers
498         when one operand is a function pointer.  Use POINTER_TYPE_P and
499         FUNC_OR_METHOD_TYPE_P.
500         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
501         * fold-const.c (build_range_check): Likewise.
502         * match.pd (simple_comparison): Likewise.
503
504 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
505
506         PR c/82967
507         * spellcheck.c (get_edit_distance_cutoff): New function.
508         (selftest::test_edit_distance_unit_test_oneway): Rename to...
509         (selftest::test_get_edit_distance_one_way): ...this.
510         (selftest::test_get_edit_distance_unit): Rename to...
511         (selftest::test_get_edit_distance_both_ways): ...this.
512         (selftest::test_edit_distances): Move tests to this new function,
513         and test some more pairs of strings.  Update for above renaming.
514         (selftest::get_old_cutoff): New function.
515         (selftest::test_get_edit_distance_cutoff): New function.
516         (selftest::assert_suggested_for): New function.
517         (ASSERT_SUGGESTED_FOR): New macro.
518         (selftest::assert_not_suggested_for): New function.
519         (ASSERT_NOT_SUGGESTED_FOR): New macro.
520         (selftest::test_suggestions): New function.
521         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
522         tests to selftest::test_edit_distances and call it.  Add calls to
523         selftest::test_get_edit_distance_cutoff and
524         selftest::test_suggestions.
525         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
526         (best_match::consider): Replace hard-coded cutoff calculation with
527         a call to...
528         (best_match::get_cutoff): New declaration.
529         (best_match::get_best_meaningful_candidate): Likewise.
530
531 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
532
533         * builtins.c (fold_builtin_strlen): Remove TODO comment.
534
535 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
536
537         revert:
538         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
539
540         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
541         terminated string literal.
542
543 2018-09-14  Martin Sebor  <msebor@redhat.com>
544
545         * builtins.c (unterminated_array): Handle ARRAY_REF.
546         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
547         * builtins.h (unterminated_array): Declare extern.
548         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
549         arrays.
550         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
551         calls.
552
553 2018-09-14  Martin Sebor  <msebor@redhat.com>
554             Jeff Law  <law@redhat.com>
555
556         * builtins.c (unterminated_array): New.
557         (expand_builtin_strcpy): Adjust.
558         (expand_builtin_strcpy_args): Detect unterminated arrays.
559         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
560         unterminated arrays.
561         * gimple-fold.h (get_maxval_strlen): Add argument.
562         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
563
564         * gimple-fold.c (get_range_strlen): Add argument.
565         (get_maxval_strlen): Adjust.
566         * gimple-fold.h (get_range_strlen): Add argument.
567
568 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
569
570         * config/i386/movdirintrin.h: Fix copyright year.
571
572 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
573
574         * reg-stack.c: Include regs.h.
575         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
576         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
577         FIRST_STACK_REG, not DFmode.
578         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
579         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
580         (change stack): Default register mode to the reg_raw_mode of
581         FIRST_STACK_REG, not DFmode.
582         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
583         (*swapxf): Rename from swapxf.
584
585 2018-09-14  Carl Love  <cel@us.ibm.com>
586
587         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
588         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
589
590 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
591
592         PR target/87224
593         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
594         alternatives.
595
596 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
597
598         PR target/85628
599         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
600
601 2018-09-14  Jason Merrill  <jason@redhat.com>
602
603         Fix --enable-gather-detailed-mem-stats.
604         * hash-table.c (hash_table_usage): Change from variable to function.
605         * hash-table.h: Adjust.
606         * Makefile.in: Add missing dependencies on hash-table.h.
607
608 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
609
610         PR tree-optimization/87259
611         PR lto/87283
612         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
613         execute_cse_reciprocals_1 has tried transforming.
614
615 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
616
617         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
618         VR_VARYING for PLUS/MINUS_EXPR.
619
620 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
621
622         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
623         formatting.
624
625 2018-09-14  Richard Biener  <rguenther@suse.de>
626
627         PR middle-end/63155
628         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
629         bits for the merged partition.
630
631 2018-09-13  Martin Sebor  <msebor@redhat.com>
632             Bernd Edlinger  <bernd.edlinger@hotmail.de>
633
634         * builtins.h (c_srlen): Add argument.
635         * builtins.c (warn_string_no_nul): New function.
636         (c_strlen): Add argument and use it.  Update recursive calls.
637         Pass DECL argument to string_constant to get info on non
638         terminated strings.  Update *NONSTR as needed.
639         (fold_builtin_strlen): Add argument to calls to c_strlen.
640         Warn for unterminated arrays.
641         (warn_string_no_null): Add prototype.
642         * expr.c (string_constant): Update arguments.  Update recursive
643         calls appropriately.  Detect missing NUL terminator and outermost
644         declaration its missing in.
645         Improve checks for arrays with nonzero lower bound or elements
646         that are not a single byte.  Simplify offset computation.
647         Simplify checks for non-NUL terminated strings.
648         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
649         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
650
651 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
652
653         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
654         correctly.
655         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
656         argument.  Fix range checks.
657         * fold-const.h (c_getstr): Adjust protoype.
658         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
659         string is constant but contains no NUL byte.
660
661         * expr.c (string_constant): Adjust function comment.
662         Remove bogus check for zero termination.
663
664         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
665
666         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
667         (get_constant_size): Don't make STRING_CSTs larger than they are.
668         (check_string_literal): New check function for STRING_CSTs.
669         (output_constant): Use it.
670
671 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
672
673         PR target/86812
674         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
675
676 2018-09-13  Richard Biener  <rguenther@suse.de>
677
678         PR tree-optimization/87263
679         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
680         (struct unwind_state): Add max_rpo field.
681         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
682         Compute max_rpo, the max RPO number a block can be backwards reached
683         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
684         separating it from the iterating mode.
685
686 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
687
688         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
689         (rfs_decision): New scheduling decision.
690
691 2018-09-13  Richard Biener  <rguenther@suse.de>
692
693         PR bootstrap/87134
694         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
695         (vn_nary_op_insert_pieces_predicated): Do not write useless
696         valid_dominated_by_p entry outside of the allocated storage.
697
698 2018-09-13  Omar Sandoval  <osandov@osandov.com>
699             Tom de Vries  <tdevries@suse.de>
700
701         PR debug/86985
702         * dwarf2out.c (is_c): New function.
703         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
704
705 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
706
707         PR target/85628
708         * config/aarch64/aarch64.md (*aarch64_bfxil):
709         Define.
710         * config/aarch64/constraints.md (Ulc): Define.
711         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
712         Define.
713         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
714         New function.
715
716 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
717
718         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
719         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
720         aarch64_layout_frame call. 
721         (aarch64_expand_epilogue): Likewise.
722         (aarch64_initial_elimination_offset): Likewise.
723         (aarch64_get_separate_components): Likewise.
724         (aarch64_use_return_insn_p): Likewise.
725         (aarch64_layout_frame): Remove unneeded check.
726
727 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
728
729         * configure.ac: Only append
730         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
731         gcc_config_arguments if it was never reconfigured or last reconfigure
732         was with different arguments.
733         * configure: Regenerated.
734
735 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
736             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
737
738         PR middle-end/87290
739         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
740         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
741
742 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
743
744         PR tree-optimization/87287
745         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
746         X % C == 0 to X % (unsigned) C == 0 optimization to ...
747         * match.pd (X % C == 0): ... here.  New optimization.
748
749 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
750
751         PR middle-end/82853
752         * expr.h (maybe_optimize_mod_cmp): Declare.
753         * expr.c (mod_inv): New function.
754         (maybe_optimize_mod_cmp): New function.
755         (do_store_flag): Use it.
756         * cfgexpand.c (expand_gimple_cond): Likewise.
757
758 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
759             Julian Brown  <julian@codesourcery.com>
760
761         PR middle-end/86336
762         * gimplify.c (gimplify_scan_omp_clauses): Set
763         target_firstprivatize_array_bases in OpenACC parallel and kernels
764         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
765         OpenACC data regions.
766
767 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
768
769         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
770         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
771
772 2018-09-12  Richard Biener  <rguenther@suse.de>
773
774         PR tree-optimization/87280
775         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
776         edge but unreachable target.
777         (do_rpo_vn): For conservatively handling a PHI only mark
778         the backedge executable but not the block reachable.
779
780 2018-09-12  Richard Biener  <rguenther@suse.de>
781
782         PR tree-optimization/87266
783         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
784         visited blocks.
785
786 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
787
788         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
789         constants.
790         ("trunc<BFP:mode><DFP_ALL:mode>2")
791         ("trunc<DFP_ALL:mode><BFP:mode>2")
792         ("extend<BFP:mode><DFP_ALL:mode>2")
793         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
794         according to the target operand type.
795
796 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
797             Andreas Krebbel  <krebbel@linux.ibm.com>
798
799         PR tree-optimization/86844
800         * gimple-ssa-store-merging.c
801         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
802         there are any overlapping stores in between them, make sure they are
803         also coalesced or we give up completely.
804
805 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
806
807         PR middle-end/87248
808         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
809         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
810
811 2018-09-12  Tom de Vries  <tdevries@suse.de>
812
813         * common.opt (gdescribe-dies): Add option.
814         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
815         attribute for artifical and nameless decls.
816         (dwarf2out_register_external_die): Add description attribute to
817         external reference die.
818         (add_desc_attribute): New functions.
819         (gen_subprogram_die): Add description attribute to
820         DW_TAG_call_site_parameter.
821         * tree-pretty-print.c (print_generic_expr_to_str): New function.
822         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
823         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
824         -gno-describe-dies.
825         (@item -gdescribe-dies): Add.
826
827 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
828
829         * tree-vrp.c (vrp_shift_undefined_p): Remove.
830         (extract_range_from_binary_expr_1: Call
831         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
832         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
833         depend on sign.
834
835 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
836
837         * gimple-ssa-warn-alloca.c
838         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
839         field for ALLOCA_BOUND_*_LARGE.
840
841 2018-09-11  Nathan Sidwell  <nathan@acm.org>
842
843         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
844
845 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
846
847         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
848         for clobbers.  Remove obsolete comment.
849
850 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
851
852         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
853         mpxchk, mpxld and mpxst types.
854         (define_attr length_immediate): Remove all processing of mpx types.
855         (define_attr prefix_0f): Ditto.
856         (define_attr memory): Ditto.
857
858 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
859
860         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
861         (log<mode>2): Change operand 1 predicate to general_operand.
862         Extend operand 1 to XFmode and generate logxf3 insn.
863         (log10<mode>2): Change operand 1 predicate to general_operand.
864         Extend operand 1 to XFmode and generate log10xf3 insn.
865         (log2<mode>2): Change operand 1 predicate to general_operand.
866         Extend operand 1 to XFmode and generate log2xf3 insn.
867         (fyl2xp1_extend<mode>xf3_i387): Remove.
868         (log1p<mode>2): Change operand 1 predicate to general_operand.
869         Extend operand 1 to XFmode and generate log1pxf3 insn.
870         (fxtract_extend<mode>xf3_i387): Remove.
871         (logb<mode>2): Change operand 1 predicate to general_operand.
872         Extend operand 1 to XFmode and generate logbxf3 insn.
873         (ilogb<mode>2): Change operand 1 predicate to general_operand.
874         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
875         (significand<mode>2): Change operand 1 predicate to general_operand.
876         Extend operand 1 to XFmode and generate significandxf3 insn.
877
878 2018-09-11  Nathan Sidwell  <nathan@acm.org>
879
880         * gcc.c (perror_with_name, pfatal_with_name): Delete.
881         (load_specs): Use fatal_error.
882         (DELETE_IF_ORDINARY, process_command): Use error.
883         (execute, run_attempt): Use fatal_error.
884
885 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
886
887         * diagnostic-core.h (sorry_at): New prototype.
888         * diagnostic.c (sorry_at): New function.
889
890 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
891
892         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
893         by zero as VR_UNDEFINED.
894
895 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
896
897         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
898         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
899         (<sincos>mode2): New expander.
900         (sincos_extend<mode>xf3_i387): Remove insn pattern.
901         (sincos -> sin, cos splitters): Remove splitter patterns.
902         (sincos<mode>3): Change operand 2 predicate to general_operand.
903         Extend operand 2 to XFmode and generate sincosxf3 insn.
904         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
905         Change operand 3 predicate to const1_operand.
906         (fptan_extend<mode>xf4_i387): Remove insn pattern.
907         (tanxf2): Update operands in the call to fptanxf4_i387.
908         (tan<mode>2): Change operand 1 predicate to general_operand.
909         Extend operand 1 to XFmode and generate tanxf3 insn.
910         (atan2xf3): Rename from *fpatanxf3_i387.
911         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
912         (atan2xf3): Remove expander.
913         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
914         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
915         (atan<mode>2): Change operand 1 predicate to general_operand.
916         Extend operand 1 to XFmode and generate atanxf3 insn.
917
918 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
919
920         * config/i386/i386.md (x87/SSE constant load splitter): Use
921         memory_operand instead of nonimmediate_operand for input operand
922         predicate.
923
924 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
925
926         * config/i386/i386.md (float partial SSE register stall splitter): Move
927         splitter near its instruction pattern.
928         (float_extend partial SSE register stall splitter): Ditto.
929         (float_truncate partial SSE register stall splitter): Ditto.
930
931 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
932
933         PR target/86794
934         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
935         to speculation_safe_value_not_needed.
936
937         PR target/85666
938         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
939         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
940         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
941         leaf_function_p, instead use has_hard_reg_initial_val.
942
943 2018-09-09  Nathan Sidwell  <nathan@acm.org>
944
945         * gcc.h (pfatal_with_name): Don't declare here.
946         * gcc.c (pfatal_with_name): Make static.
947
948 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
949
950         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
951         earlyclobber.
952
953 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
954
955         PR rtl-optimization/85458
956         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
957         priority hook to reduce the priority of EXPR.
958
959 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
960
961         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
962         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
963         with X87_ENABLE_FLOAT.  Remove preparation code.
964         (*float<SWI48:mode><MODEF:mode>2): Rename from
965         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
966         math using "enabled" attribute.
967         (*floatdi<MODEF:mode>2_i387): Rename from
968         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
969         enable for 32bit targets only.
970         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
971         splitter.
972         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
973         as operand 1 predicate.  Rewrite as define_insn_and_split.
974         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
975
976 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
977
978         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
979         to fallthru to FLOAT case.
980
981 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
982
983         PR target/86731
984         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
985         around folding of vec_sl to handle out of range shift values.
986
987 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
988
989         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
990         Update callers to gen_fix_trunc<mode>_i387_fisttp
991         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
992         nonimmediate_operand.
993         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
994         and corresponding splitters.
995         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
996         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
997         (fix_truncdi_i387_with_temp): Remove insn pattern
998         and corresponding splitters.
999         (fix_trunc<mode>_i387): Change operand 0 predicate to
1000         nonimmediate_operand.
1001         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
1002         and corresponding splitters.
1003         (*fistdi2_1): Remove.
1004         (fistdi2): Ditto.
1005         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
1006         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
1007         (*fist<mode>2_1): Remove.
1008         (fist<mode>2): Ditto.
1009         (fist<mode>2_with_temp): Remove insn pattern and corresponding
1010         splitters.
1011         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
1012         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
1013         (fistdi2_<rounding>): Change operand 0 predicate to
1014         nonimmediate_operand.
1015         (fistdi2_<rounding>_with_temp): Remove insn pattern
1016         and corresponding splitters.
1017         (fist<mode>2_<rounding>): Change operand 0 predicate to
1018         nonimmediate_operand.
1019         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
1020         and corresponding splitters.
1021
1022         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
1023
1024 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1025
1026         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1027         the init value.
1028
1029 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
1030
1031         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1032         early gimple folding of vec_splat().
1033         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
1034         * gimple-fold.h: Add an extern define for tree_vec_extract().
1035
1036 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
1037
1038         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1039         wrappers around TREE_TYPE comparisons.
1040
1041 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1042
1043         PR target/80080
1044         * config/s390/predicates.md: Add nonsym_memory_operand.
1045         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1046         contains a SYMBOL_REF, load it into an intermediate pseudo.
1047         (s390_emit_compare_and_swap): Legitimize operand.
1048         * config/s390/s390.md: Use the new nonsym_memory_operand
1049         with UNSPECV_CAS patterns.
1050
1051 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1052
1053         PR target/80080
1054         * config/s390/s390-passes.def: New file.
1055         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1056         declaration.
1057         (make_pass_s390_early_mach): Add declaration.
1058         * config/s390/s390.c (make_pass_s390_early_mach):
1059         (s390_option_override): Remove dynamic registration.
1060         * config/s390/t-s390: Add s390-passes.def.
1061
1062 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1063
1064         * config/s390/s390.c (s390_decompose_constant_pool_ref):
1065         Remove UNSPEC_LTREL_BASE check.
1066         (annotate_constant_pool_refs): Likewise.
1067         (find_constant_pool_ref): Likewise.
1068         (find_ltrel_base): Removed.
1069         (replace_ltrel_base): Removed.
1070         (s390_mainpool_finish): Remove replace_ltrel_base call.
1071         (s390_chunkify_start): Remove pending LTREL_BASE logic.
1072         (s390_chunkify_finish): Remove replace_ltrel_base call.
1073         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1074
1075 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
1076
1077         PR target/86779
1078         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1079         to speculation_safe_value_not_needed.
1080
1081 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
1082             Bernd Schmidt  <bernds_cb1@t-online.de>
1083
1084         * config/nvptx/nvptx-opts.h: New file.
1085         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1086         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1087         (ASM_SPEC): Define.
1088         (TARGET_SM35): New macro.
1089         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1090         correct predicate.
1091         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1092         values.
1093         (misa=): New option.
1094         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1095
1096 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
1097
1098         * config/i386/i386.md (truncdfsf2): Remove expander.
1099         (truncdfsf2_with_temp): Ditto.
1100         (truncxf<mode>2): Ditto.
1101         (*truncdfsf_fast_mixed): Remove insn pattern.
1102         (*truncdfsf_fast_i387): Ditto.
1103         (*truncdfsf_mixed): Ditto.
1104         (*truncdfsf_i387): Ditto.
1105         (*truncdfsf2_i387_1): Ditto.
1106         (*truncxfsf2_mixed): Ditto.
1107         (*truncxfdf2_mixed): Ditto.
1108         (*truncxf<mode>2_i387_noop): Ditto. Update callers
1109         to call gen_truncxf<mode>2 instead.
1110         (*truncxf<mode>2_i387): Remove.
1111         (reg->reg splitters): Remove splitter pattern.
1112         (reg->mem splitters): Ditto.
1113
1114         (truncdfsf2): New insn pattern.
1115         (truncxf<mode>2): Ditto.
1116
1117 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1118
1119         * tree-ssa-math-opts.c (is_mult_by): New function.
1120         (is_square_of): Use the above.
1121         (optimize_recip_sqrt): New function.
1122         (pass_cse_reciprocals::execute): Use the above.
1123
1124 2018-09-05  Richard Biener  <rguenther@suse.de>
1125
1126         PR bootstrap/87134
1127         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1128         to zero-init the emplaced vec.
1129
1130 2018-09-05  Martin Liska  <mliska@suse.cz>
1131
1132         PR tree-optimization/87205
1133         * tree-switch-conversion.c (pass_lower_switch::execute):
1134         Group cases for switch statements.
1135
1136 2018-09-05  Richard Biener  <rguenther@suse.de>
1137
1138         PR tree-optimization/87217
1139         * tree-ssa-sccvn.c (vuse_valueize): New.
1140         (vn_reference_lookup_pieces): Use it.
1141         (vn_reference_lookup): Likewise.
1142
1143 2018-09-05  Nathan Sidwell  <nathan@acm.org>
1144
1145         PR c++/87137
1146         * stor-layout.c (place_field): Scan forwards to check last
1147         bitfield when ms_bitfield_placement is in effect.
1148
1149 2018-09-05  Richard Biener  <rguenther@suse.de>
1150
1151         PR bootstrap/87225
1152         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1153         return.
1154
1155 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
1156             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
1157
1158         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1159         * config.gcc (extra_objs): Build it.
1160         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1161         Likewise.
1162         * config/aarch64/aarch64-passes.def
1163         (pass_tag_collision_avoidance): New pass.
1164         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1165         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1166         (aarch64_classify_address): Remove static qualifier.
1167         (aarch64_address_info, aarch64_address_type): Move to...
1168         * config/aarch64/aarch64-protos.h: ... here.
1169         (make_pass_tag_collision_avoidance): New function.
1170         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1171         New tuning flag.
1172
1173 2018-09-05  Martin Liska  <mliska@suse.cz>
1174
1175         * doc/gcov.texi: Update documentation of humar
1176         readable mode.
1177         * gcov.c (format_count): Print one decimal place, it provides
1178         more fine number of situations like '1G' vs. '1.4G'.
1179
1180 2018-09-05  Martin Liska  <mliska@suse.cz>
1181
1182         PR target/87164
1183         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1184         * optc-gen.awk: Allow 'Var' for Deprecated options in order
1185         to generate a MASK value.
1186
1187 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
1188
1189         PR debug/86593
1190         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1191         if frame pointer isn't used.
1192         (compute_frame_pointer_to_fb_displacement): Likewise.
1193
1194 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
1195
1196         PR target/87198
1197         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1198         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1199         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1200         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1201         and OPTION_MASK_ISA_XSAVEC_UNSET.
1202
1203 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
1204
1205         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1206         XOR operations in NAND case.
1207
1208 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
1209
1210         * wide-int-range.cc (wide_int_range_convert): New.
1211         * wide-int-range.h (wide_int_range_convert): New.
1212         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1213         code into wide_int_range_convert.
1214         (extract_range_into_wide_ints): Do not munge anti range constants
1215         into the entire domain.  Just return the range back.
1216
1217 2018-09-04  Martin Liska  <mliska@suse.cz>
1218
1219         * genmatch.c (output_line_directive): Add new argument
1220         fnargs.
1221         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1222
1223 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
1224
1225         * doc/invoke.texi (Option Summary): Add whitespace.
1226
1227         * doc/invoke.texi (Option Summary): Add -Waligned-new.
1228
1229 2018-09-04  Richard Biener  <rguenther@suse.de>
1230
1231         PR tree-optimization/87211
1232         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1233         backedge value we're supposed to treat as VARYING also number
1234         the PHI to VARYING in case it got a different value-number already.
1235
1236 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
1237
1238         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
1239         (extract_range_from_binary_expr_1): Do not call
1240         vrp_can_optimize_bit_op.
1241         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
1242         static.
1243         (wide_int_range_get_mask_and_bounds): New.
1244         (wide_int_range_optimize_bit_op): New.
1245         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
1246         (wide_int_range_bit_and): Same.
1247         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
1248         (wide_int_range_optimize_bit_op): New.
1249         (wide_int_range_get_mask_and_bounds): New.
1250
1251 2018-09-04  Richard Biener  <rguenther@suse.de>
1252
1253         PR tree-optimization/87176
1254         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
1255         variable.  When value-numbering a virtual PHI node make sure
1256         to not value-number to the backedge value.
1257
1258 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
1259
1260         * doc/extend.texi (Long Long, Hex Floats): Document support for
1261         long long and hex floats in more recent versions of ISO C++.
1262
1263 2018-09-03  Richard Biener  <rguenther@suse.de>
1264
1265         PR tree-optimization/87177
1266         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
1267         cleanup.
1268
1269 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1270
1271         * bb-reorder.c (edge_order): Convert to C-qsort-style
1272         tri-state comparator.
1273         (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
1274
1275 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1276
1277         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
1278         tri-state comparator.
1279         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
1280
1281 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1282
1283         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
1284         (mergesort): ... here as maximum count for using netsort.
1285         (gcc_qsort): Set nlim to 3 if stable sort is requested.
1286         (gcc_stablesort): New.
1287         * system.h (gcc_stablesort): Declare.
1288
1289 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
1290
1291         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
1292         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
1293         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
1294
1295 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
1296
1297         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
1298         lxsdx and stxsdx alternatives.
1299         (*mov<mode>_hardfloat64): Ditto.
1300         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
1301
1302 2018-09-03  Richard Biener  <rguenther@suse.de>
1303
1304         PR tree-optimization/87200
1305         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1306         simplify result.
1307
1308 2018-09-03  Martin Liska  <mliska@suse.cz>
1309
1310         PR tree-optimization/87201
1311         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1312         Fix parenthesis in an expression.
1313
1314 2018-09-03  Richard Biener  <rguenther@suse.de>
1315
1316         PR tree-optimization/87197
1317         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1318         visited.  CSE the VN_INFO hashtable lookup.
1319
1320         PR tree-optimization/87169
1321         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1322         iterating make sure there's no extra backedges from irreducible
1323         regions feeding the header.  Mark the destination block
1324         executable.
1325
1326 2018-09-03  Martin Liska  <mliska@suse.cz>
1327
1328         PR driver/83193
1329         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1330         * common/common-targhooks.c (default_get_valid_option_values):
1331         New function.
1332         * common/common-targhooks.h (default_get_valid_option_values):
1333         Likewise.
1334         * common/config/i386/i386-common.c: Move processor_target_table
1335         from i386.c.
1336         (ix86_get_valid_option_values): New function.
1337         (TARGET_GET_VALID_OPTION_VALUES): New macro.
1338         * config/i386/i386.c (struct ptt): Move to i386-common.c.
1339         (PTA_*): Move all defined masks into i386-common.c.
1340         (ix86_function_specific_restore): Use new processor_cost_table.
1341         * config/i386/i386.h (struct ptt): Moved from i386.c.
1342         (struct pta): Likewise.
1343         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1344         * doc/tm.texi.in: Likewise.
1345         * opt-suggestions.c (option_proposer::suggest_option):
1346         Pass prefix to build_option_suggestions.
1347         (option_proposer::get_completions): Likewise.
1348         (option_proposer::build_option_suggestions): Use the new target
1349         hook.
1350         * opts.c (struct option_help_tuple): New struct.
1351         (print_filtered_help): Use the new target hook.
1352
1353 2018-09-03  Martin Liska  <mliska@suse.cz>
1354
1355         PR middle-end/59521
1356         * predict.c (set_even_probabilities): Add likely_edges
1357         argument and handle cases where we have precisely one
1358         likely edge.
1359         (combine_predictions_for_bb): Catch also likely_edges.
1360         (tree_predict_by_opcode): Handle gswitch statements.
1361         * tree-cfg.h (find_case_label_for_value): New declaration.
1362         (find_taken_edge_switch_expr): Likewise.
1363         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1364         Find pivot in decision tree based on probabily, not by number of
1365         nodes.
1366
1367 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
1368
1369         * doc/standards.texi (Standards): Update Objective-C reference.
1370
1371 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1372
1373         * doc/install.texi (Prerequisites): Update link for MPC.
1374
1375 2018-09-01  Michael Matz  <matz@suse.de>
1376
1377         PR tree-optimization/87074
1378         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1379         PHIs for outer-loop uses.
1380
1381 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1382
1383         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1384         * doc/invoke.texi (C Dialect Options): Ditto.
1385
1386 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1387
1388         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1389
1390 2018-08-31  Richard Biener  <rguenther@suse.de>
1391
1392         PR tree-optimization/87168
1393         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1394         (rpo_elim::eliminate_avail): When OP was not visited it must
1395         be available.
1396
1397 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
1398
1399         * tree-vrp.c (copy_value_range): Convert param "from" from
1400         "value_range *" to "const value_range *".
1401         (range_is_null): Likewise for param "vr".
1402         (range_int_cst_p): Likewise.
1403         (range_int_cst_singleton_p): Likewise.
1404         (symbolic_range_p): Likewise.
1405         (value_ranges_intersect_p): Likewise for both params.
1406         (value_range_nonnegative_p): Likewise for param "vr".
1407         (value_range_constant_singleton): Likewise.
1408         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1409         (extract_range_into_wide_ints): Likewise for param "vr".
1410         (extract_range_from_multiplicative_op): Likewise for params "vr0"
1411         and "vr1".
1412         (vrp_can_optimize_bit_op): Likewise.
1413         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1414         "vr1_".
1415         (extract_range_from_unary_expr): Likewise.
1416         (debug_value_range): Likewise for param "vr".
1417         (value_range::dump): Add "const" qualifier.
1418         (vrp_prop::check_array_ref): Convert local "vr" from
1419         "value_range *" to "const value_range *".
1420         (vrp_prop::check_mem_ref): Likewise.
1421         (vrp_prop::visit_stmt): Likewise for local "old_vr".
1422         (vrp_intersect_ranges_1): Likewise for param "vr_1".
1423         (vrp_intersect_ranges): Likewise.
1424         (simplify_stmt_for_jump_threading): Likewise for local "vr".
1425         (vrp_prop::vrp_finalize): Likewise.
1426         * tree-vrp.h (value_range::dump): Add "const" qualifier.
1427         (vrp_intersect_ranges): Add "const" qualifier to params as above.
1428         (extract_range_from_unary_expr): Likewise.
1429         (value_range_constant_singleton): Likewise.
1430         (symbolic_range_p): Likewise.
1431         (copy_value_range): Likewise.
1432         (extract_range_from_binary_expr_1): Likewise.
1433         (range_int_cst_p): Likewise.
1434         (vrp_set_zero_nonzero_bits): Likewise.
1435         (range_int_cst_singleton_p): Likewise.
1436
1437 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
1438
1439         * config/aarch64/arm_neon.h (vabsd_s64): New.
1440         (vnegd_s64): Likewise.
1441
1442 2018-08-31  Martin Jambor  <mjambor@suse.cz>
1443
1444         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1445
1446 2018-08-31  Martin Liska  <mliska@suse.cz>
1447
1448         * ipa-icf.c (sem_item::add_type): Use
1449         sem_item::m_type_hash_cache.
1450         * ipa-icf.h: Move the cache from sem_item_optimizer
1451         to sem_item.
1452
1453 2018-08-31  Nathan Sidwell  <nathan@acm.org>
1454
1455         * doc/extend.texi (Backwards Compatibility): Remove implicit
1456         extern C leeway of () being (...).
1457
1458 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1459
1460         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1461
1462 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
1463
1464         PR target/86684
1465         PR target/87149
1466         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1467
1468 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
1469
1470         PR middle-end/87138
1471         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1472         gen_int_mode.  Formatting fixes.
1473
1474 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
1475
1476         * target.def (custom_function_descriptors): Improve documentation.
1477         * doc/tm.texi.in (Trampolines): Expand discussion of function
1478         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1479         beginning of the section.
1480         * doc/tm.texi: Regenerated.
1481
1482 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
1483
1484         * cfg.h (class auto_edge_flag): Spell out the template-id of the
1485         base class in the initializer list.  This is a workaround for
1486         building with older GCC.
1487         (class auto_bb_flag): Likewise.
1488
1489 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
1490
1491         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1492         (altivec_vcmpequ<VI_char>_p): Remove star.
1493         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1494         vector load modes.
1495         (expand_strncmp_vec_sequence): New function.
1496         (emit_final_str_compare_vec): New function.
1497         (expand_strn_compare): Add support for vector strncmp.
1498         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1499         length specification to bytes.
1500         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1501         (vcmpnezb_p): New pattern.
1502         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1503         for option -mstring-compare-inline-limit.
1504
1505 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
1506
1507         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1508         (PTA_SKYLAKE): Add PTA_AES.
1509         (PTA_GOLDMONT): Likewise.
1510
1511 2018-08-29  Jan Hubicka  <jh@suse.cz>
1512
1513         PR lto/86517
1514         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1515         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1516
1517 2018-08-29  Jan Hubicka  <jh@suse.cz>
1518
1519         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1520         TYPE_STUB_DECL.
1521         (hash_tree): Do not visit TYPE_STUB_DECL.
1522         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1523         stream TYPE_STUB_DECL.
1524         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1525         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1526         after free_lang_data.
1527         (type_in_anonymous_namespace_p): Likewise.
1528
1529 2018-08-29  Jan Hubicka  <jh@suse.cz>
1530
1531         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1532         comment that it has to be even number.
1533         (class sreal): Change m_sig type to int32_t.
1534         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1535         int64_t for temporary calculations.
1536         (sreal_verify_basics): Drop one bit from minimum and maximum.
1537
1538 2018-08-30  Richard Biener  <rguenther@suse.de>
1539
1540         PR tree-optimization/87147
1541         * tree-ssa-sccvn.c (SSA_VISITED): New function.
1542         (visit_phi): When the degenerate result is from the backedge and
1543         we didn't visit its definition yet drop to VARYING.
1544         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1545
1546 2018-08-29  Jan Hubicka  <jh@suse.cz>
1547
1548         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1549         DECL_VINDEX.
1550         (hash_tree): Likewise.
1551
1552 2018-08-29  Jan Hubicka  <jh@suse.cz>
1553
1554         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1555         and TYPE_NEXT_REF_TO.
1556
1557 2018-08-29  Jan Hubicka  <jh@suse.cz>
1558
1559         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1560         comment that it has to be even number.
1561         (class sreal): Change m_sig type to int32_t.
1562         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1563         int64_t for temporary calculations.
1564         (sreal_verify_basics): Drop one bit from minimum and maximum.
1565
1566 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
1567
1568         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1569
1570 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
1571
1572         PR middle-end/86995
1573         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1574         if to_add is negative.
1575
1576 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1577
1578         PR middle-end/87053
1579         * builtins.c (c_strlen): Improve range checks.
1580
1581 2018-08-29  Martin Sebor  <msebor@redhat.com>
1582             Jeff Law  <law@redhat.com>
1583
1584         PR tree-optimization/86714
1585         PR tree-optimization/86711
1586         * builtins.c (c_strlen): Add arguments to call to string_constant.
1587         * expr.c (string_constant): Add argument.  Detect missing nul
1588         terminator and outermost declaration it's missing in.
1589         * expr.h (string_constant): Add argument.
1590         * fold-const.c (read_from_constant_string): Add arguments to call to
1591         string_constant.
1592         (c_getstr): Likewise.
1593         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1594         to string_constant.
1595         * tree-ssa-strlen.c (get_stridx): Likewise.
1596
1597 2018-08-29  Jan Hubicka  <jh@suse.cz>
1598
1599         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1600         Do not stream DECL_VINDEX.
1601         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1602         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1603         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1604
1605 2018-08-29  Richard Biener  <rguenther@suse.de>
1606
1607         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1608         virtual operands that are not default defs to honor region
1609         boundaries.
1610         (rpo_vn_valueize): Remove ineffective code here.
1611
1612 2018-08-29  Richard Biener  <rguenther@suse.de>
1613
1614         PR tree-optimization/87132
1615         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1616         when skipping defs reachable over backedges.
1617
1618 2018-08-29  Richard Biener  <rguenther@suse.de>
1619
1620         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1621         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1622         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1623         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1624         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1625         (vn_reference_lookup_3): Remove use of const_parms.
1626         (free_rpo_vn): Do not free const_parms.
1627         (do_rpo_vn): Do not call init_const_parms.
1628         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1629         SSA_NAME_POINTS_TO_READONLY_MEMORY.
1630         (call_may_clobber_ref_p_1): Likewise.
1631
1632 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
1633
1634         PR other/86726
1635         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1636         (-O): Ditto.
1637         (-ftree-scev-cprop): Document.
1638
1639 2018-08-29  Jan Hubicka  <jh@suse.cz>
1640
1641         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1642         parameters.
1643         (sreal constructor): Update.
1644         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1645         sreal:operator/): Update.
1646
1647 2018-08-29  Martin Liska  <mliska@suse.cz>
1648
1649         * tree-switch-conversion.c (switch_conversion::expand):
1650         Strenghten assumption about gswitch statements.
1651
1652 2018-08-29  Richard Biener  <rguenther@suse.de>
1653
1654         PR tree-optimization/87117
1655         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1656         re-value-number released SSA VDEFs.
1657
1658 2018-08-29  Richard Biener  <rguenther@suse.de>
1659
1660         PR tree-optimization/87126
1661         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1662
1663 2018-08-28  Jim Wilson  <jimw@sifive.com>
1664
1665         * config/riscv/pic.md: Rewrite.
1666         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1667         invalid address.
1668         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1669         (SOFTF, default_load, softload, softstore): New.
1670
1671 2018-08-28  Jeff Law  <law@redhat.com>
1672
1673         * fold-const.c (fold_binary_loc): Remove recently added assert.
1674
1675 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
1676
1677         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1678         to OP parmeter of generated function.
1679
1680 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
1681
1682         PR tree-optimization/87009
1683         * match.pd: Add boolean optimizations.
1684
1685 2018-08-28  Martin Sebor  <msebor@redhat.com>
1686
1687         PR middle-end/86631
1688         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1689         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1690         (pass_walloca::gate): Use it.
1691         (alloca_call_type): Same.
1692         (pass_walloca::execute): Same.
1693         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1694
1695 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
1696
1697         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1698         GCC_VERSION for usage of "__gcc_dump_printf__" format from
1699         >= 3005 to >= 9000.
1700
1701 2018-08-28  Richard Biener  <rguenther@suse.de>
1702
1703         PR tree-optimization/87124
1704         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1705         constants before looking up avail.
1706
1707 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
1708
1709         PR middle-end/87099
1710         * calls.c (maybe_warn_nonstring_arg): Punt early if
1711         warn_stringop_overflow is zero.  Don't call get_range_strlen
1712         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1713         Swap comparison operands to have constants on rhs.  Only use
1714         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
1715         increment lenrng[0].
1716
1717 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
1718
1719         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1720         use of tree_to_shwi.  Remove duplicated test for the size being
1721         a whole number of bytes.
1722
1723 2018-08-28  Richard Biener  <rguenther@suse.de>
1724
1725         PR tree-optimization/87117
1726         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1727         Handle removed stmt without LHS (GIMPLE_NOP).
1728
1729 2018-08-28  Richard Biener  <rguenther@suse.de>
1730
1731         PR tree-optimization/87117
1732         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1733         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1734
1735 2018-08-28  Richard Biener  <rguenther@suse.de>
1736
1737         PR tree-optimization/87117
1738         * tree-ssa-pre.c (compute_avail): Do not make expressions
1739         with predicated values available.
1740         (get_expr_value_id): Assert we do not run into predicated value
1741         expressions.
1742
1743 2018-08-28  Richard Biener  <rguenther@suse.de>
1744
1745         PR tree-optimization/87117
1746         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1747         get virtual operands.
1748         (get_expr_operands): Handle STRING_CST like other decls.
1749
1750 2018-08-28  Martin Liska  <mliska@suse.cz>
1751
1752         * tree.h: Update documentation of fndecl_built_in_p
1753         functions.
1754
1755
1756 2018-08-27  Jeff Law  <law@redhat.com>
1757         PR tree-optimization/87110
1758         * tree-ssa-dse.c (compute_trims): Handle non-constant
1759         TYPE_SIZE_UNIT.
1760
1761 2018-08-27  Martin Sebor  <msebor@redhat.com>
1762
1763         PR tree-optimization/86914
1764         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1765
1766 2018-08-27  Martin Sebor  <msebor@redhat.com>
1767
1768         PR tree-optimization/87112
1769         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1770         the type of the bound argument.
1771
1772 2018-08-27  Jeff Law  <law@redhat.com>
1773
1774         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1775         type does not have a TYPE_SIZE_UNIT.
1776
1777 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
1778
1779         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1780         with include of backend.h.
1781
1782 2018-08-27  Richard Biener  <rguenther@suse.de>
1783
1784         PR tree-optimization/86927
1785         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1786         use const cond reduction code.
1787
1788 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
1789
1790         PR tree-optimization/85758
1791         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1792
1793 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
1794
1795         PR c++/87091
1796         * diagnostic-show-locus.c (class layout_range): Update for
1797         conversion of show_caret_p to a tri-state.
1798         (layout_range::layout_range): Likewise.
1799         (make_range): Likewise.
1800         (layout::maybe_add_location_range): Likewise.
1801         (layout::should_print_annotation_line_p): Don't show annotation
1802         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1803         (layout::get_state_at_point): Update for conversion of
1804         show_caret_p to a tri-state.  Bail out early for
1805         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1806         underlining or source colorization.
1807         (gcc_rich_location::add_location_if_nearby): Update for conversion
1808         of show_caret_p to a tri-state.
1809         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1810         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1811         Likewise.
1812         (selftest::test_one_liner_labels): Likewise.
1813         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1814         conversion of show_caret_p to a tri-state.
1815         * pretty-print.c (text_info::set_location): Likewise.
1816         * pretty-print.h (text_info::set_location): Likewise.
1817         * substring-locations.c (format_warning_n_va): Likewise.
1818         * tree-diagnostic.c (default_tree_printer): Likewise.
1819         * tree-pretty-print.c (newline_and_indent): Likewise.
1820
1821 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
1822
1823         PR c++/87091
1824         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1825         line above for line-insertion fix-it hints.
1826         (selftest::test_fixit_insert_containing_newline): Update the
1827         expected results, and add a test with line-numbering enabled.
1828
1829 2018-08-27  Martin Liska  <mliska@suse.cz>
1830
1831         PR sanitizer/86962
1832         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1833         params with DECL_HAS_VALUE_EXPR_P.
1834
1835 2018-08-27  Martin Liska  <mliska@suse.cz>
1836
1837         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1838         selected expansion strategy.
1839
1840 2018-08-27  Martin Liska  <mliska@suse.cz>
1841
1842         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1843         * builtins.c (is_builtin_fn): Likewise.
1844         * attribs.c (diag_attr_exclusions): Use new function
1845         fndecl_built_in_p and remove check for FUNCTION_DECL if
1846         possible.
1847         (builtin_mathfn_code): Likewise.
1848         (fold_builtin_expect): Likewise.
1849         (fold_call_expr): Likewise.
1850         (fold_builtin_call_array): Likewise.
1851         (fold_call_stmt): Likewise.
1852         (set_builtin_user_assembler_name): Likewise.
1853         (is_simple_builtin): Likewise.
1854         * calls.c (gimple_alloca_call_p): Likewise.
1855         (maybe_warn_nonstring_arg): Likewise.
1856         * cfgexpand.c (expand_call_stmt): Likewise.
1857         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1858         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1859         (cgraph_node::verify_node): Likewise.
1860         * cgraphclones.c (build_function_decl_skip_args): Likewise.
1861         (cgraph_node::create_clone): Likewise.
1862         * config/arm/arm.c (arm_insert_attributes): Likewise.
1863         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1864         * dse.c (scan_insn): Likewise.
1865         * expr.c (expand_expr_real_1): Likewise.
1866         * fold-const.c (operand_equal_p): Likewise.
1867         (fold_binary_loc): Likewise.
1868         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1869         * gimple-low.c (lower_stmt): Likewise.
1870         * gimple-pretty-print.c (dump_gimple_call): Likewise.
1871         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1872         * gimple.c (gimple_build_call_from_tree): Likewise.
1873         (gimple_call_builtin_p): Likewise.
1874         (gimple_call_combined_fn): Likewise.
1875         * gimplify.c (gimplify_call_expr): Likewise.
1876         (gimple_boolify): Likewise.
1877         (gimplify_modify_expr): Likewise.
1878         (gimplify_addr_expr): Likewise.
1879         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1880         * ipa-cp.c (determine_versionability): Likewise.
1881         * ipa-fnsummary.c (compute_fn_summary): Likewise.
1882         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1883         * ipa-split.c (visit_bb): Likewise.
1884         (split_function): Likewise.
1885         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1886         * lto-cgraph.c (input_node): Likewise.
1887         * lto-streamer-out.c (write_symbol): Likewise.
1888         * omp-low.c (setjmp_or_longjmp_p): Likewise.
1889         (lower_omp_1): Likewise.
1890         * predict.c (strip_predict_hints): Likewise.
1891         * print-tree.c (print_node): Likewise.
1892         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1893         * trans-mem.c (is_tm_irrevocable): Likewise.
1894         (is_tm_load): Likewise.
1895         (is_tm_simple_load): Likewise.
1896         (is_tm_store): Likewise.
1897         (is_tm_simple_store): Likewise.
1898         (is_tm_abort): Likewise.
1899         (tm_region_init_1): Likewise.
1900         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1901         * tree-cfg.c (verify_gimple_call): Likewise.
1902         (move_stmt_r): Likewise.
1903         (stmt_can_terminate_bb_p): Likewise.
1904         * tree-eh.c (lower_eh_constructs_2): Likewise.
1905         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1906         * tree-inline.c (remap_gimple_stmt): Likewise.
1907         (copy_bb): Likewise.
1908         (estimate_num_insns): Likewise.
1909         (fold_marked_statements): Likewise.
1910         * tree-sra.c (scan_function): Likewise.
1911         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1912         (optimize_stack_restore): Likewise.
1913         (pass_fold_builtins::execute): Likewise.
1914         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1915         (mark_all_reaching_defs_necessary_1): Likewise.
1916         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1917         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1918         (pass_forwprop::execute): Likewise.
1919         * tree-ssa-loop-im.c (stmt_cost): Likewise.
1920         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1921         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1922         * tree-ssa-strlen.c (get_string_length): Likewise.
1923         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1924         (find_func_aliases_for_call): Likewise.
1925         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1926         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1927         * tree-tailcall.c (find_tail_calls): Likewise.
1928         * tree.c (need_assembler_name_p): Likewise.
1929         (free_lang_data_in_decl): Likewise.
1930         (get_call_combined_fn): Likewise.
1931         * ubsan.c (is_ubsan_builtin_p): Likewise.
1932         * varasm.c (incorporeal_function_p): Likewise.
1933         * tree.h (DECL_BUILT_IN): Remove and replace with
1934         fndecl_built_in_p.
1935         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1936         (fndecl_built_in_p): New.
1937
1938 2018-08-27  Martin Liska  <mliska@suse.cz>
1939
1940         PR tree-optimization/86847
1941         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1942         Dump also subtree probability.
1943         (switch_decision_tree::do_jump_if_equal): New function.
1944         (switch_decision_tree::emit_case_nodes): Handle special
1945         situations in balanced tree that can be emitted much simpler.
1946         Fix calculation of probabilities that happen in tree expansion.
1947         * tree-switch-conversion.h (struct cluster): Add
1948         is_single_value_p.
1949         (struct simple_cluster): Likewise.
1950         (struct case_tree_node): Add new function has_child.
1951         (do_jump_if_equal): New.
1952
1953 2018-08-27  Martin Liska  <mliska@suse.cz>
1954
1955         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1956         Add new argument to bit_test_cluster constructor.
1957         (bit_test_cluster::emit): Set bits really number of values
1958         handlel by a test.
1959         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1960         probability argument.
1961         * tree-switch-conversion.h (struct bit_test_cluster):
1962         Add m_handles_entire_switch.
1963
1964 2018-08-27  Martin Liska  <mliska@suse.cz>
1965
1966         PR tree-optimization/86702
1967         * tree-switch-conversion.c (jump_table_cluster::emit):
1968         Make probabilities even for values in jump table
1969         according to number of cases handled.
1970         (switch_decision_tree::compute_cases_per_edge): Pass
1971         argument to reset_out_edges_aux function.
1972         (switch_decision_tree::analyze_switch_statement): Likewise.
1973         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1974         Make it static.
1975
1976 2018-08-27  Martin Liska  <mliska@suse.cz>
1977
1978         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1979         cfun argument explicitly.
1980         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1981         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1982         function gimple_switch_default_bb.
1983         (convert_switch_statements):
1984         (expand_builtins):
1985         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1986         * stmt.c (label_to_block_fn): Use label_to_block and pass
1987         cfun argument explicitly and use gimple_switch_label_bb.
1988         (expand_case): Likewise.
1989         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1990         cfun argument explicitly. Likewise.
1991         (make_edges_bb): Likewise.
1992         (make_cond_expr_edges): Likewise.
1993         (get_cases_for_edge): Likewise.
1994         (make_gimple_switch_edges): Likewise.
1995         (label_to_block_fn): Likewise.
1996         (label_to_block): Likewise.
1997         (make_goto_expr_edges): Likewise.
1998         (make_gimple_asm_edges): Likewise.
1999         (main_block_label): Likewise.
2000         (group_case_labels_stmt): Likewise.
2001         (find_taken_edge_computed_goto): Likewise.
2002         (find_taken_edge_switch_expr): Likewise.
2003         (gimple_verify_flow_info): Likewise.
2004         (gimple_redirect_edge_and_branch): Likewise.
2005         (gimple_switch_label_bb): New function.
2006         (gimple_switch_default_bb): Likewise.
2007         (gimple_switch_edge): Likewise.
2008         (gimple_switch_default_edge): Likewise.
2009         * tree-cfg.h (label_to_block_fn): Remove and replace ...
2010         (label_to_block): ... with this.
2011         (gimple_switch_label_bb): New.
2012         (gimple_switch_default_bb): Likewise.
2013         (gimple_switch_edge): Likewise.
2014         (gimple_switch_default_edge): Likewise.
2015         * tree-cfgcleanup.c (convert_single_case_switch): Use
2016         new gimple functions and pass new argument to label_to_block.
2017         (cleanup_control_flow_bb):
2018         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
2019         cfun argument explicitly.
2020         (make_eh_edges): Likewise.
2021         (redirect_eh_dispatch_edge): Likewise.
2022         (lower_resx): Likewise.
2023         (lower_eh_dispatch): Likewise.
2024         (maybe_remove_unreachable_handlers): Likewise.
2025         (unsplit_eh): Likewise.
2026         (cleanup_empty_eh): Likewise.
2027         (verify_eh_edges): Likewise.
2028         (verify_eh_dispatch_edge): Likewise.
2029         * tree-ssa-dom.c (record_edge_info): Likewise.
2030         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2031         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
2032         (thread_through_normal_block): Likewise.
2033         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
2034         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2035         * tree-switch-conversion.c (switch_conversion::collect): Use new
2036         gimple functions.
2037         (switch_conversion::check_final_bb): Likewise.
2038         (switch_conversion::gather_default_values): Pass new argument
2039         to label_to_block.
2040         (switch_conversion::build_constructors): Likewise.
2041         (switch_decision_tree::compute_cases_per_edge): Use new
2042         gimple_switch_edge function.
2043         (switch_decision_tree::analyze_switch_statement): Pass new argument
2044         to label_to_block.
2045         (switch_decision_tree::try_switch_expansion): Use
2046         gimple_switch_default_edge.
2047         * tree-vrp.c (find_switch_asserts): Pass new argument
2048         to label_to_block.
2049         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2050         (vr_values::simplify_switch_using_ranges): Likewise.
2051
2052 2018-08-27  Richard Biener  <rguenther@suse.de>
2053
2054         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2055         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2056
2057         * tree-ssa-sccvn.h (struct vn_pval): New structure.
2058         (struct vn_nary_op_s): Add unwind_to member.  Add
2059         predicated_values flag and put result into a union together
2060         with a linked list of vn_pval.
2061         (struct vn_ssa_aux): Add name member to make maintaining
2062         a map of SSA name to vn_ssa_aux possible.  Remove no longer
2063         needed info, dfsnum, low, visited, on_sccstack, use_processed
2064         and range_info_anti_range_p members.
2065         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2066         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2067         New functions.
2068         (vn_valueize): New global.
2069         (vn_context_bb): Likewise.
2070         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2071         VN_INFO_PTR_INFO): Remove.
2072         * tree-ssa-sccvn.c: ... (rewrite)
2073         (pass_fre::execute): For -O2+ initialize loops and run
2074         RPO VN in optimistic mode (iterating).  For -O1 and -Og
2075         run RPO VN in non-optimistic mode.
2076         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2077         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2078         * doc/invoke.texi (sccvn-max-scc-size): Remove.
2079         (rpo-vn-max-loop-depth): Document.
2080         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2081         when valuezing the VUSE signals we walked out of the region.
2082         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2083         (phi_translate): Set VN context block to use for availability
2084         lookup.
2085         (compute_avail): Likewise.
2086         (pre_valueize): New function.
2087         (pass_pre::execute): Adjust to the RPO VN API.
2088
2089         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2090         (propagate_constants_for_unrolling): Remove.
2091         (tree_unroll_loops_completely): Perform value-numbering
2092         on the unrolled bodies loop parent.
2093
2094 2018-08-27  Richard Biener  <rguenther@suse.de>
2095
2096         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2097         for partial antic compute.
2098
2099 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
2100
2101         PR rtl-optimization/87065
2102         * combine.c (simplify_if_then_else): Formatting fix.
2103         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2104         check.
2105         (known_cond): Don't return const_true_rtx for vector modes.  Use
2106         CONST0_RTX instead of const0_rtx.  Formatting fixes.
2107
2108 2018-08-27  Martin Liska  <mliska@suse.cz>
2109
2110         PR gcov-profile/87069
2111         * gcov.c (process_file): Record files already processed
2112         and warn about a file being processed multiple times.
2113
2114 2018-08-27  Martin Liska  <mliska@suse.cz>
2115
2116         PR driver/83193
2117         * config/aarch64/aarch64.c (aarch64_override_options_internal):
2118         Set default values for x_aarch64_*_string strings.
2119         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2120         prefix.  For -mabi do not print '=ABI' in help and use
2121         <option_value> format for -msve-vector-bits and -moverride
2122         options.
2123
2124 2018-08-26  Jeff Law  <law@redhat.com>
2125
2126         * config/mips/frame-header-opt.c: Include "backend.h" rather than
2127         "cfg.h"
2128
2129 2018-08-26  Marek Polacek  <polacek@redhat.com>
2130
2131         PR c++/87029, Implement -Wredundant-move.
2132         * doc/invoke.texi: Document -Wredundant-move.
2133
2134 2018-08-25  Martin Sebor  <msebor@redhat.com>
2135
2136         PR tree-optimization/87059
2137         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2138         to the same type as the other.
2139         * fold-const.c (fold_binary_loc): Assert expectation.
2140
2141 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
2142
2143         * config/darwin.c (machopic_legitimize_pic_address): Clean up
2144         extraneous parentheses, dead code section and formatting.
2145
2146 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
2147
2148         PR c++/87091
2149         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2150         wide enough for jumps in the line-numbering to be visible.
2151         (layout::print_gap_in_line_numbering): New member function.
2152         (layout::calculate_line_spans): When using line numbering, merge
2153         line spans that are only 1 line apart.
2154         (diagnostic_show_locus): When printing line numbers, show gaps in
2155         line numbering directly, rather than printing headers.
2156         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2157         line-numbering with multiple line spans.
2158         (selftest::test_fixit_insert_containing_newline_2): Add test of
2159         line-numbering, in which the spans are close enough to be merged.
2160
2161 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
2162
2163         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2164         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
2165         * tree-vrp.c (range_is_nonnull): Remove.
2166         (range_includes_zero_p): Accept value_range instead of min/max.
2167         (extract_range_from_binary_expr_1): Do not early bail on
2168         POINTER_PLUS_EXPR.
2169         Use range_includes_zero_p instead of range_is_nonnull.
2170         (extract_range_from_unary_expr): Use range_includes_zero_p instead
2171         of range_is_nonnull.
2172         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
2173         special case VR_ANTI_RANGE.
2174         (vrp_finalize): Same.
2175         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2176         instead of min/max.
2177         (range_is_nonnull): Remove.
2178         * vr-values.c (vrp_stmt_computes_nonzero): Use
2179         range_includes_zero_p instead of range_is_nonnull.
2180         (extract_range_basic): Pass value_range to range_includes_zero_p
2181         instead of range_is_nonnull.
2182
2183 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
2184
2185         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2186         * emit-rtl.h (rtl_data): Remove return_bnd.
2187         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2188         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2189         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2190         (POINTER_BOUNDS_MODE): Remove definition.
2191         (make_pointer_bounds_mode): Remove.
2192         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2193         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2194         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2195         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2196         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2197         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2198         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2199
2200         * config/i386/i386-modes.def (BND32, BND64): Remove.
2201         * config/i386/i386.c (dbx_register_map): Remove bound registers.
2202         (dbx64_register_map): Ditto.
2203         (svr4_dbx_register_map): Ditto.
2204         (indirect_thunk_bnd_needed): Remove.
2205         (indirect_thunks_bnd_used): Ditto.
2206         (indirect_return_bnd_needed): Ditto.
2207         (indirect_return_via_cx_bnd): Ditto.
2208         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2209         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2210         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
2211         (output_indirect_thunk_function): Remove handling of
2212         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2213         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2214         (ix86_save_reg): Remove handling of crtl->return_bnd.
2215         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2216         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2217         and UNSPEC_BNDLX_ADDR.
2218         (ix86_output_indirect_branch_via_reg): Remove handling of
2219         indirect_thunk_prefix_bnd.
2220         (ix86_output_indirect_branch_via_push): Ditto.
2221         (ix86_output_function_return): Ditto.
2222         (ix86_output_indirect_function_return): Ditto.
2223         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2224         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2225         (CALL_USED_REGISTERS): Ditto.
2226         (REG_ALLOC_ORDER): Update for removal of bound registers.
2227         (HI_REGISTER_NAMES): Ditto.
2228         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2229         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2230         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2231         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2232         (FIRST_PSEUDO_REG): Update.
2233         (BND): Remove mode iterator.
2234         * config/i386/predicates.md (bnd_mem_operator): Remove.
2235
2236 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
2237
2238         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
2239         vectors.
2240
2241 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
2242
2243         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
2244         the case in which the permute needs only a single element and
2245         repeats for every vector of the result.  Extend that case to
2246         handle variable-length vectors.
2247         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2248
2249 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
2250
2251         PR debug/79342
2252         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
2253         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
2254
2255 2018-08-24  Richard Biener  <rguenther@suse.de>
2256
2257         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
2258         bb_flags_allocated members.
2259         (auto_flag): New RAII class for allocating flags.
2260         (auto_edge_flag): New RAII class for allocating edge flags.
2261         (auto_bb_flag): New RAII class for allocating bb flags.
2262         * cfgloop.c (verify_loop_structure): Allocate temporary edge
2263         flag dynamically.
2264         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
2265         in favor of temporarily allocated BB flag.
2266         * hsa-brig.c: Re-order includes.
2267         * hsa-dump.c: Likewise.
2268         * hsa-regalloc.c: Likewise.
2269         * print-rtl.c: Likewise.
2270         * profile-count.c: Likewise.
2271
2272 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
2273
2274         PR target/86989
2275         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2276         the TOC register.
2277
2278 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
2279
2280         PR 87073/bootstrap
2281         * wide-int-range.cc (wide_int_range_div): Do not ignore result
2282         from wide_int_range_multiplicative_op.
2283
2284 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2285
2286         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
2287         "permutaion".
2288
2289 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
2290
2291         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
2292         to 'expanded'.
2293
2294 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
2295
2296         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
2297         full GENERIC expression used for replacement.
2298
2299 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
2300
2301         * tree-vrp.c (abs_extent_range): Remove.
2302         (extract_range_into_wide_ints): Pass wide ints by reference.
2303         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2304         Pass wide ints by reference in all calls to
2305         extract_range_into_wide_ints.
2306         * wide-int-range.cc (wide_int_range_div): New.
2307         * wide-int-range.h (wide_int_range_div): New.
2308         (wide_int_range_includes_zero_p): New.
2309         (wide_int_range_zero_p): New.
2310
2311 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
2312
2313         * config/aarch64/aarch64.md (arches): New enum.
2314         (arch): New enum attr.
2315         (arch_enabled): New attr.
2316         (enabled): Now uses arch_enabled only.
2317         (simd, sve, fp16): Removed attribute.
2318         (fp): Attr now defined in terms of 'arch'.
2319         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2320         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2321         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2322         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2323         attributes into 'arch'.
2324         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2325         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2326         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2327         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2328         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2329         'simd' attribute into 'arch'.
2330         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2331         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2332         Convert use of 'fp' attribute to 'arch'.
2333         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2334         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2335         into 'arch'.
2336         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2337         (different modes) Merge 'fp' and 'simd' into 'arch'.
2338         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2339         'simd' into 'arch'.
2340
2341 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
2342
2343         PR rtl-optimization/87026
2344         * expmed.c (canonicalize_comparison): If we can no longer create
2345         pseudoregisters, don't.
2346
2347 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
2348
2349         PR target/86951
2350         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2351         prototype.
2352         * config/arm/arm.c (speculation_barrier_libfunc): New static
2353         variable.
2354         (arm_init_libfuncs): Initialize it.
2355         (arm_emit_speculation_barrier): New function.
2356         * config/arm/arm.md (speculation_barrier): Call
2357         arm_emit_speculation_barrier for architectures that do not have
2358         DSB or ISB.
2359         (speculation_barrier_insn): Only match on Armv7 or later.
2360
2361 2018-08-23  Richard Biener  <rguenther@suse.de>
2362
2363         PR middle-end/87024
2364         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2365         calls.
2366
2367 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
2368
2369         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2370         of single-vector TBLs.
2371         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2372         one input is given.
2373
2374 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
2375
2376         PR target/85910
2377         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2378         aarch64_evpc_tbl guard.
2379
2380 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2381
2382         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2383         behaviour.
2384
2385 2018-08-22  Martin Sebor  <msebor@redhat.com>
2386
2387         PR middle-end/87052
2388         * tree-pretty-print.c (pretty_print_string): Add argument.
2389         (dump_generic_node): Call to pretty_print_string with string size.
2390
2391 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
2392
2393         PR rtl-optimization/86771
2394         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2395         of two SETs into those two SETs, one to be placed at i2, if that SETs
2396         destination is modified between i2 and i3.
2397
2398 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2399
2400         PR tree-optimization/86725
2401         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2402         function.
2403         (vect_analyze_scalar_cycles_1): Check it.
2404
2405 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2406
2407         PR tree-optimization/86725
2408         * tree-vect-loop.c (vect_is_simple_reduction): When treating
2409         an outer loop phi as a double reduction, make sure that the
2410         single user of the phi result is an inner loop phi.
2411
2412 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2413
2414         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2415         grouped stores with gaps to a strided group.
2416
2417 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2418
2419         * tree-vect-stmts.c (get_group_load_store_type)
2420         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2421         first statement in a group.
2422
2423 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2424
2425         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2426         the sequence used in gcc/gcc.c.
2427
2428 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2429
2430         PR other/704
2431         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2432         building it.
2433
2434 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2435
2436         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2437         Darwin10-specific unwinder-shim.
2438         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2439         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC): 
2440         New to cater for Darwin10 Rosetta.
2441
2442 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2443
2444         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2445         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2446
2447 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2448
2449         PR bootstrap/81033
2450         PR target/81733
2451         PR target/52795
2452         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2453         (dwarf2out_switch_text_section): Generate a local label for the
2454         second function sub-section and apply it as the second FDE start
2455         label.
2456         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2457         second sub-section start.
2458
2459 2018-08-22  Richard Biener  <rguenther@suse.de>
2460
2461         PR tree-optimization/86988
2462         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2463
2464 2018-08-22  Richard Biener  <rguenther@suse.de>
2465
2466         PR tree-optimization/86945
2467         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2468
2469 2018-08-22  Alexandre Oliva <oliva@adacore.com>
2470
2471         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2472         a comment about how uses of r2 for .sdata2 come about.
2473
2474 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
2475
2476         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2477
2478 2018-08-21  Marek Polacek  <polacek@redhat.com>
2479
2480         PR c++/86981, Implement -Wpessimizing-move.
2481         * doc/invoke.texi: Document -Wpessimizing-move.
2482
2483 2018-08-21  Jan Hubicka  <jh@suse.cz>
2484
2485         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2486         * tree.h (is_redundant_typedef): Remove.
2487         * dwarf2out.c (is_redundant_typedef): Turn into static function.
2488
2489 2018-08-21  Jan Hubicka  <jh@suse.cz>
2490
2491         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2492         when possible.
2493
2494 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
2495
2496         * expmed.c (extract_low_bits): Reject invalid subregs early.
2497
2498 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2499
2500         PR middle-end/86121
2501         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2502         behaviour.
2503
2504 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2505
2506         * config/vxworks.h: Guard vxworks_asm_out_constructor and
2507         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2508         * config/vxworks.c: Likewise.
2509
2510 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2511
2512         * config/vxworks.c: Set targetm.have_ctors_dtors
2513         if HAVE_INITFINI_ARRAY_SUPPORT.
2514         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2515         if HAVE_INITFINI_ARRAY_SUPPORT.
2516
2517 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2518
2519         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2520         default search path for VxWorks < 7.
2521
2522 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2523
2524         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2525         (get_format_string): Refer to c_getstr.
2526
2527 2018-08-21  Tom de Vries  <tdevries@suse.de>
2528
2529         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2530         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2531         (debuginfo_early_stop): Declare.
2532         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2533         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2534         (debuginfo_early_stop): New function.
2535         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2536         and debuginfo_early_stop.
2537         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2538         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2539         (general_init): Call debuginfo_early_init.
2540         (finalize): Call debuginfo_fini.
2541         (do_compile): Call debuginfo_init.
2542         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2543         -fdump-early-debug.
2544         (@item -fdump-debug, @item -fdump-earlydebug): Add.
2545
2546 2018-08-21  Tom de Vries  <tdevries@suse.de>
2547
2548         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2549         flag_dump_noaddr and flag_dump_unnumbered.
2550
2551 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
2552
2553         * wide-int-range.cc (wide_int_range_abs): New.
2554         (wide_int_range_order_set): Rename from wide_int_range_min_max.
2555         * wide-int-range.h (wide_int_range_abs): New.
2556         (wide_int_range_min_max): New.
2557         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2558         case to call wide_int_range_abs.
2559         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2560         (extract_range_from_abs_expr): Delete.
2561
2562 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
2563
2564         PR target/87033
2565         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2566         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2567         for indexed loads.
2568
2569 2018-08-20  Nathan Sidwell  <nathan@acm.org>
2570             Jeff Law <law@redhat.com>
2571
2572         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2573         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2574
2575 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
2576
2577         PR other/84889
2578         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2579         (decl_attributes): Likewise.
2580         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2581         instance.
2582         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2583         * diagnostic-core.h (class auto_diagnostic_group): New class.
2584         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2585         (diagnostic_report_diagnostic): Handle the first diagnostics within
2586         a group.
2587         (emit_diagnostic): Add auto_diagnostic_group instance.
2588         (inform): Likewise.
2589         (inform_n): Likewise.
2590         (warning): Likewise.
2591         (warning_at): Likewise.
2592         (warning_n): Likewise.
2593         (pedwarn): Likewise.
2594         (permerror): Likewise.
2595         (error): Likewise.
2596         (error_n): Likewise.
2597         (error_at): Likewise.
2598         (sorry): Likewise.
2599         (fatal_error): Likewise.
2600         (internal_error): Likewise.
2601         (internal_error_no_backtrace): Likewise.
2602         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2603         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2604         * diagnostic.h (struct diagnostic_context): Add fields
2605         "diagnostic_group_nesting_depth",
2606         "diagnostic_group_emission_count", "begin_group_cb",
2607         "end_group_cb".
2608         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2609         Add auto_diagnostic_group instance(s).
2610         (find_explicit_erroneous_behavior): Likewise.
2611         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2612         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2613         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2614         (gimplify_va_arg_expr): Likewise.
2615         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2616         (HSA_SORRY_AT): Likewise.
2617         * ipa-devirt.c (compare_virtual_tables): Likewise.
2618         (warn_odr): Likewise.
2619         * multiple_target.c (expand_target_clones): Likewise.
2620         * opts-common.c (cmdline_handle_error): Likewise.
2621         * reginfo.c (globalize_reg): Likewise.
2622         * substring-locations.c (format_warning_n_va): Likewise.
2623         * tree-inline.c (expand_call_inline): Likewise.
2624         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2625         * tree-ssa-loop-niter.c
2626         (do_warn_aggressive_loop_optimizations): Likewise.
2627         * tree-ssa-uninit.c (warn_uninit): Likewise.
2628         * tree.c (warn_deprecated_use): Likewise.
2629
2630 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
2631
2632         PR target/87014
2633         * config/i386/i386.md (eh_return): Always update EH return
2634         address in word_mode.
2635
2636 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
2637
2638         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2639         TARGET_SPLIT_COMPLEX_ARG is defined.
2640
2641 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2642
2643         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2644
2645 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2646
2647         PR target/86984
2648         * expr.c (expand_assignment): Assert that bitpos is positive.
2649         (store_field): Likewise
2650         (expand_expr_real_1): Make sure that bitpos is positive.
2651         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2652         integer overflow.
2653
2654 2018-08-20  Nathan Sidwell  <nathan@acm.org>
2655
2656         * Makefile.in (CPP_ID_DATA_H): Delete.
2657         (CPP_INTERNAL_H): Don't add it.
2658         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2659         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2660
2661 2018-08-20  Richard Biener  <rguenther@suse.de>
2662
2663         PR tree-optimization/78655
2664         * tree-vrp.c (extract_range_from_binary_expr_1): Make
2665         pointer + offset nonnull if either operand is nonnull work.
2666
2667 2018-08-20  Tom de Vries  <tdevries@suse.de>
2668
2669         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2670         unless the referenced die describes the added property using
2671         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
2672         Otherwise, add a DW_AT_location to the referenced die.
2673
2674 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
2675
2676         PR target/86994
2677         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2678         register_operand when calling ix86_set_reg_reg_cost.
2679         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2680         Set *total to 0 for operands that satisfy x86_64_immediate_operand
2681         predicate and to 1 otherwise.
2682
2683 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
2684
2685         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2686         emit a diagnostic that it is not supported and reset the option.
2687         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2688         supported and consume the option.  (ASM_FINAL_SPEC): New.
2689
2690 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
2691
2692         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2693         a sentence.
2694
2695 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
2696
2697         C-SKY port: Documentation
2698
2699         * doc/extend.texi (C-SKY Function Attributes): New section.
2700         * doc/invoke.texi (Option Summary): Add C-SKY options.
2701         (C-SKY Options): New section.
2702         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2703
2704 2018-08-17  Jojo  <jijie_rong@c-sky.com>
2705             Huibin Wang  <huibin_wang@c-sky.com>
2706             Sandra Loosemore  <sandra@codesourcery.com>
2707             Chung-Lin Tang  <cltang@codesourcery.com>
2708
2709         C-SKY port: Backend implementation
2710
2711         * config/csky/*: New.
2712         * common/config/csky/*: New.
2713
2714 2018-08-17  Jojo  <jijie_rong@c-sky.com>
2715             Huibin Wang  <huibin_wang@c-sky.com>
2716             Sandra Loosemore  <sandra@codesourcery.com>
2717             Chung-Lin Tang  <cltang@codesourcery.com>
2718             Andrew Jenner  <andrew@codesourcery.com>
2719
2720         C-SKY port: Configury
2721
2722         * config.gcc (csky-*-*): New.
2723         * configure.ac: Add csky to targets for dwarf2 debug_line support.
2724         * configure: Regenerated.
2725
2726 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
2727
2728         * dump-context.h: Include "dumpfile.h".
2729         (dump_context::dump_printf_va): Convert final param from va_list
2730         to va_list *.  Convert from ATTRIBUTE_PRINTF to
2731         ATTRIBUTE_GCC_DUMP_PRINTF.
2732         (dump_context::dump_printf_loc_va): Likewise.
2733         * dumpfile.c: Include "stringpool.h".
2734         (make_item_for_dump_printf_va): Delete.
2735         (make_item_for_dump_printf): Delete.
2736         (class dump_pretty_printer): New class.
2737         (dump_pretty_printer::dump_pretty_printer): New ctor.
2738         (dump_pretty_printer::emit_items): New member function.
2739         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2740         function.
2741         (dump_pretty_printer::emit_item): New member function.
2742         (dump_pretty_printer::stash_item): New member function.
2743         (dump_pretty_printer::format_decoder_cb): New member function.
2744         (dump_pretty_printer::decode_format): New member function.
2745         (dump_context::dump_printf_va): Reimplement in terms of
2746         dump_pretty_printer.
2747         (dump_context::dump_printf_loc_va): Convert final param from va_list
2748         to va_list *.
2749         (dump_context::begin_scope): Reimplement call to
2750         make_item_for_dump_printf.
2751         (dump_printf): Update for change to dump_printf_va.
2752         (dump_printf_loc): Likewise.
2753         (selftest::test_capture_of_dump_calls): Convert "stmt" from
2754         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
2755         with %T, %E, and %G.
2756         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2757         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2758         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2759         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2760         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2761         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2762         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2763         within a dump_printf_loc call to "%wu".
2764         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2765         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
2766         missing space after "=".
2767         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2768         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2769         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2770         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2771         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
2772         duplicate "vectorized" from message.
2773
2774 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
2775
2776         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2777         polyNxK_t element's TYPE_STRING_FLAG.
2778
2779 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
2780
2781         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2782         (they were unnamed before).  Fix comments.
2783
2784 2018-08-17  Nathan Sidwell  <nathan@acm.org>
2785
2786         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2787
2788 2018-08-17  Richard Biener  <rguenther@suse.de>
2789
2790         PR tree-optimization/86841
2791         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2792
2793 2018-08-17  Martin Liska  <mliska@suse.cz>
2794
2795         * common.opt: Remove Warn, Init and Report for options with
2796         Ignore/Deprecated flag. Warning is done automatically for
2797         Deprecated flags.
2798         * config/i386/i386.opt: Likewise.
2799         * config/ia64/ia64.opt: Likewise.
2800         * config/rs6000/rs6000.opt: Likewise.
2801         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2802         Remove usage of flag_check_pointer_bounds.
2803         * lto-wrapper.c (merge_and_complain): Do not handle
2804         OPT_fcheck_pointer_bounds.
2805         (append_compiler_options): Likewise.
2806         * opt-functions.awk: Do not handle Deprecated.
2807         * optc-gen.awk: Check that Var, Report and Init are not
2808         used for an option with Ignore/Deprecated flag.
2809         * opts-common.c (decode_cmdline_option): Do not report
2810         CL_ERR_DEPRECATED.
2811         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2812         options.
2813         * opts.h (struct cl_option): Remove cl_deprecated flag.
2814         (CL_ERR_DEPRECATED): Remove error enum value.
2815
2816 2018-08-17  Richard Biener  <rguenther@suse.de>
2817
2818         PR middle-end/86505
2819         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2820         across a va-arg-pack using call adjust its return value accordingly.
2821
2822 2018-08-16  Martin Sebor  <msebor@redhat.com>
2823
2824         PR tree-optimization/86853
2825         * gimple-ssa-sprintf.c (struct format_result): Rename member.
2826         (struct fmtresult): Add member and initialize it in ctors.
2827         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
2828         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
2829         (format_directive): Set POSUNDER4K when MAYFAIL is set.
2830         (parse_directive): Handle %C same as %c.
2831         (sprintf_dom_walker::compute_format_length): Adjust.
2832         (is_call_safe): Adjust.
2833
2834 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2835
2836         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
2837         for determining how to count the elements.
2838         * builtins.h (c_strlen): Adjust prototype.
2839         * expr.c (string_constant): Add new parameter mem_size.
2840         Set *mem_size appropriately.
2841         * expr.h (string_constant): Adjust protoype.
2842         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2843         * gimple-fold.h (get_range_strlen): Adjust prototype.
2844         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2845         (format_string): Call get_string_length with eltsize.
2846
2847 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
2848
2849         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2850         to emit the span, rather than setting it as the prefix.
2851
2852 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
2853
2854         * diagnostic-show-locus.c (layout::start_annotation_line): Add
2855         "margin_char" parameter, defaulting to space.  Use it in place
2856         of pp_space for the initial part of the margin.
2857         (layout::print_leading_fixits): Use '+' when filling the margin
2858         of line-insertion fix-it hints.
2859
2860 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2861
2862         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2863         Delete.
2864
2865 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2866
2867         * config/rs6000/altivec.md: Don't set length attribute to the default
2868         value.
2869         * config/rs6000/darwin.md: Ditto.
2870         * config/rs6000/dfp.md: Ditto.
2871         * config/rs6000/htm.md: Ditto.
2872         * config/rs6000/rs6000.md: Ditto.
2873         * config/rs6000/sync.md: Ditto.
2874         * config/rs6000/vsx.md: Ditto.
2875
2876 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2877
2878         * config/rs6000/altivec.md: Don't set length attribute to the default
2879         value, for branch instructions.
2880         * config/rs6000/darwin.md: Ditto.
2881         * config/rs6000/rs6000.md: Ditto.
2882
2883 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2884
2885         * config/rs6000/rs6000.md (length): Always define as const_int 4.
2886         (unnamed conditional branch define_insn): Set length to 4 or 8
2887         depending on offset.
2888         (<bd>_<mode>): Similar, for alternative 0.
2889         (<bd>tf_<mode>): Ditto.
2890
2891 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
2892
2893         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2894
2895 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
2896
2897         * doc/rtl.texi: Replace old RTX class names with new names.
2898
2899
2900 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
2901
2902         * expmed.h (canonicalize_comparison): New declaration.
2903         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2904         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2905         * optabs.c (prepare_cmp_insn): Likewise.
2906         * rtl.h (unsigned_condition_p): New function which checks if a
2907         comparison operator is unsigned.
2908
2909 2018-08-16  Nathan Sidwell  <nathan@acm.org>
2910
2911         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2912         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2913
2914 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
2915
2916         PR target/84711
2917         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2918         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2919         (mov<mov>): ..this and enable unconditionally.
2920
2921 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
2922
2923         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2924
2925 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
2926
2927         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2928         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2929         "Common" with "Target".
2930
2931 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
2932
2933         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2934         * doc/invoke.texi (mmitigate-rop): Remove.
2935         * config/i386/i386.c: Do not include "regrename.h".
2936         (ix86_rop_should_change_byte_p, reg_encoded_number)
2937         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2938         Remove.
2939         (ix86_reorg): Remove call to ix86_mitigate_rop.
2940         * config/i386/i386.md (attr "modrm_class"): Remove.
2941         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2942         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2943         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2944
2945 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
2946
2947         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2948         allow folding of mergeh() and mergel() for the float and double types.
2949         (fold_mergehl_helper): Rework to handle building a permute tree
2950         for float vectors.
2951
2952 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
2953
2954         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2955         for TARGET_SSE.
2956
2957 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
2958
2959         * common.opt (fdiagnostics-show-labels): New option.
2960         * diagnostic-show-locus.c (class layout_range): Add field
2961         "m_label".
2962         (class layout): Add field "m_show_labels_p".
2963         (layout_range::layout_range): Add param "label" and use it to
2964         initialize m_label.
2965         (make_range): Pass in NULL for new "label" param of layout_range's
2966         ctor.
2967         (layout::layout): Initialize m_show_labels_p.
2968         (layout::maybe_add_location_range): Pass in loc_range->m_label
2969         when constructing layout_range instances.
2970         (struct line_label): New struct.
2971         (layout::print_any_labels): New member function.
2972         (layout::print_line): Call it if label-printing is enabled.
2973         (selftest::test_one_liner_labels): New test.
2974         (selftest::test_diagnostic_show_locus_one_liner): Call it.
2975         * diagnostic.c (diagnostic_initialize): Initialize
2976         context->show_labels_p.
2977         * diagnostic.h (struct diagnostic_context): Add field
2978         "show_labels_p".
2979         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2980         -fno-diagnostics-show-labels.
2981         * dwarf2out.c (gen_producer_string): Add
2982         OPT_fdiagnostics_show_labels to the ignored options.
2983         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2984         param.
2985         (gcc_rich_location::maybe_add_expr): Likewise.
2986         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2987         label" param, defaulting to NULL.
2988         (gcc_rich_location::add_expr): Add "label" param.
2989         (gcc_rich_location::maybe_add_expr): Likewise.
2990         (class text_range_label): New class.
2991         (class range_label_for_type_mismatch): New class.
2992         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2993         of format_warning_va.
2994         (fmtwarn_n): Likewise for new params of format_warning_n_va.
2995         * lto-wrapper.c (merge_and_complain): Add
2996         OPT_fdiagnostics_show_labels to the "pick one setting" options.
2997         (append_compiler_options): Likewise to the dropped options.
2998         (append_diag_options): Likewise to the passed-on options.
2999         * opts.c (common_handle_option): Handle the new option.
3000         * selftest-diagnostic.c
3001         (test_diagnostic_context::test_diagnostic_context): Enable
3002         show_labels_p.
3003         * substring-locations.c: Include "gcc-rich-location.h".
3004         (format_warning_n_va): Add "fmt_label" and "param_label" params
3005         and use them as appropriate.
3006         (format_warning_va): Add "fmt_label" and "param_label" params,
3007         passing them on to format_warning_n_va.
3008         (format_warning_at_substring): Likewise.
3009         (format_warning_at_substring_n): Likewise.
3010         * substring-locations.h (format_warning_va): Add "fmt_label" and
3011         "param_label" params.
3012         (format_warning_n_va): Likewise.
3013         (format_warning_at_substring): Likewise.
3014         (format_warning_at_substring_n): Likewise.
3015         * toplev.c (general_init): Initialize global_dc->show_labels_p.
3016
3017 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
3018
3019         PR testsuite/86519
3020         * builtins.c (expand_builtin_memcmp): Do not expand the call
3021         when overflow is detected.
3022
3023 2018-08-15  Martin Sebor  <msebor@redhat.com>
3024
3025         PR tree-optimization/71625
3026         * config/aarch64/aarch64-builtins.c
3027         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3028
3029 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
3030
3031         * config/s390/s390.c (s390_reorg): Remove loop.
3032
3033 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
3034
3035         * config/darwin.c
3036          (darwin_function_switched_text_sections): Delete.
3037         * gcc/config/darwin.h
3038          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3039
3040 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
3041
3042         PR target/81685
3043         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3044         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
3045         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3046
3047 2018-08-15  Martin Liska  <mliska@suse.cz>
3048
3049         PR tree-optimization/86925
3050         * predict.c (expr_expected_value_1): When taking
3051         later predictor, assign also probability.
3052         Use fold_build2_initializer_loc in order to fold
3053         the expression in -frounding-math.
3054
3055 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
3056
3057         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3058         patterns.
3059         (expand_vec_perm_1): Try the new method.
3060
3061 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
3062
3063         PR target/86547
3064         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3065         Check whether lra_live_max_point is 0 before dividing.
3066
3067 2018-08-14  Martin Sebor  <msebor@redhat.com>
3068
3069         PR tree-optimization/86650
3070         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3071         (vrp_prop::check_mem_ref): Same.
3072
3073 2018-08-13  Liu Hao <lh_mouse@126.com>
3074
3075         * pretty-print.c (eat_esc_sequence): Swap the foreground and
3076         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3077         and clear it thereafter, as it only works for DBCS.
3078
3079 2018-08-13  Liu Hao <lh_mouse@126.com>
3080
3081         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3082         handle returned by _get_osf_handle().
3083
3084 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
3085
3086         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3087         for folding vec_perm.
3088
3089 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
3090
3091         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3092         Add support for gimple-folding of vec_pack() and vec_unpack()
3093         intrinsics.
3094
3095 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
3096
3097         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3098         vec_xst variants to the list.
3099         (rs6000_gimple_fold_builtin): Add support for folding unaligned
3100         vector loads and stores.
3101
3102 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
3103
3104         * config.gcc (rs6000-ibm-aix4.x): Delete.
3105         (rs6000-ibm-aix5.1): Delete.
3106         (rs6000-ibm-aix5.2): Delete.
3107         (rs6000-ibm-aix5.3): Delete.
3108         * config/rs6000/aix43.h: Delete.
3109         * config/rs6000/aix51.h: Delete.
3110         * config/rs6000/aix52.h: Delete.
3111         * config/rs6000/t-aix43: Delete.
3112
3113 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
3114
3115         * config/s390/s390.c (s390_decompose_constant_pool_ref):
3116         New function.
3117         (s390_decompose_address): Factor out constant pool ref
3118         decomposition.
3119
3120 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3121
3122         * config/nds32/nds32-predicates.c
3123         (nds32_can_use_bclr_p): Change return type as bool.
3124         (nds32_can_use_bset_p): Ditto.
3125         (nds32_can_use_btgl_p): Ditto.
3126         (nds32_can_use_bitci_p): Ditto.
3127         * config/nds32/nds32-protos.h
3128         (nds32_can_use_bclr_p): Change declaration.
3129         (nds32_can_use_bset_p): Ditto.
3130         (nds32_can_use_btgl_p): Ditto.
3131         (nds32_can_use_bitci_p): Ditto.
3132
3133 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3134
3135         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3136         Support -msched-prolog-epilog option.
3137         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3138
3139 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3140
3141         * common/config/nds32/nds32-common.c
3142         (nds32_option_optimization_table): Enalbe -malways-align.
3143
3144 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
3145
3146         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3147         extra_headers.
3148         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3149         OPT_misr_secure_ case.
3150         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3151         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3152         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3153         secure attribute.
3154         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3155         (nds32_isr_info): New field security_level.
3156         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3157         * config/nds32/nds32.md (return_internal): Consider critical attribute.
3158         * config/nds32/nds32.opt (misr-secure): New option.
3159         * config/nds32/nds32_init.inc: New file.
3160         * config/nds32/nds32_isr.h: New file.
3161
3162 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
3163
3164         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3165         Update comment for atomic instructions.
3166         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3167         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3168         Remove.
3169         (atomic_loaddi): Revise fence expansion to only emit fence prior to
3170         load for __ATOMIC_SEQ_CST model.
3171         (atomic_loaddi_1): Remove float register target.
3172         (atomic_storedi): Handle CONST_INT values.
3173         (atomic_storedi_1): Remove float register source.  Add special case
3174         for zero value.
3175         (memory_barrier): New expander and insn.
3176
3177 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
3178
3179         PR tree-optimization/86835
3180         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3181         new_stmt after def_gsi, make sure to insert new_square_stmt after
3182         that stmt, not 2 stmts before it.
3183
3184 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
3185
3186         PR target/82418
3187         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3188         instead of SWI48.
3189
3190 2018-08-10  Martin Liska  <mliska@suse.cz>
3191
3192         PR target/83610
3193         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3194         function type.
3195         * builtins.c (expand_builtin_expect_with_probability):
3196         New function.
3197         (expand_builtin_expect_with_probability): New function.
3198         (build_builtin_expect_predicate): Add new argumnet probability
3199         for BUILT_IN_EXPECT_WITH_PROBABILITY.
3200         (fold_builtin_expect):
3201         (fold_builtin_2):
3202         (fold_builtin_3):
3203         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3204         * builtins.h (fold_builtin_expect): Set new argument.
3205         * doc/extend.texi: Document __builtin_expect_with_probability.
3206         * doc/invoke.texi: Likewise.
3207         * gimple-fold.c (gimple_fold_call): Pass new argument.
3208         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3209         also BUILT_IN_EXPECT_WITH_PROBABILITY.
3210         * predict.c (get_predictor_value): New function.
3211         (expr_expected_value): Add new argument probability. Assume
3212         that predictor and probability are always non-null.
3213         (expr_expected_value_1): Likewise.  For __builtin_expect and
3214         __builtin_expect_with_probability set probability.  Handle
3215         combination in binary expressions.
3216         (tree_predict_by_opcode): Simplify code by simply calling
3217         get_predictor_value.
3218         (pass_strip_predict_hints::execute): Add handling of
3219         BUILT_IN_EXPECT_WITH_PROBABILITY.
3220         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3221         new predictor.
3222         * tree.h (DECL_BUILT_IN_P): New function.
3223
3224 2018-08-10  Martin Liska  <mliska@suse.cz>
3225
3226         PR tree-optimization/85799
3227         * passes.def: Add argument for pass_strip_predict_hints.
3228         * predict.c (class pass_strip_predict_hints): Add new argument
3229         early_p.
3230         (strip_predictor_early): New function.
3231         (pass_strip_predict_hints::execute): Call the function to
3232         strip predictors.
3233         (strip_predict_hints): New function.
3234         * predict.def: Fix comment.
3235
3236 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
3237
3238         * Makefile.in: Clarify which tm.texi to copy over to assert the
3239         right to grant a GFDL license for all.
3240
3241 2018-08-09  Jeff Law  <law@redhat.com>
3242
3243         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
3244         unused variable.
3245
3246 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
3247
3248         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
3249         prototype.
3250
3251 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3252
3253         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
3254         reductions for variable-length vectors.
3255
3256 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
3257
3258         PR other/84889
3259         * common.opt (fdiagnostics-show-line-numbers): New option.
3260         * diagnostic-show-locus.c (class layout): Add fields
3261         "m_show_line_numbers_p" and "m_linenum_width";
3262         (num_digits): New function.
3263         (test_num_digits): New function.
3264         (layout::layout): Initialize new fields.  Update m_x_offset
3265         logic to handle any left margin.
3266         (layout::print_source_line): Print line number when requested.
3267         (layout::start_annotation_line): New member function.
3268         (layout::print_annotation_line): Call it.
3269         (layout::print_leading_fixits): Likewise.
3270         (layout::print_trailing_fixits): Likewise.  Update calls to
3271         move_to_column for new parameter.
3272         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
3273         it to potentially call start_annotation_line.
3274         (layout::show_ruler): Call start_annotation_line.
3275         (selftest::test_line_numbers_multiline_range): New selftest.
3276         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
3277         and selftest::test_line_numbers_multiline_range.
3278         * diagnostic.c (diagnostic_initialize): Initialize
3279         show_line_numbers_p.
3280         * diagnostic.h (struct diagnostic_context): Add field
3281         "show_line_numbers_p".
3282         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3283         -fno-diagnostics-show-line-numbers.
3284         * dwarf2out.c (gen_producer_string): Add
3285         OPT_fdiagnostics_show_line_numbers to the ignored options.
3286         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
3287         one setting" options.
3288         (append_compiler_options): Likewise to the dropped options.
3289         (append_diag_options): Likewise to the passed-on options.
3290         * opts.c (common_handle_option): Handle the new option.
3291         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
3292
3293 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3294
3295         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
3296         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
3297         third argument of type "const signed char" to descriptions of
3298         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
3299         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
3300         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
3301         __builtin_bcdsub_ov functions.
3302
3303 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3304
3305         PR tree-optimization/86858
3306         * tree-vect-loop.c (vect_is_simple_reduction): Restore
3307         flow_bb_inside_loop_p calls.
3308
3309 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3310
3311         PR tree-optimization/86871
3312         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3313         instead of gimple_assign_lhs.
3314
3315 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
3316
3317         PR target/86887
3318         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3319         register constraint to operand 0.
3320         (add<mode>3_carryinC): Likewise.
3321         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3322
3323 2018-08-09  Martin Liska  <mliska@suse.cz>
3324
3325         PR c/86895
3326         * common.opt: Remove extra line.
3327
3328 2018-08-09  Martin Liska  <mliska@suse.cz>
3329
3330         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3331         at the end of a line, make first letter capital and end up
3332         a sentence with a dot.
3333         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3334         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3335         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3336         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3337         (PARAM_MAX_ISL_OPERATIONS): Likewise.
3338         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3339         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3340         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3341         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3342         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3343         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3344         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3345         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3346         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3347         (PARAM_TREE_REASSOC_WIDTH): Likewise.
3348         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3349         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3350         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3351
3352 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
3353
3354         PR target/84332
3355         * config/s390/s390.c (s390_option_override_internal): Reduce the
3356         stack-clash-protection-probe-interval param if it would be too big
3357         for z900.
3358
3359 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
3360
3361         PR target/46179
3362         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3363         * config/m68k/m68k.c (handle_move_double): Don't call
3364         m68k_final_prescan_insn.
3365         (m68k_adjust_decorated_operand): Renamed from
3366         m68k_final_prescan_insn, remove first and third operand and
3367         simplify.
3368         (print_operand): Call it.
3369         (print_operand_address): Call it.
3370
3371 2018-08-08  Nathan Sidwell  <nathan@acm.org>
3372
3373         * diagnostic.c (diagnostic_report_current_module): Use
3374         linemap_included_from & linemap_included_from_linemap.
3375
3376 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
3377
3378         * config/aarch64/aarch64-cores.def: Add phecda core.
3379         * config/aarch64/aarch64-tune.md: Regenerate.
3380         * doc/invoke.texi: Add phecda core.
3381
3382 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
3383
3384         PR target/85295
3385         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3386         definitions.
3387         * config/s390/s390.md ("movti"): Add more alternatives for
3388         constant to GPR copies.
3389
3390 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
3391
3392         * config/s390/s390.c: Fix whitespace damage throughout the file.
3393         * config/s390/s390.h: Likewise.
3394         * config/s390/tpf.h: Likewise.
3395
3396 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
3397
3398         * config/s390/s390.c (s390_loadrelative_operand_p):
3399         Remove TARGET_CPU_ZARCH usages.
3400         (s390_rtx_costs): Likewise.
3401         (s390_legitimate_constant_p): Likewise.
3402         (s390_cannot_force_const_mem): Likewise.
3403         (legitimate_reload_constant_p): Likewise.
3404         (s390_preferred_reload_class): Likewise.
3405         (legitimize_pic_address): Likewise.
3406         (legitimize_tls_address): Likewise.
3407         (s390_split_branches): Removed.
3408         (s390_add_execute): Removed.
3409         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3410         (s390_mainpool_start): Likewise.
3411         (s390_mainpool_finish): Likewise.
3412         (s390_mainpool_cancel): Removed.
3413         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3414         (s390_chunkify_cancel): Likewise.
3415         (s390_return_addr_rtx): Likewise.
3416         (s390_register_info): Remove split_branches_pending_p uages.
3417         (s390_optimize_register_info): Likewise.
3418         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3419         split_branches_pending_p usages.
3420         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3421         (s390_load_got): Likewise.
3422         (s390_expand_split_stack_prologue): Likewise.
3423         (output_asm_nops): Likewise.
3424         (s390_function_profiler): Likewise.
3425         (s390_emit_call): Likewise.
3426         (s390_conditional_register_usage): Likewise.
3427         (s390_optimize_prologue): Likewise.
3428         (s390_reorg): Remove TARGET_CPU_ZARCH and
3429         split_branches_pending_p usages.
3430         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3431         usages.
3432         (s390_output_indirect_thunk_function): Likewise.
3433         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3434         (TARGET_CPU_ZARCH_P): Removed.
3435         (struct machine_function): Remove split_branches_pending_p.
3436         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3437
3438 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
3439
3440         * common/config/s390/s390-common.c (processor_flags_table):
3441         Remove flags.
3442         * config.gcc: Remove with_arch/with_tune support.
3443         * config/s390/2064.md: Remove cpu attribute comparisons.
3444         * config/s390/driver-native.c (s390_host_detect_local_cpu):
3445         Remove MTN.
3446         * config/s390/linux.h (ASM_SPEC):
3447         Remove -march support.
3448         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3449         Use a table to get an arch level.
3450         * config/s390/s390-opts.h (enum processor_type):
3451         Remove enum values.
3452         * config/s390/s390.c
3453         (processor_table): Remove entries, add arch_level values.
3454         (s390_issue_rate): Remove cases.
3455         (s390_option_override): Adjust
3456         s390_option_override_internal() call.
3457         (s390_option_override_internal): Remove deprecation warning.
3458         (s390_valid_target_attribute_tree): Adjust
3459         s390_option_override_internal() call.
3460         * config/s390/s390.h (struct s390_processor):
3461         Share with s390-c.c, add arch_level field.
3462         * config/s390/s390.md:
3463         Remove occurrences in cpu attribute.
3464         * config/s390/s390.opt: Remove -march/-mtune support.
3465         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3466         * doc/invoke.texi: Remove deprecation warning.
3467
3468 2018-08-08  Luis Machado  <luis.machado@linaro.org>
3469
3470         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3471         global.
3472         (qdf24xx_tunings): Set vector cost structure to
3473         qdf24xx_vector_cost.
3474
3475         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3476         <register_sextend>: Set to 3.
3477
3478 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
3479
3480         PR target/86838
3481         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3482         * config/aarch64/aarch64-simd.md
3483         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3484         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3485         (aarch64_frecpx<mode>): ...this new pattern.
3486         * config/aarch64/aarch64-simd-builtins.def: Remove comment
3487         about aarch64_frecp<FRECP:frecp_suffix><mode>.
3488
3489 2018-08-07  Martin Liska  <mliska@suse.cz>
3490
3491         PR middle-end/83023
3492         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3493         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3494         * predict.def (PRED_MALLOC_NONNULL): New predictor.
3495         * doc/extend.texi: Document that malloc attribute adds
3496         hit to compiler.
3497
3498 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
3499
3500         PR target/86785
3501         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3502         Define to speculation_safe_value_not_needed.
3503
3504 2018-08-06  Jeff Law  <law@redhat.com>
3505
3506         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3507         the vr_values instance to cprop_into_stmt.
3508         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3509         (cprop_operand): Also query EVRP to determine if OP is a constant.
3510
3511 2018-08-06  Nathan Sidwell  <nathan@acm.org>
3512
3513         * diagnostic.c (diagnostic_report_current_module): Reroll
3514         included-at loop.  Translate text.
3515
3516 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
3517
3518         * function-tests.c (selftest::test_expansion_to_rtl): Call
3519         free_after_compilation.
3520
3521 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3522
3523         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3524
3525 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
3526
3527         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3528         loops with memory block operations from getting unrolled.
3529
3530 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
3531
3532         PR target/86807
3533         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3534         Define to speculation_safe_value_not_needed.
3535
3536 2018-08-06  Jeff Law  <law@redhat.com>
3537
3538         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3539         assert.
3540
3541 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3542
3543         PR target/86662
3544         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3545         with all enabled __intN types.
3546
3547         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3548
3549 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3550
3551         * alias.c (record_set): Check for clobber high.
3552         * cfgexpand.c (expand_gimple_stmt): Likewise.
3553         * combine-stack-adj.c (single_set_for_csa): Likewise.
3554         * combine.c (find_single_use_1): Likewise.
3555         (set_nonzero_bits_and_sign_copies): Likewise.
3556         (get_combine_src_dest): Likewise.
3557         (is_parallel_of_n_reg_sets): Likewise.
3558         (try_combine): Likewise.
3559         (record_dead_and_set_regs_1): Likewise.
3560         (reg_dead_at_p_1): Likewise.
3561         (reg_dead_at_p): Likewise.
3562         * dce.c (deletable_insn_p): Likewise.
3563         (mark_nonreg_stores_1): Likewise.
3564         (mark_nonreg_stores_2): Likewise.
3565         * df-scan.c (df_find_hard_reg_defs): Likewise.
3566         (df_uses_record): Likewise.
3567         (df_get_call_refs): Likewise.
3568         * dwarf2out.c (mem_loc_descriptor): Likewise.
3569         * haifa-sched.c (haifa_classify_rtx): Likewise.
3570         * ira-build.c (create_insn_allocnos): Likewise.
3571         * ira-costs.c (scan_one_insn): Likewise.
3572         * ira.c (equiv_init_movable_p): Likewise.
3573         (rtx_moveable_p): Likewise.
3574         (interesting_dest_for_shprep): Likewise.
3575         * jump.c (mark_jump_label_1): Likewise.
3576         * postreload-gcse.c (record_opr_changes): Likewise.
3577         * postreload.c (reload_cse_simplify): Likewise.
3578         (struct reg_use): Add source expr.
3579         (reload_combine): Check for clobber high.
3580         (reload_combine_note_use): Likewise.
3581         (reload_cse_move2add): Likewise.
3582         (move2add_note_store): Likewise.
3583         * print-rtl.c (print_pattern): Likewise.
3584         * recog.c (decode_asm_operands): Likewise.
3585         (store_data_bypass_p): Likewise.
3586         (if_test_bypass_p): Likewise.
3587         * regcprop.c (kill_clobbered_value): Likewise.
3588         (kill_set_value): Likewise.
3589         * reginfo.c (reg_scan_mark_refs): Likewise.
3590         * reload1.c (maybe_fix_stack_asms): Likewise.
3591         (eliminate_regs_1): Likewise.
3592         (elimination_effects): Likewise.
3593         (mark_not_eliminable): Likewise.
3594         (scan_paradoxical_subregs): Likewise.
3595         (forget_old_reloads_1): Likewise.
3596         * reorg.c (find_end_label): Likewise.
3597         (try_merge_delay_insns): Likewise.
3598         (redundant_insn): Likewise.
3599         (own_thread_p): Likewise.
3600         (fill_simple_delay_slots): Likewise.
3601         (fill_slots_from_thread): Likewise.
3602         (dbr_schedule): Likewise.
3603         * resource.c (update_live_status): Likewise.
3604         (mark_referenced_resources): Likewise.
3605         (mark_set_resources): Likewise.
3606         * rtl.c (copy_rtx): Likewise.
3607         * rtlanal.c (reg_referenced_p): Likewise.
3608         (single_set_2): Likewise.
3609         (noop_move_p): Likewise.
3610         (note_stores): Likewise.
3611         * sched-deps.c (sched_analyze_reg): Likewise.
3612         (sched_analyze_insn): Likewise.
3613
3614 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3615
3616         * cse.c (invalidate_reg): New function extracted from...
3617         (invalidate): ...here.
3618         (canonicalize_insn): Check for clobber high.
3619         (invalidate_from_clobbers): invalidate clobber highs.
3620         (invalidate_from_sets_and_clobbers): Likewise.
3621         (count_reg_usage): Check for clobber high.
3622         (insn_live_p): Likewise.
3623         * cselib.c (cselib_expand_value_rtx_1):Likewise.
3624         (cselib_invalidate_regno): Check for clobber in setter.
3625         (cselib_invalidate_rtx): Pass through setter.
3626         (cselib_invalidate_rtx_note_stores):
3627         (cselib_process_insn): Check for clobber high.
3628         * cselib.h (cselib_invalidate_rtx): Add operand.
3629
3630 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3631
3632         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3633         (mark_not_eliminable): Likewise.
3634         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3635         * lra-lives.c (process_bb_lives): Check for clobber high.
3636         * lra.c (new_insn_reg): Remember clobber highs.
3637         (collect_non_operand_hard_regs): Check for clobber high.
3638         (lra_set_insn_recog_data): Likewise.
3639         (add_regs_to_insn_regno_info): Likewise.
3640         (lra_update_insn_regno_info): Likewise.
3641
3642 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3643
3644         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3645         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3646
3647 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3648
3649         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3650         (copy_insn_1): Likewise.
3651         (gen_hard_reg_clobber_high): New gen function.
3652         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3653         * genemit.c (gen_exp): Likewise.
3654         (gen_emit_seq): Pass through info.
3655         (gen_insn): Check for CLOBBER_HIGH.
3656         (gen_expand): Pass through info.
3657         (gen_split): Likewise.
3658         (output_add_clobbers): Likewise.
3659         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3660         (remove_clobbers): Likewise.
3661         * rtl.h (gen_hard_reg_clobber_high): New declaration.
3662
3663 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3664
3665         * doc/rtl.texi (clobber_high): Add.
3666         (parallel): Add in clobber high
3667         * rtl.c (rtl_check_failed_code3): Add function.
3668         * rtl.def (CLOBBER_HIGH): Add expression.
3669         * rtl.h (RTL_CHECKC3): Add macro.
3670         (rtl_check_failed_code3): Add declaration.
3671         (XC3EXP): Add macro.
3672
3673 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
3674
3675         PR target/86386
3676         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3677         cfun->machine->max_used_stack_alignment if needed.
3678
3679 2018-08-04  Martin Sebor  <msebor@redhat.com>
3680
3681         PR tree-optimization/86571
3682         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3683         NaN output to 4.
3684
3685 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
3686
3687         PR target/86799
3688         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3689         Define.
3690
3691 2018-08-03  Jeff Law  <law@redhat.com>
3692
3693         PR target/86795
3694         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3695         Define to speculation_safe_value_not_needed.
3696
3697 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
3698
3699         * doc/gcov.texi (-x): Remove duplicate "to".
3700         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3701         (-Wif-not-aligned): Remove duplicate "is".
3702         (-flto): Remove duplicate "the".
3703         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3704         duplicate "v5.00.b".
3705         (MSP430 Options): Remove duplicate "and" from the description
3706         of "-mgprel-sec=regexp".
3707         (x86 Options): Remove duplicate copies of "vmldLog102" and
3708         vmlsLog104 from description of "-mveclibabi=type".
3709
3710 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
3711
3712         * internal-fn.h (first_commutative_argument): Declare.
3713         * internal-fn.c (first_commutative_argument): New function.
3714         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3715         restrictions for pattern statements.  Use first_commutative_argument
3716         to look for commutative operands in calls to internal functions.
3717
3718 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
3719
3720         * Makefile.in (wide-int-range.o): New.
3721         * tree-vrp.c: Move all the wide_int_* functions to...
3722         * wide-int-range.cc: ...here.
3723         * tree-vrp.h: Move all the wide_int_* prototypes to...
3724         * wide-int-range.h: ...here.
3725
3726 2018-08-03  Tom de Vries  <tdevries@suse.de>
3727
3728         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3729         UI_NONE.
3730         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3731         * except.c (output_function_exception_table): Do early exit if
3732         targetm_common.except_unwind_info (&global_options) == UI_NONE.
3733
3734 2018-08-03  Martin Liska  <mliska@suse.cz>
3735
3736         * predict.c (dump_prediction): Change to 2 digits
3737         in fraction part.
3738
3739 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
3740
3741         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3742         neon_dup_q to...
3743         (falkor_am_1_gtov_gtov): ... a new insn reservation.
3744
3745 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3746
3747         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3748         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3749         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3750         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3751         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3752
3753 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
3754
3755         * diagnostic-show-locus.c (diagnostic_show_locus): Use
3756         pp_take_prefix when saving the existing prefix.
3757         * diagnostic.c (diagnostic_append_note): Likewise.
3758         * langhooks.c (lhd_print_error_function): Likewise.
3759         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3760         param's type.  Free the existing prefix.
3761         (pp_take_prefix): New function.
3762         (pretty_printer::pretty_printer): Drop the prefix parameter.
3763         Rename the length parameter to match the comment.
3764         (pretty_printer::~pretty_printer): Free the prefix.
3765         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3766         parameter.
3767         (struct pretty_printer): Drop the "const" from "prefix" field's
3768         type and clarify memory management.
3769         (pp_set_prefix): Drop the "const" from the 2nd param.
3770         (pp_take_prefix): New decl.
3771
3772 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3773
3774         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3775         for word_mode_ok here instead of passing as argument.
3776         (expand_block_compare): Change select_block_compare_mode() call.
3777         (expand_strncmp_gpr_sequence): New function.
3778         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3779
3780 2018-08-02  Jeff Law  <law@redhat.com>
3781
3782         PR target/86790
3783         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3784         Define to speculation_safe_value_not_needed.
3785
3786         PR target/86784
3787         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3788         Define to speculation_safe_value_not_needed.
3789
3790 2018-08-02  Tom de Vries  <tdevries@suse.de>
3791
3792         PR target/86660
3793         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3794         function.  Return UI_TARGET unconditionally.
3795         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3796         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3797
3798 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
3799
3800         * genemit.c (print_overload_test): Fix typo.
3801
3802 2018-08-02  Richard Biener  <rguenther@suse.de>
3803
3804         PR tree-optimization/86816
3805         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3806         which checks for value availability before querying it.
3807         (gvn_uses_equal): Use it.
3808         (same_succ_hash): Likewise.
3809         (gimple_equal_p): Likewise.
3810
3811 2018-08-02  Nick Clifton  <nickc@redhat.com>
3812
3813         PR target/86813
3814         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3815         Define to speculation_safe_value_not_needed.
3816
3817         PR target/86810
3818         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3819         Define to speculation_safe_value_not_needed.
3820
3821         PR target/86810
3822         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3823         Define to speculation_safe_value_not_needed.
3824
3825         PR target/86803
3826         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3827         Define to speculation_safe_value_not_needed.
3828
3829         PR target/86797
3830         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3831         Define to speculation_safe_value_not_needed.
3832
3833         PR target/86791
3834         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3835         Define to speculation_safe_value_not_needed.
3836
3837         PR target/86789
3838         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3839         Define to speculation_safe_value_not_needed.
3840
3841         PR target/86787
3842         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3843         Define to speculation_safe_value_not_needed.
3844
3845         PR target/86782
3846         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3847         speculation_safe_value_not_needed.
3848
3849         PR target/86781
3850         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3851         to speculation_safe_value_not_needed.
3852
3853 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
3854
3855         * doc/md.texi: Expand the documentation of instruction names
3856         to mention port-local uses.  Document '@' in pattern names.
3857         * read-md.h (overloaded_instance, overloaded_name): New structs.
3858         (mapping): Declare.
3859         (md_reader::handle_overloaded_name): New member function.
3860         (md_reader::get_overloads): Likewise.
3861         (md_reader::m_first_overload): New member variable.
3862         (md_reader::m_next_overload_ptr): Likewise.
3863         (md_reader::m_overloads_htab): Likewise.
3864         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3865         m_next_overload_ptr and m_overloads_htab.
3866         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3867         (get_mode_token, get_code_token, get_int_token): New functions.
3868         (map_attr_string): Add an optional argument that passes back
3869         the associated iterator.
3870         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3871         (md_reader::handle_overloaded_name, add_overload_instance): New
3872         functions.
3873         (apply_iterators): Handle '@' names.  Report an error if '@'
3874         is used without iterators.
3875         (initialize_iterators): Initialize the new iterator_group fields.
3876         * genopinit.c (handle_overloaded_code_for)
3877         (handle_overloaded_gen): New functions.
3878         (main): Use them to print declarations of maybe_code_for_* and
3879         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3880         * genemit.c (print_overload_arguments, print_overload_test)
3881         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3882         (main): Use it to print definitions of maybe_code_for_* and
3883         maybe_gen_* functions.
3884         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3885         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3886         instead of explicit mode checks.
3887         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3888         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3889         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3890         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3891         (aarch64_expand_compare_and_swap): Likewise
3892         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3893         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3894         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3895         (aarch64_constant_pool_reload_icode): Delete.
3896         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3897         instead of aarch64_constant_pool_reload_icode.  Use
3898         code_for_aarch64_reload_mov instead of explicit mode checks.
3899         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3900         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3901         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3902         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3903         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3904         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3905         (aarch64_atomic_load_op_code): Delete.
3906         (aarch64_emit_atomic_load_op): Likewise.
3907         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3908         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
3909         instead of aarch64_emit_atomic_load_op.
3910         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3911         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3912         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3913         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3914         character before the pattern name.
3915         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3916         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3917         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3918         (aarch64_frecps<mode>): Likewise.
3919         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3920         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3921         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3922         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3923         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3924
3925 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
3926
3927         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3928         Allow HFmode constants if TARGET_FP_F16INST.
3929
3930 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
3931
3932         PR target/86014
3933         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3934         No longer check last store for clobber of address register.
3935
3936 2018-08-02  Martin Liska  <mliska@suse.cz>
3937
3938         PR gcov-profile/86817
3939         * gcov.c (process_all_functions): New function.
3940         (main): Call it.
3941         (process_file): Move functions processing to
3942         process_all_functions.
3943
3944 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
3945
3946         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3947         "const" to the "gimple *" and "rtx_insn *" parameters.
3948         * dumpfile.h (dump_user_location_t::dump_user_location_t):
3949         Likewise.
3950         (dump_location_t::dump_location_t): Likewise.
3951
3952 2018-08-01  Martin Sebor  <msebor@redhat.com>
3953
3954         PR tree-optimization/86650
3955         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3956         rather than a "gcall *".  Directly pass the data of interest
3957         to percent_K_format, rather than building a temporary CALL_EXPR
3958         to hold it.
3959         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3960         (gimple_fold_builtin_strncat): Adjust.
3961         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3962         gcall* argument with gimple*.
3963         * gimple-ssa-warn-restrict.c (check_call): Same.
3964         (wrestrict_dom_walker::before_dom_children): Same.
3965         (builtin_access::builtin_access): Same.
3966         (check_bounds_or_overlap): Same
3967         (maybe_diag_overlap): Same.
3968         (maybe_diag_offset_bounds): Same.
3969         * tree-diagnostic.c (default_tree_printer): Move usage of
3970         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3971         to this callsite.
3972         * tree-pretty-print.c (percent_K_format): Add argument.
3973         * tree-pretty-print.h: Add argument.
3974         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3975         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3976         (maybe_diag_stxncpy_trunc): Same.
3977         (handle_builtin_stxncpy): Same.
3978         (handle_builtin_strcat): Same.
3979
3980 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3981
3982         * match.pd: Optimise pointer range checks.
3983
3984 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3985
3986         PR tree-optimization/86758
3987         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3988         to remove pattern statements.
3989
3990 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3991
3992         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3993         result of dfs_enumerate_from when constructing stmt_vec_infos,
3994         instead of additionally calling get_loop_body.
3995
3996 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3997
3998         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3999         parameter.
4000         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
4001         When creating an iv, assert that the step is not known to be zero.
4002         (vect_setup_realignment): Update call accordingly.
4003         * tree-vect-stmts.c (vectorizable_store): Likewise.
4004         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
4005
4006 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4007
4008         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
4009         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
4010         (vectorizable_reduction): Likewise.
4011         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4012         (vect_detect_hybrid_slp_stmts): Likewise.
4013         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4014
4015 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
4016
4017         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
4018         (wide_int_set_zero_nonzero_bits): ...this.
4019         (zero_nonzero_bits_from_vr): Rename to...
4020         (vrp_set_zero_nonzero_bits): ...this.
4021         (extract_range_from_multiplicative_op_1): Abstract wide int
4022         code...
4023         (wide_int_range_multiplicative_op): ...here.
4024         (extract_range_from_binary_expr_1): Extract wide int binary
4025         operations into their own functions.
4026         (wide_int_range_lshift): New.
4027         (wide_int_range_can_optimize_bit_op): New.
4028         (wide_int_range_shift_undefined_p): New.
4029         (wide_int_range_bit_xor): New.
4030         (wide_int_range_bit_ior): New.
4031         (wide_int_range_bit_and): New.
4032         (wide_int_range_trunc_mod): New.
4033         (extract_range_into_wide_ints): New.
4034         (vrp_shift_undefined_p): New.
4035         (extract_range_from_multiplicative_op): New.
4036         (vrp_can_optimize_bit_op): New.
4037         * tree-vrp.h (value_range::dump): New.
4038         (wide_int_range_multiplicative_op): New.
4039         (wide_int_range_lshift):New.
4040         (wide_int_range_shift_undefined_p): New.
4041         (wide_int_range_bit_xor): New.
4042         (wide_int_range_bit_ior): New.
4043         (wide_int_range_bit_and): New.
4044         (wide_int_range_trunc_mod): New.
4045         (zero_nonzero_bits_from_bounds): Rename to...
4046         (wide_int_set_zero_nonzero_bits): ...this.
4047         (zero_nonzero_bits_from_vr): Rename to...
4048         (vrp_set_zero_nonzero_bits): ...this.
4049         (range_easy_mask_min_max): Rename to...
4050         (wide_int_range_can_optimize_bit_op): this.
4051         * vr-values.c (simplify_bit_ops_using_ranges): Rename
4052         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4053
4054 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4055
4056         * tree-vectorizer.h (vect_orig_stmt): New function.
4057         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4058         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4059         (vect_create_epilog_for_reduction): Likewise.
4060         (vectorizable_live_operation): Likewise.
4061         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4062         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4063         * tree-vect-stmts.c (vectorizable_call): Likewise.
4064         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4065
4066 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4067
4068         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4069         argument.
4070         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4071         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4072         (vect_transform_loop): Likewise.
4073         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4074
4075 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4076
4077         * tree-vectorizer.h (vect_schedule_slp): Return void.
4078         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4079         (vect_schedule_slp): Likewise.
4080
4081 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4082
4083         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4084         argument.
4085         (vect_transform_loop): Update calls accordingly.  Schedule SLP
4086         instances before the main loop, if any exist.
4087
4088 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4089
4090         PR tree-optimization/86749
4091         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4092         If the lhs is used in a COND_EXPR, check that it is being used
4093         as the "then" or "else" value.
4094
4095 2018-08-01  Tom de Vries  <tdevries@suse.de>
4096
4097         PR target/86800
4098         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4099         speculation_safe_value_not_needed.
4100
4101 2018-08-01  Richard Biener  <rguenther@suse.de>
4102
4103         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4104         as base and offset.
4105
4106 2018-08-01  Martin Liska  <mliska@suse.cz>
4107
4108         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4109         format how successful transformation is dumped.
4110         (gimple_mod_pow2_value_transform): Likewise.
4111         (gimple_mod_subtract_transform): Likewise.
4112         (gimple_stringops_transform): Likewise.
4113
4114 2018-08-01  Martin Liska  <mliska@suse.cz>
4115
4116         PR value-prof/35543
4117         * value-prof.c (interesting_stringop_to_profile_p):
4118         Simplify the code and add BUILT_IN_MEMMOVE.
4119         (gimple_stringops_transform): Likewise.
4120
4121 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
4122
4123         * config/aarch64/aarch64-simd.md
4124         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4125         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4126         use GPI iterator instead of SI mode.
4127
4128 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
4129
4130         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4131         rs6000_speculation_barrier.
4132         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4133         new barrier pattern name.
4134
4135 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
4136
4137         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4138         (speculation_barrier): New insn.
4139
4140 2018-08-01  Richard Biener  <rguenther@suse.de>
4141
4142         PR bootstrap/86724
4143         * graphite.h: Include isl/id.h and isl/space.h to allow build
4144         with ISL 0.20.
4145
4146 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
4147
4148         PR target/86651
4149         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4150         mode for COFF targets.
4151         * defaults.h (TARGET_COFF): Define.
4152         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4153         TARGET_COFF): Define.
4154         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4155         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4156         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4157
4158 2018-07-31  Alexandre Oliva <oliva@adacore.com>
4159
4160         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4161         * gimple-streamer-out.c (output_bb): Save it.
4162         * lto-streamer-in.c (input_struct_function_base): Restore
4163         instance discriminator if available.  Create map on demand.
4164         * lto-streamer-out.c (output_struct_function_base): Save it if
4165         available.
4166         * final.c (decl_to_instance_map): Document LTO strategy.
4167
4168 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
4169             Olivier Hainque  <hainque@adacore.com>
4170
4171         * debug.h (decl_to_instance_map_t): New type.
4172         (decl_to_instance_map): Declare.
4173         (maybe_create_decl_to_instance_map): New inline function.
4174         * final.c (bb_discriminator, last_bb_discriminator): New statics,
4175         to track basic block discriminators.
4176         (final_start_function_1): Initialize them.
4177         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4178         bb_discriminator.
4179         (decl_to_instance_map): New variable.
4180         (map_decl_to_instance, maybe_set_discriminator): New functions.
4181         (notice_source_line): Set discriminator.
4182
4183 2018-07-31  Ian Lance Taylor  <iant@golang.org>
4184
4185         * targhooks.c (default_have_speculation_safe_value): Add
4186         ATTRIBUTE_UNUSED.
4187
4188 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
4189
4190         * dump-context.h: Include "pretty-print.h".
4191         (dump_context::refresh_dumps_are_enabled): New decl.
4192         (dump_context::emit_item): New decl.
4193         (class dump_context): Add fields "m_test_pp" and
4194         "m_test_pp_flags".
4195         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4196         (temp_dump_context::get_dumped_text): New decl.
4197         (class temp_dump_context): Add field "m_pp".
4198         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4199         (dump_context::refresh_dumps_are_enabled): ...and add a test for
4200         m_test_pp.
4201         (set_dump_file): Update for above change.
4202         (set_alt_dump_file): Likewise.
4203         (dump_loc): New overload, taking a pretty_printer *.
4204         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
4205         to any test pretty-printer.
4206         (make_item_for_dump_gimple_stmt): New function, adapted from
4207         optinfo::add_gimple_stmt.
4208         (dump_context::dump_gimple_stmt): Call it, and use the result,
4209         eliminating the direct usage of dump_file and alt_dump_file in
4210         favor of indirectly using them via emit_item.
4211         (make_item_for_dump_gimple_expr): New function, adapted from
4212         optinfo::add_gimple_expr.
4213         (dump_context::dump_gimple_expr): Call it, and use the result,
4214         eliminating the direct usage of dump_file and alt_dump_file in
4215         favor of indirectly using them via emit_item.
4216         (make_item_for_dump_generic_expr): New function, adapted from
4217         optinfo::add_tree.
4218         (dump_context::dump_generic_expr): Call it, and use the result,
4219         eliminating the direct usage of dump_file and alt_dump_file in
4220         favor of indirectly using them via emit_item.
4221         (make_item_for_dump_printf_va): New function, adapted from
4222         optinfo::add_printf_va.
4223         (make_item_for_dump_printf): New function.
4224         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4225         and use the result, eliminating the direct usage of dump_file and
4226         alt_dump_file in favor of indirectly using them via emit_item.
4227         (make_item_for_dump_dec): New function.
4228         (dump_context::dump_dec): Call it, and use the result,
4229         eliminating the direct usage of dump_file and alt_dump_file in
4230         favor of indirectly using them via emit_item.
4231         (make_item_for_dump_symtab_node): New function, adapted from
4232         optinfo::add_symtab_node.
4233         (dump_context::dump_symtab_node): Call it, and use the result,
4234         eliminating the direct usage of dump_file and alt_dump_file in
4235         favor of indirectly using them via emit_item.
4236         (dump_context::begin_scope): Reimplement, avoiding direct usage
4237         of dump_file and alt_dump_file in favor of indirectly using them
4238         via emit_item.
4239         (dump_context::emit_item): New member function.
4240         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4241         Set up test pretty-printer on the underlying context.  Call
4242         refresh_dumps_are_enabled.
4243         (temp_dump_context::~temp_dump_context): Call
4244         refresh_dumps_are_enabled.
4245         (temp_dump_context::get_dumped_text): New member function.
4246         (selftest::verify_dumped_text): New function.
4247         (ASSERT_DUMPED_TEXT_EQ): New macro.
4248         (selftest::test_capture_of_dump_calls): Run all tests twice, with
4249         and then without optinfo enabled.  Add uses of
4250         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
4251         * dumpfile.h: Update comment for the dump_* API.
4252         * optinfo-emit-json.cc
4253         (selftest::test_building_json_from_dump_calls): Update for new
4254         param for temp_dump_context ctor.
4255         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
4256         and "m_owned" field.
4257         (optinfo_item::~optinfo_item): Likewise.
4258         (optinfo::add_item): New member function.
4259         (optinfo::emit): Update comment.
4260         (optinfo::add_string): Delete.
4261         (optinfo::add_printf): Delete.
4262         (optinfo::add_printf_va): Delete.
4263         (optinfo::add_gimple_stmt): Delete.
4264         (optinfo::add_gimple_expr): Delete.
4265         (optinfo::add_tree): Delete.
4266         (optinfo::add_symtab_node): Delete.
4267         (optinfo::add_dec): Delete.
4268         * optinfo.h (class dump_context): New forward decl.
4269         (optinfo::add_item): New decl.
4270         (optinfo::add_string): Delete.
4271         (optinfo::add_printf): Delete.
4272         (optinfo::add_printf_va): Delete.
4273         (optinfo::add_gimple_stmt): Delete.
4274         (optinfo::add_gimple_expr): Delete.
4275         (optinfo::add_tree): Delete.
4276         (optinfo::add_symtab_node): Delete.
4277         (optinfo::add_dec): Delete.
4278         (optinfo::add_poly_int): Delete.
4279         (optinfo_item::optinfo_item): Remove "owned" param.
4280         (class optinfo_item): Remove field "m_owned".
4281
4282 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4283
4284         PR middle-end/86705
4285         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
4286         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
4287         requested variable alignment.
4288         (expand_one_ssa_partition): Likewise.
4289         (expand_one_var): Likewise.
4290
4291 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4292
4293         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4294         to speculation_safe_value_not_needed.
4295
4296 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4297
4298         * targhooks.h (speculation_safe_value_not_needed): New prototype.
4299         * targhooks.c (speculation_safe_value_not_needed): New function.
4300         * target.def (have_speculation_safe_value): Update documentation.
4301         * doc/tm.texi: Regenerated.
4302
4303 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4304
4305         * config/aarch64/iterators.md (ALLI_TI): New iterator.
4306         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4307         expand.
4308         (despeculate_copy<ALLI:mode>_insn): New insn.
4309         (despeculate_copyti_insn): New insn.
4310         (despeculate_simple<ALLI:mode>): New insn
4311         (despeculate_simpleti): New insn.
4312         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4313         function.
4314         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4315         aarch64_speculation_safe_value.
4316         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4317
4318 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4319
4320         * config/aarch64/aarch64-speculation.cc: New file.
4321         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4322         before pass_reorder_blocks.
4323         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4324         prototype.
4325         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4326         X14 and X15 when tracking speculation.
4327         * config/aarch64/aarch64.md (register name constants): Add
4328         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4329         (unspec): Add UNSPEC_SPECULATION_TRACKER.
4330         (speculation_barrier): New insn attribute.
4331         (cmp<mode>): Allow SP in comparisons.
4332         (speculation_tracker): New insn.
4333         (speculation_barrier): Add speculation_barrier attribute.
4334         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4335         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4336         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4337
4338 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4339
4340         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4341         aarch64_track_speculation is true.
4342         (tb<optab><mode>1): Likewise.
4343         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4344         generate CB[N]Z when tracking speculation.
4345         (aarch64_split_compare_and_swap): Likewise.
4346         (aarch64_split_atomic_op): Likewise.
4347
4348 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4349
4350         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4351
4352 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4353
4354         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4355         (speculation_barrier): New insn.
4356
4357 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4358
4359         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4360         * config/arm/arm.md (speculation_barrier): New expand.
4361         (speculation_barrier_insn): New pattern.
4362
4363 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4364
4365         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4366         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4367         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4368         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4369         list.
4370         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4371         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4372         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4373         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4374         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4375         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4376         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4377         * builtins.c (expand_speculation_safe_value): New function.
4378         (expand_builtin): Call it.
4379         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4380         * doc/extend.texi: Document __builtin_speculation_safe_value.
4381         * doc/md.texi: Document "speculation_barrier" pattern.
4382         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4383         TARGET_HAVE_SPECULATION_SAFE_VALUE.
4384         * doc/tm.texi: Regenerated.
4385         * target.def (have_speculation_safe_value, speculation_safe_value): New
4386         hooks.
4387         * targhooks.c (default_have_speculation_safe_value): New function.
4388         (default_speculation_safe_value): New function.
4389         * targhooks.h (default_have_speculation_safe_value): Add prototype.
4390         (default_speculation_safe_value): Add prototype.
4391
4392 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
4393
4394         * dump-context.h (dump_context::dump_loc): New decl.
4395         * dumpfile.c (dump_context::dump_loc): New member function.
4396         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4397         and dump_gimple_stmt.
4398         (dump_context::dump_gimple_expr_loc): Likewise, using
4399         dump_gimple_expr.
4400         (dump_context::dump_generic_expr_loc): Likewise, using
4401         dump_generic_expr.
4402         (dump_context::dump_printf_loc_va): Likewise, using
4403         dump_printf_va.
4404         (dump_context::begin_scope): Explicitly using the global function
4405         "dump_loc", rather than the member function.
4406
4407 2018-07-31  Martin Sebor  <msebor@redhat.com>
4408
4409         PR tree-optimization/86741
4410         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4411
4412 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
4413
4414         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4415         depend on whether prefetch instructions will be emitted or not.
4416         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4417         will be emitted or not.
4418         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4419         (TARGET_SETMEM_PFD): New macros.
4420
4421 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4422
4423         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4424         (NULL_STMT_VEC_INFO): Delete.
4425         (stmt_vec_info::operator*): Likewise.
4426         (stmt_vec_info::operator gimple *): Likewise.
4427         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4428         of NULL_STMT_VEC_INFO.
4429         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4430         (vect_reassociating_reduction_p): Likewise.
4431         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4432         (vectorizable_store): Likewise.
4433         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4434         (vec_info::free_stmt_vec_infos): Likewise.
4435
4436 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4437
4438         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4439         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4440
4441 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4442
4443         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4444         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4445         (vec_info::free_stmt_vec_info): New private member functions.
4446         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4447         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4448         * tree-parloops.c (gather_scalar_reductions): Remove calls to
4449         set_stmt_vec_info_vec and free_stmt_vec_infos.
4450         * tree-vect-loop.c (_loop_vec_info): Remove call to
4451         set_stmt_vec_info_vec.
4452         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4453         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4454         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4455         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4456         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
4457         assignments in {vec_info::,}new_stmt_vec_info that are redundant
4458         with the clearing in the xcalloc.
4459         (stmt_vec_info_vec): Delete.
4460         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4461         (vectorize_loops): Likewise.
4462         (vec_info::~vec_info): Remove argument from call to
4463         free_stmt_vec_infos.
4464         (vec_info::add_stmt): Remove vinfo argument from call to
4465         new_stmt_vec_info.
4466
4467 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4468
4469         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4470         rather than a gimple stmt.
4471         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
4472         information for pattern statements when passed the original
4473         statement; instead wait to be passed the pattern statement itself.
4474         Don't call set_vinfo_for_stmt here.
4475         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4476         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4477         stmt_vec_infos here.
4478         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4479         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4480         stmt_vec_infos entry.
4481
4482 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4483
4484         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4485         * tree-vectorizer.c (vec_info::replace_stmt): New function.
4486         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4487         * tree-vect-stmts.c (vectorizable_call): Likewise.
4488         (vectorizable_simd_clone_call): Likewise.
4489
4490 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4491
4492         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4493         * tree-vectorizer.c (vec_info::remove_stmt): New function.
4494         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4495         * tree-vect-loop.c (vect_transform_loop): Likewise.
4496         * tree-vect-slp.c (vect_schedule_slp): Likewise.
4497         * tree-vect-stmts.c (vect_remove_stores): Likewise.
4498
4499 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4500
4501         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4502         (vect_dr_stmt): Delete.
4503         * tree-vectorizer.c (vec_info::lookup_dr): New function.
4504         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4505         of DR_VECT_AUX.
4506         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4507         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4508         (vect_verify_datarefs_alignment, vect_peeling_supportable)
4509         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4510         (vect_analyze_data_refs): Likewise.
4511         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
4512         argument.
4513         (vect_find_same_alignment_drs): Likewise.
4514         (vect_slp_analyze_node_dependences): Update calls accordingly.
4515         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
4516         instead of DR_VECT_AUX.
4517         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4518         of a vector data references.  Use vec_info::lookup_dr instead of
4519         DR_VECT_AUX.
4520         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4521         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
4522         instead of DR_VECT_AUX.
4523
4524 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4525
4526         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4527         dr_vec_info.
4528         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4529         accordingly.
4530         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4531         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4532         (vect_gen_prolog_loop_niters): Likewise.
4533
4534 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4535
4536         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4537         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4538         (vect_known_alignment_in_bytes, vect_dr_behavior)
4539         (vect_get_scalar_dr_size): Take references as dr_vec_infos
4540         instead of data_references.  Update calls to other routines for
4541         which the same change has been made.
4542         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4543         dr_vec_infos instead of stmt_vec_infos.
4544         (vect_analyze_data_ref_dependence): Update call accordingly.
4545         (vect_slp_analyze_data_ref_dependence)
4546         (vect_record_base_alignments): Use DR_VECT_AUX.
4547         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4548         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4549         (vector_alignment_reachable_p, vect_get_data_access_cost)
4550         (vect_peeling_supportable, vect_analyze_group_access_1)
4551         (vect_analyze_group_access, vect_analyze_data_ref_access)
4552         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4553         (vect_compile_time_alias, vect_small_gap_p)
4554         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4555         (vect_supportable_dr_alignment): Take references as dr_vec_infos
4556         instead of data_references.  Update calls to other routines for
4557         which the same change has been made.
4558         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4559         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4560         (vect_slp_analyze_and_verify_node_alignment)
4561         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4562         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4563         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
4564         above changes.
4565         (_vect_peel_info::dr): Replace with...
4566         (_vect_peel_info::dr_info): ...this new field.
4567         (vect_peeling_hash_get_most_frequent)
4568         (vect_peeling_hash_choose_best_peeling): Update accordingly.
4569         (vect_peeling_hash_get_lowest_cost):
4570         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
4571         routines for which the same change has been made.
4572         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
4573         data_reference.
4574         * tree-vect-loop-manip.c (get_misalign_in_elems)
4575         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
4576         above changes.
4577         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4578         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4579         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4580         (get_group_load_store_type, get_negative_load_store_type)
4581         (vect_get_data_ptr_increment, vectorizable_store)
4582         (vectorizable_load): Likewise.
4583         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4584         Update calls to other routines for which the same change has been made.
4585
4586 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4587
4588         * tree-vectorizer.h (vec_info::move_dr): New member function.
4589         (dataref_aux): Rename to...
4590         (dr_vec_info): ...this and add "dr" and "stmt" fields.
4591         (_stmt_vec_info::dr_aux): Update accordingly.
4592         (_stmt_vec_info::data_ref_info): Delete.
4593         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4594         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4595         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4596         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4597         of data_ref.
4598         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
4599         (STMT_VINFO_DR_INFO): New macro.
4600         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4601         (set_dr_misalignment): Update after rename of dataref_aux.
4602         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
4603         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4604         initialization of STMT_VINFO_DATA_REF.
4605         * tree-vectorizer.c (vec_info::move_dr): New function.
4606         * tree-vect-patterns.c (vect_recog_bool_pattern)
4607         (vect_recog_mask_conversion_pattern)
4608         (vect_recog_gather_scatter_pattern): Use it.
4609         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4610         the "dr" and "stmt" fields of dr_vec_info instead of
4611         STMT_VINFO_DATA_REF.
4612
4613 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4614
4615         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4616         (is_pattern_stmt_p): Use it.
4617         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4618         on pattern statements.
4619
4620 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4621
4622         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4623         original stmt as a stmt_vec_info rather than a gimple stmt.
4624         (vect_pattern_recog_1): Take the statement directly as a
4625         stmt_vec_info, rather than via a gimple_stmt_iterator.
4626         Update call to vect_mark_pattern_stmts.
4627         (vect_pattern_recog): Update calls accordingly.
4628
4629 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4630
4631         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4632         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4633         a vect_def_type for the first argument.
4634         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4635         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
4636         operand if it isn't defined by a vectorized statement.
4637         (vect_build_gather_load_calls): Remove the mask_dt argument and
4638         update calls to vect_get_vec_def_for_stmt_copy.
4639         (vectorizable_bswap): Likewise the dt argument.
4640         (vectorizable_call): Update calls to vectorizable_bswap and
4641         vect_get_vec_def_for_stmt_copy.
4642         (vectorizable_simd_clone_call, vectorizable_assignment)
4643         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4644         (vectorizable_comparison): Update calls to
4645         vect_get_vec_def_for_stmt_copy.
4646         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
4647         vect_is_simple_use.
4648         (vect_get_loop_based_defs): Remove dt argument and update call
4649         to vect_get_vec_def_for_stmt_copy.
4650         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4651         and vect_get_vec_def_for_stmt_copy.
4652         (vectorizable_load): Update calls to vect_build_gather_load_calls
4653         and vect_get_vec_def_for_stmt_copy.
4654         * tree-vect-loop.c (vect_create_epilog_for_reduction)
4655         (vectorizable_reduction, vectorizable_live_operation): Update calls
4656         to vect_get_vec_def_for_stmt_copy.
4657
4658 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4659
4660         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4661         and gimple stmt arguments with a stmt_vec_info.
4662         (vect_record_base_alignments): Update calls accordingly.
4663         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4664         and gimple stmt arguments with a stmt_vec_info.
4665         (vect_build_slp_tree_1): Remove vinfo argument and update call
4666         to vect_record_max_nunits.
4667         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4668         and vect_record_max_nunits.
4669
4670 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4671
4672         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4673         file and take a stmt_vec_info instead of a gimple stmt.
4674         (supportable_widening_operation, vect_finish_replace_stmt)
4675         (vect_finish_stmt_generation, vect_get_store_rhs)
4676         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4677         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4678         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4679         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4680         (vect_create_data_ref_ptr, bump_vector_ptr)
4681         (vect_permute_store_chain, vect_setup_realignment)
4682         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4683         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4684         (vectorizable_reduction, vectorizable_induction)
4685         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4686         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4687         than gimple stmts as arguments.
4688         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4689         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4690         (can_group_stmts_p, vect_check_gather_scatter)
4691         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4692         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4693         (vect_permute_load_chain, vect_shift_permute_load_chain)
4694         (vect_transform_grouped_load)
4695         (vect_record_grouped_load_vectors): Likewise.
4696         * tree-vect-loop.c (vect_fixup_reduc_chain)
4697         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4698         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4699         (vectorizable_reduction, vectorizable_induction)
4700         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4701         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4702         (vect_get_load_store_mask): Likewise.
4703         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4704         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4705         * tree-vect-stmts.c (vect_mark_relevant)
4706         (is_simple_and_all_uses_invariant)
4707         (exist_non_indexing_operands_for_use_p, process_use)
4708         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4709         (vect_get_vec_def_for_operand, vect_get_vec_defs)
4710         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4711         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4712         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4713         (get_negative_load_store_type, get_load_store_type)
4714         (vect_check_load_store_mask, vect_check_store_rhs)
4715         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4716         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4717         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4718         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4719         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4720         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4721         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4722         (supportable_widening_operation): Likewise.
4723
4724 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4725
4726         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4727         a stmt_vec_info instead of a gcall.
4728         (vect_check_gather_scatter): Update call accordingly.
4729         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4730         of a gphi.
4731         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4732         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4733         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4734         instead of a gimple stmt.
4735         (vect_transform_loop): Update calls accordingly.
4736         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4737         stmt_vec_infos instead of gimple stmts.
4738         (vect_analyze_slp_instance): Update use accordingly.
4739         * tree-vect-stmts.c (read_vector_array, write_vector_array)
4740         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4741         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4742         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4743         (vect_gen_widened_results_half, vect_get_loop_based_defs)
4744         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4745         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4746         down to subroutines.
4747
4748 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4749
4750         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4751         of the worklist from a vector of gimple stmts to a vector of
4752         stmt_vec_infos.
4753         * tree-vect-stmts.c (vect_mark_relevant, process_use)
4754         (vect_mark_stmts_to_be_vectorized): Likewise
4755
4756 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4757
4758         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4759         statement before passing it to vect_analyze_stmt.
4760         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4761         the chain of phi vector definitions.  Track the exit phi via its
4762         stmt_vec_info.
4763         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4764         STMT_VINFO_REDUC_DEF.
4765         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4766         stmt_vec_infos to handle the statement chains.
4767         (vect_get_slp_defs): Record the first statement in the node
4768         using a stmt_vec_info.
4769         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4770         statements here and pass their stmt_vec_info down to subroutines.
4771         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4772         down to vect_finish_stmt_generation.
4773         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4774         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4775         stmt_vec_infos to subroutines.
4776         (vect_remove_stores): Use stmt_vec_infos to handle the statement
4777         chains.
4778
4779 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4780
4781         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4782         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4783         (vect_permute_store_chain, vect_setup_realignment)
4784         (vect_permute_load_chain, vect_shift_permute_load_chain)
4785         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4786         stmts internally, and when passing values to other vectorizer routines.
4787         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4788         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4789         (vect_analyze_loop_operations, get_initial_def_for_reduction)
4790         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4791         (vectorizable_reduction, vectorizable_induction)
4792         (vectorizable_live_operation, vect_transform_loop_stmt)
4793         (vect_transform_loop): Likewise.
4794         * tree-vect-patterns.c (vect_reassociating_reduction_p)
4795         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4796         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4797         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4798         (vect_slp_analyze_node_operations_1): Likewise.
4799         * tree-vect-stmts.c (vect_mark_relevant, process_use)
4800         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4801         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4802         (vect_finish_stmt_generation_1, get_group_load_store_type)
4803         (get_load_store_type, vect_build_gather_load_calls)
4804         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4805         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4806         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4807         (vectorizable_store, vectorizable_load, vectorizable_condition)
4808         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4809         (supportable_widening_operation): Likewise.
4810         (vect_get_vector_types_for_stmt): Likewise.
4811         * tree-vectorizer.h (vect_dr_behavior): Likewise.
4812
4813 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4814
4815         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4816         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4817         (vect_permute_store_chain, vect_permute_load_chain)
4818         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4819         repeated stmt_vec_info lookups.
4820         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4821         (vect_update_ivs_after_vectorizer): Likewise.
4822         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4823         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4824         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4825         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4826         (vect_bb_slp_scalar_cost): Likewise.
4827         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4828
4829 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4830
4831         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4832         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4833         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4834         to get gassigns and gcalls, rather than operating on generc gimple
4835         stmts.
4836         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4837         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4838         (vectorizable_load, vect_analyze_stmt): Likewise.
4839         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4840
4841 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4842
4843         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4844         return stmt_vec_infos rather than gimple stmts.  Do not accept
4845         null arguments.
4846         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4847         of a gimple stmt.
4848         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4849         Update use of get_later_stmt.
4850         (vect_get_constant_vectors): Update call accordingly.
4851         (vect_schedule_slp_instance): Likewise
4852         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4853         (vect_slp_analyze_instance_dependence): Likewise.
4854         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4855
4856 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4857
4858         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4859         (stmt_info_for_cost::stmt_info): ...this new field.
4860         (add_stmt_costs): Update accordingly.
4861         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4862         (vect_get_known_peeling_cost): Likewise.
4863         (vect_estimate_min_profitable_iters): Likewise.
4864         * tree-vect-stmts.c (record_stmt_cost): Likewise.
4865
4866 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4867
4868         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4869         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4870         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4871         accordingly.
4872         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4873
4874 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4875
4876         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4877         a gimple stmt to a stmt_vec_info.
4878         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4879
4880 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4881
4882         * tree-vectorizer.h (vec_info::grouped_stores): Change from
4883         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4884         (_loop_vec_info::reduction_chains): Likewise.
4885         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4886         accordingly.
4887         * tree-vect-slp.c (vect_analyze_slp): Likewise.
4888
4889 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4890
4891         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4892         a gimple stmt to a stmt_vec_info.
4893         (_stmt_vec_info::next_element): Likewise.
4894         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4895         (vect_slp_analyze_and_verify_node_alignment)
4896         (vect_analyze_group_access_1, vect_analyze_group_access)
4897         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4898         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4899         (vect_supportable_dr_alignment): Update accordingly.
4900         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4901         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4902         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4903         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4904         * tree-vect-slp.c (vect_build_slp_tree_1)
4905         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4906         (vect_split_slp_store_group, vect_analyze_slp_instance)
4907         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4908         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4909         (get_group_load_store_type, get_load_store_type)
4910         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4911         (vect_transform_stmt, vect_remove_stores): Likewise.
4912
4913 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4914
4915         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4916         than a gimple stmt.
4917         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4918         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4919         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4920         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4921         (vector_alignment_reachable_p, vect_get_data_access_cost)
4922         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4923         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4924         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4925         (vect_analyze_group_access_1, vect_analyze_group_access)
4926         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4927         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4928         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4929         result of vect_dr_stmt and use the stmt_vec_info instead of
4930         the associated gimple stmt.
4931         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4932         (vect_gen_prolog_loop_niters): Likewise.
4933         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4934
4935 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4936
4937         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4938         to a vec<stmt_vec_info>.
4939         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4940         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4941         vec<stmt_vec_info>.
4942         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4943         to a vec<stmt_vec_info>.
4944         (bst_traits::value_type, bst_traits::value_type): Likewise.
4945         (bst_traits::hash): Update accordingly.
4946         (vect_get_and_check_slp_defs): Change the stmts parameter from
4947         a vec<gimple *> to a vec<stmt_vec_info>.
4948         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4949         (vect_build_slp_tree): Likewise.
4950         (vect_build_slp_tree_2): Likewise.  Update uses of
4951         SLP_TREE_SCALAR_STMTS.
4952         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4953         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4954         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4955         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4956         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4957         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4958         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4959         (vect_get_constant_vectors, vect_get_slp_defs)
4960         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4961         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4962         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4963         instead of gimple stmts.
4964         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4965         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4966         (vect_slp_analyze_instance_dependence): Update uses of
4967         SLP_TREE_SCALAR_STMTS.
4968         (vect_slp_analyze_and_verify_node_alignment): Likewise.
4969         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4970         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4971         (get_initial_defs_for_reduction): Likewise.
4972         (vect_create_epilog_for_reduction): Likewise.
4973         (vectorize_fold_left_reduction): Likewise.
4974         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4975         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4976         (can_vectorize_live_stmts): Likewise.
4977
4978 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4979
4980         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4981         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4982         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4983         than gimple stmts.
4984         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4985         of a gimple stmt.
4986         (gather_scalar_reductions): Update after above interface changes.
4987         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4988         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4989         than gimple stmts.
4990         (vect_force_simple_reduction): Likewise.
4991         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4992         LOOP_VINFO_REDUCTIONS.
4993         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4994
4995 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4996
4997         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4998         a gimple stmt to a stmt_vec_info.
4999         * tree-vect-loop.c (vect_active_double_reduction_p)
5000         (vect_force_simple_reduction, vectorizable_reduction): Update
5001         accordingly.
5002
5003 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5004
5005         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
5006         vec<gimple *> to a vec<stmt_vec_info>.
5007         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
5008         the reduction_phis argument from a vec<gimple *> to a
5009         vec<stmt_vec_info>.
5010         (vectorizable_reduction): Likewise the phis local variable that
5011         is passed to vect_create_epilog_for_reduction.  Update for new type
5012         of SLP_TREE_VEC_STMTS.
5013         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
5014         (vectorizable_live_operation): Likewise.
5015         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
5016         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
5017
5018 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5019
5020         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
5021         a gimple stmt to a stmt_vec_info.
5022         (vectorizable_condition, vectorizable_live_operation)
5023         (vectorizable_reduction, vectorizable_induction): Pass back the
5024         vectorized statement as a stmt_vec_info.
5025         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5026         use of STMT_VINFO_VEC_STMT.
5027         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5028         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5029         as stmt_vec_infos rather than gimple stmts.
5030         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5031         to a stmt_vec_info.
5032         (vectorizable_live_operation): Likewise.
5033         (vectorizable_reduction, vectorizable_induction): Likewise,
5034         updating use of STMT_VINFO_VEC_STMT.
5035         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5036         of STMT_VINFO_VEC_STMT.
5037         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5038         (vectorizable_simd_clone_call, vectorizable_conversion)
5039         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5040         (vectorizable_store, vectorizable_load, vectorizable_condition)
5041         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5042         from a gimple stmt to a stmt_vec_info.
5043         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
5044         pointer to a stmt_vec_info to the vectorizable_* routines.
5045
5046 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5047
5048         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5049         a gimple stmt to a stmt_vec_info.
5050         (is_pattern_stmt_p): Update accordingly.
5051         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5052         (vect_record_grouped_load_vectors): Likewise.
5053         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5054         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5055         (vect_model_reduction_cost): Likewise.
5056         (vect_create_epilog_for_reduction): Likewise.
5057         (vectorizable_reduction, vectorizable_induction): Likewise.
5058         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5059         Return the stmt_vec_info for the pattern statement.
5060         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5061         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5062         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5063         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5064         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5065         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5066         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5067         (free_stmt_vec_info, vect_is_simple_use): Likewise.
5068
5069 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5070
5071         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5072         (vect_finish_stmt_generation): Likewise.
5073         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5074         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5075         (vect_build_gather_load_calls): Use the return value of the above
5076         functions instead of a separate call to vinfo_for_stmt.  Use narrow
5077         scopes for the input gimple stmt and wider scopes for the associated
5078         stmt_vec_info.  Use vec_info::lookup_def when setting these
5079         stmt_vec_infos from an SSA_NAME definition.
5080         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5081         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5082         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5083         (vectorizable_store, vectorizable_load, vectorizable_condition)
5084         (vectorizable_comparison): Likewise.
5085         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5086         (vectorizable_reduction): Likewise.
5087
5088 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5089
5090         * tree-vectorizer.h (vect_is_simple_use): Add an optional
5091         stmt_vec_info * parameter before the optional gimple **.
5092         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5093         (process_use, vect_get_vec_def_for_operand_1): Update callers.
5094         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5095         * tree-vect-loop.c (vectorizable_reduction): Likewise.
5096         (vectorizable_live_operation): Likewise.
5097         * tree-vect-patterns.c (type_conversion_p): Likewise.
5098         (vect_look_through_possible_promotion): Likewise.
5099         (vect_recog_rotate_pattern): Likewise.
5100         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5101
5102 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5103
5104         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5105         a typedef to a wrapper class.
5106         (NULL_STMT_VEC_INFO): New macro.
5107         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5108         (stmt_vec_info::operator*): New function.
5109         (stmt_vec_info::operator gimple *): Likewise.
5110         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5111         (add_stmt_costs): Likewise.
5112         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5113         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5114         (vect_get_known_peeling_cost): Likewise.
5115         (vect_estimate_min_profitable_iters): Likewise.
5116         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5117         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5118         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5119         (vectorizable_store, free_stmt_vec_infos): Likewise.
5120         (new_stmt_vec_info): Change return type of xcalloc to
5121         _stmt_vec_info *.
5122
5123 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5124
5125         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5126         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5127         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5128         a single_imm_use-based sequence.
5129         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5130
5131 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5132
5133         * tree-vectorizer.h (vec_info::lookup_def): Declare.
5134         * tree-vectorizer.c (vec_info::lookup_def): New function.
5135         * tree-vect-patterns.c (vect_get_internal_def): Use it.
5136         (vect_widened_op_tree): Likewise.
5137         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5138         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5139         (vectorizable_reduction): Likewise.
5140         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5141         of a gimple *.
5142         (vect_is_slp_reduction): Update calls accordingly.  Use
5143         vec_info::lookup_def.
5144         (vect_is_simple_reduction): Likewise
5145         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5146
5147 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5148
5149         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5150         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5151         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5152         of vinfo_for_stmt.
5153         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5154         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5155         (vect_update_vf_for_slp, vect_analyze_loop_operations)
5156         (vect_is_slp_reduction, vectorizable_induction)
5157         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5158         * tree-vect-patterns.c (vect_init_pattern_stmt):
5159         (vect_determine_min_output_precision_1, vect_determine_precisions)
5160         (vect_pattern_recog): Likewise.
5161         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5162         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5163         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5164         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5165         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5166         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
5167         info field from a loop to a loop_vec_info.
5168
5169 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5170
5171         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5172         (vec_info::add_stmt): Declare.
5173         * tree-vectorizer.c (vec_info::add_stmt): New function.
5174         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5175         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5176         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5177         (vectorizable_induction): Likewise.
5178         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5179         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5180         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5181         (vectorizable_load): Likewise.
5182         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5183         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5184         (vect_recog_gather_scatter_pattern): Likewise.
5185         (append_pattern_def_seq): Likewise.  Remove a check that is
5186         performed by add_stmt itself.
5187
5188 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5189
5190         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5191         which make_ssa_name was called with new_stmt before new_stmt
5192         had been created.
5193
5194 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5195
5196         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5197         split out from...
5198         (vect_is_slp_reduction): ...here...
5199         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
5200         that are already known to be false.
5201
5202 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5203
5204         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5205         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
5206         STMT_VINFO_NUM_SLP_USES when it's true.
5207         (vect_free_slp_instance): Add a final_p parameter and pass it to
5208         vect_free_slp_tree.
5209         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5210         (vect_analyze_slp_instance): Likewise.
5211         (vect_slp_analyze_operations): Likewise.
5212         (vect_slp_analyze_bb_1): Likewise.
5213         * tree-vectorizer.c (vec_info): Likewise.
5214         * tree-vect-loop.c (vect_transform_loop): Likewise.
5215
5216 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5217
5218         * tree-vect-loop.c (vectorizable_reduction): Assert that the
5219         function is not called for second and subsequent members of
5220         a reduction group.
5221
5222 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5223
5224         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5225         cases for nested loops from here to ...
5226         (vect_create_epilog_for_reduction): ...here.  Only call
5227         vect_is_simple_use for inner-loop reductions.
5228
5229 2018-07-31  Martin Liska  <mliska@suse.cz>
5230
5231         PR gcov-profile/85338
5232         PR gcov-profile/85350
5233         PR gcov-profile/85372
5234         * profile.c (struct location_triplet): New.
5235         (struct location_triplet_hash): Likewise.
5236         (output_location): Do not output a BB that
5237         is already recorded for a line.
5238         (branch_prob): Use streamed_locations.
5239
5240 2018-07-31  Martin Liska  <mliska@suse.cz>
5241
5242         PR gcov-profile/85370
5243         * coverage.c (coverage_begin_function): Do not mark target
5244         clones as artificial functions.
5245
5246 2018-07-31  Martin Liska  <mliska@suse.cz>
5247
5248         PR gcov-profile/83813
5249         PR gcov-profile/84758
5250         PR gcov-profile/85217
5251         PR gcov-profile/85332
5252         * profile.c (branch_prob): Do not record GOTO expressions
5253         for GIMPLE statements which locations are already streamed.
5254
5255 2018-07-31  Olivier Hainque  <hainque@adacore.com>
5256
5257         * gcc.c (handle_spec_function): Accept a soft_matched_part
5258         argument, as do_spec_1.  Pass it down to ...
5259         (eval_spec_function): Accept a soft_matched_part argument,
5260         and pass it down to ...
5261         (do_spec_2): Accept a soft_matched_part argument, and pass
5262         it down to do_spec_1.
5263         (do_spec_1): Pass soft_matched_part to handle_spec_function.
5264         (handle_braces): Update call to handle_spec_function.
5265         (driver::set_up_specs): Update calls to do_spec_2.
5266         (compare_debug_dump_opt_spec_function): Likewise.
5267         (compare_debug_self_opt_spec_function): Likewise.
5268
5269 2018-07-31  Olivier Hainque  <hainque@adacore.com>
5270
5271         * common.opt (nolibc): New option.
5272         * doc/invoke.texi (Link Options): Document it.
5273         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
5274         * config/alpha/linux.h: Likewise.
5275         * config/arc/elf.h: Likewise.
5276         * config/arm/uclinux-elf.h: Likewise.
5277         * config/arm/unknown-elf.h: Likewise.
5278         * config/avr/avrlibc.h: Likewise.
5279         * config/bfin/bfin.h: Likewise.
5280         * config/bfin/linux.h: Likewise.
5281         * config/bfin/uclinux.h: Likewise.
5282         * config/darwin.h: Likewise.
5283         * config/darwin10.h: Likewise.
5284         * config/darwin12.h: Likewise.
5285         * config/gnu-user.h: Likewise.
5286         * config/lm32/uclinux-elf.h: Likewise.
5287         * config/pa/pa-hpux11.h: Likewise.
5288         * config/pa/pa64-hpux.h: Likewise.
5289         * config/sparc/sparc.h: Likewise.
5290
5291 2018-07-31  Olivier Hainque  <hainque@adacore.com>
5292
5293         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
5294         undefined variables.
5295
5296 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
5297
5298         PR target/86640
5299         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
5300         instead of GEN_INT.
5301
5302 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5303
5304         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5305         terminated string literal.
5306
5307 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
5308
5309         PR rtl-optimization/85160
5310         * combine.c (is_just_move): New function.
5311         (try_combine): Allow combining two instructions into two if neither of
5312         the original instructions was a move.
5313
5314 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
5315
5316         PR target/86673
5317         * doc/extend.texi (Global Register Variables): Discourage use of type
5318         qualifiers.
5319         (Local Register Variables): Likewise.
5320
5321 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
5322
5323         PR tree-optimization/86506
5324         * hwint.h (ceil_log2): Resync with hwint.c implementation.
5325
5326 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
5327
5328         PR target/86547
5329         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5330         hard_regno, make sure no insn between `from` and `to` clobbers it.
5331
5332 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
5333             Tom de Vries  <tdevries@suse.de>
5334
5335         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5336         (PTX_DEFAULT_RUNTIME_DIM): ... this.
5337         (nvptx_goacc_validate_dims): Set default worker and gang dims to
5338         PTX_DEFAULT_RUNTIME_DIM.
5339         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5340
5341 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
5342
5343         * config/pa/pa.c (pa_output_addr_vec): Align address table.
5344         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5345         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5346
5347 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
5348
5349         * config/rs6000/constraints.md (wG constraint): Delete, no longer
5350         used.
5351         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5352         predicate to reflect toc fusion has been deleted.
5353         (toc_fusion_mem_raw): Delete, no longer used.
5354         (toc_fusion_mem_wrapped): Likewise.
5355         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5356         fusion mask bit.
5357         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5358         Delete, no longer used.
5359         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5360         meant to be used for toc fusion.
5361         (rs6000_debug_print_mode): Delete toc fusion debugging.
5362         (rs6000_debug_reg_global): Likewise.
5363         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5364         fusion and secondary reload support that were never used.
5365         (rs6000_option_override_internal): Delete TOC fusion, that was only
5366         partially defined, and it did not work unless you also used the
5367         -mcmodel= switch.
5368         (rs6000_legitimate_address_p): Delete TOC fusion support.
5369         (rs6000_opt_masks): Likewise.
5370         (fusion_wrap_memory_address): Delete function, no longer used.
5371         (fusion_split_address); Delete TOC fusion support.
5372         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5373         longer used with toc fusion being deleted.
5374         (TARGET_TOC_FUSION_FP): Likewise.
5375         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5376         UNSPEC.
5377         (toc fusion spliter): Delete TOC fusion support.
5378         (toc_fusionload_<mode>): Likewise.
5379         (toc_fusionload_di): Likewise.
5380         (fusion_gpr_load_<mode>): Delete generator function, this insn no
5381         longer needs to be named.  Rename predicate to delete TOC fusion.
5382         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5383         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5384         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5385         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5386         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5387
5388 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5389
5390         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5391         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
5392         __int128 in built-in function prototypes.
5393         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5394         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5395
5396 2018-07-27  Martin Sebor  <msebor@redhat.com>
5397
5398         PR tree-optimization/86696
5399         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5400         types, including enums.
5401         (handle_char_store): Be prepared for the above function to fail.
5402
5403 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
5404
5405         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5406         when optimization level is lower than 2 or optimize for size.
5407
5408 2018-07-26  Martin Sebor  <msebor@redhat.com>
5409
5410         PR tree-optimization/86043
5411         PR tree-optimization/86042
5412         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5413         (get_string_cst_length): Rename...
5414         (get_min_string_length): ...to this.  Add argument.
5415         (handle_char_store): Extend to handle multi-character stores by
5416         MEM_REF.
5417         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
5418         * tree.h (initializer_zerop): Add argument.
5419
5420 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
5421
5422         PR middle-end/86660
5423         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5424         declare target to variables if they have always,{to,from,tofrom} map
5425         kinds.
5426
5427 2018-07-26  Martin Liska  <mliska@suse.cz>
5428
5429         PR lto/86548
5430         * lto-wrapper.c: Add linker_output as prefix
5431         for ltrans_output_file.
5432
5433 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
5434
5435         PR rtl-optimization/85805
5436         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5437         value for hard registers if that was written in the same mode.
5438
5439 2018-07-26  Martin Liska  <mliska@suse.cz>
5440
5441         PR gcov-profile/86536
5442         * gcov.c (format_gcov): Use printf format %.*f directly
5443         and do not handle special values.
5444
5445 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5446
5447         * common/config/arc/arc-common.c (arc_option_optimization_table):
5448         Update default optimizations for size.
5449
5450 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5451
5452         * config/arc/arc.md (movsf_insn): Add short instruction selection.
5453         * config/arc/constraints.md (CfZ): New constraint.
5454         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5455         (subsf3_fpu): Likewise.
5456         (cmpsf_fpu): Likewise.
5457         (cmpsf_fpu_uneq): Likewise.
5458
5459 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5460
5461         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5462         accesses as well.
5463         (arc_is_uncached_mem_p): uncached applies to both the variable and
5464         the pointer.
5465
5466 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5467
5468         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5469         register names.
5470
5471 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
5472
5473         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5474         field "m_scopes" from vec to auto_vec.
5475
5476 2018-07-25  Martin Liska  <mliska@suse.cz>
5477
5478         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5479         return type.
5480
5481 2018-07-25  Richard Biener  <rguenther@suse.de>
5482
5483         PR debug/86654
5484         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5485         special wrt context_die late.
5486         (gen_subprogram_die): Re-use DIEs in local scope.
5487
5488 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
5489
5490         PR tree-optimization/86644
5491         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
5492
5493 2018-07-25  Martin Liska  <mliska@suse.cz>
5494
5495         PR middle-end/86645
5496         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5497         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5498
5499 2018-07-25  Martin Liska  <mliska@suse.cz>
5500
5501         PR sanitizer/79635
5502         * params.def: Explain ASan abbreviation and provide
5503         a documentation link.
5504
5505 2018-07-24  Martin Sebor  <msebor@redhat.com>
5506
5507         PR tree-optimization/86622
5508         PR tree-optimization/86532
5509         * builtins.h (string_length): Declare.
5510         * builtins.c (c_strlen): Correct handling of non-constant offsets.
5511         (check_access): Be prepared for non-constant length ranges.
5512         (string_length): Make extern.
5513         * expr.c (string_constant): Only handle the minor non-constant
5514         array index.  Use string_constant to compute the length of
5515         a generic string constant.
5516
5517 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
5518
5519         PR tree-optimization/86618
5520         * tree-vect-stmts.c (vectorizable_call): Don't take the address
5521         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5522
5523 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
5524
5525         PR tree-optimization/86636
5526         * json.cc (json::object::set): Fix comment.  Add assertions.
5527         (json::array::append): Move here from json.h.  Add comment and an
5528         assertion.
5529         (json::string::string): Likewise.
5530         * json.h (json::array::append): Move to json.cc.
5531         (json::string::string): Likewise.
5532         * optinfo-emit-json.cc
5533         (optrecord_json_writer::impl_location_to_json): Assert that we
5534         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5535         wrapper around it.  Expand the location once, rather than three
5536         times.
5537         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5538         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5539         wrappers.
5540         (optrecord_json_writer::optinfo_to_json): Likewise, in four
5541         places.  Fix some overlong lines.
5542
5543 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
5544
5545         * config/aarch64/aarch64-simd.md
5546         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5547         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5548         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5549         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5550         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5551         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5552         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5553         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5554         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5555
5556 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
5557
5558         PR middle-end/86627
5559         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5560         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
5561         and abs_d == d, do the power of two handling if profitable.
5562
5563 2018-07-24  Richard Biener  <rguenther@suse.de>
5564
5565         * match.pd: Add BIT_FIELD_REF canonicalizations.
5566
5567 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5568
5569         PR c/86617
5570         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5571
5572 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5573
5574         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5575         terminated STRING_CST object.
5576
5577 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5578
5579         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5580
5581 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
5582
5583         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5584         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5585         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5586         the elements into a register.
5587         (rs6000_split_v4si_init_di_reg): Delete.
5588         (rs6000_split_v4si_init): Delete.
5589         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5590         (vsx_init_v4si): Rewrite as a define_expand.
5591
5592 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
5593
5594         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5595         zero_extend argument from memory): New.
5596
5597 2018-07-22  Martin Sebor  <msebor@redhat.com>
5598
5599         PR bootstrap/86621
5600         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5601         diagnosing calls with unknown arguments unless -Walloca-larger-than
5602         is restricted to less than PTRDIFF_MAX bytes.
5603
5604 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
5605
5606         * doc/gcov.texi (Invoking Gcov): Editorial changes.
5607
5608 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
5609
5610         * pretty-print.c (text_info::set_location): Remove redundant
5611         "line_table" parameter from call to rich_location::set_range.
5612
5613 2018-07-20  Martin Sebor  <msebor@redhat.com>
5614
5615         PR middle-end/82063
5616         * builtins.c (expand_builtin_alloca): Adjust.
5617         * calls.c (alloc_max_size): Simplify.
5618         * cgraphunit.c (cgraph_node::expand): Adjust.
5619         * common.opt (larger_than_size, warn_frame_larger_than): Remove
5620         variables.
5621         (frame_larger_than_size): Same.
5622         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5623         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5624         Initialize.
5625         * doc/invoke.texi (GCC Command Options): Document option arguments.
5626         Explain byte-size arguments and suffixes.
5627         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5628         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5629         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5630         * doc/options.texi (UInteger): Expand.
5631         (Host_Wide_Int, ByteSize): Document new properties.
5632         * final.c (final_start_function_1): Include sizes in an error message.
5633         * function.c (frame_offset_overflow): Same.
5634         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5635         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5636         Diagnose unbounded alloca calls only for limits of less than
5637         PTRDIFF_MAX.
5638         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
5639         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
5640         for alloca(0).
5641         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
5642         only for limits of less than PTRDIFF_MAX.
5643         * langhooks-def.h (lhd_handle_option): Change function argument
5644         to HOST_WIDE_INT.
5645         * langhooks.c (lhd_handle_option): Same.
5646         * langhooks.h (handle_option): Same.
5647         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5648         ByteSize flags.
5649         (var_type, var_type_struct): Same.
5650         (var_set): Handle ByteSize flag.
5651         * optc-gen.awk: Add comments to output to ease debugging.  Make
5652         use of HOST_WIDE_INT where appropriate.
5653         * opts-gen-save.awk:  Use %lx to format unsigned long.
5654         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5655         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5656         arguments.  Parse bytes-size suffixes.
5657         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5658         (enum_value_to_arg): Same.
5659         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
5660         (handle_option): Adjust.
5661         (generate_option): Change function argument to HOST_WIDE_INT.
5662         (cmdline_handle_error): Adjust.
5663         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5664         (set_option): Change function argument to HOST_WIDE_INT.
5665         (option_enabled): Handle cl_host_wide_int.
5666         (get_option_state): Handle CLVC_SIZE.
5667         (control_warning_option): Same.
5668         * opts.c (common_handle_option): Change function argument to
5669         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
5670         OPT_Wvla_larger_than_.
5671         * opts.h (enum cl_var_type): Add an enumerator.
5672         * stor-layout.c (layout_decl): Print a more meaningful warning.
5673         * toplev.c (output_stack_usage): Adjust.
5674
5675 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
5676
5677         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5678         call to inline_expand_builtin_string_cmp.
5679         (expand_builtin_strcmp): Likewise.
5680         (expand_builtin_strncmp): Likewise.
5681         (inline_string_cmp): Delete the last parameter, change char_type_node
5682         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5683         two operands.
5684         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5685         the inlining expansion on target where the type of the call has same or
5686         narrower precision than unsigned char.
5687
5688 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
5689
5690         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5691         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5692         * common.opt (fsave-optimization-record): New option.
5693         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5694         * doc/invoke.texi (-fsave-optimization-record): New option.
5695         * dumpfile.c: Include "optinfo-emit-json.h".
5696         (struct kv_pair): Move to coretypes.h.
5697         (optgroup_options): Make non-static.
5698         (dump_context::end_scope): Call
5699         optimization_records_maybe_pop_dump_scope.
5700         * dumpfile.h (optgroup_options): New decl.
5701         * json.cc: New file.
5702         * json.h: New file.
5703         * optinfo-emit-json.cc: New file.
5704         * optinfo-emit-json.h: New file.
5705         * optinfo.cc: Include "optinfo-emit-json.h".
5706         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5707         (optinfo_enabled_p): Check optimization_records_enabled_p.
5708         (optinfo_wants_inlining_info_p): Likewise.
5709         * optinfo.h: Update comment.
5710         * profile-count.c (profile_quality_as_string): New function.
5711         * profile-count.h (profile_quality_as_string): New decl.
5712         (profile_count::quality): New accessor.
5713         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5714         and optinfo_emit_json_cc_tests.
5715         * selftest.h (selftest::json_cc_tests): New decl.
5716         (selftest::optinfo_emit_json_cc_tests): New decl.
5717         * toplev.c: Include "optinfo-emit-json.h".
5718         (compile_file): Call optimization_records_finish.
5719         (do_compile): Call optimization_records_start.
5720         * tree-ssa-live.c: Include optinfo.h.
5721         (remove_unused_scope_block_p): Retain inlining information if
5722         optinfo_wants_inlining_info_p returns true.
5723
5724 2018-07-20  Richard Biener  <rguenther@suse.de>
5725
5726         PR debug/86585
5727         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5728         to cover -flto-partition=none.
5729
5730 2018-07-20  Martin Liska  <mliska@suse.cz>
5731
5732         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5733         (get_decl_source_range): Remove unused function.
5734
5735 2018-07-20  Richard Biener  <rguenther@suse.de>
5736
5737         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5738         (struct vn_phi_s): Likewise.
5739         (struct vn_reference_s): Likewise.
5740         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5741         for searching the slot of an entry known to be in the hash itself.
5742         (vn_phi_hasher::equal): Likewise.
5743         (vn_reference_hasher::equal): Likewise.
5744         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5745         globals.
5746         (optimistic_info, current_info): Remove, keeping only valid_info.
5747         (vn_reference_lookup_1): Remove fallback lookup.
5748         (vn_reference_lookup_2): Likewise.
5749         (vn_nary_op_lookup_1): Likewise.
5750         (vn_phi_lookup): Likewise.
5751         (vn_nary_build_or_lookup_1): Make sure to not chain the built
5752         hash element.
5753         (vn_reference_insert): Adjust, chain the inserted hash element
5754         at last_inserted_ref.
5755         (vn_reference_insert_pieces): Likewise.
5756         (visit_reference_op_call): Likewise.
5757         (vn_nary_op_insert_into): Chain the inserted hash element at
5758         last_inserted_nary.
5759         (vn_nary_op_insert_pieces): Adjust.
5760         (vn_nary_op_insert): Likewise.
5761         (vn_nary_op_insert_stmt): Likewise.
5762         (vn_phi_insert): Adjust, chain the inserted hash element at
5763         last_inserted_phi.
5764         (process_scc): Remove clearing and copying the optimistic
5765         table.  Instead remove elements inserted during an optimistic
5766         iteration from the single table we maintain.
5767         (init_scc_vn): Adjust.
5768         (free_scc_vn): Likewise.
5769         (sccvn_dom_walker::record_cond): Likewise.
5770         (sccvn_dom_walker::after_dom_children): Likewise.
5771
5772 2018-07-19  Martin Sebor  <msebor@redhat.com>
5773
5774         PR tree-optimization/84047
5775         PR tree-optimization/83776
5776         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5777         (check_array_bounds): Call it.
5778
5779 2018-07-19  Martin Sebor  <msebor@redhat.com>
5780
5781         * align.h (align_flags): Use member initialization.
5782
5783 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
5784
5785         * Makefile.in (OBJS): Add optinfo.o.
5786         * coretypes.h (class symtab_node): New forward decl.
5787         (struct cgraph_node): New forward decl.
5788         (class varpool_node): New forward decl.
5789         * dump-context.h: New file.
5790         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5791         "tree-pass.h".
5792         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5793         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5794         (set_alt_dump_file): Likewise.
5795         (dump_context::~dump_context): New dtor.
5796         (dump_gimple_stmt): Move implementation to...
5797         (dump_context::dump_gimple_stmt): ...this new member function.
5798         Add the stmt to any pending optinfo, creating one if need be.
5799         (dump_gimple_stmt_loc): Move implementation to...
5800         (dump_context::dump_gimple_stmt_loc): ...this new member function.
5801         Start a new optinfo and add the stmt to it.
5802         (dump_gimple_expr): Move implementation to...
5803         (dump_context::dump_gimple_expr): ...this new member function.
5804         Add the stmt to any pending optinfo, creating one if need be.
5805         (dump_gimple_expr_loc): Move implementation to...
5806         (dump_context::dump_gimple_expr_loc): ...this new member function.
5807         Start a new optinfo and add the stmt to it.
5808         (dump_generic_expr): Move implementation to...
5809         (dump_context::dump_generic_expr): ...this new member function.
5810         Add the tree to any pending optinfo, creating one if need be.
5811         (dump_generic_expr_loc): Move implementation to...
5812         (dump_context::dump_generic_expr_loc): ...this new member
5813         function.  Add the tree to any pending optinfo, creating one if
5814         need be.
5815         (dump_printf): Move implementation to...
5816         (dump_context::dump_printf_va): ...this new member function.  Add
5817         the text to any pending optinfo, creating one if need be.
5818         (dump_printf_loc): Move implementation to...
5819         (dump_context::dump_printf_loc_va): ...this new member function.
5820         Start a new optinfo and add the stmt to it.
5821         (dump_dec): Move implementation to...
5822         (dump_context::dump_dec): ...this new member function.  Add the
5823         value to any pending optinfo, creating one if need be.
5824         (dump_context::dump_symtab_node): New member function.
5825         (dump_context::get_scope_depth): New member function.
5826         (dump_context::begin_scope): New member function.
5827         (dump_context::end_scope): New member function.
5828         (dump_context::ensure_pending_optinfo): New member function.
5829         (dump_context::begin_next_optinfo): New member function.
5830         (dump_context::end_any_optinfo): New member function.
5831         (dump_context::s_current): New global.
5832         (dump_context::s_default): New global.
5833         (dump_scope_depth): Delete global.
5834         (dumpfile_ensure_any_optinfo_are_flushed): New function.
5835         (dump_symtab_node): New function.
5836         (get_dump_scope_depth): Reimplement in terms of dump_context.
5837         (dump_begin_scope): Likewise.
5838         (dump_end_scope): Likewise.
5839         (selftest::temp_dump_context::temp_dump_context): New ctor.
5840         (selftest::temp_dump_context::~temp_dump_context): New dtor.
5841         (selftest::verify_item): New function.
5842         (ASSERT_IS_TEXT): New macro.
5843         (ASSERT_IS_TREE): New macro.
5844         (ASSERT_IS_GIMPLE): New macro.
5845         (selftest::test_capture_of_dump_calls): New test.
5846         (selftest::dumpfile_c_tests): Call it.
5847         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5848         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5849         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5850         descriptive comment.
5851         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5852         (dump_node, dump_bb): Move these unrelated decls.
5853         (class dump_manager): Add leading comment.
5854         * optinfo.cc: New file.
5855         * optinfo.h: New file.
5856
5857 2018-07-19  Michael Collison  <michael.collison@arm.com>
5858             Richard Henderson <rth@redhat.com>
5859
5860         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5861         (subti): Handle op1 zero.
5862         (subvti4, usub4ti4): New.
5863         (*sub<GPI>3_compare1_imm): New.
5864         (sub<GPI>3_carryinCV): New.
5865         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5866         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5867
5868 2018-07-19  Michael Collison  <michael.collison@arm.com>
5869             Richard Henderson <rth@redhat.com>
5870
5871         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5872         (addti3): Create simpler code if low part is already known to be 0.
5873         (addvti4, uaddvti4): New.
5874         (*add<GPI>3_compareC_cconly_imm): New.
5875         (*add<GPI>3_compareC_cconly): New.
5876         (*add<GPI>3_compareC_imm): New.
5877         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5878         handle constants within this pattern..
5879         (*add<GPI>3_compareV_cconly_imm): New.
5880         (*add<GPI>3_compareV_cconly): New.
5881         (*add<GPI>3_compareV_imm): New.
5882         (add<GPI>3_compareV): New.
5883         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5884         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5885         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5886         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5887         with 'comparison' operator.
5888         (*add<GPI>3_compareV_cconly_imm): Ditto.
5889         (*add<GPI>3_compareV_cconly): Ditto.
5890         (*add<GPI>3_compareV_imm): Ditto.
5891         (add<GPI>3_compareV): Ditto.
5892         (add<mode>3_carryinC): Ditto.
5893         (*add<mode>3_carryinC_zero): Ditto.
5894         (*add<mode>3_carryinC): Ditto.
5895         (add<mode>3_carryinV): Ditto.
5896         (*add<mode>3_carryinV_zero): Ditto.
5897         (*add<mode>3_carryinV): Ditto.
5898
5899 2018-07-19  Michael Collison  <michael.collison@arm.com>
5900             Richard Henderson <rth@redhat.com>
5901
5902         * config/aarch64/aarch64-modes.def (CC_V): New.
5903         * config/aarch64/aarch64-protos.h
5904         (aarch64_addti_scratch_regs): Declare
5905         (aarch64_subvti_scratch_regs): Declare.
5906         (aarch64_expand_subvti): Declare.
5907         (aarch64_gen_unlikely_cbranch): Declare
5908         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5909         for signed overflow using CC_Vmode.
5910         (aarch64_get_condition_code_1): Handle CC_Vmode.
5911         (aarch64_gen_unlikely_cbranch): New function.
5912         (aarch64_addti_scratch_regs): New function.
5913         (aarch64_subvti_scratch_regs): New function.
5914         (aarch64_expand_subvti): New function.
5915
5916 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5917
5918         * config/aarch64/aarch64-option-extensions.def: New entry for profile
5919         extension.
5920         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5921         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5922         extension.
5923
5924 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5925
5926         PR target/83009
5927         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5928         address check not strict.
5929
5930 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5931
5932         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5933         Umq with Umn.
5934         (store_pair_lanes<mode>): Likewise.
5935         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5936         enum value 'ADDR_QUERY_LDP_STP_N'.
5937         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5938         (aarch64_print_address_internal): Add declaration.
5939         (aarch64_print_ldpstp_address): Remove.
5940         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5941         (aarch64_print_operand): Change printing of 'y'.
5942         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5943         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5944         'true' rather than '1'.
5945         * config/aarch64/constraints.md (Uml): Likewise.
5946         (Uml): Rename to Umn.
5947         (Umq): Remove.
5948
5949 2018-07-19  Richard Biener  <rguenther@suse.de>
5950
5951         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5952         a trailing array.
5953         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5954         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5955         (vn_reference_hasher): Likewise.
5956         (struct vn_tables_s): Remove obstack and alloc-pool members.
5957         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5958         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5959         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5960         (vn_reference_insert_pieces): Likewise.
5961         (alloc_vn_nary_op_noinit): Adjust.
5962         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5963         (vn_phi_eq): Adjust.
5964         (shared_lookup_phiargs): Remove.
5965         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5966         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5967         (visit_reference_op_call): Likewise.
5968         (copy_nary, copy_phi, copy_reference): Remove.
5969         (process_scc): Rewind the obstack when iterating.  Do not
5970         copy the elements to valid_info but just move them from one
5971         hashtable to the other.
5972         (allocate_vn_table): Adjust.
5973         (free_vn_table): Likewise.
5974         (init_scc_vn): Likewise.
5975         (free_scc_vn): Likewise.
5976
5977 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
5978
5979         PR target/86560
5980         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5981         indirect_return as function type attribute.
5982         (ix86_attribute_table): Change indirect_return to function
5983         type attribute.
5984         * doc/extend.texi: Update indirect_return attribute.
5985
5986 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
5987
5988         * wide-int.h (widest2_int): New.
5989         * gimple-fold.c (arith_overflowed_p): Use it.
5990         * tree.h (widest2_int_cst): New.
5991         * tree-vrp.c (wide_int_binop_overflow): Rename from
5992         vrp_int_const_binop.
5993         Rewrite to work on trees.
5994         (extract_range_from_multiplicative_op_1): Abstract code to...
5995         (wide_int_range_min_max): ...here.
5996         (wide_int_range_cross_product): ...and here.
5997         (extract_range_from_binary_expr_1): Abstract overflow code to...
5998         (wide_int_range_mult_wrapping): ...here.
5999         * tree-vrp.h (wide_int_range_cross_product): New.
6000         (wide_int_range_mult_wrapping): New.
6001
6002 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
6003             Julia Koval  <julia.koval@intel.com>
6004
6005         * config/i386/x86-tune-costs.h (skylake_memcpy,
6006         skylake_memset): Replace rep_prefix with unrolling for size 512.
6007
6008 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
6009
6010         PR middle-end/86544
6011         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
6012         comparision with EQ_EXPR in last stmt.
6013
6014 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6015
6016         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
6017         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
6018         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
6019         previously known as "PowerPC AltiVec Built-in Functions".  Move
6020         some material to new subsubsections "PowerPC AltiVec Built-in
6021         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
6022         ISA 2.07".
6023         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
6024         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
6025         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6026         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6027
6028 2018-07-18  Richard Biener  <rguenther@suse.de>
6029
6030         PR tree-optimization/86557
6031         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
6032         EXACT_DIV_EXPR.
6033
6034 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
6035
6036         * config/s390/s390.c (s390_function_profiler): Generate CFI.
6037
6038 2018-07-17  Jeff Law  <law@redhat.com>
6039
6040         * config/arm/arm.c (get_label_padding): Update for recent
6041         changes to label_to_alignment.
6042
6043         PR tree-optimization/86010
6044         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6045
6046         * config/mips/mips.c (vr4130_align_insns): Update for recent
6047         changes to label_to_alignment.
6048
6049         * config/frv/frv.c (frv_label_align): Update for recent changes
6050         to label_to_alignment.
6051
6052         * config/nios2/nios2.c (nios2_label_align): Update for recent
6053         changes which dropped ALIGN_LABELS_LOG.
6054
6055 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
6056
6057         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6058         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6059
6060 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
6061
6062         * config/arc/arc.c (arc_label_align): Use align_labels instead of
6063         deprecated align_labels_log.
6064
6065 2018-07-17  Richard Biener  <rguenther@suse.de>
6066
6067         PR lto/86456
6068         * dwarf2out.c (init_sections_and_labels): Always generate
6069         a debug_line_str_section for early LTO debug.
6070         (dwarf2out_finish): Reset debug_line_str_hash output early.
6071         Bump counter for extra dwarf5 .debug_loc labels to not conflict
6072         with fat LTO part.
6073         (dwarf2out_early_finish): Output debug_line_str.
6074
6075 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
6076
6077         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6078         index register on z196 or later.
6079
6080 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
6081
6082         * config/s390/s390.c (s390_default_align): Set default function
6083         alignment to 16.
6084         (s390_override_options_after_change): Call s390_default align.
6085         (s390_option_override_internal): Call s390_default_align.
6086         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6087
6088 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
6089
6090         PR middle-end/86542
6091         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6092         to _looptemp_ clauses, other than the first two.
6093
6094 2018-07-17  Martin Liska  <mliska@suse.cz>
6095
6096         * opts.c: Do not enable OPT_falign_* for -Os.
6097
6098 2018-07-17  Martin Liska  <mliska@suse.cz>
6099
6100         * align.h (MAX_CODE_ALIGN): New.
6101         (MAX_CODE_ALIGN_VALUE): New.
6102         * common/config/i386/i386-common.c (ix86_handle_option):
6103         (MAX_CODE_ALIGN): Moved to align.h.
6104         * final.c (MAX_CODE_ALIGN): Likewise.
6105         * opts.c (parse_and_check_align_values):
6106         (MAX_CODE_ALIGN): Likewise.
6107         (MAX_CODE_ALIGN_VALUE): Likewise.
6108
6109 2018-07-17  Martin Liska  <mliska@suse.cz>
6110
6111         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6112         in order to fulfil coding style.
6113         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6114         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6115         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6116         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6117         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6118         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6119         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6120         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6121         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6122
6123 2018-07-17  Martin Liska  <mliska@suse.cz>
6124
6125         * align.h: New file.
6126         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6127         directly.
6128         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6129         align_flags of label_to_alignment.
6130         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6131         align_flags class.
6132         * config/m68k/m68k.c: Do not use removed align_labels_value and
6133         align_loops_value.
6134         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6135         (LOOP_ALIGN): Likewise.
6136         (LABEL_ALIGN): Likewise.
6137         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6138         Remove not used macro.
6139         (rs6000_loop_align): Change return type to align_flags.
6140         (rs6000_loop_align_max_skip): Remove.
6141         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6142         Change return type to align_flags.
6143         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6144         Remove not used macro.
6145         (rs6000_loop_align):  Change return type to align_flags.
6146         (rs6000_loop_align_max_skip): Remove.
6147         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6148         * config/rx/rx-protos.h (rx_align_for_label): Make it
6149         static function.
6150         * config/rx/rx.c (rx_align_for_label): Change return type
6151         to align_flags.
6152         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6153         macro definitions.
6154         into align_flags class.
6155         (LABEL_ALIGN): Likewise.
6156         (LOOP_ALIGN): Likewise.
6157         * config/s390/s390.c (s390_label_align): Use align_flags
6158         class member.
6159         (s390_asm_output_function_label): Likewise.
6160         * config/sh/sh.c (sh_override_options_after_change):
6161         Use align_flags class directly without macros.
6162         (find_barrier): Likewise.
6163         (barrier_align): Likewise.
6164         (sh_loop_align): Likewise.
6165         * config/spu/spu.c (spu_option_override):
6166         Use align_flags_tuple::get_value instead of removed macros.
6167         (spu_sched_init): Likewise.
6168         * config/spu/spu.h (GTY): Likewise.
6169         * config/visium/visium.c (visium_option_override):
6170         Set "8" as default secondary alignment.
6171         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6172         in order to guarantee secondary alignment of 8.
6173         * coretypes.h: Include align.h header file.
6174         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6175         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6176         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6177         * doc/tm.texi.in: Likewise.
6178         * final.c (struct label_alignment): Remove not used structure.
6179         (LABEL_ALIGN): Change type to align_flags.
6180         (LOOP_ALIGN): Likewise.
6181         (JUMP_ALIGN): Likewise.
6182         (default_loop_align_max_skip): Remove.
6183         (default_label_align_max_skip): Likewise.
6184         (default_jump_align_max_skip): Likewise.
6185         (default_label_align_after_barrier_max_skip):
6186         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6187         (LABEL_TO_MAX_SKIP): Remove.
6188         (label_to_alignment): Return align_flags type instead of integer.
6189         (label_to_max_skip): Remove.
6190         (align_fuzz): Use align_flags type.
6191         (compute_alignments): Use align_flags type and use align_flags::max
6192         to combine multiple alignments.
6193         (grow_label_align): Grow vec instead of C array.
6194         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6195         (shorten_branches):  Use align_flags type and use align_flags::max
6196         to combine multiple alignments.
6197         (final_scan_insn_1): Remove usage of secondary alignment that comes
6198         from label alignment, but instead use proper secondary alignment
6199         which is computed in grow_label_align.
6200         * flags.h (struct align_flags_tuple): Move to align.h.
6201         (struct align_flags): Likewise.
6202         (state_align_loops): Rename to align_loops.
6203         (state_align_jumps): Rename to align_jumps.
6204         (state_align_labels): Rename to align_labels.
6205         (state_align_functions): Rename to align_functions.
6206         (align_loops_log): Remove.
6207         (align_jumps_log): Remove.
6208         (align_labels_log): Remove.
6209         (align_functions_log): Remove.
6210         (align_loops_max_skip): Remove.
6211         (align_jumps_max_skip): Remove.
6212         (align_labels_max_skip): Remove.
6213         (align_functions_max_skip): Remove.
6214         (align_loops_value): Remove.
6215         (align_jumps_value): Remove.
6216         (align_labels_value): Remove.
6217         (align_functions_value): Remove.
6218         * output.h (label_to_alignment): Change return type to align_flags.
6219         (label_to_max_skip): Remove.
6220         * target.def: Remove loop_align_max_skip, label_align_max_skip,
6221         jump_align_max_skip macros.
6222         * targhooks.h (default_loop_align_max_skip): Remove.
6223         (default_label_align_max_skip): Likewise.
6224         (default_jump_align_max_skip): Likewise.
6225         (default_label_align_after_barrier_max_skip): Remove.
6226         * toplev.c (read_log_maxskip): Use ::normalize function.
6227         (parse_N_M): Remove not used argument and also call ::normalize.
6228         (parse_alignment_opts): Do not pass unused arguments.
6229         * varasm.c (assemble_start_function): Use directly align_functions
6230         instead of removed macros.
6231         * system.h: Do not poison removed macros.
6232
6233 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
6234
6235         PR middle-end/86539
6236         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
6237         and cond temporaries don't have reference type if iterator has
6238         pointer type.  For init use &for_pre_body instead of pre_p if
6239         for_pre_body is non-empty.
6240
6241 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
6242
6243         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
6244         double-double modes to SFmode directly directly.
6245         (trunc<mode>sf2_fprs): Delete.
6246
6247 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
6248
6249         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
6250         for conversions between IFmode and the decimal floating point modes.
6251         (init_float128_ieee): Use the correct names for conversions between
6252         KFmode and the decimal floating point modes.
6253
6254 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
6255
6256         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
6257         for the conversions between TDmode and IFmode.
6258         (init_float128_ieee): Use more correct names for the conversions
6259         between TDmode and KFmode.
6260
6261 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
6262
6263         PR tree-optimization/86526
6264         * builtins.c (expand_builtin_memcmp): Formatting fixes.
6265         (inline_expand_builtin_string_cmp): Likewise.
6266         (inline_string_cmp): Likewise.  Use c_readstr instead of
6267         builtin_memcpy_read_str.  Add unit_mode temporary.
6268
6269 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6270
6271         PR middle-end/86528
6272         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
6273         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
6274
6275 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6276
6277         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
6278         Alphabetize prototypes of built-in functions, separating out
6279         built-in functions that are listed in this section but should be
6280         described elsewhere.
6281
6282 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
6283
6284         PR target/86511
6285         * expmed.c (emit_store_flag): Do not emit setcc followed by a
6286         conditional move when trapping comparison was split to a
6287         non-trapping one (and vice versa).
6288
6289 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
6290
6291         * config/s390/s390.c (s390_function_profiler): Generate nops
6292         instead of profiler call sequences.
6293         * config/s390/s390.opt: Add the new option.
6294
6295 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
6296
6297         * config/s390/s390.c (s390_function_profiler): Generate
6298         __mcount_loc section.
6299         * config/s390/s390.opt: Add the new option.
6300
6301 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
6302
6303         * common.opt: Add the new warning.
6304         * config/s390/s390.c (s390_function_profiler): Emit "brasl
6305         %r0,__fentry__" when -mfentry is specified.
6306         (s390_option_override_internal): Disallow -mfentry for 31-bit
6307         CPUs.
6308         * config/s390/s390.opt: Add the new option.
6309
6310 2018-07-16  Richard Biener  <rguenther@suse.de>
6311
6312         PR lto/86523
6313         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6314         for function-local FUNCTION_DECL and RESULT_DECL immediately.
6315
6316 2018-07-16  Martin Liska  <mliska@suse.cz>
6317
6318         PR ipa/86529
6319         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6320         to ::get_create.
6321
6322 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
6323
6324         * config/arc/arcHS.md: Update ARCHS scheduling rules.
6325
6326 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
6327
6328         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6329         for ARCHS4x.
6330         * config/arc/arc-cpus.def (hs4x): New cpu.
6331         (hs4xd): Likewise.
6332         * config/arc/arc-tables.opt: Regenerate.
6333         * config/arc/arc.c (arc_sched_issue_rate): New function.
6334         (TARGET_SCHED_ISSUE_RATE): Define.
6335         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6336         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6337         fpu_cvt.
6338         (attr tune): Add ARCHS4x tune values.
6339         (attr tune_dspmpy): Define.
6340         (*tst): Correct instruction type.
6341         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6342         * config/arc/arcHS4x.md: New file.
6343         * config/arc/fpu.md: Update instruction type attributes.
6344         * config/arc/t-multilib: Regenerate.
6345
6346 2018-07-16  Tom de Vries  <tdevries@suse.de>
6347
6348         PR debug/86455
6349         * var-tracking.c (vt_initialize): Fix pre_dec handling.
6350
6351 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6352
6353         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6354         early clobber.
6355
6356 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
6357
6358         PR tree-optimization/86514
6359         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6360         conversion to a boolean type from a type with greater precision.
6361
6362 2018-07-16  Tom de Vries  <tdevries@suse.de>
6363
6364         * var-tracking.c (vt_initialize): Print adjusted insn slim if
6365         dump_flags request TDF_SLIM.
6366
6367 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
6368
6369         * fold-const.c (int_const_binop_1): Abstract...
6370         (wide_int_binop): ...wide int code here.
6371         (poly_int_binop): ...poly int code here.
6372         Abstract the rest of int_const_binop_1 into int_const_binop.
6373         * fold-const.h (wide_int_binop): New.
6374         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6375         Remove useless PLUS/MINUS_EXPR case.
6376         (zero_nonzero_bits_from_vr): Move wide int code...
6377         (zero_nonzero_bits_from_bounds): ...here.
6378         (extract_range_from_binary_expr_1): Move mask optimization code...
6379         (range_easy_mask_min_max): ...here.
6380         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6381         (range_easy_mask_min_max): New.
6382
6383 2018-07-15  Jeff Law  <law@redhat.com>
6384
6385         PR target/85993
6386         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6387         block.
6388
6389 2018-07-14  Jim Wilson  <jimw@sifive.com>
6390
6391         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6392
6393 2018-07-14  Paul Koning  <ni1d@arrl.net>
6394
6395         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6396
6397 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
6398
6399         * lto-streamer-out.c (copy_function_or_variable): Dump info about
6400         copying section.
6401
6402 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
6403             Steve Munroe  <munroesj52@gmail.com>
6404
6405         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6406         (_mm_andnot_si128): Likewise.
6407         (_mm_or_si128): Likewise.
6408         (_mm_xor_si128): Likewise.
6409
6410 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
6411
6412         PR middle-end/78809
6413         * builtins.c (expand_builtin_memcmp): Inline the calls first
6414         when result_eq is false.
6415         (expand_builtin_strcmp): Inline the calls first.
6416         (expand_builtin_strncmp): Likewise.
6417         (inline_string_cmp): New routine. Expand a string compare
6418         call by using a sequence of char comparison.
6419         (inline_expand_builtin_string_cmp): New routine. Inline expansion
6420         a call to str(n)cmp/memcmp.
6421         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6422         New option.
6423         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6424
6425 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
6426
6427         * config/arm/driver-arm.c: Include arm-native.h.
6428         (host_detect_local_cpu): Use auto-generated data tables.
6429         (vendors, arm_cpu_table): Delete.  Move part information to ...
6430         * config/arm/arm-cpus.in: ... here.
6431         * config/arm/parsecpu.awk (gen_native): New function.
6432         (vendor, part): New CPU fields.
6433         (END): Add support for building the native CPU detection tables.
6434         * config/arm/t-arm (arm-native.h): Add build rule.
6435         (driver-arm.o): Add dependency on arm-native.h.
6436
6437 2018-07-13  Richard Biener  <rguenther@suse.de>
6438
6439         PR middle-end/85974
6440         * match.pd (addr1 - addr2): Allow either of the operand to
6441         have a conversion.
6442
6443 2018-07-13  Tom de Vries  <tdevries@suse.de>
6444
6445         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6446         in remap_ssa_name.
6447
6448 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
6449
6450         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6451         arrays instead of numbered variables.
6452
6453 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
6454
6455         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6456         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6457
6458 2018-07-13  Richard Biener  <rguenther@suse.de>
6459
6460         PR debug/86452
6461         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6462         instead of get_context_die.
6463
6464 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
6465             Richard Biener  <rguenther@suse.de>
6466
6467         PR middle-end/86489
6468         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6469         that the loop latch destination where phi is defined.
6470
6471 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
6472
6473         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6474         (riscv_expand_epilogue): Add assertion to check interrupt mode.
6475         (riscv_set_current_function): Extract getting interrupt type to new
6476         function.
6477         (riscv_get_interrupt_type): New function.
6478         (riscv_merge_decl_attributes): New function, checking interrupt type is
6479         same.
6480         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6481
6482 2018-07-12  Paul Koning  <ni1d@arrl.net>
6483
6484         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6485         directive.
6486
6487 2018-07-12  Paul Koning  <ni1d@arrl.net>
6488
6489         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6490         zero reference, add doloop_end instead.
6491         * doc/md.texi (decrement_and_branch_until_zero): Remove.
6492         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
6493         detail for doloop_end.
6494
6495 2018-07-12  Martin Sebor  <msebor@redhat.com>
6496
6497         PR c/86453
6498         * attribs.c (decl_attributes): Reject conflicting attributes before
6499         calling attribute handlers.
6500
6501 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
6502
6503         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6504          parameter.
6505         (gcc::dump_manager::get_dump_file_name): likewise.
6506         (dump_begin): Likewise.
6507         * dumpfile.h (dump_begin): Update prototype.
6508         (gcc::dump_manager::get_dump_file_name,
6509         gcc::dump_manager::get_dump_file_name): Update prototype.
6510
6511 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6512
6513         * internal-fn.h (vectorizable_internal_fn_p): New function.
6514         * tree-vect-slp.c (compatible_calls_p): Likewise.
6515         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
6516         to internal functions.
6517         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6518
6519 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6520
6521         * fold-const.h (inverse_conditions_p): Declare.
6522         * fold-const.c (inverse_conditions_p): New function.
6523         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
6524         that test the inverse condition of a conditional internal function.
6525         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6526         * internal-fn.c (internal_fn_mask_index): Handle conditional
6527         internal functions.
6528         (vectorized_internal_fn_supported_p): New function.
6529         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6530         (any_pred_load_store): Replace with...
6531         (need_to_predicate): ...this new variable.
6532         (redundant_ssa_names): New variable.
6533         (ifcvt_can_use_mask_load_store): Move initial checks to...
6534         (ifcvt_can_predicate): ...this new function.  Handle tree codes
6535         for which a conditional internal function exists.
6536         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6537         instead of ifcvt_can_use_mask_load_store.  Update after variable
6538         name change.
6539         (predicate_load_or_store): New function, split out from
6540         predicate_mem_writes.
6541         (check_redundant_cond_expr): New function.
6542         (value_available_p): Likewise.
6543         (predicate_rhs_code): Likewise.
6544         (predicate_mem_writes): Rename to...
6545         (predicate_statements): ...this.  Use predicate_load_or_store
6546         and predicate_rhs_code.
6547         (combine_blocks, tree_if_conversion): Update after above name changes.
6548         (ifcvt_local_dce): Handle redundant_ssa_names.
6549         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6550         general conditional functions.
6551         * tree-vect-stmts.c (vectorizable_call): Likewise.
6552
6553 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6554             Alan Hayward  <alan.hayward@arm.com>
6555             David Sherwood  <david.sherwood@arm.com>
6556
6557         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6558         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6559         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6560         plus and minus and convert them into IFN_COND_FMA-based sequences.
6561         (convert_mult_to_fma): Handle conditional plus and minus.
6562
6563 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6564
6565         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6566         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6567         (cond_fnms_optab): New optabs.
6568         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6569         internal functions.
6570         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6571         * internal-fn.h (get_conditional_internal_fn): Declare.
6572         (get_unconditional_internal_fn): Likewise.
6573         * internal-fn.c (cond_ternary_direct): New macro.
6574         (expand_cond_ternary_optab_fn): Likewise.
6575         (direct_cond_ternary_optab_supported_p): Likewise.
6576         (FOR_EACH_COND_FN_PAIR): Likewise.
6577         (get_conditional_internal_fn): New function.
6578         (get_unconditional_internal_fn): Likewise.
6579         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6580         (gimple_match_op::gimple_match_op): Add a new overload for 5
6581         operands.
6582         (gimple_match_op::set_op): Likewise.
6583         (gimple_resimplify5): Declare.
6584         * genmatch.c (decision_tree::gen): Generate simplifications for
6585         5 operands.
6586         * gimple-match-head.c (gimple_simplify): Define an overload for
6587         5 operands.  Handle calls with 5 arguments in the top-level overload.
6588         (convert_conditional_op): Handle conversions from unconditional
6589         internal functions to conditional ones.
6590         (gimple_resimplify5): New function.
6591         (build_call_internal): Pass a fifth operand.
6592         (maybe_push_res_to_seq): Likewise.
6593         (try_conditional_simplification): Try converting conditional
6594         internal functions to unconditional internal functions.
6595         Handle 3-operand unconditional forms.
6596         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6597         Define ternary equivalents of the current rules for binary conditional
6598         internal functions.
6599         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6600         ternary operations.
6601         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6602         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6603         (optab): Handle them.
6604         (SVE_COND_FP_TERNARY): New int iterator.
6605         (sve_fmla_op, sve_fmad_op): New int attributes.
6606         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6607         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6608         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6609
6610 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6611
6612         * target.def (preferred_else_value): New target hook.
6613         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6614         * doc/tm.texi: Regenerate.
6615         * targhooks.h (default_preferred_else_value): Declare.
6616         * targhooks.c (default_preferred_else_value): New function.
6617         * internal-fn.h (conditional_internal_fn_code): Declare.
6618         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6619         (get_conditional_internal_fn): Use it.
6620         (conditional_internal_fn_code): New function.
6621         * gimple-match.h (gimple_match_cond): New struct.
6622         (gimple_match_op): Add a cond member function.
6623         (gimple_match_op::gimple_match_op): Update all forms to take a
6624         gimple_match_cond.
6625         * genmatch.c (expr::gen_transform): Use the same condition as res_op
6626         for the suboperation, but don't specify a particular else_value.
6627         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6628         (visit_nary_op, visit_reference_op_load): Pass
6629         gimple_match_cond::UNCOND to the gimple_match_op constructor.
6630         * gimple-match-head.c: Include tree-eh.h
6631         (convert_conditional_op): New function.
6632         (maybe_resimplify_conditional_op): Likewise.
6633         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6634         (gimple_resimplify2): Likewise.
6635         (gimple_resimplify3): Likewise.
6636         (gimple_resimplify4): Likewise.
6637         (maybe_push_res_to_seq): Return null for conditional operations.
6638         (try_conditional_simplification): New function.
6639         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
6640         constructor.
6641         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6642         IFN_COND_* call.
6643         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6644         function.
6645         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6646
6647 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
6648
6649         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6650         DECL_FCONTEXT
6651         (hash_tree): Do not hash DECL_FCONTEXT
6652         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6653         Do not stream DECL_FCONTEXT.
6654         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6655         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6656
6657 2018-07-12  Richard Biener  <rguenther@suse.de>
6658
6659         PR debug/86462
6660         * dwarf2out.c (gen_block_die): Only output blocks when they have
6661         at least one !DECL_IGNORED_P variable.
6662
6663 2018-07-12  Richard Biener  <rguenther@suse.de>
6664
6665         PR target/84829
6666         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6667         Remove -mieee-fp handling.
6668
6669 2018-07-12  Richard Biener  <rguenther@suse.de>
6670
6671         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6672         left-over from last patch.
6673
6674 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
6675
6676         PR tree-optimization/86492
6677         * gimple-ssa-store-merging.c
6678         (imm_store_chain_info::coalesce_immediate_stores): Call
6679         check_no_overlap even for the merge_overlapping case.  Formatting fix.
6680
6681 2018-07-12  Richard Biener  <rguenther@suse.de>
6682
6683         PR middle-end/86479
6684         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6685         move possibly trapping operations into the conditional.
6686
6687 2018-07-12  Richard Biener  <rguenther@suse.de>
6688
6689         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6690         (vn_lookup_simplify_result): Remove recursion limit applied
6691         here.
6692         (vn_nary_build_or_lookup_1): Adjust.
6693         (try_to_simplify): Likewise.
6694         * gimple-match-head.c (gimple_resimplify1): Instead apply one
6695         here.
6696         (gimple_resimplify2): Likewise.
6697         (gimple_resimplify3): Likewise.
6698         (gimple_resimplify4): Likewise.
6699
6700 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
6701
6702         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6703         Use __mmask64 type instead of __mmask8 for __M argument.
6704         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6705         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6706         __U argument.
6707         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6708         __mmask16 for __M argument.
6709         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6710         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6711         to __mmask16 instead of __mmask8.
6712         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6713         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6714         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6715         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6716         instead of __mmask16 for __U argument.
6717         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6718         __mmask16 instead of __mmask8 for __U argument.
6719         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6720         __U argument.
6721         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6722         __mmask16.
6723         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6724         argument.
6725         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6726         __U argument.
6727         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6728         __mmask16.
6729         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6730         of __mmask16.
6731         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6732         __U argument.
6733         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6734         __U argument.
6735         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6736         __U argument.
6737         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6738         __U argument.
6739         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6740         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6741         return type as well as __M argument type and all casts from __mmask8
6742         to __mmask32.
6743         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6744         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6745         return type as well as __M argument type and all casts from __mmask8
6746         to __mmask16.
6747         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6748         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6749         return type as well as __M argument type and all casts from __mmask8
6750         to __mmask32.
6751         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6752         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6753         return type as well as __M argument type and all casts from __mmask8
6754         to __mmask16.
6755         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6756         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6757         __mmask16.
6758
6759 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
6760
6761         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6762         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6763         for __U argument.
6764
6765 2018-07-11  Paul Koning  <ni1d@arrl.net>
6766
6767         * doc/md.texi (define_subst): Document how multiple occurrences of
6768         the same argument in the replacement pattern are handled.
6769
6770 2018-07-11  Paul Koning  <ni1d@arrl.net>
6771
6772         * doc/extend.texi (Common Variable Attributes): Move "mode" into
6773         alphabetical order.
6774         (Common Type Attributes): Add "mode" attribute.
6775
6776 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
6777
6778         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6779         stream DECL_ORIGINAL_TYPE.
6780         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6781         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6782         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6783         Do not walk original type.
6784         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6785         external decls.
6786         (write_ts_decl_non_common_tree_pointers): Do not stream
6787         DECL_ORIGINAL_TYPE
6788         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6789         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6790
6791 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
6792
6793         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6794         thread twice from the same starting edge.
6795
6796 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
6797
6798         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6799         * gimple.c (gimple_call_nonnull_result_p): ...here...
6800         (gimple_call_nonnull_arg): ...and here.
6801         * gimple.h (gimple_call_nonnull_result_p): New.
6802         (gimple_call_nonnull_arg): New.
6803
6804 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
6805
6806         * config/arm/arm-cpus.in: Move information from fpu field of each
6807         cpu definition to the isa field.
6808         * config/arm/parsecpu.awk (fpu): Delete match rule.
6809         (gen_comm_data): Don't add bits from the CPU's FPU entry.
6810
6811 2018-07-11  Richard Biener  <rguenther@suse.de>
6812
6813         PR debug/86457
6814         * dwarf2out.c (init_sections_and_labels): Use
6815         output_asm_line_debug_info consistently.
6816         (dwarf2out_early_finish): Likewise.
6817         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6818         type units.
6819
6820 2018-07-11  Richard Biener  <rguenther@suse.de>
6821
6822         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6823         Rework father_bb setting in a way to avoid propagating constants
6824         multiple times on a loop body.
6825
6826 2018-07-10  Mark Wielaard  <mark@klomp.org>
6827
6828         PR debug/86459
6829         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6830
6831 2018-07-10  Richard Biener  <rguenther@suse.de>
6832
6833         * hash-map.h (hash_map::iterator::operator*): Return
6834         references to key and value.
6835
6836 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
6837
6838         PR c++/86443
6839         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6840         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6841         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6842         (gimplify_omp_for): For composite loops, move outer
6843         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6844         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
6845         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
6846         TREE_LIST for both the original class iterator and the "last" helper
6847         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
6848         loop, remember has_decl_expr from outer composite loops for the
6849         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6850
6851 2018-07-09  Martin Sebor  <msebor@redhat.com>
6852
6853         PR middle-end/77357
6854         PR middle-end/86428
6855         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6856         accessing implicitly initialized array elements.
6857         * expr.c (string_constant): Handle string initializers of
6858         character arrays within aggregates.
6859         * gimple-fold.c (fold_array_ctor_reference): Add argument.
6860         Store element offset.  As a special case, handle zero size.
6861         (fold_nonarray_ctor_reference): Same.
6862         (fold_ctor_reference): Add argument.  Store subobject offset.
6863         * gimple-fold.h (fold_ctor_reference): Add argument.
6864
6865 2018-07-09  Paul Koning  <ni1d@arrl.net>
6866
6867         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6868         (pdp11_insn_cost): New function.
6869         (pdp11_md_asm_adjust): New function.
6870         (TARGET_INVALID_WITHIN_DOLOOP): Define.
6871         (pdp11_rtx_costs): Update to match machine better.
6872         (output_addr_const_pdp11): Correct format mismatch warnings.
6873         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6874         * config/pdp11/pdp11.md: General change to add base_cost and/or
6875         length attributes for use by new pdp11_insn_cost function.
6876         (MIN_BRANCH): Correct definition.
6877         (MIN_SOB): Ditto.
6878         (doloop_end): Use standard pattern name for looping pattern.
6879         (doloop_end_nocc): New.
6880         (movsf): Add another constraint alternative.
6881         (zero_extendqihi2): Add constraint alternatives for not in place
6882         extend.
6883         (zero_extendhisi2): Remove.
6884         (shift patterns): Add CC handling variants.
6885         (bswaphi2): New.
6886         (bswapsi2): New.
6887         (rothi3): New.
6888         (define_peephole2): New peephole to recognize mov that sets CC for
6889         subsequent test.
6890
6891 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6892
6893         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6894         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6895         wi::add.
6896
6897 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
6898
6899         PR c/86420
6900         * real.c (real_nextafter): Return true if result is denormal.
6901
6902 2018-07-09  Martin Liska  <mliska@suse.cz>
6903
6904         * common.opt: Add back wrongly removed attribute.
6905
6906 2018-07-09  Richard Biener  <rguenther@suse.de>
6907
6908         PR debug/86413
6909         * dwarf2out.c (gen_block_die): For an early generated DIE
6910         always output high/low PC attributes.
6911
6912 2018-07-09  Tom de Vries  <tdevries@suse.de>
6913
6914         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6915         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6916         onto VAR_DECL with abstract origin.
6917
6918 2018-07-07  Jim Wilson  <jimw@sifive.com>
6919
6920         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6921
6922 2018-07-07  Tom de Vries  <tdevries@suse.de>
6923
6924         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6925
6926 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
6927
6928         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6929         overflow_type.
6930         (combine_bound): Use wide-int overflow calculation instead of
6931         rolling our own.
6932         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6933         overflow_type.
6934         * fold-const.c (int_const_binop_2): Same.
6935         (extract_muldiv_1): Same.
6936         (fold_div_compare): Same.
6937         (fold_abs_const): Same.
6938         * match.pd: Same.
6939         * poly-int.h (add): Same.
6940         (sub): Same.
6941         (neg): Same.
6942         (mul): Same.
6943         * predict.c (predict_iv_comparison): Same.
6944         * profile-count.c (slow_safe_scale_64bit): Same.
6945         * simplify-rtx.c (simplify_const_binary_operation): Same.
6946         * tree-chrec.c (tree_fold_binomial): Same.
6947         * tree-data-ref.c (split_constant_offset_1): Same.
6948         * tree-if-conv.c (idx_within_array_bound): Same.
6949         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6950         * tree-ssa-phiopt.c (minmax_replacement): Same.
6951         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6952         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6953         * vr-values.c (vr_values::adjust_range_with_scev): Same.
6954         * wide-int.cc (wi::add_large): Same.
6955         (wi::mul_internal): Same.
6956         (wi::sub_large): Same.
6957         (wi::divmod_internal): Same.
6958         * wide-int.h: Change overflow type to overflow_type for neg, add,
6959         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6960         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6961         mul_internal, divmod_internal.
6962         (overflow_type): New enum.
6963         (accumulate_overflow): New.
6964
6965 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
6966
6967         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6968         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6969
6970 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
6971
6972         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6973         argument is checked for zero before entering loop, avoid checking again.
6974
6975 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
6976
6977         * gimplify.h (generic_expr_could_trap_p): Set as global function.
6978         * gimplify.h (generic_expr_could_trap_p): Likwise.
6979         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6980
6981 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
6982
6983         PR tree-optimization/86401
6984         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6985         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6986         (fold_bit_and_mask): ... here.  New helper function for match.pd.
6987         * fold-const.h (fold_bit_and_mask): Declare.
6988         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6989
6990 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
6991
6992         PR target/86324
6993         * target.def (translate_mode_attribute): New hook.
6994         * targhooks.h (default_translate_mode_attribute): Declare.
6995         * targhooks.c (default_translate_mode_attribute): New function.
6996         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6997         * doc/tm.texi: Regenerate.
6998         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6999         (rs6000_translate_mode_attribute): New function.
7000
7001 2018-07-06  Paul Koning  <ni1d@arrl.net>
7002
7003         * doc/md.texi (define_split): Document DONE and FAIL.
7004         (define_peephole2): Ditto.
7005
7006 2018-07-05  Jeff Law  <law@redhat.com>
7007
7008         PR tree-optimization/86010
7009         * tree-ssa-dse.c (compute_trims): More aggressively trim at
7010         both the head and tail of mem* and str* calls.
7011
7012 2018-07-05  Jim Wilson  <jimw@sifive.com>
7013
7014         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
7015
7016 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
7017
7018         * config/aarch64/aarch64-simd.md: correct flags text for
7019         MIN_EXPR replacement.
7020
7021 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
7022
7023         * configure: Regenerated.
7024
7025 2018-07-05  Carl Love  <cel@us.ibm.com>
7026
7027         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7028         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7029         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7030         VSX_BUILTIN_DOUBLEL_V4SF.
7031
7032 2018-07-05  Martin Sebor  <msebor@redhat.com>
7033
7034         PR c++/86400
7035         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7036         than its domain to compute its the upper bound of a char array.
7037
7038 2018-07-05  Nathan Sidwell  <nathan@acm.org>
7039
7040         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7041         * doc/cpp.texi: Update comment.
7042         * doc/tm.texi: Rebuilt.
7043         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7044         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7045         * doc/extend.texi (Backwards Compatibility): Clarify it is system
7046         headers affected by extern "C".
7047         * system.h: Poison NO_IMPLICIT_EXTERN_C.
7048         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7049         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7050         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7051         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7052         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7053         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7054         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7055         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7056         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7057         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7058         config/sparc/sp64-elf.h, config/spu/spu.h,
7059         config/stormy16/stormy16.h, config/v850/v850.h,
7060         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7061         define NO_IMPLICIT_EXTERN_C.
7062         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7063
7064 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
7065
7066         PR target/84711
7067         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7068         instead of GET_MODE_SIZE when comparing Units.
7069
7070 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
7071
7072         PR target/84711
7073         * rtlanal.c (set_noop_p): Constrain on mode change,
7074         include hard-reg-set.h
7075
7076 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
7077
7078         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7079
7080 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
7081
7082         Revert
7083         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
7084
7085         PR sanitizer/84250
7086         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7087         libasan.
7088         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7089
7090 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
7091
7092         PR sanitizer/84250
7093         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7094         libasan.
7095         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7096
7097 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
7098
7099         PR middle-end/86380
7100         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7101
7102 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
7103
7104         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7105         neg_*_op* variables.
7106
7107 2018-07-04  Martin Liska  <mliska@suse.cz>
7108
7109         * tree-switch-conversion.c: Define
7110         max_ratio_for_speed and max_ratio_for_size constants.
7111
7112 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
7113             Martin Liska  <mliska@suse.cz>
7114
7115         PR middle-end/66240
7116         PR target/45996
7117         PR c/84100
7118         * common.opt: Rename align options with 'str_' prefix.
7119         * common/config/i386/i386-common.c (set_malign_value): New
7120         function.
7121         (ix86_handle_option): Use it to set -falign-* options/
7122         * config/aarch64/aarch64-protos.h (struct tune_params): Change
7123         type from int to string.
7124         * config/aarch64/aarch64.c: Update default values from int
7125         to string.
7126         * config/alpha/alpha.c (alpha_override_options_after_change):
7127         Likewise.
7128         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7129         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7130         max skip conditionally.
7131         * config/i386/freebsd.h (SUBALIGN_LOG): New.
7132         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7133         max skip conditionally.
7134         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7135         max skip conditionally.
7136         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7137         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7138         max skip conditionally.
7139         * config/i386/i386.c (struct ptt): Change type from int to
7140         string.
7141         (ix86_default_align): Set default values.
7142         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7143         max skip conditionally.
7144         * config/i386/iamcu.h (SUBALIGN_LOG): New.
7145         (ASM_OUTPUT_MAX_SKIP_ALIGN):
7146         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7147         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7148         max skip conditionally.
7149         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7150         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7151         * config/i386/x86-64.h (SUBALIGN_LOG): New.
7152         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7153         max skip conditionally.
7154         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7155         * config/ia64/ia64.c (ia64_option_override): Set default values
7156         for alignment options.
7157         * config/m68k/m68k.c: Handle new str_align_* options.
7158         * config/mips/mips.c (mips_set_compression_mode): Change
7159         type of constants.
7160         (mips_option_override): Set default values for options.
7161         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7162         Likewise.
7163         * config/rs6000/rs6000.c (rs6000_option_override_internal):
7164         Likewise.
7165         * config/rx/rx.c (rx_option_override): Likewise.
7166         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7167         (LABEL_ALIGN): Use align_labels_log.
7168         (LOOP_ALIGN): Use align_loops_align.
7169         * config/s390/s390.c (s390_asm_output_function_label): Use new
7170         macros.
7171         * config/sh/sh.c (sh_override_options_after_change):
7172         Change type of constants.
7173         * config/spu/spu.c (spu_sched_init): Likewise.
7174         * config/sparc/sparc.c (sparc_option_override): Set default
7175         values for options.
7176         * config/visium/visium.c (visium_option_override): Likewise.
7177         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7178         emit p2align format with last argument if it's not needed.
7179         * doc/invoke.texi: Document extended format of -falign-*.
7180         * final.c: Use align_labels alignment.
7181         * flags.h (struct target_flag_state): Change type to use
7182         align_flags.
7183         (struct align_flags_tuple): New.
7184         (struct align_flags): Likewise.
7185         (align_loops_log): Redefine macro to use new types.
7186         (align_loops_max_skip): Redefine macro to use new types.
7187         (align_jumps_log): Redefine macro to use new types.
7188         (align_jumps_max_skip): Redefine macro to use new types.
7189         (align_labels_log): Redefine macro to use new types.
7190         (align_labels_max_skip): Redefine macro to use new types.
7191         (align_functions_log): Redefine macro to use new types.
7192         (align_loops): Redefine macro to use new types.
7193         (align_jumps): Redefine macro to use new types.
7194         (align_labels): Redefine macro to use new types.
7195         (align_functions): Redefine macro to use new types.
7196         (align_functions_max_skip): Redefine macro to use new types.
7197         (align_loops_value): New macro.
7198         (align_jumps_value): New macro.
7199         (align_labels_value): New macro.
7200         (align_functions_value): New macro.
7201         * function.c (invoke_set_current_function_hook): Propagate
7202         alignment values from flags to global variables default in
7203         topleev.h.
7204         * ipa-icf.c (sem_function::equals_wpa): Use
7205         cl_optimization_option_eq instead of memcmp.
7206         * lto-streamer.h (cl_optimization_stream_out): Support streaming
7207         of string types.
7208         (cl_optimization_stream_in): Likewise.
7209         * optc-save-gen.awk: Support strings in cl_optimization.
7210         * opth-gen.awk: Likewise.
7211         * opts.c (finish_options): Remove error checking of invalid
7212         value ranges.
7213         (MAX_CODE_ALIGN): Remove.
7214         (MAX_CODE_ALIGN_VALUE): Likewise.
7215         (parse_and_check_align_values): New function.
7216         (check_alignment_argument): Likewise.
7217         (common_handle_option): Use check_alignment_argument.
7218         * opts.h (parse_and_check_align_values): Declare.
7219         * toplev.c (init_alignments): Remove.
7220         (read_log_maxskip): New.
7221         (parse_N_M): Likewise.
7222         (parse_alignment_opts): Likewise.
7223         (backend_init_target): Remove usage of init_alignments.
7224         * toplev.h (parse_alignment_opts): Declare.
7225         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7226         argument.
7227         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7228         * tree.c (cl_option_hasher::equal): New.
7229         * varasm.c: Use new global macros.
7230
7231 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
7232
7233         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7234         Use a simpler align directive also if MAXSKIP = ALIGN-1.
7235         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7236         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7237         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7238         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7239         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
7240         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
7241         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
7242         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7243         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7244         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7245         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7246
7247 2018-07-04  Martin Liska  <mliska@suse.cz>
7248             Jonathan Wakely  <jwakely@redhat.com>
7249
7250         * coverage.c: Use correct type.
7251         * doc/invoke.texi: Language correction.
7252
7253 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
7254
7255         PR target/85620
7256         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
7257         ENDBRANCH for non-tail call which may return via indirect branch.
7258         * doc/extend.texi: Document indirect_return attribute.
7259
7260 2018-07-03  Martin Sebor  <msebor@redhat.com>
7261
7262         PR tree-optimization/86274
7263         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
7264         precondition.
7265         (format_floating): Correct handling of infinities and NaNs.
7266
7267 2018-07-03  Martin Sebor  <msebor@redhat.com>
7268
7269         * print-tree.c (print_real_cst): New function.
7270         (print_node_brief): Call it.
7271         (print_node): Ditto.
7272
7273 2018-07-03  Jeff Law  <law@redhat.com>
7274
7275         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
7276         into a single pattern.
7277
7278         * config/h8300/h8300.md (ors code_iterator): New.
7279         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
7280         a single pattern and single splitter.
7281         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
7282         (iorqi3_1, xorqi3_1): Likewise.
7283         (iorqi3, xorqi3 expanders): Similarly.
7284
7285         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
7286         (movmd_internal) into a single pattern using the P mode iterator.
7287         (movmd splitters): Similarly.
7288         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
7289         (movsd splitters): Similarly.
7290
7291         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
7292         ADDB, ADDW and ADDL into a single ADD attribute which selects the
7293         right table based on the size of the operand.
7294         * config/h8300/h8300.md (length_table): Corresponding changes. All
7295         references to "addb", "addw" and "addl" changed to "add".
7296         (btst patterns): Merge two variants into a single pattern.
7297         (tstqi, tsthi): Likewise.
7298         (addhi3_incdec, addsi3_incdec): Likewise.
7299         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
7300         (mulhi3, mulsi3): Likewise.
7301         (udivhi3, udivsi3): Likewise.
7302         (divhi3, divsi3): Likewise.
7303         (andorqi3, andorhi3, andorsi3): Likewise.
7304
7305 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
7306
7307         PR target/85694
7308         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7309         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7310
7311 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7312
7313         PR tree-optimization/85694
7314         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7315         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7316         UNSPEC_URHADD.
7317         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7318         (<u>avg<mode>3_ceil): New patterns.
7319
7320 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
7321
7322         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7323         scan-tree-dump directive.
7324         * gcc.dg/vect/slp-perm-2.c: Likewise.
7325         * gcc.dg/vect/slp-perm-3.c: Likewise.
7326         * gcc.dg/vect/slp-perm-5.c: Likewise.
7327         * gcc.dg/vect/slp-perm-6.c: Likewise.
7328         * gcc.dg/vect/slp-perm-7.c: Likewise.
7329         * gcc.dg/vect/slp-perm-8.c: Likewise.
7330
7331 2018-07-03  Marek Polacek  <polacek@redhat.com>
7332
7333         PR middle-end/86202
7334         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7335
7336 2018-07-03  Richard Biener  <rguenther@suse.de>
7337
7338         PR ipa/86389
7339         * tree-ssa-structalias.c (find_func_clobbers): Properly
7340         handle indirect calls.
7341
7342 2018-07-03  Jeff Law  <law@redhat.com>
7343
7344         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7345         (shifts): New code iterator.
7346         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7347         expander.  Fix HImode handling on H8/SX.
7348         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7349         (subqi3, subhi3, subsi3 expanders): Likewise.
7350         (andqi3, andhi3, andsi3 expanders): Likewise.
7351         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7352         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7353         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7354         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7355         (zero_extendqihi2, zero_extendqisi2): Likewise.
7356         (extendqihi2, extendqisi2): Likewise.
7357         (rotlqi3, rotlhi3, rotlsi3): Likewise.
7358         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7359         (rotlqi3_1, rotlhi3_1): Likewise.
7360         (logicalhi3_sn, logicalsi3_sn): Likewise.
7361         (logicalhi3, logicalsi3): Likewise.
7362
7363 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7364
7365         * tree-vect-patterns.c (vect_recog_rotate_pattern)
7366         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7367         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7368         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7369         type to append_pattern_def_seq instead of creating a stmt_vec_info
7370         directly.
7371         (build_mask_conversion): Likewise.  Remove vinfo argument.
7372         (vect_add_conversion_to_patterm): Likewise, renaming to...
7373         (vect_add_conversion_to_pattern): ...this.
7374         (vect_recog_mask_conversion_pattern): Update call to
7375         build_mask_conversion.  Pass the vector type to
7376         append_pattern_def_seq here too.
7377         (vect_recog_gather_scatter_pattern): Update call to
7378         vect_add_conversion_to_pattern.
7379
7380 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7381
7382         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7383         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7384         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7385         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7386         STMT_VINFO_PATTERN_DEF_SEQ to null here.
7387         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7388         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7389         append_pattern_def_seq instead of new_pattern_def_seq.
7390         (vect_recog_divmod_pattern): Do both of the above.
7391         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7392         is null.
7393
7394 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7395
7396         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7397         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7398         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7399         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7400         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7401         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7402         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7403         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7404         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7405         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7406         parameter with a single stmt_vec_info.
7407         (vect_recog_func_ptr): Likewise.
7408         (vect_recog_gather_scatter_pattern): Likewise, folding in...
7409         (vect_try_gather_scatter_pattern): ...this.
7410         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7411         the stmt_vec_info of the statement to be matched.  Don't clear
7412         STMT_VINFO_RELATED_STMT.
7413         (vect_pattern_recog): Update call accordingly.
7414
7415 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7416
7417         PR tree-optimization/85694
7418         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7419         (uavgM3_ceil): Document new optabs.
7420         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7421         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7422         functions.
7423         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7424         (savg_ceil_optab): New optabs.
7425         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7426         (vect_vect_recog_func_ptrs): Add it.
7427         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7428         constant directly from the associated lhs.
7429
7430 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7431
7432         * tree-vect-patterns.c (vect_split_statement): New function.
7433         (vect_convert_input): Use it to try to split an existing cast.
7434
7435 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7436
7437         * poly-int.h (print_hex): New function.
7438         * dumpfile.h (dump_dec, dump_hex): Declare.
7439         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7440         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7441         min_input_precision, operation_precision and operation_sign.
7442         * tree-vect-patterns.c (vect_get_range_info): New function.
7443         (vect_same_loop_or_bb_p, vect_single_imm_use)
7444         (vect_operation_fits_smaller_type): Delete.
7445         (vect_look_through_possible_promotion): Add an optional
7446         single_use_p parameter.
7447         (vect_recog_over_widening_pattern): Rewrite to use new
7448         stmt_vec_info infomration.  Handle one operation at a time.
7449         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7450         (vect_truncatable_operation_p, vect_set_operation_type)
7451         (vect_set_min_input_precision): New functions.
7452         (vect_determine_min_output_precision_1): Likewise.
7453         (vect_determine_min_output_precision): Likewise.
7454         (vect_determine_precisions_from_range): Likewise.
7455         (vect_determine_precisions_from_users): Likewise.
7456         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7457         (vect_vect_recog_func_ptrs): Put over_widening first.
7458         Add cast_forwprop.
7459         (vect_pattern_recog): Call vect_determine_precisions.
7460
7461 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7462
7463         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7464         statements that have been replaced by further pattern statements.
7465         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7466
7467 2018-07-03  Richard Biener  <rguenther@suse.de>
7468
7469         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7470         always set *dt.  Dump vectype in vectype overload.
7471         * dumpfile.h (dump_gimple_expr): New function.
7472         (dump_gimple_expr_loc): Likewise.
7473         * dumpfile.c (dump_gimple_expr): New function.
7474         (dump_gimple_expr_loc): Likewise.
7475
7476 2018-07-02  Jeff Law  <law@redhat.com>
7477
7478         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7479         the H8/300, H8/300H and H8/S variants into a single pattern.
7480         (movhi_h8300, movqi_h8300hs): Similarly.
7481         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7482         (QHI mode iterator): New.
7483
7484         * config/h8300/h8300.md: Remove trailing whitespace.
7485
7486 2018-07-02  Jim Wilson  <jimw@sifive.com>
7487
7488         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7489         instead of emit_insn for interrupt returns.
7490         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7491         (riscv_sret, riscv_uret): Likewise.
7492
7493 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
7494
7495         * pretty-print.c (selftest::test_pp_format): Move save and restore
7496         of quotes to class auto_fix_quotes, and add an instance.
7497         * selftest.c: Include "intl.h".
7498         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7499         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7500         * selftest.h (selftest::auto_fix_quotes): New class.
7501
7502 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
7503
7504         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7505         (aarch64_sve_prepare_conditional_op): Remove.
7506         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7507         Allow aarch64_simd_reg_or_zero as select operand; remove
7508         the aarch64_sve_prepare_conditional_op call.
7509         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7510         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7511         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7512         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7513         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7514         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7515         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7516         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7517         and a splitters to match all of the *_any patterns.
7518         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7519
7520         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7521         (SVE_COND_FP_BINARY_REV): Remove.
7522         (sve_int_op_rev, sve_fp_op_rev): New.
7523         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7524         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7525         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7526         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7527         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7528         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7529         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7530         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7531         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7532
7533         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7534         Remove match_dup 1 from the inner unspec.
7535         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7536
7537         * config/aarch64/aarch64.md (movprfx): New attr.
7538         (length): Default movprfx to 8.
7539         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7540         (*madd<SVE_I>, *msub<SVE_I): Likewise.
7541         (*<su>mul<SVE_I>3_highpart): Likewise.
7542         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7543         (*v<ASHIFT><SVE_I>3): Likewise.
7544         (*<su><MAXMIN><SVE_I>3): Likewise.
7545         (*<su><MAXMIN><SVE_F>3): Likewise.
7546         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7547         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7548         (*div<SVE_F>4): Likewise.
7549
7550 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
7551
7552         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7553         in dump string.
7554
7555 2018-07-02  Richard Biener  <rguenther@suse.de>
7556
7557         PR tree-optimization/86363
7558         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7559         memset argument refers to a non-variable address.
7560
7561 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
7562
7563         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7564         {PLUS,MINUS}_EXPR code to...
7565         (adjust_symbolic_bound): ...here,
7566         (combine_bound): ...here,
7567         (set_value_range_with_overflow): ...and here.
7568
7569 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
7570
7571         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7572         code...
7573         (extract_range_from_abs_expr): ...here.
7574
7575 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
7576
7577         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7578         -fno-omit-frame-pointer when not optimizing.
7579
7580 2018-07-02  Martin Liska  <mliska@suse.cz>
7581
7582         PR ipa/86279
7583         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7584         (propagate_nothrow): Likewise.
7585
7586 2018-07-02  Martin Liska  <mliska@suse.cz>
7587
7588         PR ipa/86323
7589         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7590
7591 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
7592
7593         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7594         function in r262149, changing "loc" param from source_location to
7595         const dump_location_t &.
7596         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7597         declaration, as above.
7598
7599 2018-07-01  Paul Koning  <ni1d@arrl.net>
7600
7601         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7602         -munit-asm, -mgnu-asm, -mdec-asm.
7603         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7604         (pdp11_output_labelref): New.
7605         (pdp11_output_def): New.
7606         (pdp11_output_addr_vec_elt): New.
7607         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
7608         %# and %@ format codes.
7609         (pdp11_option_override): New.
7610         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7611         (pdp11_output_ident): New.
7612         (pdp11_asm_named_section): New.
7613         (pdp11_asm_init_sections): New.
7614         (pdp11_file_start): New.
7615         (pdp11_file_end): New.
7616         (output_ascii): Use .ascii/.asciz for -mdec-asm.
7617         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
7618         %o, like %c but octal.
7619         (pdp11_option_override): New.
7620         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7621         -mdec-asm.
7622         (DATA_SECTION_ASM_OP): Ditto.
7623         (READONLY_DATA_SECTION_ASM_OP): New.
7624         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7625         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7626         (ASM_OUTPUT_LABELREF): Ditto.
7627         (ASM_OUTPUT_DEF): Ditto.
7628         (ASM_OUTPUT_EXTERNAL): New.
7629         (ASM_OUTPUT_SOURCE_FILENAME): New.
7630         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7631         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7632         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
7633         %# and %@ format codes.
7634         * config/pdp11/pdp11.opt (mgnu-asm): New.
7635         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7636         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7637         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7638
7639 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
7640
7641         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7642         dereferencing path[] beyond its length.
7643         (debug_path): New.
7644         (debug_all_paths): New.
7645         (rewire_first_differing_edge): New.
7646         (adjust_paths_after_duplication): New.
7647         (duplicate_thread_path): Call adjust_paths_after_duplication.
7648         Add new argument.
7649         (thread_through_all_blocks): Add new argument to
7650         duplicate_thread_path.
7651
7652 2018-06-30  Jim Wilson  <jimw@sifive.com>
7653
7654         * config/riscv/predicates.md (p2m1_shift_operand): New.
7655         (high_mask_shift_operand): New.
7656         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7657         pattern using p2m1_shift_operand.
7658         (lshsi3_zero_extend_3+2): New combiner pattern using
7659         high_mask_shift_operand.
7660
7661 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7662
7663         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7664         split out from...
7665         (vect_recog_rotate_pattern): ...here.
7666         (vect_convert_input): Try to insert casts of invariants in the
7667         preheader.
7668         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7669         preheader to be empty.
7670
7671 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7672
7673         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7674         vector type.  If given, install it in the new statement's
7675         STMT_VINFO_VECTYPE.
7676         (vect_element_precision): New function.
7677         (vect_unpromoted_value): New struct.
7678         (vect_unpromoted_value::vect_unpromoted_value): New function.
7679         (vect_unpromoted_value::set_op): Likewise.
7680         (vect_look_through_possible_promotion): Likewise.
7681         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7682         (vect_widened_op_tree, vect_convert_input): Likewise.
7683         (vect_convert_inputs, vect_convert_output): Likewise.
7684         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7685         to handle the optional cast of the multiplication result and
7686         vect_widened_op_tree to detect the widened multiplication itself.
7687         Do not require the input and output of promotion casts to have
7688         the same sign, but base the signedness of the operation on the
7689         input rather than the result.  If the pattern includes two
7690         promotions, check that those promotions have the same sign.
7691         Do not restrict the MULT_EXPR handling to a double-width result;
7692         handle quadruple-width results and wider.  Use vect_convert_inputs
7693         to convert the inputs to the common type.
7694         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
7695         to handle the optional cast of the ABS result.  Also allow a sign
7696         change or a sign extension between the ABS and MINUS.
7697         Use vect_widened_op_tree to detect the widened subtraction and use
7698         vect_convert_inputs to convert the inputs to the common type.
7699         (vect_handle_widen_op_by_const): Delete.
7700         (vect_recog_widen_op_pattern): New function.
7701         (vect_recog_widen_mult_pattern): Use it.
7702         (vect_recog_widen_shift_pattern): Likewise.
7703         (vect_recog_widen_sum_pattern): Use
7704         vect_look_through_possible_promotion to handle the promoted
7705         PLUS_EXPR operand.
7706
7707 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7708
7709         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7710         the containing gimple_seq *.
7711         * gimple-iterator.h (gsi_for_stmt): Declare it.
7712         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7713         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7714         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7715         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7716         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7717         checks.
7718         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7719         split out from...
7720         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
7721         statement being replaced is part of an existing pattern
7722         definition sequence, inserting the new pattern statements before
7723         the original one.
7724         (vect_pattern_recog_1): Don't return a bool.  If the statement
7725         is already part of a pattern, instead apply pattern matching
7726         to the pattern definition statements.  Don't clear the
7727         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7728         (vect_pattern_recog): Don't break after the first match;
7729         continue processing the pattern definition statements instead.
7730         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7731
7732 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7733
7734         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7735         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7736         (vect_recog_widen_sum_pattern): Use it.
7737
7738 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7739
7740         * tree-vect-loop.c (vectorizable_reduction): Assert that the
7741         phi is not a pattern statement and has not been replaced by
7742         a pattern statement.
7743         * tree-vect-patterns.c (type_conversion_p): Don't check
7744         STMT_VINFO_IN_PATTERN_P.
7745         (vect_recog_vector_vector_shift_pattern): Likewise.
7746         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7747         the pattern statement rather than the original statement; check
7748         directly for a WIDEN_MULT_EXPR here.
7749         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7750         vect_is_simple_use to return the pattern statement rather
7751         than the original statement; use is_pattern_stmt_p to check
7752         for such a pattern statement.
7753         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7754         to return the pattern statement rather than the original statement;
7755         don't do the same transformation here.
7756         (vect_is_simple_use): If the defining statement has been replaced
7757         by a pattern statement, return the pattern statement instead.
7758         Remove the corresponding (local) transformation from the vectype
7759         overload.
7760
7761 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7762
7763         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7764         end and default to null.
7765         * tree-vect-loop.c (vect_create_epilog_for_reduction)
7766         (vectorizable_reduction): Update calls accordingly, dropping the
7767         gimple ** argument if the passed-back statement isn't needed.
7768         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7769         (vect_recog_rotate_pattern): Likewise.
7770         (vect_recog_mask_conversion_pattern): Likewise.
7771         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7772         (vect_mask_constant_operand_p): Likewise.
7773         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7774         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7775         (get_group_load_store_type, get_load_store_type): Likewise.
7776         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7777         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7778         (vectorizable_conversion, vectorizable_assignment): Likewise.
7779         (vectorizable_shift, vectorizable_operation): Likewise.
7780         (vectorizable_store, vect_is_simple_cond): Likewise.
7781         (vectorizable_condition, vectorizable_comparison): Likewise.
7782         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7783         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7784         and move it to the end.  Cope with null def_stmt_outs.
7785
7786 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7787
7788         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7789
7790 2018-06-29  Jeff Law  <law@redhat.com>
7791
7792         * config/v850/v850.c (v850_legitimate_address_p): Handle large
7793         displacements for TARGET_V850E2V3 and newer.
7794         (TARGET_LRA_P): Remove.  Defaults to LRA now.
7795         * config/v850/v850.md (sign23byte_load): Remove.
7796         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7797         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7798
7799 2018-06-29  Martin Liska  <mliska@suse.cz>
7800
7801         PR lto/85759
7802         * coverage.c (coverage_init): Mangle full path name.
7803         * doc/invoke.texi: Document the change.
7804         * gcov-io.c (mangle_path): New.
7805         * gcov-io.h (mangle_path): Likewise.
7806         * gcov.c (mangle_name): Use mangle_path for path mangling.
7807
7808 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7809
7810         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7811         if starting source register is not even.
7812
7813 2018-06-29  Martin Liska  <mliska@suse.cz>
7814
7815         PR tree-optimization/86263
7816         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7817         Make edge redirection.
7818
7819 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
7820
7821         * dumpfile.c (dump_loc): Add indentation based on scope depth.
7822         (dump_scope_depth): New variable.
7823         (get_dump_scope_depth): New function.
7824         (dump_begin_scope): New function.
7825         (dump_end_scope): New function.
7826         * dumpfile.h (get_dump_scope_depth): New declaration.
7827         (dump_begin_scope): New declaration.
7828         (dump_end_scope): New declaration.
7829         (class auto_dump_scope): New class.
7830         (AUTO_DUMP_SCOPE): New macro.
7831         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7832         AUTO_DUMP_SCOPE.
7833
7834 2018-06-29  Richard Biener  <rguenther@suse.de>
7835
7836         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7837         compute_all_dependences succeeds.
7838         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7839         exceed --param loop-max-datarefs-for-datadeps.
7840
7841 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
7842
7843         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7844
7845 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
7846
7847         PR target/86348
7848         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7849         alternative 0 in preferred_for_speed attribute.
7850
7851 2018-06-28  Paul Koning  <ni1d@arrl.net>
7852
7853         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7854         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7855         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7856         * config/pdp11/pdp11.md: Correct "length" attribute calculation
7857         for shift insn patterns.
7858
7859 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
7860
7861         * cgraph.c (cgraph_node::get_body): Replace assignments to
7862         "dump_file" with calls to set_dump_file.
7863         * dumpfile.c (alt_dump_file): Make static, and group with...
7864         (alt_flags): ...this definition.
7865         (dumps_are_enabled): New variable.
7866         (refresh_dumps_are_enabled): New function.
7867         (set_dump_file): New function.
7868         (set_alt_dump_file): New function.
7869         (gcc::dump_manager::dump_start): Replace assignments to
7870         "dump_file" and "alt_dump_file" with calls to set_dump_file and
7871         set_alt_dump_file.
7872         (gcc::dump_manager::dump_finish): Likewise.
7873         * dumpfile.h (alt_dump_file): Delete decl.
7874         (dumps_are_enabled): New variable decl.
7875         (set_dump_file): New function decl.
7876         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7877         global.
7878         * tree-nested.c (lower_nested_functions): Replace assignments to
7879         "dump_file" with calls to set_dump_file.
7880
7881 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
7882
7883         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7884         goto_locus of each outgoing edge of each basic block.
7885
7886 2018-06-28  Richard Biener  <rguenther@suse.de>
7887
7888         * dwarf2out.c (decl_scope_table): Remove.
7889         (push_decl_scope): Likewise.
7890         (pop_decl_scope): Likewise.
7891         (gen_type_die_for_member): Do not call push/pop_decl_scope.
7892         (gen_struct_or_union_type_die): Likewise.
7893         (gen_tagged_type_die): Likewise.
7894         (dwarf2out_init): Do not initialize decl_scope_table.
7895         (dwarf2out_c_finalize): Do not free it.
7896
7897 2018-06-28  Richard Biener  <rguenther@suse.de>
7898
7899         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7900         deciding whether to not re-use a DIE.
7901
7902 2018-06-28  Richard Biener  <rguenther@suse.de>
7903
7904         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7905         DW_AT_abstract_origin attribute.
7906
7907 2018-06-28  Martin Liska  <mliska@suse.cz>
7908
7909         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7910         Use newly introduced constants.
7911         * tree-switch-conversion.h (struct jump_table_cluster):
7912         Define max_ratio_for_size and max_ratio_for_speed.
7913
7914 2018-06-28  Martin Liska  <mliska@suse.cz>
7915
7916         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7917         Add new checking assert to catch invalid state.
7918         (jump_table_cluster::can_be_handled): Handle single case
7919         clusters.
7920         (jump_table_cluster::is_beneficial): Bail out for such case.
7921         (bit_test_cluster::find_bit_tests):
7922         Add new checking assert to catch invalid state.
7923         (bit_test_cluster::can_be_handled): Handle single case
7924         clusters.
7925         (bit_test_cluster::is_beneficial): Bail out for such case.
7926         (switch_decision_tree::analyze_switch_statement):
7927         Fix comment.
7928
7929 2018-06-28  Martin Liska  <mliska@suse.cz>
7930
7931         * common.opt: Introduce -completion option.
7932         * gcc.c (driver_handle_option): Handle it.
7933         (driver::main): Print completions if completion
7934         is set.
7935         * opt-suggestions.c (option_proposer::get_completions):
7936         New function.
7937         (option_proposer::suggest_completion): Likewise.
7938         (option_proposer::find_param_completions): Likewise.
7939         (verify_autocompletions): Likewise.
7940         (test_completion_valid_options): Likewise.
7941         (test_completion_valid_params): Likewise.
7942         (in_completion_p): Likewise.
7943         (empty_completion_p): Likewise.
7944         (test_completion_partial_match): Likewise.
7945         (test_completion_garbage): Likewise.
7946         (opt_proposer_c_tests): Likewise.
7947         * opt-suggestions.h: Declare new functions.
7948         * opts.c (common_handle_option): Handle OPT__completion_.
7949         * selftest-run-tests.c (selftest::run_tests): Add
7950         opt_proposer_c_tests.
7951         * selftest.c (assert_str_startswith): New.
7952         * selftest.h (assert_str_startswith): Likewise.
7953         (opt_proposer_c_tests): New.
7954         (ASSERT_STR_STARTSWITH): Likewise.
7955
7956 2018-06-28  Martin Liska  <mliska@suse.cz>
7957
7958         * Makefile.in: Add opt-suggestions.o.
7959         * gcc-main.c: Include opt-suggestions.h.
7960         * gcc.c (driver::driver): Likewise.
7961         (driver::~driver): Remove m_option_suggestions.
7962         (driver::build_option_suggestions): Moved to option_proposer.
7963         (driver::suggest_option): Likewise.
7964         (driver::handle_unrecognized_options): Use option_proposer.
7965         * gcc.h (class driver): Add new memver m_option_proposer.
7966         * opt-suggestions.c: New file.
7967         * opt-suggestions.h: New file.
7968
7969 2018-06-28  Martin Liska  <mliska@suse.cz>
7970
7971         * vec.h (class auto_string_vec): New (moved from auto_argvec).
7972         (auto_string_vec::~auto_string_vec): Likewise.
7973
7974 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
7975
7976         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7977         prevent_decl_creation_for_types fields up and add reset_location field.
7978         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7979         statement if id->reset_location is true.
7980         (copy_edges_for_bb): Do not set goto_locus on the new edges if
7981         id->reset_location is true.
7982         (copy_phis_for_bb): Force input_location on the arguments if
7983         id->reset_location is true.
7984         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7985         is set on the function to be inlined.
7986
7987 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
7988
7989         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7990
7991 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
7992
7993         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7994         registers for Pmode.
7995         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7996         hard registers for the clobbered pseudo.
7997
7998 2018-06-27  Paul Koning  <ni1d@arrl.net>
7999
8000         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8001         mutually exclusive options.
8002         * config/pdp11/constraints.md (h): New constraint.
8003         (O): Update definition to match shift code generation.
8004         (D): New constraint.
8005         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
8006         (CCFP): Remove.
8007         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
8008         function.
8009         (output_jump): Change arguments.
8010         (pdp11_fixed_cc_regs): New function.
8011         (pdp11_cc_mode): Ditto.
8012         (pdp11_expand_shift): Ditto.
8013         (pdp11_assemble_shift): Ditto.
8014         (pdp11_small_shift): Ditto.
8015         (pdp11_branch_cost): Remove.
8016         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
8017         from output.
8018         (pdp11_register_move_cost): Update for CC registers.
8019         (pdp11_rtx_costs): Add case for LSHIFTRT.
8020         (pdp11_output_jump): Add CCNZ mode conditional branches.
8021         (notice_update_cc_on_set): Remove.
8022         (pdp11_cc_mode): New function.
8023         (simple_memory_operand): Correct pre/post decrement case.
8024         (no_side_effect_operand): New function.
8025         (pdp11_regno_reg_class): Add CC_REGS class.
8026         (pdp11_fixed_cc_regs): New function.
8027         (pdp11_small_shift): New function.
8028         (pdp11_expand_shift): New function to expand shift insns.
8029         (pdp11_assemble_shift): New function to output shifts.
8030         (pdp11_branch_cost): Remove.
8031         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
8032         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
8033         (WCHAR_TYPE): Ditto.
8034         (PTRDIFF_TYPE): Ditto.
8035         (ADJUST_INSN_LENGTH): New macro.
8036         (FIXED_REGISTERS): Add CC registers.
8037         (CALL_USED_REGISTERS): Ditto.
8038         (reg_class): Ditto.
8039         (REG_CLASS_NAMES): Ditto.
8040         (REG_CLASS_CONTENTS): Ditto.
8041         (SELECT_CC_MODE): Use new function.
8042         (TARGET_FLAGS_REGNUM): New macro.
8043         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8044         (cc0_reg_rtx): Remove.
8045         (CC_STATUS_MDEP): Remove.
8046         (CC_STATUS_MDEFP_INIT): Remove.
8047         (CC_IN_FPU): Remove.
8048         (NOTICE_UPDATE_CC): Remove.
8049         (REGISTER_NAMES): Add CC registers.
8050         (BRANCH_COST): Change to constant 1.
8051         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8052         handling.
8053         * config/pdp11/pdp11.opt (mbcopy): Remove.
8054         (mbcopy-builtin): Remove.
8055         (mbranch-cheap): Remove.
8056         (mbranch-expensive): Remove.
8057         * config/pdp11/predicates.md (expand_shift_operand): Update to
8058         match shift code generation.
8059         (ccnz_operator): New predicate.
8060         * doc/invoke.texi (PDP-11 Options): Remove deleted options
8061         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8062         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
8063         exclusive options.
8064         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
8065         description of O constraint.
8066
8067 2018-06-27  Jeff Law  <law@redhat.com>
8068             Austin Law  <austinklaw@gmail.com>
8069
8070         * config/v850/v850.md (addsi3_set_flags): New pattern.
8071         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8072         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8073         (zero_extendhisi2_v850_set_flags): Likewise.
8074         (zero_extendqisi2_v850_set_flags): Likewise.
8075         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8076         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8077         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8078
8079         * config/v850/v850-protos.h (notice_update_cc): Remove.
8080         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8081         (v850_print_operand): Handle 'D' and "d".
8082         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8083         Add handling of arithmetic/logical operations compared against zero.
8084         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8085         Do not look at v850_compare_op, instead get mode from last argument.
8086         (v850_gen_compare): Remove
8087         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8088         after reload for prologue insns.
8089         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8090         patterns.
8091         (construct_save_jarl): Likewise.
8092         (TARGET_FLAGS_REGNUM): Define.
8093         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8094         (NOTICE_UPDATE_CC): Remove.
8095         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8096         than cc0.  Conditionalize on reload_completed.
8097         (cmpsi_insn, setfcc_insn): Likewise.
8098         (tst1 splitter): Turn into define_and_split which sets the flags
8099         after reload.
8100         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8101         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8102         (cstoresf4, cstoredf4): Clobber the flags.
8103         (cmpsi, cmpsf, cmpdf): Remove expanders.
8104         (setf_insn): Remove pattern.
8105         (addsi3): Turn into define_and_split which clobbers the flags after
8106         reload and a suitable pattern (addsi3_clobber_flags) for use after
8107         reload.
8108         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8109         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8110         (ashrsi3, ashrsi3_v850e2): Likewise.
8111         (bins): Clobber the flags.
8112         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8113         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8114         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8115         (call_value_internal_short, call_value_internal_long): Likewise.
8116         (callt_save_interrupt, callt_return_interrupt): Likewise.
8117         (save_interrupt, return_interrupt): Likewise.
8118         (callt_save_all_interrupt, save_all_interrupt): Likewise.
8119         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8120         (restore_all_interrupt, _restore_all_interrupt): Likewise.
8121         (All FP comparisons): Only allow after reload has completed.
8122         (trfsr): Likewise.
8123         (divh, divhu): Tweak output template.
8124         (branch_z_normal, branch_z_invert): Remove
8125         (branch_nz_normal, branch_nz_invert): Likewise.
8126         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8127
8128         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8129         * config/v850/v850.c (notice_update_cc): Remove.
8130         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
8131         (CC_NO_CARRY): Likewise.
8132         (NOTICE_UPDATE_CC): Define to nothing.
8133         * config/v850/v850.md: Remove block comment on cc0 handling
8134         Remove "cc" attribute from all patterns.  Remove cc_status handling
8135         from all patterns.  Minor formatting fixes.
8136
8137 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8138
8139         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8140         (cortex-a76.cortex-a55): Likewise.
8141         * config/aarch64/aarch64-tune.md: Regenerate.
8142         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8143         cortex-a76.cortex-a55.
8144
8145 2018-06-27  Jeff Law  <law@redhat.com>
8146
8147         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8148         (MULTILIB_DIRNAMES): Similarly.
8149
8150 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
8151
8152         * gimple.h (gimple_return_retbnd): Delete.
8153         (gimple_return_set_retbnd): Likewise.
8154         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8155         gimple_return_set_retbnd.
8156         * gimple-pretty-print.c (dump_gimple_return): Remove call to
8157         gimple_return_retbnd and adjust.
8158         * tree-inline.h (struct copy_body_data): Remove retbnd field.
8159         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8160         Explicitly return NULL in a couple more cases.  Move assertion
8161         on debug statements and remove unreachable code.
8162         (reset_debug_binding): Do not test id->retbnd.
8163         (expand_call_inline): Do not set it.
8164
8165 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8166
8167         * configure.ac: Add --disable-gcov option.
8168         * configure: Regenerate.
8169         * Makefile.in: Honour @enable_gcov@.
8170         * doc/install.texi: Document --disable-gcov.
8171
8172 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8173
8174         * config/arm/arm-cpus.in (cortex-a76): New entry.
8175         (cortex-a76.cortex-a55): Likewise.
8176         * config/arm/arm-tables.opt: Regenerate.
8177         * config/arm/arm-tune.md: Likewise.
8178         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8179         * doc/invoke.texi (ARM Options): Document cortex-a76 and
8180         cortex-a76.cortex-a55.
8181
8182 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
8183
8184         PR target/85769
8185         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8186
8187 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
8188
8189         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8190         comment.
8191         (EPILOGUE_USES): Likewise.
8192
8193 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
8194
8195         * tree-inline.c (remap_location): New function extracted from...
8196         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
8197         (copy_phis_for_bb): ...here.  Call remap_location.
8198         (copy_cfg_body): Adjust call to copy_edges_for_bb.
8199
8200 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
8201
8202         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8203         unaligned vsx for 16B memset.
8204
8205 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
8206
8207         PR target/86285
8208         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8209         ieee128_float_type_node to long_double_type_node unless
8210         TARGET_LONG_DOUBLE_128 is set.
8211
8212 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
8213
8214         * cfgloop.c (get_loop_location): Convert return type from
8215         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8216         by implicit construction from rtx_insn *, and using
8217         dump_user_location_t::from_function_decl for the fallback case.
8218         * cfgloop.h (get_loop_location): Convert return type from
8219         location_t to dump_user_location_t.
8220         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8221         dump_printf_loc to pass in a dump_location_t rather than a
8222         location_t, via the gimple stmt.
8223         * coverage.c (get_coverage_counts): Update calls to
8224         dump_printf_loc to pass in dump_location_t rather than a
8225         location_t.
8226         * doc/optinfo.texi (Dump types): Convert example of
8227         dump_printf_loc from taking "locus" to taking "insn".  Update
8228         description of the "_loc" calls to cover dump_location_t.
8229         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8230         "selftest.h".
8231         (dump_user_location_t::dump_user_location_t): New constructors,
8232         from gimple *stmt and rtx_insn *.
8233         (dump_user_location_t::from_function_decl): New function.
8234         (dump_loc): Make static.
8235         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
8236         const dump_location_t &.
8237         (dump_generic_expr_loc): Delete.
8238         (dump_printf_loc): Convert param "loc" from location_t to
8239         const dump_location_t &.
8240         (selftest::test_impl_location): New function.
8241         (selftest::dumpfile_c_tests): New function.
8242         * dumpfile.h: Include "profile-count.h".
8243         (class dump_user_location_t): New class.
8244         (struct dump_impl_location_t): New struct.
8245         (class dump_location_t): New class.
8246         (dump_printf_loc): Convert 2nd param from source_location to
8247         const dump_location_t &.
8248         (dump_generic_expr_loc): Delete.
8249         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
8250         const dump_location_t &.
8251         * gimple-fold.c (fold_gimple_assign): Update call to
8252         dump_printf_loc to pass in a dump_location_t rather than a
8253         location_t, via the gimple stmt.
8254         (gimple_fold_call): Likewise.
8255         * gimple-loop-interchange.cc
8256         (loop_cand::analyze_iloop_reduction_var): Update for change to
8257         check_reduction_path.
8258         (tree_loop_interchange::interchange): Update for change to
8259         find_loop_location.
8260         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
8261         change in return-type of find_loop_location.
8262         (graphite_regenerate_ast_isl): Likewise.
8263         * graphite-optimize-isl.c (optimize_isl): Likewise.
8264         * graphite.c (graphite_transform_loops): Likewise.
8265         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
8266         pass in a dump_location_t rather than a location_t, via the
8267         gimple stmt.
8268         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8269         * ipa.c (walk_polymorphic_call_targets): Likewise.
8270         * loop-unroll.c (report_unroll): Convert "locus" param from
8271         location_t to dump_location_t.
8272         (decide_unrolling): Update for change to get_loop_location's
8273         return type.
8274         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
8275         location_t to dump_user_location_t.
8276         (grid_find_single_omp_among_assignments_1): Updates calls to
8277         dump_printf_loc to pass in a dump_location_t rather than a
8278         location_t, via the gimple stmt.
8279         (grid_parallel_clauses_gridifiable): Convert "tloc" from
8280         location_t to dump_location_t.  Updates calls to dump_printf_loc
8281         to pass in a dump_location_t rather than a location_t, via the
8282         gimple stmt.
8283         (grid_inner_loop_gridifiable_p): Likewise.
8284         (grid_dist_follows_simple_pattern): Likewise.
8285         (grid_gfor_follows_tiling_pattern): Likewise.
8286         (grid_target_follows_gridifiable_pattern): Likewise.
8287         (grid_attempt_target_gridification): Convert initialization
8288         of local "grid" from memset to zero-initialization; FIXME: does
8289         this require C++11?  Update call to dump_printf_loc to pass in a
8290         optinfo_location rather than a location_t, via the gimple stmt.
8291         * profile.c (read_profile_edge_counts): Updates call to
8292         dump_printf_loc to pass in a dump_location_t rather than a
8293         location_t
8294         (compute_branch_probabilities): Likewise.
8295         * selftest-run-tests.c (selftest::run_tests): Call
8296         dumpfile_c_tests.
8297         * selftest.h (dumpfile_c_tests): New decl.
8298         * tree-loop-distribution.c (pass_loop_distribution::execute):
8299         Update for change in return type of find_loop_location.
8300         * tree-parloops.c (parallelize_loops): Likewise.
8301         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
8302         "locus" from location_t to dump_user_location_t.
8303         (canonicalize_loop_induction_variables): Likewise.
8304         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8305         for change in return type of find_loop_location.
8306         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8307         to dump_printf_loc to pass in a dump_location_t rather than a
8308         location_t, via the stmt.
8309         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8310         Likewise.
8311         * tree-vect-loop-manip.c (find_loop_location): Convert return
8312         type from source_location to dump_user_location_t.
8313         (vect_do_peeling): Update for above change.
8314         (vect_loop_versioning): Update for change in type of
8315         vect_location.
8316         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8317         from location_t to dump_user_location_t.
8318         (vect_estimate_min_profitable_iters): Update for change in type
8319         of vect_location.
8320         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8321         location_t to dump_location_t.
8322         (vect_slp_bb): Update for change in type of vect_location.
8323         * tree-vectorizer.c (vect_location): Convert from source_location
8324         to dump_user_location_t.
8325         (try_vectorize_loop_1): Update for change in vect_location's type.
8326         (vectorize_loops): Likewise.
8327         (increase_alignment): Likewise.
8328         * tree-vectorizer.h (vect_location): Convert from source_location
8329         to dump_user_location_t.
8330         (find_loop_location): Convert return type from source_location to
8331         dump_user_location_t.
8332         (check_reduction_path): Convert 1st param from location_t to
8333         dump_user_location_t.
8334         * value-prof.c (check_counter): Update call to dump_printf_loc to
8335         pass in a dump_user_location_t rather than a location_t; update
8336         call to error_at for change in type of "locus".
8337         (check_ic_target): Update call to dump_printf_loc to
8338         pass in a dump_user_location_t rather than a location_t, via the
8339         call_stmt.
8340
8341 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
8342
8343         * config/s390/s390.h (enum processor_flags): Do not use
8344         default tune parameter when -march was specified.
8345
8346 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
8347
8348         PR target/86314
8349         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8350         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8351         operands.
8352
8353 2018-06-26  Richard Biener  <rguenther@suse.de>
8354
8355         PR tree-optimization/86287
8356         PR bootstrap/86316
8357         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8358         (vect_analyze_loop): Initialize n_stmts.
8359
8360 2018-06-26  Richard Biener  <rguenther@suse.de>
8361
8362         PR middle-end/86271
8363         * fold-const.c (fold_convertible_p): Pointer extension
8364         isn't valid.
8365
8366 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
8367
8368         PR debug/86064
8369         * dwarf2out.c (loc_list_has_views): Adjust comments.
8370         (dw_loc_list): Split single cross-partition range with
8371         nonzero locview.
8372
8373 2018-06-25  Jeff Law  <law@redhat.com>
8374
8375         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8376         on -mbig-switch by default.
8377
8378         * config/v850/predicates.md (const_float_1_operand): Fix match_code
8379         test.
8380         (const_float_0_operand): Remove unused predicate.
8381         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8382         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8383         (recipsf2): New expander.  Original pattern now called
8384         (recipsf2_insn).
8385         (recipdf2, recipdf2_insn): Similarly.
8386         (rsqrtsf2, rsqrtsf2_insn): Similarly
8387         (rsqrtdf2, rsqrtdf2_insn): Similarly
8388
8389 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
8390
8391         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8392         Simplify logic for FreeBSD (twice).
8393
8394 2018-06-25  Martin Sebor  <msebor@redhat.com>
8395
8396         PR tree-optimization/86204
8397         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8398         a strnlen result if it's less than the length of the string.
8399
8400 2018-06-25  Martin Sebor  <msebor@redhat.com>
8401
8402         PR tree-optimization/85700
8403         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8404         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8405         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8406
8407 2018-06-25  Martin Sebor  <msebor@redhat.com>
8408
8409         * doc/extend.texi (Zero-length arrays): Update and clarify.
8410
8411 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
8412
8413         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8414         added IEEE/IBM long double multilib support on PowerPC little
8415         endian Linux systems.
8416         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8417         (MULTILIB_DEFAULTS): Likewise.
8418         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8419         Likewise.
8420         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8421         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8422         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8423
8424 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
8425
8426         PR middle-end/86311
8427         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8428         (REORDER_45): Likewise.
8429
8430 2018-06-25  Jeff Law  <law@redhat.com>
8431
8432         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8433         dividend to 32 bits.  Adjust length.
8434         (udivmodhi4): Cleanup output template.  Fix length.
8435
8436 2018-06-25  Carl Love  <cel@us.ibm.com>
8437
8438         * config/rs6000/vsx.md: Change word selector to prefered location.
8439
8440 2018-06-25  Richard Biener  <rguenther@suse.de>
8441
8442         PR tree-optimization/86304
8443         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8444         epilogue-if-converted loops as well.
8445
8446 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
8447
8448         * lto-section-out.c (lto_begin_section): Do not print section
8449         name for noaddr and unnumbered dumps.
8450
8451 2018-06-25  Richard Biener  <rguenther@suse.de>
8452
8453         * tree-vectorizer.h (struct vec_info_shared): New structure
8454         with parts split out from struct vec_info and loop_nest from
8455         struct _loop_vec_info.
8456         (struct vec_info): Adjust accordingly.
8457         (struct _loop_vec_info): Likewise.
8458         (LOOP_VINFO_LOOP_NEST): Adjust.
8459         (LOOP_VINFO_DATAREFS): Likewise.
8460         (LOOP_VINFO_DDRS): Likewise.
8461         (struct _bb_vec_info): Likewise.
8462         (BB_VINFO_DATAREFS): Likewise.
8463         (BB_VINFO_DDRS): Likewise.
8464         (struct _stmt_vec_info): Add dr_aux member.
8465         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8466         (DR_MISALIGNMENT_UNINITIALIZED): New.
8467         (set_dr_misalignment): Adjust.
8468         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8469         (vect_analyze_loop): Adjust prototype.
8470         (vect_analyze_loop_form): Likewise.
8471         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8472         Compute dependences lazily.
8473         (vect_record_base_alignments): Use shared datarefs/ddrs.
8474         (vect_verify_datarefs_alignment): Likewise.
8475         (vect_analyze_data_refs_alignment): Likewise.
8476         (vect_analyze_data_ref_accesses): Likewise.
8477         (vect_analyze_data_refs): Likewise.
8478         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8479         constructor parameter for shared part.
8480         (vect_analyze_loop_form): Pass in shared part and adjust.
8481         (vect_analyze_loop_2): Pass in storage for the number of
8482         stmts.  Move loop nest finding to the caller.  Compute
8483         datarefs lazily.
8484         (vect_analyze_loop): Pass in shared part.
8485         (vect_transform_loop): Verify shared datarefs are unchanged.
8486         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8487         constructor parameter for shared part.
8488         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8489         (vect_slp_bb): Verify shared datarefs are unchanged before
8490         transform.
8491         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8492         change.
8493         (new_stmt_vec_info): Initialize DR_AUX misalignment to
8494         DR_MISALIGNMENT_UNINITIALIZED.
8495         * tree-vectorizer.c (vec_info::vec_info): Add constructor
8496         parameter for shared part.
8497         (vec_info::~vec_info): Adjust.
8498         (vec_info_shared::vec_info_shared): New.
8499         (vec_info_shared::~vec_info_shared): Likewise.
8500         (vec_info_shared::save_datarefs): Likewise.
8501         (vec_info_shared::check_datarefs): Likewise.
8502         (try_vectorize_loop_1): Construct shared part live for analyses
8503         of a single loop for multiple vector sizes.
8504         * tree-parloops.c (gather_scalar_reductions): Adjust.
8505
8506 2018-06-25  Richard Biener  <rguenther@suse.de>
8507
8508         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8509         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8510         (vect_analyze_data_refs): Remove similar code from here and
8511         simplify accordingly.
8512
8513 2018-06-25  Richard Biener  <rguenther@suse.de>
8514
8515         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8516         for reverse storage order accesses rather than asserting
8517         they cannot happen here.
8518
8519 2018-06-25  Tom de Vries  <tdevries@suse.de>
8520
8521         PR debug/86257
8522         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8523         Use data16 instead of .byte for insn prefix.
8524
8525 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
8526
8527         PR C++/86082
8528         * parser.c (make_char_string_pack): Pass this literal chars
8529         through cpp_interpret_string.
8530         (cp_parser_userdef_numeric_literal): Check the result of
8531         make_char_string_pack.
8532
8533 2018-06-24  Maya Rashish  <coypu@sdf.org>
8534
8535         * ginclude/stddef.h: Simplify conditions around avoiding
8536         re-definition of __size_t.
8537
8538 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8539
8540         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8541         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8542
8543 2018-06-22  Maya Rashish  <coypu@sdf.org>
8544
8545         * doc/invoke.texi (mno-fancy-math-387): Update for changes
8546         made to OpenBSD and NetBSD through the years.
8547
8548 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8549
8550         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8551         behavior of vec_pack (vector double, vector double) to match
8552         behavior of vec_float2 (vector double, vector double).
8553
8554 2018-06-22  Olivier Hainque  <hainque@adacore.com>
8555
8556         * gimplify.c (gimplify_function_tree): Prevent creation
8557         of a trampoline for the address of the current function
8558         passed to entry/exit instrumentation hooks.
8559
8560 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
8561
8562         PR target/86222
8563         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8564         correctly.
8565
8566 2018-06-22  Martin Liska  <mliska@suse.cz>
8567
8568         PR tree-optimization/86263
8569         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8570         Bail out if is_enabled is false.
8571         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8572         New declaration.
8573         (jump_table_cluster::is_enabled): New function.
8574
8575 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8576
8577         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8578         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8579         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8580         (lto_input_ts_binfo_tree_pointers): Likewise.
8581         * tree-streamer-out.c (streamer_write_tree_bitfields,
8582         write_ts_binfo_tree_pointers): Likewise.
8583         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8584
8585 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8586
8587         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8588
8589 2018-06-22  Martin Liska  <mliska@suse.cz>
8590
8591         * symbol-summary.h (get): Make it pure and inline move
8592         functionality from ::get function.
8593         (get): Remove and inline into ::get and ::get_create.
8594         (get_create): Move code from ::get function.
8595
8596 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8597
8598         PR target/85994
8599         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8600         -x assembler-with-cpp.
8601
8602 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8603
8604         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8605         _FILE_OFFSET_BITS=64 for C++.
8606
8607 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
8608
8609         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8610         conversion insn that shows up when pr85657-3.c is compiled using
8611         IEEE 128-bit long double.
8612         (neg<mode>2_internal): Use the correct mode to check whether the
8613         mode is IBM extended.
8614         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8615         multiply and divide external functions from being created more
8616         than once.
8617
8618 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8619
8620         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8621         functions.
8622         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
8623         the edge can be forwarded.
8624         (cfg_layout_merge_blocks): Likewise.
8625
8626 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8627
8628         * except.c (finish_eh_generation): Commit edge insertions only after
8629         the EH edges have been redirected from post-landing to landing pads.
8630
8631 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8632
8633         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8634         create_tmp_var_for to create the FRAME decl.
8635         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8636
8637 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8638
8639         * tree-inline.c (copy_edges_for_bb): Minor tweak.
8640         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8641         debug statement when resetting its value.
8642         (expand_call_inline): Copy the locus of the call onto the assignment
8643         of the return value, if any.  Use local variable in more cases.
8644
8645 2018-06-21  Martin Liska  <mliska@suse.cz>
8646
8647         * ipa-pure-const.c (propagate_nothrow): Use
8648         funct_state_summaries->get.
8649         (dump_malloc_lattice): Likewise.
8650         (propagate_malloc): Likewise.
8651
8652 2018-06-21  Richard Biener  <rguenther@suse.de>
8653
8654         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8655         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8656         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8657         comment.
8658         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8659         BLOCK_ABSTRACT_ORIGIN unconditionally.
8660
8661 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
8662
8663         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8664         deleting it.
8665         * ipa-reference.c (ipa_reference_c_finalize): Delete
8666         ipa_ref_opt_sum_summaries and set it to NULL.
8667
8668 2018-06-21  Tom de Vries  <tdevries@suse.de>
8669
8670         PR tree-optimization/85859
8671         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8672         test with comment from bb_no_side_effects_p.
8673
8674 2018-06-21  Richard Biener  <rguenther@suse.de>
8675
8676         PR tree-optimization/86232
8677         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8678         max for constant niter.
8679
8680 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8681
8682         * config/aarch64/aarch64-simd.md
8683         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8684
8685 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8686
8687         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8688         Make opernads of the unspec commutative.
8689
8690 2018-06-21  Richard Biener  <rguenther@suse.de>
8691
8692         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8693         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8694         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8695         (vect_analyze_data_ref_dependence): Re-order checks to deal with
8696         NULL DR_STEP.
8697         (vect_record_base_alignments): Do not record base alignment
8698         for gathers or scatters.
8699         (vect_compute_data_ref_alignment): Drop return value that is always
8700         true.  Bail out early for gathers or scatters.
8701         (vect_enhance_data_refs_alignment): Bail out early for gathers
8702         or scatters.
8703         (vect_find_same_alignment_drs): Likewise.
8704         (vect_analyze_data_refs_alignment): Remove dead code.
8705         (vect_slp_analyze_and_verify_node_alignment): Likewise.
8706         (vect_analyze_data_refs): For possible gathers or scatters do
8707         not create an alternate DR, just check their possible validity
8708         and mark them.  Adjust DECL_NONALIASED handling to not rely
8709         on DR_BASE_ADDRESS.
8710         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8711         update inits of gathers or scatters.
8712         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8713         Also copy gather/scatter flag to pattern vinfo.
8714
8715 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8716
8717         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8718         behavior of vec_packsu (vector unsigned long long, vector unsigned
8719         long long) to match behavior of vec_packs with same signature.
8720
8721 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
8722             Thomas Schwinge <thomas@codesourcery.com>
8723             Cesar Philippidis  <cesar@codesourcery.com>
8724
8725         * gimplify.c (gimplify_scan_omp_clauses): Add support for
8726         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8727         (gimplify_adjust_omp_clauses): Likewise.
8728         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8729         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8730         (gimplify_omp_target_update): Update handling of acc update and
8731         enter/exit data.
8732         * omp-low.c (install_var_field): Remove unused parameter
8733         base_pointers_restrict.
8734         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8735         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8736         FINALIZE}
8737         (omp_target_base_pointers_restrict_p): Delete.
8738         (scan_omp_target): Update call to scan_sharing_clauses.
8739         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8740         FINALIZE}.
8741         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8742         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8743         (convert_local_omp_clauses): Likewise.
8744         * tree-pretty-print.c (dump_omp_clause): Likewise.
8745         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
8746         FINALIZE}.
8747         (omp_clause_code_name): Likewise.
8748
8749 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
8750
8751         PR debug/86194
8752         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8753         be narrowed.
8754
8755         PR tree-optimization/86231
8756         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
8757         anti-range don't overwrite *vr0min before using it to compute *vr0max.
8758
8759 2018-06-20  Tom de Vries  <tdevries@suse.de>
8760
8761         PR tree-optimization/86097
8762         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8763         iv type if signedness of iv type is not the same as that of *nit.
8764
8765 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
8766
8767         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
8768         EDGE_EH edges, verify they are all EDGE_EH.
8769
8770 2018-06-20  Maya Rashish  <coypu@sdf.org>
8771
8772         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8773
8774 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8775
8776         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8777         * config/aarch64/aarch64.c (xgene1_tunings): Add
8778         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8779         (aarch64_mode_valid_for_sched_fusion_p):
8780         Allow 16-byte modes.
8781         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8782         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8783         128-bit modes.
8784         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8785         New pattern.
8786         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8787         * config/aarch64/iterators.md (VQ2): New mode iterator.
8788
8789 2018-06-20  Martin Liska  <mliska@suse.cz>
8790
8791         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8792         Change default ratio from 10 to 8.
8793
8794 2018-06-20  Martin Liska  <mliska@suse.cz>
8795
8796         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8797         New.
8798         (bit_test_cluster::find_bit_tests): Likewise.
8799         (switch_decision_tree::analyze_switch_statement): Find clusters.
8800         * tree-switch-conversion.h (struct jump_table_cluster): Document
8801         hierarchy.
8802
8803 2018-06-20  Martin Liska  <mliska@suse.cz>
8804
8805         * tree-switch-conversion.c (switch_conversion::collect):
8806         Record m_uniq property.
8807         (switch_conversion::expand): Bail out for special conditions.
8808         (group_cluster::~group_cluster): New.
8809         (group_cluster::group_cluster): Likewise.
8810         (group_cluster::dump): Likewise.
8811         (jump_table_cluster::emit): New.
8812         (switch_decision_tree::fix_phi_operands_for_edges): New.
8813         (struct case_node): Remove struct.
8814         (jump_table_cluster::can_be_handled): New.
8815         (case_values_threshold): Moved to header.
8816         (reset_out_edges_aux): Likewise.
8817         (jump_table_cluster::is_beneficial): New.
8818         (bit_test_cluster::can_be_handled): Likewise.
8819         (add_case_node): Remove.
8820         (bit_test_cluster::is_beneficial): New.
8821         (case_bit_test::cmp): New.
8822         (bit_test_cluster::emit): New.
8823         (expand_switch_as_decision_tree_p): Remove.
8824         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8825         (fix_phi_operands_for_edge): Likewise.
8826         (switch_decision_tree::analyze_switch_statement): New.
8827         (compute_cases_per_edge): Move ...
8828         (switch_decision_tree::compute_cases_per_edge): ... here.
8829         (try_switch_expansion): Likewise.
8830         (switch_decision_tree::try_switch_expansion): Likewise.
8831         (record_phi_operand_mapping): Likewise.
8832         (switch_decision_tree::record_phi_operand_mapping): Likewise.
8833         (emit_case_decision_tree): Likewise.
8834         (switch_decision_tree::emit): Likewise.
8835         (balance_case_nodes): Likewise.
8836         (switch_decision_tree::balance_case_nodes): Likewise.
8837         (dump_case_nodes): Likewise.
8838         (switch_decision_tree::dump_case_nodes): Likewise.
8839         (emit_jump): Likewise.
8840         (switch_decision_tree::emit_jump): Likewise.
8841         (emit_cmp_and_jump_insns): Likewise.
8842         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8843         (emit_case_nodes): Likewise.
8844         (switch_decision_tree::emit_case_nodes): Likewise.
8845         (conditional_probability): Remove.
8846         * tree-switch-conversion.h (enum cluster_type): New.
8847         (PRINT_CASE): New.
8848         (struct cluster): Likewise.
8849         (cluster::cluster): Likewise.
8850         (struct simple_cluster): Likewise.
8851         (simple_cluster::simple_cluster): Likewise.
8852         (struct group_cluster): Likewise.
8853         (struct jump_table_cluster): Likewise.
8854         (struct bit_test_cluster): Likewise.
8855         (struct min_cluster_item): Likewise.
8856         (struct case_tree_node): Likewise.
8857         (case_tree_node::case_tree_node): Likewise.
8858         (jump_table_cluster::case_values_threshold): Likewise.
8859         (struct case_bit_test): Likewise.
8860         (struct switch_decision_tree): Likewise.
8861         (struct switch_conversion): Likewise.
8862         (switch_decision_tree::reset_out_edges_aux): Likewise.
8863
8864 2018-06-20  Martin Liska  <mliska@suse.cz>
8865
8866         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8867         (hoist_edge_and_branch_if_true): Likewise.
8868         (expand_switch_using_bit_tests_p): Likewise.
8869         (struct case_bit_test): Likewise.
8870         (case_bit_test_cmp): Likewise.
8871         (emit_case_bit_tests): Likewise.
8872         (switch_conversion::switch_conversion): New class.
8873         (struct switch_conv_info): Remove old struct.
8874         (collect_switch_conv_info): More to ...
8875         (switch_conversion::collect): ... this.
8876         (check_range): Likewise.
8877         (switch_conversion::check_range): Likewise.
8878         (check_all_empty_except_final): Likewise.
8879         (switch_conversion::check_all_empty_except_final): Likewise.
8880         (check_final_bb): Likewise.
8881         (switch_conversion::check_final_bb): Likewise.
8882         (create_temp_arrays): Likewise.
8883         (switch_conversion::create_temp_arrays): Likewise.
8884         (free_temp_arrays): Likewise.
8885         (gather_default_values): Likewise.
8886         (switch_conversion::gather_default_values): Likewise.
8887         (build_constructors): Likewise.
8888         (switch_conversion::build_constructors): Likewise.
8889         (constructor_contains_same_values_p): Likewise.
8890         (switch_conversion::contains_same_values_p): Likewise.
8891         (array_value_type): Likewise.
8892         (switch_conversion::array_value_type): Likewise.
8893         (build_one_array): Likewise.
8894         (switch_conversion::build_one_array): Likewise.
8895         (build_arrays): Likewise.
8896         (switch_conversion::build_arrays): Likewise.
8897         (gen_def_assigns): Likewise.
8898         (switch_conversion::gen_def_assigns): Likewise.
8899         (prune_bbs): Likewise.
8900         (switch_conversion::prune_bbs): Likewise.
8901         (fix_phi_nodes): Likewise.
8902         (switch_conversion::fix_phi_nodes): Likewise.
8903         (gen_inbound_check): Likewise.
8904         (switch_conversion::gen_inbound_check): Likewise.
8905         (process_switch): Use the newly created class.
8906         (switch_conversion::expand): New.
8907         (switch_conversion::~switch_conversion): New.
8908         * tree-switch-conversion.h: New file.
8909
8910 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8911
8912         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8913         tree-vect-patterns.c.
8914         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8915         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
8916         (vect_recog_sad_pattern): Likewise.
8917         (vect_recog_widen_sum_pattern): Likewise.
8918         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
8919         (vect_recog_widen_shift_pattern): Remove the type_in argument.
8920         (vect_recog_rotate_pattern): Likewise.
8921         (vect_recog_mult_pattern): Likewise.
8922         (vect_recog_vector_vector_shift_pattern): Likewise.
8923         (vect_recog_divmod_pattern): Likewise.
8924         (vect_recog_mixed_size_cond_pattern): Likewise.
8925         (vect_recog_bool_pattern): Likewise.
8926         (vect_recog_mask_conversion_pattern): Likewise.
8927         (vect_try_gather_scatter_pattern): Likewise.
8928         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
8929         (vect_recog_over_widening_pattern): Likewise.
8930         (vect_recog_gather_scatter_pattern): Likewise.
8931         (vect_recog_func_ptr): Move from tree-vectorizer.h
8932         (vect_vect_recog_func_ptrs): Move further down the file.
8933         (vect_recog_func): Likewise.  Remove the third argument.
8934         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8935         (vect_pattern_recog_1): Expect the pattern function to do any
8936         necessary target tests.  Also expect it to provide a vector type.
8937         Remove the type_in handling.
8938
8939 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8940
8941         * tree-vect-patterns.c (vect_pattern_detected): New function.
8942         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8943         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8944         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8945         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8946         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8947         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8948         (vect_recog_mask_conversion_pattern)
8949         (vect_try_gather_scatter_pattern): Likewise.
8950
8951 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8952
8953         * tree-vect-patterns.c (vect_get_internal_def): New function.
8954         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8955         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8956         (search_type_for_mask_1): Use it.
8957
8958 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8959
8960         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8961         redundant WIDEN_SUM_EXPR handling.
8962         (vect_recog_sad_pattern): Likewise.
8963
8964 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8965
8966         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8967         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8968         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8969         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8970         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8971
8972 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8973
8974         * tree-vect-stmts.c (vectorizable_call): Make sure that we
8975         use the stmt_vec_info of the original bb statement for the
8976         new zero assignment, even if the call is part of a pattern.
8977
8978 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8979
8980         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8981         that the sequence is attached to the original statement rather
8982         than the pattern statement.
8983         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8984         PATTERN_DEF_SEQ from the original statement rather than
8985         the main pattern statement.
8986         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8987         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8988         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8989
8990 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8991
8992         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8993         definition statements before the early exit for statements that aren't
8994         live or relevant.
8995         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8996         split out from...
8997         (vect_transform_loop): ...here.  Process pattern definition
8998         statements without first checking whether the main pattern
8999         statement is live or relevant.
9000
9001 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
9002
9003         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
9004         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
9005
9006 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
9007
9008         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
9009         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
9010         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
9011         (expand_block_compare): Change select_block_compare_mode call.
9012         (expand_strncmp_align_check): Use new functions, fix comment.
9013         (emit_final_str_compare_gpr): New function.
9014         (expand_strn_compare): Refactor and clean up code.
9015         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
9016
9017 2018-06-19  Tony Reix  <tony.reix@atos.com>
9018             Damien Bergamini  <damien.bergamini@atos.com>
9019             David Edelsohn  <dje.gcc@gmail.com>
9020
9021         * collect2.c (static_obj): New variable.
9022         (static_libs): New variable.
9023         (is_in_list): Uncomment declaration.
9024         (main): Track AIX libraries linked statically.
9025         (is_in_list): Uncomment definition.
9026         (scan_prog_file): Don't add AIX shared libraries initializer
9027         to constructor list if linking statically.
9028
9029 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
9030
9031         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
9032         constant.
9033         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9034
9035 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
9036
9037         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9038         blocks.
9039
9040 2018-06-19  Martin Liska  <mliska@suse.cz>
9041
9042         * config/i386/i386.c (ix86_can_inline_p): Do not use
9043         ipa_fn_summaries::get_create.
9044         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9045         get.
9046         (devirtualization_time_bonus): Likewise.
9047         (ipcp_propagate_stage): Likewise.
9048         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9049         (edge_set_predicate): Likewise.
9050         (evaluate_conditions_for_known_args): Likewise.
9051         (evaluate_properties_for_edge): Likewise.
9052         (ipa_call_summary::reset): Tranform to ...
9053         (ipa_call_summary::~ipa_call_summary): ... this.
9054         (ipa_fn_summary::reset): Transform to ...
9055         (ipa_fn_summary::~ipa_fn_summary): ... this.
9056         (ipa_fn_summary_t::remove): Rename to ...
9057         (ipa_fn_summary_t::remove_callees): ... this.
9058         (ipa_fn_summary_t::duplicate): Use placement new
9059         instead of memory copy.
9060         (ipa_call_summary_t::duplicate): Likewise.
9061         (ipa_call_summary_t::remove): Remove.
9062         (dump_ipa_call_summary): Change get_create to get.
9063         (ipa_dump_fn_summary): Dump only when summary exists.
9064         (analyze_function_body): Use symbol_summary::get instead
9065         of get_create.
9066         (compute_fn_summary): Likewise.
9067         (estimate_edge_devirt_benefit): Likewise.
9068         (estimate_edge_size_and_time): Likewise.
9069         (inline_update_callee_summaries): Likewise.
9070         (remap_edge_change_prob): Likewise.
9071         (remap_edge_summaries): Likewise.
9072         (ipa_merge_fn_summary_after_inlining): Likewise.
9073         (write_ipa_call_summary): Likewise.
9074         (ipa_fn_summary_write): Likewise.
9075         (ipa_free_fn_summary): Likewise.
9076         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9077         (struct ipa_call_summary): Likewise.
9078         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9079         of get_create.
9080         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9081         (estimate_size_after_inlining): Likewise.
9082         (estimate_growth): Likewise.
9083         (growth_likely_positive): Likewise.
9084         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9085         (inline_call): Likewise.
9086         * ipa-inline.c (caller_growth_limits): Likewise.
9087         (can_inline_edge_p): Likewise.
9088         (can_inline_edge_by_limits_p): Likewise.
9089         (compute_uninlined_call_time): Likewise.
9090         (compute_inlined_call_time): Likewise.
9091         (want_inline_small_function_p): Likewise.
9092         (edge_badness): Likewise.
9093         (update_caller_keys): Likewise.
9094         (update_callee_keys): Likewise.
9095         (inline_small_functions): Likewise.
9096         (inline_to_all_callers_1): Likewise.
9097         (dump_overall_stats): Likewise.
9098         (early_inline_small_functions): Likewise.
9099         (early_inliner): Likewise.
9100         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9101         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9102         * ipa-pure-const.c (malloc_candidate_p): Likewise.
9103         * ipa-split.c (execute_split_functions): Likewise.
9104         * symbol-summary.h: Likewise.
9105         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9106
9107 2018-06-19  Richard Biener  <rguenther@suse.de>
9108
9109         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9110         (vectorize_loops): ... here.  Fix dbgcnt handling.
9111         (try_vectorize_loop): Wrap try_vectorize_loop_1.
9112
9113 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
9114
9115         PR target/86197
9116         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9117         ieee128 argument takes up only one (vector) register, not two (floating
9118         point) registers.
9119
9120 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
9121
9122         * gimplify.c (gimplify_init_constructor): Really never clear for an
9123         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9124
9125 2018-06-19  Richard Biener  <rguenther@suse.de>
9126
9127         PR tree-optimization/86179
9128         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9129         after failed recognition.
9130
9131 2018-06-18  Martin Sebor  <msebor@redhat.com>
9132
9133         PR middle-end/85602
9134         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9135         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9136         Handle integer subtraction.
9137         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9138         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9139
9140 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
9141
9142         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9143         param from rtx to rtx_insn *.
9144         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9145         param.
9146         (frv_ifcvt_modify_insn): Likwise.
9147         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9148         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
9149         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9150         as_a <rtx_insn *> cast to local "unprotected_region" once
9151         it's been established that it's not NULL or pc_rtx.
9152         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9153         param "sethi" from rtx to rtx_insn *.
9154         (nds32_group_float_insns): Likewise for param "insn".
9155         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9156         param.
9157         (vax_output_int_subtract): Likewise.
9158         * config/vax/vax.c (vax_output_int_add): Likewise for param
9159         "insn".
9160         (vax_output_int_subtract): Likewise.
9161         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9162         (emit_pattern_after): Likewise for param "after".
9163         (emit_insn_after): Likewise.
9164         (emit_jump_insn_after): Likewise.
9165         (emit_call_insn_after): Likewise.
9166         (emit_debug_insn_after): Likewise.
9167         (emit_pattern_before): Likewise for param "before".
9168         (emit_insn_before): Likewise.
9169         (emit_jump_insn_before): Likewise.
9170         * final.c (get_insn_template): Likewise for param "insn", removing
9171         a cast.
9172         * output.h (get_insn_template): Likewise for 2nd param.
9173         * rtl.h (emit_insn_before): Likewise.
9174         (emit_jump_insn_before): Likewise.
9175         (emit_debug_insn_before_noloc): Likewise.
9176         (emit_insn_after): Likewise.
9177         (emit_jump_insn_after): Likewise.
9178         (emit_call_insn_after): Likewise.
9179         (emit_debug_insn_after): Likewise.
9180         (set_insn_deleted): Likewise for param.
9181
9182 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
9183
9184         PR target/85358
9185         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9186         floating point modes, so that IFmode is numerically greater than
9187         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9188         to declare the ordering.  This prevents IFmode from being
9189         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9190         machine.  Include rs6000-modes.h to share the fractional values
9191         between genmodes* and the rest of the compiler.
9192         (IFmode): Likewise.
9193         (KFmode): Likewise.
9194         (TFmode): Likewise.
9195         * config/rs6000/rs6000-modes.h: New file.
9196         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9197         meaning of rs6000_long_double_size so that 126..128 selects an
9198         appropriate 128-bit floating point type.
9199         (rs6000_option_override_internal): Likewise.
9200         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9201         (TARGET_LONG_DOUBLE_128): Change the meaning of
9202         rs6000_long_double_size so that 126..128 selects an appropriate
9203         128-bit floating point type.
9204         (LONG_DOUBLE_TYPE_SIZE): Update comment.
9205         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9206         source and destination to match the standard usage.
9207         (truncifkf2): Likewise.
9208         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9209         ISA 2.07 to use an explicit clobber, instead of passing in a
9210         temporary.
9211         (copysign<mode>3_soft): Likewise.
9212
9213 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
9214
9215         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9216         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9217         (vect_slp_analyze_instance_dependence): Likewise.
9218         (vect_enhance_data_refs_alignment): Likewise.
9219         (vect_analyze_data_refs_alignment): Likewise.
9220         (vect_slp_analyze_and_verify_instance_alignment
9221         (vect_analyze_data_ref_accesses): Likewise.
9222         (vect_prune_runtime_alias_test_list): Likewise.
9223         (vect_analyze_data_refs): Likewise.
9224         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9225         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9226         (vect_analyze_scalar_cycles_1): Likewise.
9227         (vect_get_loop_niters): Likewise.
9228         (vect_analyze_loop_form_1): Likewise.
9229         (vect_update_vf_for_slp): Likewise.
9230         (vect_analyze_loop_operations): Likewise.
9231         (vect_analyze_loop): Likewise.
9232         (vectorizable_induction): Likewise.
9233         (vect_transform_loop): Likewise.
9234         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9235         * tree-vect-slp.c (vect_analyze_slp): Likewise.
9236         (vect_make_slp_decision): Likewise.
9237         (vect_detect_hybrid_slp): Likewise.
9238         (vect_slp_analyze_operations): Likewise.
9239         (vect_slp_bb): Likewise.
9240         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
9241         (vectorizable_bswap): Likewise.
9242         (vectorizable_call): Likewise.
9243         (vectorizable_simd_clone_call): Likewise.
9244         (vectorizable_conversion): Likewise.
9245         (vectorizable_assignment): Likewise.
9246         (vectorizable_shift): Likewise.
9247         (vectorizable_operation): Likewise.
9248         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
9249
9250 2018-06-18  Martin Sebor  <msebor@redhat.com>
9251
9252         PR tree-optimization/81384
9253         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
9254         * builtins.c (expand_builtin_strnlen): New function.
9255         (expand_builtin): Call it.
9256         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
9257         * builtins.def (BUILT_IN_STRNLEN): New.
9258         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
9259         Warn for bounds in excess of maximum object size.
9260         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
9261         single-value ranges.  Handle strnlen.
9262         (handle_builtin_strlen): Handle strnlen.
9263         (strlen_check_and_optimize_stmt): Same.
9264         * doc/extend.texi (Other Builtins): Document strnlen.
9265
9266 2018-06-18  Maya Rashish  <coypu@sdf.org>
9267
9268         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
9269         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
9270         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
9271
9272         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
9273         here to ...
9274         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
9275
9276 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
9277
9278         * tree.c (escaped_string::escape): Replace cast to char * by
9279         const_cast<char *> (unescaped).
9280
9281 2018-06-18  Nick Clifton  <nickc@redhat.com>
9282
9283         PR 84195
9284         * tree.c (escaped_string): New class.  Converts an unescaped
9285         string into its escaped equivalent.
9286         (warn_deprecated_use): Use the new class to convert the
9287         deprecation message, if present.
9288         (test_escaped_strings): New self test.
9289         (test_c_tests): Add test_escaped_strings.
9290         * doc/extend.texi (deprecated): Add a note that the
9291         deprecation message is affected by the -fmessage-length
9292         option, and that control characters will be escaped.
9293         (#pragma GCC error): Document this pragma.
9294         (#pragma GCC warning): Likewise.
9295         * doc/invoke.texi (-fmessage-length): Document this option's
9296         effect on the #warning and #error preprocessor directives and
9297         the deprecated attribute.
9298
9299 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
9300
9301         * tree.c (decl_value_expr_lookup): Revert latest change.
9302         (decl_value_expr_insert): Likewise.
9303
9304 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
9305
9306         * gimplify.c (nonlocal_vlas): Delete.
9307         (nonlocal_vla_vars): Likewise.
9308         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9309         referenced VLAs.
9310         (gimplify_body): Do not create and destroy nonlocal_vlas.
9311         * tree-nested.c: Include diagnostic.h.
9312         (use_pointer_in_frame): Tweak.
9313         (lookup_field_for_decl): Add assertion and declare the transformation.
9314         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9315         internal error when the reference is in a wrong context.  Do not
9316         create a debug decl by default.
9317         (note_nonlocal_block_vlas): Delete.
9318         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9319         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
9320         create a debug decl by default.
9321         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9322         call is in a wrong context.
9323         (fixup_vla_decls): New function.
9324         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
9325         debug variables were created.
9326         * tree.c (decl_value_expr_lookup): Add checking assertion.
9327         (decl_value_expr_insert): Likewise.
9328
9329 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
9330
9331         PR middle-end/82479
9332         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9333         * tree-scalar-evolution.c (interpret_expr): Likewise.
9334         (expression_expensive_p): Likewise.
9335         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9336         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9337         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9338         (ssa_defined_by_minus_one_stmt_p): New.
9339
9340 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
9341
9342         PR middle-end/64946
9343         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9344         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9345         * dojump.c (do_jump): Likewise.
9346         * expr.c (expand_expr_real_2): Check operand type's sign.
9347         * fold-const.c (const_unop): Handle ABSU_EXPR.
9348         (fold_abs_const): Likewise.
9349         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9350         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9351         (strip_sign_op_1): Likesise.
9352         * match.pd: Add new pattern to generate ABSU_EXPR.
9353         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9354         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9355         * tree-eh.c (operation_could_trap_helper_p): Likewise.
9356         * tree-inline.c (estimate_operator_cost): Likewise.
9357         * tree-pretty-print.c (dump_generic_node): Likewise.
9358         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9359         * tree.def (ABSU_EXPR): New.
9360
9361 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
9362
9363         PR middle-end/86095
9364         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9365         documented as preserved for backward compatibility only.
9366         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9367
9368         PR rtl-optimization/86108
9369         * bb-reorder.c (create_forwarder_block): Renamed to ...
9370         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
9371         jump from new landing pad to the second part.
9372         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9373         Adjust callers.
9374
9375 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
9376
9377         PR middle-end/85878
9378         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9379         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9380         Only call store_expr for halves if the mode is the same.
9381
9382         PR middle-end/86123
9383         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9384         Fix up comment formatting.
9385
9386 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9387
9388         * typed-splay-tree.h (typed_splay_tree::remove): New function.
9389         (typed_splay_tree::closure,
9390         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9391         (typed_splay_tree::typed_splay_tree,
9392         typed_splay_tree::operator =): Declared private.
9393         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9394         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9395         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9396         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9397         typed_splay_tree::splay_tree_splay,
9398         typed_splay_tree::splay_tree_foreach_helper,
9399         typed_splay_tree::splay_tree_insert,
9400         typed_splay_tree::splay_tree_remove,
9401         typed_splay_tree::splay_tree_lookup,
9402         typed_splay_tree::splay_tree_predecessor,
9403         typed_splay_tree::splay_tree_successor,
9404         typed_splay_tree::splay_tree_min,
9405         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9406         (typed_splay_tree::root, typed_splay_tree::comp,
9407         typed_splay_tree::delete_key,
9408         typed_splay_tree::delete_value): New data members.
9409         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9410         typed_splay_tree::remove.
9411
9412 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
9413
9414         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9415         -mginv and -mno-ginv to the assembler.
9416         * config/mips/mips.opt (-mcrc): New option.
9417         (-mginv): Likewise.
9418         * doc/invoke.text (-mcrc): Document.
9419         (-mginv): Likewise.
9420
9421 2018-06-15  Nick Clifton  <nickc@redhat.com>
9422
9423         PR 84195
9424         * tree.c (escaped_string): New class.  Converts an unescaped
9425         string into its escaped equivalent.
9426         (warn_deprecated_use): Use the new class to convert the
9427         deprecation message, if present.
9428         (test_escaped_strings): New self test.
9429         (test_c_tests): Add test_escaped_strings.
9430         * doc/extend.texi (deprecated): Add a note that the
9431         deprecation message is affected by the -fmessage-length
9432         option, and that control characters will be escaped.
9433         (#pragma GCC error): Document this pragma.
9434         (#pragma GCC warning): Likewise.
9435         * doc/invoke.texi (-fmessage-length): Document this option's
9436         effect on the #warning and #error preprocessor directives and
9437         the deprecated attribute.
9438
9439 2018-06-15  Richard Biener  <rguenther@suse.de>
9440
9441         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9442         here, also noting vector size used.
9443         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
9444         size used in MSG_OPTIMIZED_LOCATIONS dump.
9445         (pass_slp_vectorize::execute): Adjust.
9446
9447 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
9448
9449         PR target/85968
9450         * config/arc/arc.c (arc_return_address_register): Fix
9451         if-condition.
9452
9453 2018-06-15  Richard Biener  <rguenther@suse.de>
9454
9455         PR middle-end/86159
9456         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9457         leave useless conversion stripping to force_gimple_operand_gsi.
9458         (gimplify_build2): Likewise.
9459         (gimplify_build1): Likewise.
9460
9461 2018-06-15  Richard Biener  <rguenther@suse.de>
9462
9463         PR middle-end/86076
9464         * tree-cfg.c (move_stmt_op): unshare invariant addresses
9465         before adjusting their block.
9466
9467 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9468
9469         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9470         multilibs for *-*-rtems*.
9471         * config/riscv/t-rtems: New file.
9472
9473 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
9474
9475         PR middle-end/86122
9476         * match.pd ((A +- CST1) +- CST2): Punt if last resort
9477         unsigned_type_for returns NULL.
9478
9479         PR target/85945
9480         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9481         subregs of multi-word pseudos unless the float mode has word size.
9482
9483 2018-06-14  Richard Biener  <rguenther@suse.de>
9484
9485         PR middle-end/86139
9486         * tree-vect-generic.c (build_word_mode_vector_type): Remove
9487         duplicate and harmful type_hash_canon.
9488         * tree.c (type_hash_canon): Assert we didn't find ourselves.
9489
9490 2018-06-14  Richard Biener  <rguenther@suse.de>
9491
9492         PR ipa/86124
9493         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9494         NULL cgraph_node.
9495
9496 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9497
9498         * config/rtems.h (STDINT_LONG32): Define.
9499
9500 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
9501             Prachi Godbole  <prachi.godbole@imgtec.com>
9502
9503         * config/mips/mips-cpus.def: Define P6600.
9504         * config/mips/mips-tables.opt: Regenerate.
9505         * config/mips/mips.c (mips_ucbranch_type): New enum.
9506         (mips_rtx_cost_data): Add support for P6600.
9507         (mips_issue_rate): Likewise.
9508         (mips_multipass_dfa_lookahead): Likewise.
9509         (mips_avoid_hazard): Likewise.
9510         (mips_reorg_process_insns): Likewise.
9511         (mips_classify_branch_p6600): New function.
9512         * config/mips/mips.h (TUNE_P6600): New define.
9513         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9514         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9515         * config/mips/mips.md: Include p6600.md.
9516         (processor): Add p6600.
9517         * config/mips/p6600.md: New file.
9518         * doc/invoke.texi: Add p6600 to supported architectures.
9519
9520 2018-06-13  Martin Sebor  <msebor@redhat.com>
9521
9522         PR tree-optimization/86114
9523         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9524         of integer types.
9525         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9526
9527 2018-06-13  Richard Biener  <rguenther@suse.de>
9528
9529         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9530         Properly set vector type of the intermediate stmt.
9531         * tree-vect-stmts.c (vectorizable_operation): The destination
9532         var always has vectype_out type.
9533
9534 2018-06-13  Jeff Law  <law@redhat.com>
9535
9536         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9537         integer 0 for argument to print_rtl_with_bb.
9538         (rl78_reorg): Likewise.
9539
9540 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
9541
9542         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9543         from rtx to rtx_insn *.
9544         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9545         "label".
9546         (add_sched_insns_for_speculation): Likewise for local "target",
9547         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9548         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9549         from rtx_insn ** to rtx_code_label **.
9550         (reorg_emit_nops): Likewise.
9551         (c6x_reorg): Likewise for local "call_labels".
9552         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9553         rtx to rtx_insn *.
9554         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9555         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9556         the loops over LABEL_REFS.
9557         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9558         braf_label.
9559         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9560         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9561         (split_branches): Strengthen local "olabel" from rtx to
9562         rtx_insn *, adding a safe_as_a cast.
9563         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9564         to "rtx_insn *".
9565         (add_insn_after): Likewise for first two params.
9566         (add_insn_before): Likewise.
9567         (remove_insn): Likewise for param.
9568         (emit_pattern_before_noloc): Likewise for second and third params.
9569         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9570         (emit_call_insn_before_noloc): Likewise.
9571         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9572         to "rtx_insn *".
9573         (emit_barrier_before): Likewise.
9574         (emit_label_before): Strengthen "label" param from "rtx" to
9575         "rtx_code_label *".  Strengthen "before" param from "rtx" to
9576         "rtx_insn *".
9577         (emit_insn_after_1): Strengthen "after" param from "rtx" to
9578         "rtx_insn *".
9579         (emit_pattern_after_noloc): Likewise.
9580         (emit_insn_after_noloc): Likewise.
9581         (emit_jump_insn_after_noloc): Likewise.
9582         (emit_call_insn_after_noloc): Likewise.
9583         (emit_debug_insn_after_noloc): Likewise.
9584         (emit_barrier_after): Likewise.
9585         (emit_label_after): Likewise for both params.
9586         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
9587         "loc" param from "int" to "location_t".
9588         (emit_insn_after_setloc): Likewise.
9589         (emit_jump_insn_after_setloc): Likewise.
9590         (emit_call_insn_after_setloc): Likewise.
9591         (emit_debug_insn_after_setloc): Likewise.
9592         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
9593         "loc" param from "int" to "location_t".
9594         (emit_pattern_before): Convert NULL_RTX to NULL.
9595         (emit_insn_before_setloc): Convert "loc" param from "int" to
9596         "location_t".
9597         (emit_jump_insn_before_setloc): Likewise.
9598         (emit_call_insn_before_setloc): Likewise.
9599         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9600         rtx_insn *.  Convert "loc" param from "int" to "location_t".
9601         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9602         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9603         Convert 3rd param from "int" to "location_t".
9604         (emit_barrier_before, emit_barrier_after, next_real_insn):
9605         Strengthen param from rtx to rtx_insn *.
9606         (emit_label_before): Strengthen 1st param from "rtx" to
9607         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
9608         "rtx_insn *".
9609         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9610         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9611         Strengthen 2nd param from "rtx" to "rtx_insn *".
9612         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9613         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9614         Likewise. Convert 3rd param from "int" to "location_t".
9615         (emit_label_after): Strengthen 1st param from "rtx" to
9616         "rtx_code_label *".
9617         (next_real_insn, remove_insn): Strengthen param from "rtx" to
9618         "rtx_insn *".
9619         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9620         from "rtx" to "rtx_insn *".
9621
9622 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
9623
9624         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9625         bodies streamed in with -Q.
9626         * dumpfile.c (dump_files): Add lto-stream-out dump file.
9627         * dumpfile.h (tree_dump_index): Add lto_stream_out.
9628         * gimple-streamer-out.c: Include gimple-pretty-print.h
9629         (output_bb): Dump stmts streamed.
9630         * lto-section-out.c: Include print-tree.h
9631         (lto_begin_section): Dump sections created.
9632         (lto_output_decl_index): Dump decl encoded.
9633         * lto-streamer-out.c: Include print-tree.h
9634         (create_output_block): Dump output block created.
9635         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9636         (output_function): Dump function output.
9637         (output_constructor): Dump constructor streamed.
9638         (write_global_stream): Output indexes encoded.
9639         (produce_asm_for_decls): Dump streams encoded.
9640         * lto-streamer.c (streamer_dump_file): New global var.
9641         * lto-streamer.h (streamer_dump_file): Declare.
9642         * passes.c (ipa_write_summaries): Initialize streamer dump.
9643         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9644         in.
9645
9646 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
9647
9648         PR target/86048
9649         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9650         offsets for register save directives.  Emit a second batch of save
9651         directives, if need be, when the function accesses prior frames.
9652
9653 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9654
9655         * config/arc/fpu.md (fmasf4): Force operand to register.
9656         (fnmasf4): Likewise.
9657
9658 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9659
9660         * config/arc/arc-protos.h (arc_pad_return): Remove.
9661         * config/arc/arc.c (machine_function): Remove force_short_suffix
9662         and size_reason.
9663         (arc_print_operand): Adjust printing of '&'.
9664         (arc_verify_short): Remove conditional printing of short suffix.
9665         (arc_final_prescan_insn): Remove reference to size_reason.
9666         (pad_return): New function.
9667         (arc_reorg): Call pad_return.
9668         (arc_pad_return): Remove.
9669         (arc_init_machine_status): Remove reference to force_short_suffix.
9670         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9671         (attr length): When attribute iscompact is true force to 2
9672         regardless; in the case of maybe check if we want to force the
9673         instruction to have 4 bytes length.
9674         (nopv): Change it to generate 4 byte long nop as well.
9675         (blockage): New pattern.
9676         (simple_return): Remove call to arc_pad_return.
9677         (p_return_i): Likewise.
9678
9679 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9680
9681         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9682
9683 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9684
9685         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9686         ARC cores.
9687
9688 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9689
9690         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9691         for ARC700 and ARCv2.
9692
9693 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
9694
9695         PR target/86076
9696         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9697         operands[2] instead of operands[1].
9698
9699
9700 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
9701
9702         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9703         case, check whether the outer register overlaps an unallocatable
9704         register, not just whether it fits the required class.
9705
9706 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
9707
9708         * poly-int.h (can_div_trunc_p): Add new overload in which all values
9709         are poly_ints.
9710         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9711         (memrefs_conflict_p): Likewise.
9712         (init_alias_analysis): Likewise.
9713         * cfgexpand.c (expand_debug_expr): Likewise.
9714         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9715         * cse.c (fold_rtx): Likewise.
9716         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9717         * expr.c (emit_block_move_hints): Likewise.
9718         (clear_storage_hints, push_block, emit_push_insn): Likewise.
9719         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9720         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9721         (emit_group_store): Likewise.
9722         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
9723         to read the PRE/POST_MODIFY increment.
9724         * calls.c (store_one_arg): Use strip_offset.
9725         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9726         poly_int_rtx_p.
9727         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9728         by a VEC_SELECT.
9729         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9730         (simplify_binary_operation_1): Extend CONST_INT handling to
9731         poly_int_rtx_p.
9732         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9733         than a HOST_WIDE_INT.
9734         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9735         poly_int64.
9736         (adjust_mems, add_stores): Update accodingly.
9737         (vt_canonicalize_addr): Track polynomial offsets.
9738         (emit_note_insn_var_location): Likewise.
9739         (vt_add_function_parameter): Likewise.
9740         (vt_initialize): Likewise.
9741
9742 2018-06-12  Jeff Law  <law@redhat.com>
9743
9744         * config.gcc (alpha*-*-freebsd*): Remove.
9745         * config/alpha/freebsd.h: Remove.
9746
9747 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
9748
9749         PR other/69968
9750         * spellcheck-tree.c (levenshtein_distance): Rename to...
9751         (get_edit_distance): ...this, and update for underlying renaming.
9752         * spellcheck-tree.h (levenshtein_distance): Rename to...
9753         (get_edit_distance): ...this.
9754         * spellcheck.c (levenshtein_distance): Rename to...
9755         (get_edit_distance): ...this.  Convert from Levenshtein distance
9756         to Damerau-Levenshtein distance by supporting transpositions of
9757         adjacent characters.  Rename "v1" to "v_next" and "v0" to
9758         "v_one_ago".
9759         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9760         (selftest::test_edit_distance_unit_test_oneway): ...this, and
9761         update for underlying renaming.
9762         (selftest::levenshtein_distance_unit_test): Rename to...
9763         (selftest::test_get_edit_distance_unit): ...this, and update for
9764         underlying renaming.
9765         (selftest::test_find_closest_string): Add example from PR 69968
9766         where transposition helps
9767         (selftest::test_metric_conditions): Update for renaming.
9768         (selftest::test_metric_conditions): Likewise.
9769         (selftest::spellcheck_c_tests): Likewise.
9770         * spellcheck.h (levenshtein_distance): Rename both overloads to...
9771         (get_edit_distance): ...this.
9772         (best_match::consider): Update for renaming.
9773
9774 2018-06-12  Martin Sebor  <msebor@redhat.com>
9775
9776         PR tree-optimization/85259
9777         * builtins.c (compute_objsize): Handle constant offsets.
9778         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9779         true iff a warning has been issued.
9780         * gimple.h (gimple_nonartificial_location): New function.
9781         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9782         gimple_nonartificial_location and handle -Wno-system-headers.
9783         (handle_builtin_stxncpy): Same.
9784
9785 2018-06-12  Martin Sebor  <msebor@redhat.com>
9786
9787         PR c/85931
9788         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9789
9790 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
9791
9792         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9793         BUILTIN_VEC_XST entries for pointer to double and long long.
9794
9795 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
9796
9797         PR target/85990
9798         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9799         Update comments.
9800         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9801         Likewise.
9802
9803 2018-06-12  Martin Liska  <mliska@suse.cz>
9804
9805         * doc/options.texi: Document IntegerRange.
9806
9807 2018-06-12  Martin Liska  <mliska@suse.cz>
9808
9809         * config/i386/i386.opt: Make MPX-related options as Deprecated.
9810         * opt-functions.awk: Handle Deprecated flag.
9811         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9812         and report error.
9813         (read_cmdline_option): Report warning for a deprecated option.
9814         * opts.h (struct cl_option): Add new field cl_deprecated.
9815         (CL_ERR_DEPRECATED): New.
9816
9817 2018-06-12  Martin Liska  <mliska@suse.cz>
9818
9819         * doc/options.texi: Document Deprecated option flag.
9820
9821 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9822
9823         * config/arc/arc-arch.h (arc_extras): New enum.
9824         (arc_cpu_t):Add field extra.
9825         (arc_cpu_types): Consider the extras.
9826         * config/arc/arc-cpus.def: Add extras info.
9827         * config/arc/arc-opts.h (processor_type): Consider extra field.
9828         * config/arc/arc.c (arc_override_options): Handle extra field.
9829
9830 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9831
9832         * config/arc/arc-arch.h: Update ARC_OPTX macro.
9833         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9834         field.
9835         * config/arc/arc.c (arc_init): Update pic warning.
9836         (irq_range): Update irq range parsing warnings.
9837         (arc_override_options): Update various warning messages.
9838         (arc_handle_aux_attribute): Likewise.
9839
9840 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
9841
9842         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9843
9844 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9845
9846         * doc/sourcebuild.texi: Document usage of line number 0 in verify
9847         compiler messages directives.
9848
9849 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
9850
9851         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9852         * config/mips/mips-tables.opt: Regenerate.
9853         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9854         mips64r6.
9855         * doc/invoke.texi: Document -march=i6500.
9856
9857 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
9858
9859         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9860         (i6400_gpmul): Add cpu_unit.
9861         (i6400_gpdiv): Likewise.
9862         (i6400_msa_add_d): Update reservations.
9863         (i6400_msa_int_add) Likewise.
9864         (i6400_msa_short_logic3) Likewise.
9865         (i6400_msa_short_logic2) Likewise.
9866         (i6400_msa_short_logic) Likewise.
9867         (i6400_msa_move) Likewise.
9868         (i6400_msa_cmp) Likewise.
9869         (i6400_msa_short_float2) Likewise.
9870         (i6400_msa_div_d) Likewise.
9871         (i6400_msa_long_logic1) Likewise.
9872         (i6400_msa_long_logic2) Likewise.
9873         (i6400_msa_mult) Likewise.
9874         (i6400_msa_long_float2) Likewise.
9875         (i6400_msa_long_float4) Likewise.
9876         (i6400_msa_long_float5) Likewise.
9877         (i6400_msa_long_float8) Likewise.
9878         (i6400_fpu_fadd): Include frint type.
9879         (i6400_fpu_store): New define_insn_reservation.
9880         (i6400_fpu_load): Likewise.
9881         (i6400_fpu_move): Likewise.
9882         (i6400_fpu_fcmp): Likewise.
9883         (i6400_fpu_fmadd): Likewise.
9884         (i6400_int_mult): Include imul3nc type and update reservation.
9885         (i6400_int_div): Include idiv3 type and update reservation.
9886         (i6400_int_load): Update to check type not move_type.
9887         (i6400_int_store): Likewise.
9888         (i6400_int_prefetch): Set zero latency.
9889
9890 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
9891
9892         * gcc.c: Document new %@{...} sequence.
9893         (LINK_COMMAND_SPEC): Use it for the -L switches.
9894         (cpp_unique_options): Use it for the -I switches.
9895         (at_file_argbuf): New global variable.
9896         (in_at_file): Likewise.
9897         (alloc_args): Create at_file_argbuf.
9898         (clear_args): Truncate at_file_argbuf.
9899         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9900         (open_at_file): New function.
9901         (close_at_file): Likewise.
9902         (create_at_file): Delete.
9903         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9904         <'o'>: Likewise.
9905         <'@'>: New case.
9906         (validate_switches_from_spec): Deal with %@{...} sequence.
9907         (validate_switches): Likewise.
9908         (driver::finalize): Call clear_args.
9909
9910 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
9911
9912         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9913
9914 2018-06-11  Martin Sebor  <msebor@redhat.com>
9915
9916         * doc/invoke.texi (-Wall): List -Wc++17-compat.
9917         (Wno-class-memaccess): Add @opindex.
9918         (Wno-templates, Wno-multiple-inheritance): Same.
9919         (Wno-virtual-inheritance, Wno-namespaces): Same.
9920         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9921         (Wno-format-overflow, Wno-format-truncation): Same.
9922         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9923         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9924         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9925         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9926         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9927         (Wno-misspelled-isr): Same.
9928
9929 2018-06-11  Martin Sebor  <msebor@redhat.com>
9930
9931         * PR tree-optimization/86083
9932         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9933
9934 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
9935
9936         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9937
9938 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
9939
9940         PR target/85755
9941         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9942         on the correct operand.
9943         (*movdi_internal64): Ditto.
9944
9945 2018-06-11  Martin Liska  <mliska@suse.cz>
9946
9947         PR tree-optimization/86089
9948         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9949
9950 2018-06-11  Julia Koval  <julia.koval@intel.com>
9951
9952         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9953         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9954         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9955
9956 2018-06-11  Olivier Hainque  <hainque@adacore.com>
9957
9958         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9959         for Ada with strict dwarf2.
9960
9961 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
9962
9963         PR target/85755
9964         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9965         addresses.
9966
9967 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
9968
9969         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9970
9971 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
9972
9973         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9974         TARGET_ELF.
9975
9976 2018-06-08  Martin Liska  <mliska@suse.cz>
9977
9978         * tree-cfg.h (debug_function): Fix argument type to match
9979         implementation.
9980
9981 2018-06-08  Martin Liska  <mliska@suse.cz>
9982
9983         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9984         Remove usage of MPX-related (and removed) fields.
9985         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9986
9987 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
9988
9989         * cfg.c (debug): Use TDF_NONE rather than 0.
9990         * cfghooks.c (debug): Likewise.
9991         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9992         (struct dump_option_value_info): Convert to...
9993         (struct kv_pair): ...this template type.
9994         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9995         rather than 0.
9996         (optinfo_verbosity_options): Likewise.
9997         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9998         OPTGROUP_NONE.
9999         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
10000         than int for "optgroup_flags" param.
10001         (dump_generic_expr_loc): Use dump_flags_t rather than int for
10002         "dump_kind" param.
10003         (dump_dec): Likewise.
10004         (dump_finish): Use TDF_NONE rather than 0.
10005         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
10006         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
10007         than 0.  Update for change to option_ptr.
10008         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
10009         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
10010         0.  Update for changes to optinfo_verbosity_options and
10011         optgroup_options.
10012         (opt_info_switch_p): Convert optgroup_flags from int to
10013         optgroup_flags_t.
10014         (dump_basic_block): Use dump_flags_t rather than int
10015         for "dump_kind" param.
10016         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
10017         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
10018         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
10019         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
10020         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
10021         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
10022         TDF_NONE): Convert from macros to...
10023         (enum dump_flag): ...this new enum.
10024         (dump_flags_t): Update to use enum.
10025         (operator|, operator&, operator~, operator|=, operator&=):
10026         Implement for dump_flags_t.
10027         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10028         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10029         Convert from macros to...
10030         (enum optgroup_flag): ...this new enum.
10031         (optgroup_flags_t): New typedef.
10032         (operator|, operator|=): Implement for optgroup_flags_t.
10033         (struct dump_file_info): Convert field "alt_flags" to
10034         dump_flags_t.  Convert field "optgroup_flags" to
10035         optgroup_flags_t.
10036         (dump_basic_block): Use dump_flags_t rather than int for param.
10037         (dump_generic_expr_loc): Likewise.
10038         (dump_dec): Likewise.
10039         (dump_register): Convert param "optgroup_flags" to
10040         optgroup_flags_t.
10041         (opt_info_enable_passes): Likewise.
10042         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10043         than 0.
10044         * gimple-pretty-print.c (debug): Likewise.
10045         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10046         (merged_store_group::apply_stores): Likewise.
10047         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10048         * gimple.c (verify_gimple_pp): Likewise.
10049         * graphite-poly.c (print_pbb_body): Likewise.
10050         * passes.c (pass_manager::register_one_dump_file): Convert
10051         local "optgroup_flags" to optgroup_flags_t.
10052         * print-tree.c (print_node): Use TDF_NONE rather than 0.
10053         (debug): Likewise.
10054         (debug_body): Likewise.
10055         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10056         to optgroup_flags_t.
10057         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10058         than 0.
10059         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10060         (convert_mult_to_fma): Likewise.
10061         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10062         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10063         * tree-vect-data-refs.c (dump_lower_bound): Convert param
10064         "dump_kind" to dump_flags_t.
10065
10066 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
10067
10068         * config/rs6000/rs6000.c (min, max): Delete.
10069
10070 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
10071
10072         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10073         -mabi=spe and -mabi=no-spe.
10074
10075 2018-06-08  Martin Liska  <mliska@suse.cz>
10076
10077         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10078         where we expect an existing summary.
10079
10080 2018-06-08  Martin Liska  <mliska@suse.cz>
10081
10082         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10083         * ipa-inline.h (estimate_edge_growth): Likewise.
10084
10085 2018-06-08  Martin Liska  <mliska@suse.cz>
10086
10087         * cgraph.c (function_version_hasher::hash): Use
10088         cgraph_node::get_uid ().
10089         (function_version_hasher::equal):
10090         * cgraph.h (cgraph_node::get_uid): New method.
10091         * ipa-inline.c (update_caller_keys): Use
10092         cgraph_node::get_uid ().
10093         (update_callee_keys): Likewise.
10094         * ipa-utils.c (searchc): Likewise.
10095         (ipa_reduced_postorder): Likewise.
10096         * lto-cgraph.c (input_node): Likewise.
10097         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10098         * symbol-summary.h (symtab_insertion): Likewise.
10099         (symtab_removal): Likewise.
10100         (symtab_duplication): Likewise.
10101         * tree-pretty-print.c (dump_function_header): Likewise.
10102         * tree-sra.c (convert_callers_for_node): Likewise.
10103
10104 2018-06-08  Martin Liska  <mliska@suse.cz>
10105
10106         * cgraph.c (symbol_table::create_edge): Always assign a new
10107         unique number.
10108         (symbol_table::free_edge): Do not recycle numbers.
10109         * cgraph.h (cgraph_edge::get): New method.
10110         * symbol-summary.h (symtab_removal): Use it.
10111         (symtab_duplication): Likewise.
10112         (call_summary::hashable_uid): Remove.
10113
10114 2018-06-08  Martin Liska  <mliska@suse.cz>
10115
10116         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10117         (initialize_growth_caches): Remove.
10118         (free_growth_caches): Likewise.
10119         (do_estimate_edge_time): Use edge_growth_cache.
10120         (do_estimate_edge_size): Likewise.
10121         (do_estimate_edge_hints): Likewise.
10122         * ipa-inline.c (reset_edge_caches): Likewise.
10123         (recursive_inlining): Likewise.
10124         (inline_small_functions): Likewise.
10125         * ipa-inline.h (initialize_growth_caches): Remove.
10126         (estimate_edge_size): Likewise.
10127         (estimate_edge_time): Likewise.
10128         (estimate_edge_hints): Likewise.
10129         (reset_edge_growth_cache): Likewise.
10130         * symbol-summary.h (call_summary::remove): New method.
10131
10132 2018-06-08  Martin Liska  <mliska@suse.cz>
10133
10134         * ipa-cp.c (class edge_clone_summary): New summary.
10135         (grow_edge_clone_vectors): Remove.
10136         (ipcp_edge_duplication_hook): Remove.
10137         (class edge_clone_summary_t): New call_summary class.
10138         (ipcp_edge_removal_hook): Remove.
10139         (edge_clone_summary_t::duplicate): New function.
10140         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10141         (create_specialized_node): Likewise.
10142         (ipcp_driver): Initialize edge_clone_summaries and do not
10143         register hooks.
10144
10145 2018-06-08  Martin Liska  <mliska@suse.cz>
10146
10147         * symbol-summary.h (get): New function.
10148         (call_summary::m_initialize_when_cloning): New class member.
10149
10150 2018-06-08  Martin Liska  <mliska@suse.cz>
10151
10152         * cgraph.c (cgraph_node::remove): Do not recycle uid.
10153         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10154         (symbol_table::allocate_cgraph_symbol): Do not set uid.
10155         * passes.c (uid_hash_t): Record removed_nodes by their uids.
10156         (remove_cgraph_node_from_order): Use the removed_nodes set.
10157         (do_per_function_toporder): Likwise.
10158         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10159         instead of summary_uid.
10160         (symtab_removal): Likewise.
10161         (symtab_duplication): Likewise.
10162
10163 2018-06-08  Martin Liska  <mliska@suse.cz>
10164
10165         * ipa-cp.c (ipcp_store_bits_results): Use
10166         ipcp_transformation_sum.
10167         (ipcp_store_vr_results): Likewise.
10168         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10169         to ...
10170         (ipcp_transformation_initialize): ... this.
10171         (ipa_set_node_agg_value_chain):
10172         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10173         (write_ipcp_transformation_info): Likewise.
10174         (read_ipcp_transformation_info): Likewise.
10175         (ipcp_update_bits): Likewise.
10176         (ipcp_update_vr): Likewise.
10177         (ipcp_transform_function): Likewise.
10178         * ipa-prop.h: Rename ipcp_transformation_summary to
10179         ipcp_transformation.
10180         (class ipcp_transformation_t): New function summary.
10181         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10182         (ipa_get_agg_replacements_for_node): Likewise.
10183
10184 2018-06-08  Martin Liska  <mliska@suse.cz>
10185
10186         * ipa-pure-const.c (struct funct_state_d): Do it class instead
10187         of struct.
10188         (class funct_state_summary_t): New function_summary class.
10189         (has_function_state): Remove.
10190         (get_function_state): Likewise.
10191         (set_function_state): Likewise.
10192         (add_new_function): Likewise.
10193         (funct_state_summary_t::insert): New function.
10194         (duplicate_node_data): Remove.
10195         (remove_node_data): Remove.
10196         (funct_state_summary_t::duplicate): New function.
10197         (register_hooks): Create new funct_state_summaries.
10198         (pure_const_generate_summary): Use it.
10199         (pure_const_write_summary): Likewise.
10200         (pure_const_read_summary): Likewise.
10201         (propagate_pure_const): Likewise.
10202         (propagate_nothrow): Likewise.
10203         (dump_malloc_lattice): Likewise.
10204         (propagate_malloc): Likewise.
10205         (execute): Do not register hooks, just remove summary
10206         instead.
10207         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10208         constructor.
10209
10210 2018-06-08  Martin Liska  <mliska@suse.cz>
10211
10212         * ipa-reference.c (remove_node_data): Remove.
10213         (duplicate_node_data): Likewise.
10214         (class ipa_ref_var_info_summary_t): New class.
10215         (class ipa_ref_opt_summary_t): Likewise.
10216         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10217         (get_reference_optimization_summary): Use
10218         ipa_ref_opt_sum_summaries.
10219         (set_reference_vars_info): Remove.
10220         (set_reference_optimization_summary): Likewise.
10221         (ipa_init): Create summaries.
10222         (init_function_info): Use function summary.
10223         (ipa_ref_opt_summary_t::duplicate): New function.
10224         (ipa_ref_opt_summary_t::remove): New function.
10225         (get_read_write_all_from_node): Fix GNU coding style.
10226         (propagate): Use function summary.
10227         (write_node_summary_p): Fix GNU coding style.
10228         (stream_out_bitmap): Likewise.
10229         (ipa_reference_read_optimization_summary): Use function summary.
10230         (ipa_reference_c_finalize): Do not release hooks.
10231
10232 2018-06-08  Martin Liska  <mliska@suse.cz>
10233
10234         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10235         (analyze_function_body): Extract multiple calls of get_create.
10236         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10237         * ipa-inline.c (recursive_inlining): Use ::get method.
10238         * ipa-inline.h (estimate_edge_growth): Likewise.
10239
10240 2018-06-08  Martin Liska  <mliska@suse.cz>
10241
10242         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
10243         HSA_INVALID.
10244         (hsa_function_summary::hsa_function_summary): Use the new enum
10245         value.
10246         (hsa_gpu_implementation_p): Use hsa_summaries::get.
10247         * hsa-gen.c (hsa_get_host_function): Likewise.
10248         (get_brig_function_name): Likewise.
10249         * ipa-hsa.c (process_hsa_functions): Likewise.
10250         (ipa_hsa_write_summary): Likewise.
10251         * symbol-summary.h (symtab_duplication): Use ::get function/
10252         (get): New function.
10253
10254 2018-06-08  Martin Liska  <mliska@suse.cz>
10255
10256         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
10257         of get.
10258         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
10259         (hsa_register_kernel): Likewise.
10260         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
10261         * hsa-gen.c (hsa_get_host_function): Likewise.
10262         (get_brig_function_name): Likewise.
10263         (generate_hsa): Likewise.
10264         (pass_gen_hsail::execute): Likewise.
10265         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
10266         (devirtualization_time_bonus): Likewise.
10267         (ipcp_propagate_stage): Likewise.
10268         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10269         (edge_set_predicate): Likewise.
10270         (evaluate_conditions_for_known_args): Likewise.
10271         (evaluate_properties_for_edge): Likewise.
10272         (ipa_fn_summary::reset): Likewise.
10273         (ipa_fn_summary_t::duplicate): Likewise.
10274         (dump_ipa_call_summary): Likewise.
10275         (ipa_dump_fn_summary): Likewise.
10276         (analyze_function_body): Likewise.
10277         (compute_fn_summary): Likewise.
10278         (estimate_edge_devirt_benefit): Likewise.
10279         (estimate_edge_size_and_time): Likewise.
10280         (estimate_calls_size_and_time): Likewise.
10281         (estimate_node_size_and_time): Likewise.
10282         (inline_update_callee_summaries): Likewise.
10283         (remap_edge_change_prob): Likewise.
10284         (remap_edge_summaries): Likewise.
10285         (ipa_merge_fn_summary_after_inlining): Likewise.
10286         (ipa_update_overall_fn_summary): Likewise.
10287         (read_ipa_call_summary): Likewise.
10288         (inline_read_section): Likewise.
10289         (write_ipa_call_summary): Likewise.
10290         (ipa_fn_summary_write): Likewise.
10291         (ipa_free_fn_summary): Likewise.
10292         * ipa-hsa.c (process_hsa_functions): Likewise.
10293         (ipa_hsa_write_summary): Likewise.
10294         (ipa_hsa_read_section): Likewise.
10295         * ipa-icf.c (sem_function::merge): Likewise.
10296         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10297         (do_estimate_edge_time): Likewise.
10298         (estimate_size_after_inlining): Likewise.
10299         (estimate_growth): Likewise.
10300         (growth_likely_positive): Likewise.
10301         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10302         (inline_call): Likewise.
10303         * ipa-inline.c (caller_growth_limits): Likewise.
10304         (can_inline_edge_p): Likewise.
10305         (can_inline_edge_by_limits_p): Likewise.
10306         (compute_uninlined_call_time): Likewise.
10307         (compute_inlined_call_time): Likewise.
10308         (want_inline_small_function_p): Likewise.
10309         (edge_badness): Likewise.
10310         (update_caller_keys): Likewise.
10311         (update_callee_keys): Likewise.
10312         (recursive_inlining): Likewise.
10313         (inline_small_functions): Likewise.
10314         (inline_to_all_callers_1): Likewise.
10315         (dump_overall_stats): Likewise.
10316         (early_inline_small_functions): Likewise.
10317         (early_inliner): Likewise.
10318         * ipa-inline.h (estimate_edge_growth): Likewise.
10319         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10320         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10321         * ipa-prop.h (IPA_NODE_REF): Likewise.
10322         (IPA_EDGE_REF): Likewise.
10323         * ipa-pure-const.c (malloc_candidate_p): Likewise.
10324         (propagate_malloc): Likewise.
10325         * ipa-split.c (execute_split_functions): Likewise.
10326         * symbol-summary.h: Rename get to get_create.
10327         (get): Likewise.
10328         (get_create): Likewise.
10329         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10330
10331 2018-06-08  Martin Liska  <mliska@suse.cz>
10332
10333         * symbol-summary.h (release): Move definition out of class
10334         declaration.
10335         (symtab_removal): Likewise.
10336         (symtab_duplication): Likewise.
10337
10338 2018-06-08  Martin Liska  <mliska@suse.cz>
10339
10340         * symbol-summary.h (function_summary): Move constructor
10341         implementation out of class declaration.
10342         (release): Likewise.
10343         (symtab_insertion): Likewise.
10344         (symtab_removal): Likewise.
10345         (symtab_duplication): Likewise.
10346         (get): Likewise.
10347
10348 2018-06-08  Martin Liska  <mliska@suse.cz>
10349
10350         * Makefile.in: Remove support for MPX (macros, related functions,
10351         fields in cgraph_node, ...).
10352         * builtin-types.def (BT_BND): Likewise.
10353         (BT_FN_BND_CONST_PTR): Likewise.
10354         (BT_FN_CONST_PTR_BND): Likewise.
10355         (BT_FN_VOID_PTR_BND): Likewise.
10356         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10357         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10358         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10359         (expand_builtin_mempcpy_with_bounds): Likewise.
10360         (expand_builtin_memset_with_bounds): Likewise.
10361         (expand_builtin_memset_args): Likewise.
10362         (std_expand_builtin_va_start): Likewise.
10363         (expand_builtin): Likewise.
10364         (expand_builtin_with_bounds): Likewise.
10365         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10366         (DEF_LIB_BUILTIN_CHKP): Likewise.
10367         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10368         (DEF_CHKP_BUILTIN): Likewise.
10369         (BUILT_IN_MEMCPY): Likewise.
10370         (BUILT_IN_MEMMOVE): Likewise.
10371         (BUILT_IN_MEMPCPY): Likewise.
10372         (BUILT_IN_MEMSET): Likewise.
10373         (BUILT_IN_STPCPY): Likewise.
10374         (BUILT_IN_STRCAT): Likewise.
10375         (BUILT_IN_STRCHR): Likewise.
10376         (BUILT_IN_STRCPY): Likewise.
10377         (BUILT_IN_STRLEN): Likewise.
10378         (BUILT_IN_MEMCPY_CHK): Likewise.
10379         (BUILT_IN_MEMMOVE_CHK): Likewise.
10380         (BUILT_IN_MEMPCPY_CHK): Likewise.
10381         (BUILT_IN_MEMSET_CHK): Likewise.
10382         (BUILT_IN_STPCPY_CHK): Likewise.
10383         (BUILT_IN_STRCAT_CHK): Likewise.
10384         (BUILT_IN_STRCPY_CHK): Likewise.
10385         * calls.c (store_bounds): Likewise.
10386         (emit_call_1): Likewise.
10387         (special_function_p): Likewise.
10388         (maybe_warn_nonstring_arg): Likewise.
10389         (initialize_argument_information): Likewise.
10390         (finalize_must_preallocate): Likewise.
10391         (compute_argument_addresses): Likewise.
10392         (expand_call): Likewise.
10393         * cfgexpand.c (expand_call_stmt): Likewise.
10394         (expand_return): Likewise.
10395         (expand_gimple_stmt_1): Likewise.
10396         (pass_expand::execute): Likewise.
10397         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10398         (cgraph_node::remove): Likewise.
10399         (cgraph_node::dump): Likewise.
10400         (cgraph_node::verify_node): Likewise.
10401         * cgraph.h (chkp_function_instrumented_p): Likewise.
10402         (symtab_node::get_alias_target): Likewise.
10403         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10404         (cgraph_local_p): Likewise.
10405         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10406         (cgraph_edge::rebuild_references): Likewise.
10407         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10408         (walk_polymorphic_call_targets): Likewise.
10409         (cgraph_node::expand_thunk): Likewise.
10410         (symbol_table::output_weakrefs): Likewise.
10411         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10412         (ix86_handle_option): Likewise.
10413         * config/i386/constraints.md: Likewise.
10414         * config/i386/i386-builtin-types.def (BND): Likewise.
10415         (VOID): Likewise.
10416         (PVOID): Likewise.
10417         (ULONG): Likewise.
10418         * config/i386/i386-builtin.def (BDESC_END): Likewise.
10419         (BDESC_FIRST): Likewise.
10420         (BDESC): Likewise.
10421         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10422         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10423         * config/i386/i386.c (enum reg_class): Likewise.
10424         (ix86_target_string): Likewise.
10425         (ix86_option_override_internal): Likewise.
10426         (ix86_conditional_register_usage): Likewise.
10427         (ix86_valid_target_attribute_inner_p): Likewise.
10428         (ix86_set_indirect_branch_type): Likewise.
10429         (ix86_set_current_function): Likewise.
10430         (ix86_function_arg_regno_p): Likewise.
10431         (init_cumulative_args): Likewise.
10432         (ix86_function_arg_advance): Likewise.
10433         (ix86_function_arg): Likewise.
10434         (ix86_pass_by_reference): Likewise.
10435         (ix86_function_value_regno_p): Likewise.
10436         (ix86_function_value_1): Likewise.
10437         (ix86_function_value_bounds): Likewise.
10438         (ix86_return_in_memory): Likewise.
10439         (ix86_setup_incoming_vararg_bounds): Likewise.
10440         (ix86_va_start): Likewise.
10441         (indirect_thunk_need_prefix): Likewise.
10442         (print_reg): Likewise.
10443         (ix86_print_operand): Likewise.
10444         (ix86_expand_call): Likewise.
10445         (ix86_output_function_return): Likewise.
10446         (reg_encoded_number): Likewise.
10447         (BDESC_VERIFYS): Likewise.
10448         (ix86_init_mpx_builtins): Likewise.
10449         (ix86_init_builtins): Likewise.
10450         (ix86_emit_cmove): Likewise.
10451         (ix86_emit_move_max): Likewise.
10452         (ix86_expand_builtin): Likewise.
10453         (ix86_builtin_mpx_function): Likewise.
10454         (ix86_get_arg_address_for_bt): Likewise.
10455         (ix86_load_bounds): Likewise.
10456         (ix86_store_bounds): Likewise.
10457         (ix86_load_returned_bounds): Likewise.
10458         (ix86_store_returned_bounds): Likewise.
10459         (ix86_class_likely_spilled_p): Likewise.
10460         (ix86_hard_regno_mode_ok): Likewise.
10461         (x86_order_regs_for_local_alloc): Likewise.
10462         (ix86_mitigate_rop): Likewise.
10463         (ix86_bnd_prefixed_insn_p): Likewise.
10464         (ix86_mpx_bound_mode): Likewise.
10465         (ix86_make_bounds_constant): Likewise.
10466         (ix86_initialize_bounds): Likewise.
10467         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10468         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10469         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10470         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10471         (TARGET_CHKP_BOUND_MODE): Likewise.
10472         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10473         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10474         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10475         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10476         * config/i386/i386.h (TARGET_MPX): Likewise.
10477         (TARGET_MPX_P): Likewise.
10478         (VALID_BND_REG_MODE): Likewise.
10479         (FIRST_BND_REG): Likewise.
10480         (LAST_BND_REG): Likewise.
10481         (enum reg_class): Likewise.
10482         (BND_REG_P): Likewise.
10483         (BND_REGNO_P): Likewise.
10484         (BNDmode): Likewise.
10485         (ADJUST_INSN_LENGTH): Likewise.
10486         * config/i386/i386.md: Likewise.
10487         * config/i386/i386.opt: Likewise.
10488         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10489         (defined): Likewise.
10490         (LINK_MPX): Likewise.
10491         (MPX_SPEC): Likewise.
10492         (LIBMPX_SPEC): Likewise.
10493         (LIBMPXWRAPPERS_SPEC): Likewise.
10494         (CHKP_SPEC): Likewise.
10495         * config/i386/predicates.md: Likewise.
10496         * dbxout.c (dbxout_type): Likewise.
10497         * doc/extend.texi: Likewise.
10498         * doc/invoke.texi: Likewise.
10499         * doc/md.texi: Likewise.
10500         * doc/tm.texi: Likewise.
10501         * doc/tm.texi.in: Likewise.
10502         * dwarf2out.c (is_base_type): Likewise.
10503         (gen_formal_types_die): Likewise.
10504         (gen_subprogram_die): Likewise.
10505         (gen_type_die_with_usage): Likewise.
10506         (gen_decl_die): Likewise.
10507         (dwarf2out_late_global_decl): Likewise.
10508         * expr.c (expand_assignment): Likewise.
10509         (emit_storent_insn): Likewise.
10510         (store_expr_with_bounds): Likewise.
10511         (store_expr): Likewise.
10512         (expand_expr_real_1): Likewise.
10513         * expr.h (store_expr_with_bounds): Likewise.
10514         * function.c (use_register_for_decl): Likewise.
10515         (struct bounds_parm_data): Likewise.
10516         (assign_parms_augmented_arg_list): Likewise.
10517         (assign_parm_find_entry_rtl): Likewise.
10518         (assign_parm_is_stack_parm): Likewise.
10519         (assign_parm_load_bounds): Likewise.
10520         (assign_bounds): Likewise.
10521         (assign_parms): Likewise.
10522         (expand_function_start): Likewise.
10523         * gcc.c (CHKP_SPEC): Likewise.
10524         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10525         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10526         (wrestrict_dom_walker::check_call): Likewise.
10527         * gimple.c (gimple_build_call_from_tree): Likewise.
10528         * gimple.h (enum gf_mask): Likewise.
10529         (gimple_call_with_bounds_p): Likewise.
10530         (gimple_call_set_with_bounds): Likewise.
10531         * gimplify.c (gimplify_init_constructor): Likewise.
10532         * ipa-cp.c (initialize_node_lattices): Likewise.
10533         (propagate_constants_across_call): Likewise.
10534         (find_more_scalar_values_for_callers_subset): Likewise.
10535         * ipa-hsa.c (process_hsa_functions): Likewise.
10536         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10537         * ipa-icf.c (sem_function::merge): Likewise.
10538         * ipa-inline.c (early_inliner): Likewise.
10539         * ipa-pure-const.c (warn_function_noreturn): Likewise.
10540         (warn_function_cold): Likewise.
10541         (propagate_pure_const): Likewise.
10542         * ipa-ref.h (enum GTY): Likewise.
10543         * ipa-split.c (find_retbnd): Likewise.
10544         (consider_split): Likewise.
10545         (split_function): Likewise.
10546         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10547         * ipa.c (walk_polymorphic_call_targets): Likewise.
10548         (symbol_table::remove_unreachable_nodes): Likewise.
10549         (process_references): Likewise.
10550         (cgraph_build_static_cdtor_1): Likewise.
10551         * lto-cgraph.c (lto_output_node): Likewise.
10552         (output_refs): Likewise.
10553         (compute_ltrans_boundary): Likewise.
10554         (input_overwrite_node): Likewise.
10555         (input_node): Likewise.
10556         (input_cgraph_1): Likewise.
10557         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10558         * passes.c (pass_manager::execute_early_local_passes): Likewise.
10559         (class pass_chkp_instrumentation_passes): Likewise.
10560         (make_pass_chkp_instrumentation_passes): Likewise.
10561         * passes.def: Likewise.
10562         * rtl.h (struct GTY): Likewise.
10563         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10564         * stor-layout.c (layout_type): Likewise.
10565         * symtab.c: Likewise.
10566         * target.def: Likewise.
10567         * targhooks.c (default_chkp_bound_type): Likewise.
10568         (default_chkp_bound_mode): Likewise.
10569         (default_builtin_chkp_function): Likewise.
10570         (default_chkp_function_value_bounds): Likewise.
10571         (default_chkp_make_bounds_constant): Likewise.
10572         (default_chkp_initialize_bounds): Likewise.
10573         * targhooks.h (default_chkp_bound_type): Likewise.
10574         (default_chkp_bound_mode): Likewise.
10575         (default_builtin_chkp_function): Likewise.
10576         (default_chkp_function_value_bounds): Likewise.
10577         (default_chkp_make_bounds_constant): Likewise.
10578         (default_chkp_initialize_bounds): Likewise.
10579         * toplev.c (compile_file): Likewise.
10580         (process_options): Likewise.
10581         * tree-core.h (DEF_BUILTIN): Likewise.
10582         (DEF_BUILTIN_CHKP): Likewise.
10583         * tree-inline.c (declare_return_variable): Likewise.
10584         (remap_gimple_stmt): Likewise.
10585         (copy_bb): Likewise.
10586         (initialize_inlined_parameters): Likewise.
10587         (expand_call_inline): Likewise.
10588         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10589         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10590         (make_pass_ipa_chkp_produce_thunks): Likewise.
10591         (make_pass_chkp): Likewise.
10592         (make_pass_chkp_opt): Likewise.
10593         (make_pass_chkp_instrumentation_passes): Likewise.
10594         * tree-pretty-print.c (dump_generic_node): Likewise.
10595         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10596         * tree-ssa-dce.c (propagate_necessity): Likewise.
10597         (eliminate_unnecessary_stmts): Likewise.
10598         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10599         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10600         * tree-ssa-sccvn.h: Likewise.
10601         * tree-ssa-strlen.c (get_string_length): Likewise.
10602         (valid_builtin_call): Likewise.
10603         (adjust_last_stmt): Likewise.
10604         (handle_builtin_strchr): Likewise.
10605         (handle_builtin_strcpy): Likewise.
10606         (handle_builtin_stxncpy): Likewise.
10607         (handle_builtin_memcpy): Likewise.
10608         (handle_builtin_strcat): Likewise.
10609         (strlen_check_and_optimize_stmt): Likewise.
10610         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10611         * tree-streamer-in.c: Likewise.
10612         * tree-streamer.c (record_common_node): Likewise.
10613         * tree.c (tree_code_size): Likewise.
10614         (wide_int_to_tree_1): Likewise.
10615         (type_contains_placeholder_1): Likewise.
10616         (build_common_tree_nodes): Likewise.
10617         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10618         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10619         (POINTER_BOUNDS_P): Likewise.
10620         (BOUNDED_TYPE_P): Likewise.
10621         (BOUNDED_P): Likewise.
10622         (CALL_WITH_BOUNDS_P): Likewise.
10623         (pointer_bounds_type_node): Likewise.
10624         * value-prof.c (gimple_ic): Likewise.
10625         * var-tracking.c (vt_add_function_parameters): Likewise.
10626         * varasm.c (make_decl_rtl): Likewise.
10627         (assemble_start_function): Likewise.
10628         (output_constant): Likewise.
10629         (maybe_assemble_visibility): Likewise.
10630         * varpool.c (ctor_for_folding): Likewise.
10631         * chkp-builtins.def: Remove.
10632         * ipa-chkp.c: Remove.
10633         * ipa-chkp.h: Remove.
10634         * rtl-chkp.c: Remove.
10635         * rtl-chkp.h: Remove.
10636         * tree-chkp-opt.c: Remove.
10637         * tree-chkp.c: Remove.
10638         * tree-chkp.h: Remove.
10639
10640 2018-06-07  Carl Love  <cel@us.ibm.com>
10641
10642         * config/rs6000/vsx.md (vextract_fp_from_shorth,
10643         vextract_fp_from_shortl): Add BE support.
10644
10645 2018-06-07  Paul Koning  <ni1d@arrl.net>
10646
10647         * compare-elim.c (try_merge_compare): Don't merge compare if
10648         address contains a side effect.
10649         (try_eliminate_compare): Likewise.
10650
10651 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
10652
10653         * config.gcc: Support "tremont".
10654         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10655         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10656         PROCESSOR_TREMONT.
10657         * config/i386/i386.c (m_TREMONT): Define.
10658         (processor_target_table): Add "tremont".
10659         (PTA_TREMONT): Define.
10660         (ix86_lea_outperforms): Add TARGET_TREMONT.
10661         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10662         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10663         and M_INTEL_GOLDMONT_PLUS.
10664         (fold_builtin_cpu): Add "tremont".
10665         (ix86_add_stmt_cost): Add TARGET_TREMONT.
10666         (ix86_option_override_internal): Add "tremont".
10667         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10668         (processor_type): Add PROCESSOR_TREMONT.
10669         * config/i386/x86-tune.def: Add m_TREMONT.
10670         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10671
10672 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10673
10674         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10675         symbol defined for msp430i* devices to be lower case.
10676
10677 2018-06-07  Richard Biener  <rguenther@suse.de>
10678
10679         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10680         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
10681         Properly wrap signed arithmetic if overflow wraps.
10682
10683 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
10684
10685         PR tree-optimization/69615
10686         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10687         of a cast from a same precision integral SSA_NAME in a bb dominated
10688         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
10689         cast to utype if rhs2 has already a compatible type.
10690
10691 2018-06-07  Richard Biener  <rguenther@suse.de>
10692
10693         PR tree-optimization/85935
10694         * graphite-scop-detection.c (find_params_in_bb): Analyze
10695         condition operands with respect to the correct loop.  Assert
10696         the analysis doesn't fail.
10697
10698 2018-06-04  Carl Love  <cel@us.ibm.com>
10699
10700         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10701         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
10702         as it is slightly cheaper.
10703         (first_match_or_eos_index_<mode>):
10704         Calculate index using natural element order.
10705         (first_match_index_<mode>):
10706         Calculate index using natural element order.
10707         (first_match_or_eos_index_<mode>):
10708         Calculate index using natural order.
10709         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10710         for BE and LE modes.
10711         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10712         P9V_BUILTIN_VCLZLSBB_V16QI.
10713         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10714         specific.
10715
10716 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10717
10718         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10719         indentation and line wrap for many prototypes.  Add missing
10720         @smallexample directives around block of prototypes for vec_xl and
10721         vec_xst.
10722
10723 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
10724
10725         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10726         track if we pass or return IEEE 128-bit floating point.
10727         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10728         C++ mangling that is compatible with GCC 8.1.
10729         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10730         (init_cumulative_args): Note if we pass or return IEEE 128-bit
10731         floating point types.
10732         (rs6000_function_arg_advance_1): Likewise.
10733         (rs6000_mangle_type): Optionally generate mangled names that match
10734         what GCC 8.1 generated for IEEE 128-bit floating point types.
10735         (rs6000_globalize_decl_name): If we have an external function that
10736         passes or returns IEEE 128-bit types, generate a weak reference
10737         from the mangled name used in GCC 8.1 to the current mangled
10738         name.
10739         (rs6000_init_builtins): Make __ibm128 use the long double type if
10740         long double is IBM extended double.  Make __float128 use the long
10741         double type if long double is IEEE 128-bit.
10742
10743         PR target/85657
10744         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10745         macro for __ibm128 built-in functions.
10746         (PACK_IF): Add __ibm128 pack/unpack functions.
10747         (UNPACK_IF): Likewise.
10748         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10749         enable long double built-in functions if long double is IEEE
10750         128-bit floating point.
10751         (rs6000_invalid_builtin): Update long double built-in function
10752         error message.
10753         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10754         functions, adjust the built-in function to use the long double
10755         built-in function if __ibm128 and long double are the same type.
10756         * doc/extend.texi (PowerPC builtins): Update documention for
10757         __builtin_{,un}pack_longdouble.  Add documentation for
10758         __builtin_{,un}pack_ibm128.
10759
10760 2018-06-06  Jim Wilson  <jimw@sifive.com>
10761
10762         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10763         (struct machine_function): New field interrupt_mode.
10764         (riscv_handle_type_attribute): New function.  Add forward declaration.
10765         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10766         (riscv_expand_epilogue): Check interrupt_mode field.
10767         (riscv_set_current_function): Check interrupt attribute args and
10768         set interrupt_mode field.
10769         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10770         (riscv_sret, riscv_uret): New.
10771         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10772         new arguments to interrupt attribute.
10773
10774 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
10775
10776         PR target/63177
10777         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10778         Don't handle -mcpu=power8 if -mpower9-vector is also used.
10779
10780 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10781
10782         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10783         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10784         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10785         several redundant entries.
10786
10787 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
10788
10789         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10790         type from "rtx" to "rtx_insn *".
10791         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10792         for local "call_insn", removing cast.
10793         (ix86_expand_call): Likewise, introducing a "call_insn" local.
10794
10795 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
10796
10797         PR tree-optimization/86066
10798         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10799         for BIT_INSERT_EXPR stores.
10800
10801 2018-06-06  Richard Biener  <rguenther@suse.de>
10802
10803         PR tree-optimization/86062
10804         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10805         component refs ontop
10806         of to be offsetted base.
10807
10808 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10809
10810         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10811         to be static and remove check on interrupt attribute name.
10812
10813 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10814
10815         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10816         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10817
10818 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
10819
10820         PR target/79924
10821         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10822         second argument.
10823         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10824         Remove second argument, change how error is called.
10825         (aarch64_layout_arg): Remove second argument from
10826         aarch64_err_no_fpadvsimd call.
10827         (aarch64_init_cumulative_args): Ditto.
10828         (aarch64_gimplify_va_arg_expr): Ditto.
10829         * config/aarch64/aarch64.md (mov<mode>): Ditto.
10830
10831 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
10832
10833         * config/i386/i386.md (simple_return_indirect_internal): New expander.
10834         (*simple_return_indirect_internal<mode>): Rename from
10835         simple_return_indirect_internal.  Use W mode iterator.
10836         (rstorssp): New expander.
10837         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
10838         (clrssbsy): New expander.
10839         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
10840
10841 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10842
10843         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10844         __typeof__.
10845         (cmse_check_pointed_object): Likewise.
10846
10847 2018-06-05  Martin Liska  <mliska@suse.cz>
10848
10849         PR gcov-profile/47618
10850         * doc/invoke.texi: Document how -fprofile-dir format
10851         is extended.
10852
10853 2018-06-05  Richard Biener  <rguenther@suse.de>
10854
10855         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10856         removal pretend DOM info isn't available so we do not update
10857         it and only remove edges, not dominated blocks.  Actually free
10858         DOM info in case we removed something.  Remove unreachable blocks.
10859         (mfb_keep_latches): Work with either DOM info or marked backedges.
10860         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10861         first.  Mark backedges if DOM info isn't available.
10862         (Re-)compute DOM info after cleanup_control_flow_pre.
10863
10864 2018-06-05  Richard Biener  <rguenther@suse.de>
10865
10866         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10867         (locus_discrim_hasher::hash): Adjust.
10868         (locus_discrim_hasher::equal): Likewise.
10869         (next_discriminator_for_locus): Work on line directly.
10870         (same_line_p): Pass in expanded locus1 as well.
10871         (assign_discriminators): Avoid redundant location expansions.
10872
10873 2018-06-05  Richard Biener  <rguenther@suse.de>
10874
10875         PR tree-optimization/86046
10876         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10877         if required after clearing TREE_ADDRESSABLE.
10878
10879 2018-06-05  Richard Biener  <rguenther@suse.de>
10880
10881         PR tree-optimization/86047
10882         * tree-ssa-loop.c (for_each_index): Glob handling of all
10883         decls and constants and really handle all of them.
10884
10885 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10886
10887         PR target/81497
10888         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10889         qualifier_void_pointer and qualifier_const_void_pointer.
10890         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10891         (arm_init_builtins): Handle the above.
10892         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10893         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10894         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10895         void intrinsics.
10896
10897 2018-06-05  Martin Liska  <mliska@suse.cz>
10898
10899         * auto-profile.c (read_autofdo_file): Do not use
10900         gcov_ctr_summary struct.
10901         (afdo_callsite_hot_enough_for_early_inline): Likewise.
10902         * coverage.c (struct counts_entry): Likewise.
10903         (read_counts_file): Read just single summary entry.
10904         (get_coverage_counts): Use gcov_summary struct.
10905         * coverage.h (get_coverage_counts): Likewise.
10906         * gcov-dump.c (dump_working_sets): Likewise.
10907         (tag_summary): Dump just single summary.
10908         * gcov-io.c (gcov_write_summary): Write just histogram
10909         summary.
10910         (gcov_read_summary): Read just single summary.
10911         (compute_working_sets): Use gcov_summary struct.
10912         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10913         of GCOV_COUNTERS_SUMMABLE.
10914         (GCOV_COUNTERS_SUMMABLE): Remove.
10915         (GCOV_FIRST_VALUE_COUNTER): Replace with
10916         GCOV_COUNTER_V_INTERVAL.
10917         (struct gcov_ctr_summary): Remove.
10918         (struct gcov_summary): Directly use fields of former
10919         gcov_ctr_summary.
10920         (compute_working_sets): Use gcov_summary struct.
10921         * gcov.c (read_count_file): Do not use ctrs fields.
10922         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10923         struct.
10924         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10925         struct.
10926         * profile.c: Likewise.
10927         * profile.h: Likewise.
10928
10929 2018-06-05  Martin Liska  <mliska@suse.cz>
10930
10931         PR gcov-profile/84846
10932         * gcov.c (output_lines): Print working directory only
10933         in intermediate format.
10934
10935 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
10936
10937         * config/s390/s390-builtin-types.def: Add void function type.
10938         * config/s390/s390-builtins.def: Use the function type for the
10939         tbeginc builtin.
10940
10941 2018-06-04  Jim Wilson  <jimw@sifive.com>
10942
10943         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10944         to int.
10945         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10946         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
10947         handle EH_RETURN_DATA_REGNO registers properly.
10948         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10949         (riscv_expand_epilogue): Update comment.  Change argument name and
10950         type.  Update code to use new name and type.  Pass new args to
10951         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
10952         EXCEPTION_RETURN.
10953         * config/riscv/riscv.md (NORMAL_RETURN): New.
10954         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10955         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10956         (eh_return): Call gen_eh_return_internal and emit barrier.
10957         (eh_return_internal): Call riscv_expand_epilogue.
10958
10959 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
10960
10961         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10962         bit_insertion field and declare can_be_merged_into method.
10963         (merged_store_group::can_be_merged_into): New method.
10964         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10965         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
10966         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10967
10968 2018-06-04  Richard Biener  <rguenther@suse.de>
10969
10970         PR tree-optimization/85955
10971         * builtins.c (fold_builtin_sincos): Convert pointers to
10972         destination to appropriate type before dereferencing.
10973
10974 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
10975
10976         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10977
10978 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
10979
10980         * expr.c (expand_expr_real_1): Force the operand into memory if
10981         its TYPE_MODE is BLKmode and if there is no integer mode for
10982         the number of bits being extracted.
10983
10984 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
10985
10986         PR target/85832
10987         PR target/86036
10988         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10989         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10990
10991 2018-06-04  Richard Biener  <rguenther@suse.de>
10992
10993         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10994         (cleanup_tree_cfg_noloop): ... single caller.  Do
10995         start_recording_case_labels later.
10996
10997 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
10998
10999         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
11000         to _IMMINTRIN_H_INCLUDED.
11001         * config/i386/pconfigintrin.h: Ditto.
11002         * config/i386/waitpkgintrin.h: Ditto.
11003         * config/i386/immintrin.h: Add includes for sgxintrin.h,
11004         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
11005         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
11006         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
11007         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
11008         waitpkgintrin.h and cldemoteintrin.h.
11009
11010 2018-06-04  Richard Biener  <rguenther@suse.de>
11011
11012         PR tree-optimization/86038
11013         * tracer.c (find_best_successor): Check probability for
11014         being initialized, bail out if not.
11015
11016 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
11017
11018         PR target/86003
11019         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
11020         of bits to ignore when comparing architectures.
11021
11022 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
11023
11024         PR tree-optimization/69615
11025         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11026         maximum or minimum of the type, try to merge it also as if
11027         range1 is + [-, x - 1] or + [x + 1, -].
11028
11029         PR c++/86025
11030         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
11031
11032 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
11033
11034         PR tree-optimization/86034
11035         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11036         the unsigned bitfield type in a bit insertion sequence if it does not
11037         have a larger precision than the bitfield size.
11038         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11039
11040 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
11041
11042         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11043
11044 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
11045
11046         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11047         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11048         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11049         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11050
11051 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
11052
11053         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11054         Disable -fdelete-null-pointer-checks for ELF toolchain.
11055
11056 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
11057             Kito Cheng  <kito.cheng@gmail.com>
11058
11059         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11060         (nds32le-*-*, nds32be-*-*): Integrate checking process.
11061         (nds32*-*-*): Add glibc and uclibc conditions.
11062         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11063         (TARGET_EXCEPT_UNWIND_INFO): Define.
11064         * config/nds32/elf.h: New file.
11065         * config/nds32/linux.h: New file.
11066         * config/nds32/nds32-elf.opt: New file.
11067         * config/nds32/nds32-linux.opt: New file.
11068         * config/nds32/nds32-fp-as-gp.c
11069         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11070         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11071         TARGET_LINUX_ABI.
11072         (nds32_asm_file_end): Ditto.
11073         (nds32_print_operand): Ditto.
11074         (nds32_insert_attributes): Ditto.
11075         (nds32_init_libfuncs): New function.
11076         (TARGET_HAVE_TLS): Define.
11077         (TARGET_INIT_LIBFUNCS): Define.
11078         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11079         spec content.
11080         (TARGET_ELF): Apply different mcmodel setting.
11081         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11082         been migrated into elf.h and linux.h files.
11083         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11084         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11085         (mcmodel): The content has been migrated into nds32-elf.opt and
11086         nds32-linux.opt files.
11087         * config/nds32/t-elf: New file.
11088         * config/nds32/t-linux: New file.
11089
11090 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
11091             Shiva Chen  <shiva0217@gmail.com>
11092
11093         * config/nds32/constants.md (unspec_volatile_element): Add
11094         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11095         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11096         optimization.
11097         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11098         (make_pass_nds32_fp_as_gp): Declare.
11099         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11100         optmization pass.
11101         (nds32_asm_function_end_prologue): Remove unused asm output.
11102         (nds32_asm_function_begin_epilogue): Remove unused asm output.
11103         (nds32_asm_file_start): Output necessary fp_as_gp information.
11104         (nds32_option_override): Adjust register usage.
11105         (nds32_expand_prologue): Consider fp_as_gp situation.
11106         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11107         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11108         (epilogue): Ditto.
11109         (return): Ditto.
11110         (simple_return): Ditto.
11111         (omit_fp_begin): Output special directive for fp_as_gp.
11112         (omit_fp_end): Output special directive for fp_as_gp.
11113         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11114         mforbid-fp-as-gp): New options.
11115
11116 2018-06-01  Mark Wielaard  <mark@klomp.org>
11117
11118         * dwarf2out.c (dwarf2out_finish): Remove generation of
11119         DW_AT_loclists_base.
11120
11121 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
11122
11123         * gimple-ssa-store-merging.c: Include gimple-fold.h.
11124         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11125         (struct merged_store_group): Add bit_insertion field.
11126         (dump_char_array): Use standard hexadecimal format.
11127         (merged_store_group::merged_store_group): Set bit_insertion to false.
11128         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
11129         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
11130         also print the mask in the dump file.
11131         (pass_store_merging::gate): Minor tweak.
11132         (imm_store_chain_info::coalesce_immediate): Fix wrong association
11133         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
11134         stores with INTEGER_CST stores.
11135         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11136         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11137         and use it throughout.  Generate bit insertion sequences if need be.
11138         (pass_store_merging::process_store): Remove redundant condition.
11139         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11140
11141 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
11142
11143         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11144         the 128-bit floating point types.  Fix function comment.
11145
11146 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11147
11148         * config/aarch64/aarch64-simd.md
11149         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11150         mnemonics.
11151         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11152         mnemonics.
11153
11154 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
11155
11156         PR tree-optimization/85989
11157         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11158         variable.
11159         (backprop::intersect_uses): Check it when deciding whether this
11160         is a backedge reference.
11161         (backprop::process_block): Add each phi to m_visited_phis
11162         after visiting it, then clear it at the end.
11163
11164 2018-06-01  Richard Biener  <rguenther@suse.de>
11165
11166         * tree-vectorizer.h (vect_dr_stmt): New function.
11167         (vect_get_load_cost): Adjust.
11168         (vect_get_store_cost): Likewise.
11169         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11170         Use vect_dr_stmt instead of DR_SMTT.
11171         (vect_record_base_alignments): Likewise.
11172         (vect_calculate_target_alignment): Likewise.
11173         (vect_compute_data_ref_alignment): Likewise and make static.
11174         (vect_update_misalignment_for_peel): Likewise.
11175         (vect_verify_datarefs_alignment): Likewise.
11176         (vector_alignment_reachable_p): Likewise.
11177         (vect_get_data_access_cost): Likewise.  Pass down
11178         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11179         (vect_get_peeling_costs_all_drs): Likewise.
11180         (vect_peeling_hash_get_lowest_cost): Likewise.
11181         (vect_enhance_data_refs_alignment): Likewise.
11182         (vect_find_same_alignment_drs): Likewise.
11183         (vect_analyze_data_refs_alignment): Likewise.
11184         (vect_analyze_group_access_1): Likewise.
11185         (vect_analyze_group_access): Likewise.
11186         (vect_analyze_data_ref_access): Likewise.
11187         (vect_analyze_data_ref_accesses): Likewise.
11188         (vect_vfa_segment_size): Likewise.
11189         (vect_small_gap_p): Likewise.
11190         (vectorizable_with_step_bound_p): Likewise.
11191         (vect_prune_runtime_alias_test_list): Likewise.
11192         (vect_analyze_data_refs): Likewise.
11193         (vect_supportable_dr_alignment): Likewise.
11194         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11195         (vect_gen_prolog_loop_niters): Likewise.
11196         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11197         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11198         modify DR_STMT.
11199         (vect_recog_mask_conversion_pattern): Likewise.
11200         (vect_try_gather_scatter_pattern): Likewise.
11201         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11202         to vect_get_store_cost.
11203         (vect_get_store_cost): Get stmt_info instead of DR.
11204         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11205         (vect_get_load_cost): Get stmt_info instead of DR.
11206
11207 2018-06-01  Richard Biener  <rguenther@suse.de>
11208
11209         PR middle-end/86017
11210         * gimple-fold.c (var_decl_component_p): Also allow offsetted
11211         vars wrapped in MEM_REFs.
11212
11213 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
11214
11215         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11216         Fix subreg tests so that we only return a choice between
11217         GENERAL_REGS and FP_REGS if the original classes included both.
11218
11219 2018-06-01  Richard Biener  <rguenther@suse.de>
11220
11221         PR ipa/85960
11222         * tree-ssa-structalias.c (get_function_part_constraint):
11223         Handle NULL fi->decl.
11224         (find_func_aliases_for_call): Properly handle indirect
11225         fi from direct call.
11226         (find_func_clobbers): Likewise.
11227         (ipa_pta_execute): Likewise.
11228         (create_variable_info_for): For functions that are ifunc_resolver
11229         resolve to a varinfo that contains the result of the resolver call.
11230         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11231         aliases.
11232
11233 2018-05-31  Michael Collison  <michael.collison@arm.com>
11234
11235         * config/aarch64/aarch64.md:
11236         (*fix_to_zero_extenddfdi2): New pattern.
11237         * gcc.target/aarch64/fix_extend1.c: New testcase.
11238
11239 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
11240
11241         PR middle-end/78809
11242         PR middle-end/83026
11243         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
11244         and BUILT_IN_STRNCMP_EQ.
11245         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
11246         BUILT_IN_STRNCMP_EQ.
11247         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
11248         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11249         (gimple_fold_builtin): Likewise.
11250         * tree-ssa-strlen.c (compute_string_length): New function.
11251         (determine_min_obsize): New function.
11252         (handle_builtin_string_cmp): New function to handle calls to
11253         string compare functions.
11254         (strlen_optimize_stmt): Add handling to builtin string compare
11255         calls.
11256         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11257         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11258         * tree.c (build_common_builtin_nodes): Add new defines of
11259         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
11260
11261 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
11262
11263         PR target/85984
11264         * bb-reorder.c (pass_partition_blocks::gate): Return false for
11265         functions with naked attribute.
11266
11267 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
11268
11269         * config/i386/sse.md (avx_vec_concat<mode>):
11270         Substitute concat_tg_mode mode attribute with xtg_mode.
11271         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
11272         (concat_tg_mode): Remove mode attribute.
11273
11274 2018-05-31  Martin Sebor  <msebor@redhat.com>
11275
11276         PR c/82063
11277         * calls.c (alloc_max_size): Correct a logic error/typo.
11278         Treat excessive arguments as infinite.  Warn for invalid arguments.
11279         * doc/invoke.texi (-Walloc-size-larger-than): Update.
11280
11281 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
11282
11283         PR target/85829
11284         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
11285         and movx for Haswell.
11286
11287 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
11288             Cesar Philippidis  <cesar@codesourcery.com>
11289
11290         PR middle-end/85879
11291         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
11292         when emitting error on private/firstprivate reductions.
11293         * omp-low.c (lower_omp_target): Avoid reference-type processing
11294         on pointers for firstprivate clause.
11295
11296 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
11297
11298         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
11299         (st1x2): Likewise.
11300         (st1x3): Likewise.
11301         * config/aarch64/aarch64-simd.md
11302         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11303         (aarch64_ld1_x3_<mode>): Likewise
11304         (aarch64_st1x2<VALLDIF:mode>): Likewise
11305         (aarch64_st1_x2_<mode>): Likewise
11306         (aarch64_st1x3<VALLDIF:mode>): Likewise
11307         (aarch64_st1_x3_<mode>): Likewise
11308         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11309         (vld1_s8_x3): Likewise.
11310         (vld1_u16_x3): Likewise.
11311         (vld1_s16_x3): Likewise.
11312         (vld1_u32_x3): Likewise.
11313         (vld1_s32_x3): Likewise.
11314         (vld1_u64_x3): Likewise.
11315         (vld1_s64_x3): Likewise.
11316         (vld1_f16_x3): Likewise.
11317         (vld1_f32_x3): Likewise.
11318         (vld1_f64_x3): Likewise.
11319         (vld1_p8_x3): Likewise.
11320         (vld1_p16_x3): Likewise.
11321         (vld1_p64_x3): Likewise.
11322         (vld1q_u8_x3): Likewise.
11323         (vld1q_s8_x3): Likewise.
11324         (vld1q_u16_x3): Likewise.
11325         (vld1q_s16_x3): Likewise.
11326         (vld1q_u32_x3): Likewise.
11327         (vld1q_s32_x3): Likewise.
11328         (vld1q_u64_x3): Likewise.
11329         (vld1q_s64_x3): Likewise.
11330         (vld1q_f16_x3): Likewise.
11331         (vld1q_f32_x3): Likewise.
11332         (vld1q_f64_x3): Likewise.
11333         (vld1q_p8_x3): Likewise.
11334         (vld1q_p16_x3): Likewise.
11335         (vld1q_p64_x3): Likewise.
11336         (vst1_s64_x2): Likewise.
11337         (vst1_u64_x2): Likewise.
11338         (vst1_f64_x2): Likewise.
11339         (vst1_s8_x2): Likewise.
11340         (vst1_p8_x2): Likewise.
11341         (vst1_s16_x2): Likewise.
11342         (vst1_p16_x2): Likewise.
11343         (vst1_s32_x2): Likewise.
11344         (vst1_u8_x2): Likewise.
11345         (vst1_u16_x2): Likewise.
11346         (vst1_u32_x2): Likewise.
11347         (vst1_f16_x2): Likewise.
11348         (vst1_f32_x2): Likewise.
11349         (vst1_p64_x2): Likewise.
11350         (vst1q_s8_x2): Likewise.
11351         (vst1q_p8_x2): Likewise.
11352         (vst1q_s16_x2): Likewise.
11353         (vst1q_p16_x2): Likewise.
11354         (vst1q_s32_x2): Likewise.
11355         (vst1q_s64_x2): Likewise.
11356         (vst1q_u8_x2): Likewise.
11357         (vst1q_u16_x2): Likewise.
11358         (vst1q_u32_x2): Likewise.
11359         (vst1q_u64_x2): Likewise.
11360         (vst1q_f16_x2): Likewise.
11361         (vst1q_f32_x2): Likewise.
11362         (vst1q_f64_x2): Likewise.
11363         (vst1q_p64_x2): Likewise.
11364         (vst1_s64_x3): Likewise.
11365         (vst1_u64_x3): Likewise.
11366         (vst1_f64_x3): Likewise.
11367         (vst1_s8_x3): Likewise.
11368         (vst1_p8_x3): Likewise.
11369         (vst1_s16_x3): Likewise.
11370         (vst1_p16_x3): Likewise.
11371         (vst1_s32_x3): Likewise.
11372         (vst1_u8_x3): Likewise.
11373         (vst1_u16_x3): Likewise.
11374         (vst1_u32_x3): Likewise.
11375         (vst1_f16_x3): Likewise.
11376         (vst1_f32_x3): Likewise.
11377         (vst1_p64_x3): Likewise.
11378         (vst1q_s8_x3): Likewise.
11379         (vst1q_p8_x3): Likewise.
11380         (vst1q_s16_x3): Likewise.
11381         (vst1q_p16_x3): Likewise.
11382         (vst1q_s32_x3): Likewise.
11383         (vst1q_s64_x3): Likewise.
11384         (vst1q_u8_x3): Likewise.
11385         (vst1q_u16_x3): Likewise.
11386         (vst1q_u32_x3): Likewise.
11387         (vst1q_u64_x3): Likewise.
11388         (vst1q_f16_x3): Likewise.
11389         (vst1q_f32_x3): Likewise.
11390         (vst1q_f64_x3): Likewise.
11391         (vst1q_p64_x3): Likewise.
11392
11393 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11394
11395         * config/msp430/msp430.c (msp430_output_labelref): Prepend
11396         user_label_prefix to name.
11397
11398         * tree-core.h: Update comment about the format of NAME string
11399         passed to handler in attribute_spec.
11400
11401         * config/msp430/msp430.md: Remove erroneous subreg expression from
11402         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11403         zero_extend{q,h}isi2.
11404
11405 2018-05-30  Borislav Petkov  <bp@suse.de>
11406
11407         * doc/extend.texi: Document some architecture specific
11408         constraints and sort entries.
11409
11410 2018-05-30  Martin Sebor  <msebor@redhat.com>
11411
11412         PR middle-end/85369
11413         * builtins.c (expand_builtin_stpcpy_1): New function.
11414         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11415         only if the former succeeds.
11416
11417 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
11418
11419         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11420         in saphira.
11421
11422 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11423
11424         * doc/invoke.texi (-flinker-output): Document
11425
11426 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11427
11428         * passes.c (ipa_write_summaries): Only modify statements if body
11429         is in memory.
11430         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11431         incrementally linking.
11432         (ipa_passes): Likewise.
11433         * lto-cgraph.c (lto_output_node): When incrementally linking do not
11434         pass down resolution info.
11435         * common.opt (flag_incremental_link): Update info.
11436         * gcc.c (plugin specs): Turn flinker-output=* to
11437         -plugin-opt=-linker-output-known
11438         * toplev.c (compile_file): Also cut compilation when doing incremental
11439         link.
11440         * flag-types. (enum lto_partition_model): Add
11441         LTO_LINKER_OUTPUT_NOLTOREL.
11442         (invoke.texi): Add -flinker-output docs.
11443         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11444         link same way as WPA; do not stream in dead initializers.
11445
11446         * dwarf2out.c (dwarf2out_die_ref_for_decl,
11447         darf2out_register_external_decl): Support incremental link.
11448
11449 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11450
11451         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11452
11453 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11454
11455         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11456         it down to simple_object_copy_lto_debug_sections.
11457         (run_gcc): Determine incremental LTO link time and configure
11458         lto1 into non-wpa mode, disable renaming of debug sections.
11459
11460 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11461
11462         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11463         descriptions of various incorrectly documented functions.
11464
11465 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11466
11467         Revert:
11468         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11469         address check not strict.
11470
11471 2018-05-30  Richard Biener  <rguenther@suse.de>
11472
11473         PR tree-optimization/85964
11474         * tracer.c (better_p): Drop initialized count check, we only
11475         call the function with initialized counts now.
11476         (find_best_successor): Do find a best edge if one
11477         has uninitialized count.
11478         (find_best_predecessor): Likewise.  Do BB frequency check only
11479         if count is initialized.
11480
11481 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
11482
11483         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11484         (aarch64_ldrstr_offset_compare): New.
11485         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11486         load/store orderings.
11487         (aarch64_gen_adjusted_ldpstp): Likewise.
11488
11489 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
11490
11491         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11492         Check for subset of GENERAL_REGS and FP_REGS.
11493         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11494         r=w alternative.
11495
11496 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
11497
11498         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11499         and wi::to_poly_offset.  Add the current offset and then check
11500         whether the sum fits, rather than using an unchecked addition of
11501         a checked term.  Check for a shwi rather than a uhwi.
11502         * expr.c (get_bit_range): Use tree_to_poly_uint64.
11503         (store_constructor): Use poly_int_tree_p.
11504         (expand_expr_real_1): Likewise.
11505         * function.c (assign_temp): Likewise.
11506         * fold-const.c (const_binop): Use poly_int_tree_p and
11507         wi::to_poly_offset.
11508         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
11509         division.
11510         * ipa-icf-gimple.c (func_checker::compare_operand): Use
11511         to_poly_offset for MEM offsets.
11512         * ipa-icf.c (sem_variable::equals): Likewise.
11513         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11514         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11515         wi::to_poly_offset for BIT_FIELD_REF offsets.
11516         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11517         wi::to_poly_offset.
11518         * var-tracking.c (emit_note_insn_var_location): Use
11519         tree_to_poly_uint64.
11520
11521 2018-05-29  Jim Wilson  <jimw@sifive.com>
11522
11523         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11524
11525 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
11526
11527         PR target/85950
11528         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11529         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11530         sequence.
11531         (sse4_1_round<mode>2): Use nonimmediate_operand
11532         for operand 1 predicate.
11533
11534 2018-05-29  Martin Sebor  <msebor@redhat.com>
11535             Richard Biener  <rguenther@suse.de>
11536
11537         PR testsuite/85888
11538         * calls.c (get_size_range): Call determine_value_range instead
11539         of get_value_range..
11540         * tree-vrp.h (determine_value_range): Declared new function.
11541         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11542
11543 2018-05-29  Richard Biener  <rguenther@suse.de>
11544
11545         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11546         sure to use non-pattern stmts for get_earlier_stmt arguments.
11547         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11548         called on pattern stmts.
11549         (get_later_stmt): Likewise.
11550
11551 2018-05-29  Martin Liska  <mliska@suse.cz>
11552
11553         PR gcov-profile/85759
11554         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11555         env variables.
11556
11557 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
11558
11559         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11560         VEC_UNPACK_*_EXPR.
11561         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11562         VEC_PACK_*_EXPR.
11563
11564         PR target/85918
11565         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11566         VEC_PACK_FLOAT_EXPR): New tree codes.
11567         * tree-pretty-print.c (op_code_prio): Handle
11568         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11569         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11570         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11571         * tree-inline.c (estimate_operator_cost): Likewise.
11572         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11573         * fold-const.c (const_binop): Likewise.
11574         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11575         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11576         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11577         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11578         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11579         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11580         * expr.c (expand_expr_real_2): Likewise.
11581         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11582         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11583         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11584         optabs.
11585         * optabs.c (expand_widen_pattern_expr): For
11586         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11587         sign from result type rather than operand's type.
11588         (expand_binop_directly): For vec_packu_float_optab and
11589         vec_packs_float_optab allow result type to be different from operand's
11590         type.
11591         * optabs-tree.c (optab_for_tree_code): Handle
11592         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11593         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
11594         * tree-vect-generic.c (expand_vector_operations_1):  Handle
11595         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11596         VEC_PACK_FLOAT_EXPR.
11597         * tree-vect-stmts.c (supportable_widening_operation): Handle
11598         FIX_TRUNC_EXPR.
11599         (supportable_narrowing_operation): Handle FLOAT_EXPR.
11600         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11601         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11602         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
11603         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11604         mode attributes.
11605         (vec_pack<floatprefix>_float_<mode>): New expander.
11606         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11607         attributes.
11608         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11609         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11610         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11611         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11612         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11613         Document.
11614         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11615         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11616         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11617         VEC_PACK_FLOAT_EXPR): Document.
11618
11619 2018-05-29  Richard Biener  <rguenther@suse.de>
11620
11621         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11622         member.
11623         (stmt_vec_info_vec): Make pointer.
11624         (init_stmt_vec_info_vec): Remove.
11625         (free_stmt_vec_info_vec): Likewise.
11626         (set_stmt_vec_info_vec): New function.
11627         (free_stmt_vec_infos): Likewise.
11628         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11629         (set_vinfo_for_stmt): Likewise.
11630         (get_earlier_stmt): Likewise.
11631         (get_later_stmt): Likewise.
11632         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11633         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11634         (vec_info::~vec_info): Free stmt_vec_infos.
11635         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11636         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11637         (pass_slp_vectorize::execute): Likewise.
11638         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11639         (free_stmt_vec_info_vec): Likewise.
11640         (set_stmt_vec_info_vec): New function.
11641         (free_stmt_vec_infos): Likewise.
11642         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11643         the global stmt_vec_info_vec.
11644         * tree-parloops.c (gather_scalar_reductions): Use
11645         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11646         vector.
11647
11648 2018-05-29  Richard Biener  <rguenther@suse.de>
11649
11650         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11651
11652 2018-05-29  Martin Liska  <mliska@suse.cz>
11653             David Malcolm  <dmalcolm@redhat.com>
11654
11655         * vec.c (test_reverse): New.
11656         (vec_c_tests): Add new test.
11657         * vec.h (vl_ptr>::reverse): New function.
11658
11659 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
11660
11661         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11662
11663         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11664         and later.
11665
11666 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11667
11668         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11669
11670 2018-05-28  Richard Biener  <rguenther@suse.de>
11671
11672         PR tree-optimization/85933
11673         * tree-vect-data-refs.c (vect_record_base_alignments): Only
11674         look at stmts marked as vectorizable.
11675
11676 2018-05-28  Richard Biener  <rguenther@suse.de>
11677
11678         PR tree-optimization/85934
11679         * tree-vect-generic.c (expand_vector_operations_1): Hoist
11680         vector boolean check before scalar optimization.
11681
11682 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
11683
11684         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11685         for armv5te.
11686
11687 2018-05-28  Mark Wielaard  <mark@klomp.org>
11688
11689         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11690         if it is an expression containing a minus sign.
11691
11692 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
11693
11694         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11695
11696 2018-05-27  Paul Koning  <ni1d@arrl.net>
11697
11698         * config/pdp11/pdp11.md (truncsihi2): Remove.
11699
11700 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
11701             Chung-Ju Wu  <jasonwucj@gmail.com>
11702
11703         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11704         implementation.
11705         (unaligned_store_dw): Ditto.
11706         * config/nds32/nds32-memory-manipulation.c
11707         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11708         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11709         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11710         (emit_setmem_word_loop): Rename to ...
11711         (emit_setmem_doubleword_loop): ... this.
11712         (nds32_gen_dup_4_byte_to_word_value): New function.
11713         (nds32_gen_dup_8_byte_to_double_word_value): New function.
11714         (nds32_expand_setmem_loop): Refine implementation.
11715         (nds32_expand_setmem_loop_v3m): Ditto.
11716         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11717         pattern.
11718
11719 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
11720
11721         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11722
11723 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
11724
11725         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11726         (nds32_init_machine_status): Initialize machine->attr_naked_p and
11727         machine->attr_no_prologue_p.
11728         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11729         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11730         (nds32_expand_epilogue): Consider attr_naked_p.
11731         (nds32_expand_epilogue_v3pop): Likewise.
11732         (nds32_can_use_return_insn): Likewise.
11733         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11734         attr_no_prologue_p fields.
11735         * config/nds32/nds32.opt (mret-in-naked-func): New option.
11736
11737 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
11738
11739         PR target/85918
11740         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11741         attributes.
11742         * config/i386/sse.md
11743         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11744         Rename to ...
11745         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11746         ... this.
11747         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11748         Rename to ...
11749         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11750         ... this.
11751         (*<floatsuffix>floatv2div2sf2): Rename to ...
11752         (*float<floatunssuffix>v2div2sf2): ... this.
11753         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11754         (float<floatunssuffix>v2div2sf2_mask): ... this.
11755         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11756         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11757         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11758         to ...
11759         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11760         ... this.
11761         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11762         Rename to ...
11763         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11764         ... this.
11765         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11766         Rename to ...
11767         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11768         ... this.
11769         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11770         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11771         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11772         gen_ufix_truncv8dfv8si2.
11773         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11774         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11775         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11776         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11777         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11778         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11779         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11780         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11781
11782 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
11783
11784         PR target/85900
11785         PR target/85345
11786         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11787
11788 2018-05-25  Jim Wilson  <jimw@sifive.com>
11789
11790         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11791         * config/riscv/riscv.c (struct machine_function): Add
11792         interrupt_handler_p and attribute_checked_p fields.
11793         (riscv_attribute_table): Add interrupt.
11794         (riscv_interrupt_type_p): New.
11795         (riscv_save_reg_p): Save extra regs for interrupt handler.
11796         (riscv_use_save_libcall): Return false  for interrupt handler.
11797         (riscv_first_stack_step): Add forward declaration.
11798         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
11799         for interrupt handler with large frame.  Use it for saved reg list.
11800         (riscv_expand_prologue): Move flag_stack_usage_info support to
11801         eliminate duplication.
11802         (riscv_expand_epilogue): Generate mret for interrupt handler.
11803         (riscv_epilogue_uses): New.
11804         (riscv_can_use_return_insn): Return false for interrupt handler.
11805         (riscv_function_ok_for_sibcall): Likewise.
11806         (riscv_set_current_function): Add interrupt handler support.
11807         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11808         * config/riscv/riscv.md (UNSPECV_MRET): New.
11809         (GP_REGNUM): New.
11810         (riscv_frflags, riscv_fsflags): Use tab after opcode.
11811         (riscv_mret): New.
11812         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11813
11814 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
11815
11816         PR tree-optimization/85712
11817         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11818         this candidate has already been replaced in-situ by a copy.
11819
11820 2018-05-25  Jason Merrill  <jason@redhat.com>
11821
11822         PR c++/80485 - inline function non-zero address.
11823         * symtab.c (nonzero_address): Check DECL_COMDAT.
11824
11825 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
11826
11827         PR target/83628
11828         * config/alpha/alpha.md (ashlsi3): New insn pattern.
11829         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
11830         extension of SImode operation.  Use const123_operand predicate.
11831         (*saddsi_1): Remove.
11832         (*saddl_se_1): Ditto.
11833         (*ssubsi_1): Ditto.
11834         (*ssubl_se_1): Ditto.
11835         * config/alpha/predicates.md (const123_operand): New predicate.
11836         * config/alpha/constraints.md (P): Use IN_RANGE.
11837
11838 2018-05-25  Richard Biener  <rguenther@suse.de>
11839
11840         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11841         defaulted to true.
11842         (ref_maybe_used_by_stmt_p): Likewise.
11843         (stmt_may_clobber_ref_p): Likewise.
11844         (stmt_may_clobber_ref_p_1): Likewise.
11845         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11846         and pass it along.
11847         (ref_maybe_used_by_stmt_p): Likewise.
11848         (stmt_may_clobber_ref_p): Likewise.
11849         (stmt_may_clobber_ref_p_1): Likewise.
11850         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11851         the alias oracle to disambiguate DRs with stmts DR analysis
11852         couldn't handle.
11853         (vect_analyze_data_refs): Do not give up on not analyzable
11854         DRs for BB vectorization.  Remove code truncating the dataref
11855         vector.
11856
11857 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
11858
11859         PR target/85832
11860         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11861         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
11862         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11863
11864 2018-05-25  Richard Biener  <rguenther@suse.de>
11865
11866         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11867         function, combining stmt data ref gathering and fatal analysis
11868         parts.
11869         (vect_analyze_data_refs): Remove now redudnant code and simplify.
11870         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11871         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11872         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11873         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11874
11875 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
11876
11877         PR tree-optimization/85720
11878         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11879         SCC if all partitions are builtins.
11880         (version_loop_by_alias_check): New parameter.  Generate cancelable
11881         runtime alias check if all partitions are builtins.
11882         (distribute_loop): Update call to above function.
11883
11884 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
11885
11886         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11887         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11888         (parm_default_def_partition_arg): Ditto.
11889         (set_parm_default_def_partition): Ditto.
11890         (get_parm_default_def_partitions): Ditto and make it static.
11891         (get_undefined_value_partitions): Ditto and make it static.
11892         (remove_ssa_form): Refactor call to init_var_map here.
11893         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11894         computation for loop region.
11895         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11896         (register_default_def): Delete.
11897         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11898         (parm_default_def_partition_arg): Ditto.
11899         (set_parm_default_def_partition): Ditto.
11900         (get_parm_default_def_partitions): Ditto and make it static.
11901         (get_undefined_value_partitions): Ditto and make it static.
11902         (coalesce_with_default, coalesce_with_default): Update comment.
11903         (create_coalesce_list_for_region): New func factored out from
11904         create_outofssa_var_map.
11905         (populate_coalesce_list_for_outofssa): New func factored out from
11906         create_outofssa_var_map and coalesce_ssa_name.
11907         (create_outofssa_var_map): Delete.
11908         (coalesce_ssa_name): Refactor to support live range computation.
11909         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11910         (get_parm_default_def_partitions): Delete.
11911         (get_undefined_value_partitions): Ditto.
11912         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11913         computation for loop region.
11914         (new_tree_live_info, loe_visit_block): Ditto.
11915         (live_worklist, set_var_live_on_entry): Ditto.
11916         (calculate_live_on_exit, verify_live_on_entry): Ditto.
11917         * tree-ssa-live.h (struct _var_map): New fields.
11918         (init_var_map): Change decl.
11919         (region_contains_p): New.
11920
11921 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
11922
11923         * tree-ssa-live.h (live_merge_and_clear): Delete.
11924
11925 2018-05-25  Richard Biener  <rguenther@suse.de>
11926
11927         PR c++/85912
11928         * tree-dump.c (dequeue_and_dump): Remove access to removed
11929         operand 2 of a SWITCH_EXPR.
11930
11931 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11932
11933         * doc/sourcebuild.texi (vect_double_cond_arith): Include
11934         multiplication and division.
11935         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11936         (cond_udiv@var{m}, cond_umod@var{m}): Document.
11937         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11938         (cond_udiv_optab, cond_umod_optab): New optabs.
11939         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11940         (IFN_COND_RDIV): New internal functions.
11941         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11942         TRUNC_MOD_EXPR and RDIV_EXPR.
11943         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11944         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11945         New unspecs.
11946         (SVE_INT_BINARY): Include mult.
11947         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11948         (optab, sve_int_op): Handle mult.
11949         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11950         UNSPEC_COND_DIV.
11951         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11952         for SVE_INT_BINARY_SD.
11953
11954 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11955
11956         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11957         (optab, sve_int_op): Handle div and udiv.
11958         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11959         for SVE_INT_BINARY_SD.
11960         (*<optab><mode>3): New insn for the same.
11961
11962 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11963
11964         * tree-vect-patterns.c: Include predict.h.
11965         (vect_recog_divmod_pattern): Restrict check for division support
11966         to when optimizing for size.
11967
11968 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11969
11970         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11971         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11972         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11973         (gimple_match_op::set_op): Likewise.
11974         (gimple_resimplify4): Declare.
11975         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11976         (expr::gen_transform): Likewise.
11977         (decision_tree::gen): Generate a simplification routine for 4 operands.
11978         * gimple-match-head.c (gimple_simplify): Add an overload for
11979         4 operands.  In the top-level function, handle up to 4 call
11980         arguments and call gimple_resimplify4.
11981         (gimple_resimplify4): New function.
11982         (build_call_internal): Pass a fourth operand.
11983         (maybe_push_to_seq): Likewise.
11984         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11985         Fold VEC_COND_EXPRs of an operation and a default value into
11986         an IFN_COND_* function if possible.
11987         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11988         New unspecs.
11989         (SVE_COND_FP_BINARY): Include them.
11990         (optab, sve_fp_op): Handle them.
11991         (SVE_INT_BINARY_REV): New code iterator.
11992         (SVE_COND_FP_BINARY_REV): New int iterator.
11993         (commutative): New int attribute.
11994         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11995         Declare.
11996         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11997         function.
11998         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11999         (*cond_<optab><mode>): New patterns for reversed operands.
12000
12001 2018-05-25  Richard Biener  <rguenther@suse.de>
12002
12003         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
12004         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
12005         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
12006         (STMT_VINFO_GROUPED_ACCESS): Adjust.
12007         * tree-vect-data-refs.c (everywhere): Adjust users.
12008         * tree-vect-loop.c (everywhere): Likewise.
12009         * tree-vect-slp.c (everywhere): Likewise.
12010         * tree-vect-stmts.c (everywhere): Likewise.
12011         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12012
12013 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12014
12015         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
12016         Rename to...
12017         (gcc_cv_as_section_exclude): ... this.
12018         Try Solaris as #exclude syntax.
12019         * configure: Regenerate.
12020         * config.in: Regenerate.
12021         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
12022         SECTION_EXCLUDE.
12023         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
12024         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12025
12026         * varasm.c (default_elf_asm_named_section): Don't check if
12027         HAVE_GAS_SECTION_EXCLUDE is defined.
12028
12029 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12030
12031         * doc/md.texi: Update the documentation of the cond_* optabs
12032         to mention the new final operand.  Fix GET_MODE_NUNITS call.
12033         Describe the scalar case too.
12034         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12035         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12036         instead of 2.
12037         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12038         (get_conditional_internal_fn): Update comment.
12039         * tree-vect-loop.c (vectorizable_reduction): Pass the original
12040         accumulator value as a final argument to conditional functions.
12041         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12042         a define_expand and add an "else" operand.  Assert for now that
12043         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
12044         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12045         (*cond_<optab><mode>): New patterns.
12046         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12047         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12048         (UNSPEC_COND_EOR): Delete.
12049         (optab): Remove associated mappings.
12050         (SVE_INT_BINARY): New code iterator.
12051         (sve_int_op): Remove int attribute and add "minus" to the code
12052         attribute.
12053         (SVE_COND_INT_OP): Delete.
12054         (SVE_COND_FP_OP): Rename to...
12055         (SVE_COND_FP_BINARY): ...this.
12056
12057 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12058
12059         * optabs.c (can_reuse_operands_p): New function.
12060         (maybe_legitimize_operands): Try to reuse the results for
12061         earlier operands.
12062
12063 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
12064
12065         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12066         Add {q} suffix to insn mnemonic.
12067
12068 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12069
12070         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12071         (msp430_warn_func_return): New.
12072
12073 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
12074
12075         * fold-const.c (tree_nonzero_bits): New function.
12076         * fold-const.h (tree_nonzero_bits): Likewise.
12077         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12078         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12079
12080 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
12081
12082         PR target/85900
12083         PR target/85345
12084         * varasm.c (assemble_alias): Check ifunc_resolver only on
12085         FUNCTION_DECL.
12086
12087 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
12088
12089         PR target/85903
12090         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12091         when memory input operand is handled.
12092
12093 2018-05-24  Luis Machado  <luis.machado@linaro.org>
12094
12095         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12096         global.
12097         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12098
12099 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
12100
12101         * match.pd: Delay FMA folds until after vectorization.
12102
12103 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12104
12105         PR target/83009
12106         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12107         address check not strict.
12108
12109 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
12110
12111         * gimple-match.h (gimple_match_op): New class.
12112         (mprts_hook): Replace parameters with a gimple_match_op *.
12113         (maybe_build_generic_op): Likewise.
12114         (gimple_simplified_result_is_gimple_val): Replace parameters with
12115         a const gimple_match_op *.
12116         (gimple_simplify): Replace code_helper * and tree * parameters with
12117         a gimple_match_op * parameter.
12118         (gimple_resimplify1): Replace code_helper *, tree and tree *
12119         parameters with a gimple_match_op * parameter.
12120         (gimple_resimplify2): Likewise.
12121         (gimple_resimplify3): Likewise.
12122         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12123         parameters with a gimple_match_op * parameter.
12124         * gimple-match-head.c (gimple_simplify): Change prototypes of
12125         auto-generated functions to take a gimple_match_op * instead of
12126         separate code_helper * and tree * parameters.  Make the same
12127         change in the top-level overload and update calls to the
12128         gimple_resimplify routines.  Update calls to the auto-generated
12129         functions and to maybe_push_res_to_seq in the publicly-facing
12130         operation-specific gimple_simplify overloads.
12131         (gimple_match_op::MAX_NUM_OPS): Define.
12132         (gimple_resimplify1): Replace rcode and ops with a single res_op
12133         parameter.  Update call to gimple_simplify.
12134         (gimple_resimplify2): Likewise.
12135         (gimple_resimplify3): Likewise.
12136         (mprts_hook): Replace parameters with a gimple_match_op *.
12137         (maybe_build_generic_op): Likewise.
12138         (build_call_internal): Replace type, nargs and ops with
12139         a gimple_match_op *.
12140         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12141         with a single gimple_match_op *.  Update calls to mprts_hook,
12142         build_call_internal and gimple_simplified_result_is_gimple_val.
12143         Factor out code that is common to the tree_code and combined_fn cases.
12144         * genmatch.c (expr::gen_transform): Replace tem_code and
12145         tem_ops with a gimple_match_op called tem_op.  Update calls
12146         to the gimple_resimplify functions and maybe_push_res_to_seq.
12147         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12148         res_ops.  Update call to the gimple_resimplify functions.
12149         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12150         (decision_tree::gen): Make the functions take a gimple_match_op *
12151         called res_op instead of separate res_code and res_ops parameters.
12152         Update call accordingly.
12153         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12154         and ops with a single res_op parameter.  Update calls to
12155         maybe_build_generic_op and maybe_push_res_to_seq.
12156         (fold_stmt_1): Update calls to gimple_simplify and
12157         replace_stmt_with_simplification.
12158         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12159         and gimple_simplified_result_is_gimple_val.
12160         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12161         gimple_simplify.
12162         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12163         with a gimple_match_op *.
12164         (vn_nary_build_or_lookup): Likewise.  Update call to
12165         vn_nary_build_or_lookup_1.
12166         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12167         gimple_match_op *.  Update calls to the gimple_resimplify routines
12168         and to gimple_simplified_result_is_gimple_val.
12169         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12170         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12171         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12172         (visit_nary_op): Likewise.
12173         (visit_reference_op_load): Likewise.
12174
12175 2018-05-23  Luis Machado  <luis.machado@linaro.org>
12176
12177         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12178         modifier for printing the step amount.
12179
12180 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
12181
12182         PR target/78849
12183         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12184         types.
12185
12186 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
12187
12188         * doc/sourcebuild.texi (Endianness): New subsubsection.
12189
12190 2018-05-23  Luis Machado  <luis.machado@linaro.org>
12191
12192         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12193         <prefetch_dynamic_strides>: New const bool field.
12194         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12195         prefetch_dynamic_strides.
12196         (exynosm1_prefetch_tune): Likewise.
12197         (thunderxt88_prefetch_tune): Likewise.
12198         (thunderx_prefetch_tune): Likewise.
12199         (thunderx2t99_prefetch_tune): Likewise.
12200         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12201         false.
12202         (aarch64_override_options_internal): Update to set
12203         PARAM_PREFETCH_DYNAMIC_STRIDES.
12204         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12205         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12206         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12207         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12208         prefetch-dynamic-strides setting.
12209
12210 2018-05-23  Luis Machado  <luis.machado@linaro.org>
12211
12212         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12213         <minimum_stride>: New const int field.
12214         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12215         minimum_stride field defaulting to -1.
12216         (exynosm1_prefetch_tune): Likewise.
12217         (thunderxt88_prefetch_tune): Likewise.
12218         (thunderx_prefetch_tune): Likewise.
12219         (thunderx2t99_prefetch_tune): Likewise.
12220         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12221         <default_opt_level>: Set to 3.
12222         (aarch64_override_options_internal): Update to set
12223         PARAM_PREFETCH_MINIMUM_STRIDE.
12224         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12225         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12226         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12227         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12228         stride is constant and is below the minimum stride threshold.
12229
12230 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12231
12232         * config/arm/arm-cpus.in (mode26): Delete.
12233         (armv4): Delete mode26 reference.
12234         * config/arm/arm.c (arm_configure_build_target): Delete use of
12235         isa_bit_mode26.
12236
12237 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
12238
12239         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12240         New insn pattern.
12241         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
12242         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
12243         for non-SSE modes.
12244         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
12245         (floatunsdidf2): Ditto.
12246
12247 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
12248
12249         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
12250         (fixuns_trunc<mode>si2_avx512f): Ditto.
12251         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
12252         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
12253         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
12254
12255 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
12256
12257         PR rtl-optimization/79985
12258         * df-scan.c (df_insn_refs_collect): Remove special case for
12259         global registers and asm statements.
12260
12261 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
12262
12263         * extend.texi (Global Register Variables): Rewrite the bullet list.
12264         Note that the register is available for allocation. Note that access
12265         via inline asm must use constraints. Add note about async-signal
12266         handlers. Remove paragraph about automagic register selection.
12267
12268 2018-05-23  Richard Biener  <rguenther@suse.de>
12269
12270         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
12271         of fixed offset from memset VN.
12272
12273 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
12274
12275         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
12276         first_interp field.
12277         (alloc_cand_and_find_basis): Initialize first_interp field.
12278         (slsr_process_mul): Modify first_interp field.
12279         (slsr_process_add): Likewise.
12280         (slsr_process_cast): Modify first_interp field for each new
12281         interpretation.
12282         (slsr_process_copy): Likewise.
12283         (dump_candidate): Dump first_interp field.
12284         (replace_mult_candidate): Process all interpretations, not just
12285         subsequent ones.
12286         (replace_rhs_if_not_dup): Likewise.
12287         (replace_one_candidate): Likewise.
12288
12289 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
12290
12291         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
12292         Add new boolean.
12293         (aarch64_needs_frame_chain): New function.
12294         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
12295
12296 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
12297
12298         PR target/84882
12299         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12300         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
12301         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
12302         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12303         as true for strict-align.
12304         (aarch64_can_inline_p): Perform checks even when callee has no
12305         attributes to check for strict alignment.
12306         * doc/extend.texi (AArch64 Function Attributes): Document
12307         no-strict-align.
12308         * doc/invoke.texi: (AArch64 Options): Likewise.
12309
12310 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
12311
12312         PR tree-optimization/85853
12313         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12314         the handling of the root of the node to...
12315         (vect_slp_analyze_node_operations_1): ...this new function,
12316         and run the whole thing with the child nodes' def types
12317         set according to their SLP node's def type.
12318
12319 2018-05-23  Richard Biener  <rguenther@suse.de>
12320
12321         PR middle-end/85874
12322         * tree-data-ref.c (create_runtime_alias_checks): Defer
12323         and ignore overflow warnings.
12324
12325 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
12326
12327         PR tree-optimization/85822
12328         * tree-vrp.c (is_masked_range_test): Fix handling of negative
12329         constants.
12330
12331 2018-05-23  Richard Biener  <rguenther@suse.de>
12332
12333         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12334         memset constants via native_interpret_expr.
12335
12336 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
12337
12338         PR target/85345
12339         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12340         attribute.
12341         (cgraph_node::create_alias): Likewise.
12342         (cgraph_node::get_availability): Check ifunc_resolver instead
12343         of looking up ifunc attribute.
12344         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12345         * varasm.c (do_assemble_alias): Likewise.
12346         (assemble_alias): Likewise.
12347         (default_binds_local_p_3): Likewise.
12348         * cgraph.h (cgraph_node): Add ifunc_resolver.
12349         (cgraph_node::only_called_directly_or_aliased_p): Return false
12350         for IFUNC resolver.
12351         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12352         attribute.
12353         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12354         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12355         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12356         instead of looking up ifunc attribute.
12357
12358 2018-05-22  Luis Machado  <luis.machado@linaro.org>
12359
12360         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12361
12362 2018-05-22  Martin Sebor  <msebor@redhat.com>
12363
12364         PR middle-end/85359
12365         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12366         only when expasion succeeds.
12367         (expand_builtin_strcmp): Same.
12368         (expand_builtin_strncmp): Same.
12369
12370 2018-05-22  Martin Sebor  <msebor@redhat.com>
12371
12372         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12373
12374 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
12375             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12376
12377         * config/aarch64/aarch64-ldpstp.md: Replace uses of
12378         aarch64_mem_pair_operand with memory_operand and delete operand swapping
12379         code.
12380         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12381         Add check for legitimate_address.
12382         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12383         (aarch64_swap_ldrstr_operands): New.
12384         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12385         Define prototype.
12386
12387 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
12388             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12389
12390         * config/aarch64/aarch64.md: New patterns to generate stp
12391         and ldp.
12392         (store_pair_sw, store_pair_dw): New patterns to generate stp for
12393         single words and double words.
12394         (load_pair_sw, load_pair_dw): Likewise.
12395         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12396         Delete.
12397         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12398         Delete.
12399         * config/aarch64/aarch64-ldpstp.md: Modify peephole
12400         for different mode ldpstp and add peephole for merged zero stores.
12401         Likewise for loads.
12402         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12403         Add size check.
12404         (aarch64_gen_store_pair): Rename calls to match new patterns.
12405         (aarch64_gen_load_pair): Rename calls to match new patterns.
12406         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12407         (load_pair<DREG:mode><DREG2:mode>): ... This.
12408         (store_pair<mode>): Rename to...
12409         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12410         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12411         New mode iterators.
12412         (V_INT_EQUIV): Handle SImode.
12413         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12414         New predicate.
12415
12416 2018-05-22  Martin Sebor  <msebor@redhat.com>
12417
12418         PR c/85623
12419         * calls.c (maybe_warn_nonstring_arg): Use string length to set
12420         or ajust the presumed bound on an operation to avoid unnecessary
12421         warnings.
12422
12423 2018-05-22  Martin Sebor  <msebor@redhat.com>
12424
12425         PR tree-optimization/85826
12426         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12427         assuming that a DECL necesarily has a constant size.
12428
12429 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
12430
12431         PR middle-end/85862
12432         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12433
12434 2018-05-22  Richard Biener  <rguenther@suse.de>
12435
12436         PR tree-optimization/85834
12437         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12438         non-constant and non-zero memset arguments.
12439
12440 2018-05-22  Martin Liska  <mliska@suse.cz>
12441
12442         PR ipa/85607
12443         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12444
12445 2018-05-22  Richard Biener  <rguenther@suse.de>
12446
12447         PR tree-optimization/85863
12448         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12449         comparisons when vectype is specified.
12450         (vectorizable_condition): Do not specify vectype for
12451         vect_is_simple_cond when SLP vectorizing.
12452
12453 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
12454
12455         PR target/85657
12456         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12457         define __ibm128 as long double.
12458         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12459         as a distinct type when IEEE 128-bit support is enabled.
12460         (init_float128_ieee): Fix up conversions between IFmode and IEEE
12461         128-bit types to use the correct functions.
12462         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12463         convert between 128-bit floating point types that have different
12464         modes but the same representation, instead of using gen_lowpart to
12465         makean alias.
12466         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12467         KFmode.
12468         (IFKF_reg): New attributes to give the register constraints for
12469         IFmode and KFmode.
12470         (extend<mode>tf2_internal): New insns to mark an explicit
12471         conversion between 128-bit floating point types that have a
12472         different mode but share the same representation.
12473
12474 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
12475
12476         PR tree-optimization/85814
12477         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12478         a null return from get_strinfo when unsharing the next
12479         strinfo in the chain.
12480
12481 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
12482
12483         PR gcc/84923
12484         * varasm.c (weak_finish): Clean up weak_decls.
12485
12486 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12487
12488         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12489         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12490         UNSPEC_UADALP values.
12491         * config/aarch64/iterators.md (ABAL): New int iterator.
12492         (ABDL2): Likewise.
12493         (ADALP): Likewise.
12494         (sur): Add mappings for the above.
12495         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12496         New define_insn.
12497         (aarch64_<sur>abal<mode>_4): Likewise.
12498         (aarch64_<sur>adalp<mode>_3): Likewise.
12499         (<sur>sadv16qi): New define_expand.
12500
12501 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
12502
12503         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12504         (*movdf_internal): Ditto.
12505         (*rcpsf2_sse): Ditto.
12506         (*rsqrtsf2_sse): Ditto.
12507         (*sqrt<mode>2_sse): Ditto.
12508
12509 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
12510
12511         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12512         eor3q<mode>4.
12513         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12514         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12515         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12516         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12517         vbcaxq_s64): New.
12518         * config/aarch64/arm_neon.h: Likewise.
12519         * config/aarch64/iterators.md (VQ_I): New.
12520
12521 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
12522
12523         * config.gcc: Add arc/t-multilib-linux to tmake_file for
12524         arc*-*-linux*.
12525         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12526         MULTILIB_DIRNAMES
12527
12528 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
12529
12530         * config/nds32/constraints.md (S): New constraint.
12531         * config/nds32/nds32.md (call_internal): Use constraint S.
12532         (call_value_internal): Likewise.
12533         (sibcall_internal): Likewise.
12534         (sibcall_value_internal): Likewise.
12535
12536 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
12537             Chung-Ju Wu  <jasonwucj@gmail.com>
12538
12539         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12540         into consideration.
12541
12542 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
12543             Chung-Ju Wu  <jasonwucj@gmail.com>
12544
12545         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12546         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12547         (nds32_rtx_costs_impl): Simplify.
12548         (nds32_address_cost_impl): Simplify.
12549         (nds32_init_rtx_costs): New function.
12550         (nds32_rtx_costs_speed_prefer): Likewise.
12551         (nds32_rtx_costs_size_prefer): Likewise.
12552         (nds32_address_cost_speed_prefer): Likewise.
12553         (nds32_address_cost_speed_fwprop): Likewise.
12554         (nds32_address_cost_size_prefer): Likewise.
12555         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12556         * config/nds32/nds32.c (nds32_option_override): Use
12557         nds32_init_rtx_costs function.
12558
12559 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
12560
12561         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12562         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12563         (TARGET_PIPELINE_N8): Likewise.
12564         (TARGET_PIPELINE_N10): Likewise.
12565         (TARGET_PIPELINE_N13): Likewise.
12566         (TARGET_PIPELINE_GRAYWOLF): Likewise.
12567
12568 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
12569
12570         * config/nds32/nds32-fpu.md: Update copyright year.
12571
12572 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12573
12574         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12575
12576 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12577
12578         * config/nds32/nds32.c
12579         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12580         * config/nds32/nds32.opt (minline-asm-r15): New option.
12581
12582 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12583
12584         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12585         MASK_HW_ABS.
12586         * config/nds32/nds32.md (abssi2): New pattern.
12587
12588 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
12589
12590         * config/i386/i386.md (rex64namesuffix): New mode attribute.
12591         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12592         Merge insn pattern from sse_cvtsi2ss<round_name> and
12593         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12594         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12595         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12596         using SWI48 mode iterator.
12597         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12598         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12599         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12600         pattern from sse_cvttss2si<round_saeonly_name>
12601         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12602         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12603         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12604         using SWI48 mode iterator.
12605         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12606         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12607         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12608         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12609         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12610         using SWI48 mode iterator.
12611         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12612         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12613         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12614         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12615         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12616         SWI48 mode iterator.
12617         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12618         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12619         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12620         pattern from sse_cvttsd2si<round_saeonly_name>
12621         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12622
12623 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12624
12625         * config/nds32/nds32-md-auxiliary.c
12626         (nds32_valid_smw_lwm_base_p): Refine.
12627         (nds32_output_smw_single_word): Refine.
12628         (nds32_output_smw_double_word): New.
12629         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12630
12631 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12632
12633         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12634         (nds32_output_stack_pop): Refine.
12635         (nds32_expand_unaligned_load): Refine.
12636         (nds32_expand_unaligned_store): Refine.
12637
12638 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
12639             Chung-Ju Wu  <jasonwucj@gmail.com>
12640
12641         * config/nds32/constants.md: Add TP_REGNUM constant.
12642         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12643         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12644         UNSPEC_ADD32.
12645         * config/nds32/nds32-doubleword.md: Consider flag_pic.
12646         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12647         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12648         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12649         and PIC code generation.
12650         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12651         code generation.
12652         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12653         optimization.
12654         * config/nds32/nds32.md: Support TLS and PIC.
12655         * config/nds32/nds32.c: Support TLS and PIC.
12656         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12657         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12658         predicate.
12659
12660 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12661
12662         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12663         mode with E_ prefix.
12664
12665 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
12666             Chung-Ju Wu  <jasonwucj@gmail.com>
12667
12668         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12669         * config/nds32/nds32-md-auxiliary.c
12670         (symbolic_reference_mentioned_p): New.
12671         (nds32_legitimize_ict_address): New.
12672         (nds32_expand_ict_move): New.
12673         (nds32_indirect_call_referenced_p): New.
12674         (nds32_symbol_binds_local_p): Delete.
12675         (nds32_long_call_p): Modify.
12676         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12677         * config/nds32/nds32-protos.h
12678         (symbolic_reference_mentioned_p): Declare.
12679         (nds32_legitimize_ict_address): Declare.
12680         (nds32_expand_ict_move): Declare.
12681         (nds32_indirect_call_referenced_p): Declare.
12682         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12683         (nds32_relax_group): Use nds32_ict_const_p as condition.
12684         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12685         (nds32_asm_file_start): Output ict_model directive in asm code.
12686         (nds32_legitimate_address_p): Consider indirect call.
12687         (nds32_print_operand): Consider indirect call.
12688         (nds32_print_operand_address): Consider indirect call.
12689         (nds32_insert_attributes): Handle "indirect_call" attribute.
12690         (TARGET_LEGITIMATE_ADDRESS_P): Define.
12691         (TARGET_LEGITIMATE_CONSTANT_P): Define.
12692         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12693         (TARGET_DELEGITIMIZE_ADDRESS): Define.
12694         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12695         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12696         (TARGET_ICT_MODEL_SMALL): Define.
12697         (TARGET_ICT_MODEL_LARGE): Define.
12698         * config/nds32/nds32.md (movsi): Consider ict model.
12699         (call, call_value): Consider ict model.
12700         (sibcall, sibcall_value): Consider ict model.
12701         * config/nds32/nds32.opt (mict-model): New option.
12702         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12703         model.
12704
12705 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
12706             Monk Chiang  <sh.chiang04@gmail.com>
12707             Jim Wilson <jimw@sifive.com>
12708
12709         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12710         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
12711         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12712         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12713         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
12714         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12715         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12716         compute save_libcall_adjustment properly.
12717         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12718         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12719         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12720         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12721         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12722         (ABI_SPEC): Handle mabi=ilp32e.
12723         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12724         (RVE): Add RVE mask.
12725         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12726         <-march>: Add rv32e as an example.
12727
12728 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
12729
12730         PR c++/82899
12731         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12732         (intra_create_variable_infos): Handle C++ constructors.
12733
12734 2018-05-18  Martin Liska  <mliska@suse.cz>
12735
12736         * passes.def: Remove a redundant pass.
12737
12738 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
12739
12740         PR bootstrap/85838
12741         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12742
12743 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12744
12745         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12746         (ARMv4): Update.
12747         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12748         (ARMv6m): Update.
12749         (armv2, armv2a, armv3, armv3m): Delete architectures.
12750         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12751         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12752         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12753         Delete cpus.
12754         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12755         (*mulsidi3adddi): Likewise.
12756         (mulsidi3): Likewise.
12757         (*mulsidi3_nov6): Likewise.
12758         (umulsidi3): Likewise.
12759         (umulsidi3_nov6): Likewise.
12760         (umaddsidi4): Likewise.
12761         (*umulsidi3adddi): Likewise.
12762         (smulsi3_highpart): Likewise.
12763         (*smulsi3_highpart_nov6): Likewise.
12764         (umulsi3_highpart): Likewise.
12765         (*umulsi3_highpart_nov6): Likewise.
12766         * config/arm/arm.h (arm_arch3m): Delete.
12767         * config/arm/arm.c (arm_arch3m): Delete.
12768         (arm_option_override_internal): Update armv3-related comment.
12769         (arm_configure_build_target): Delete use of isa_bit_mode32.
12770         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12771         (arm_rtx_costs_internal): Delete check of arm_arch3m.
12772         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12773         (mulsa3): Likewise.
12774         (mulusa3): Likewise.
12775         * config/arm/arm-protos.h (arm_arch3m): Delete.
12776         * config/arm/arm-tables.opt: Regenerate.
12777         * config/arm/arm-tune.md: Likewise.
12778         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12779         deleted architectures.
12780
12781 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12782
12783         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12784         (armv5t, armv5te): New features.
12785         (ARMv5, ARMv5e): Delete fgroups.
12786         (ARMv5t, ARMv5te): Adjust for above changes.
12787         (ARMv6m): Likewise.
12788         (armv5, armv5e): Delete arches.
12789         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12790         arm_arch5.
12791         (*call_reg_arm): Likewise.
12792         (*call_value_reg_armv5): Likewise.
12793         (*call_value_reg_arm): Likewise.
12794         (*call_symbol): Likewise.
12795         (*call_value_symbol): Likewise.
12796         (*sibcall_insn): Likewise.
12797         (*sibcall_value_insn): Likewise.
12798         (clzsi2): Likewise.
12799         (prefetch): Likewise.
12800         (define_split and define_peephole2 dependent on arm_arch5):
12801         Likewise.
12802         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12803         arm_arch5e.
12804         (TARGET_ARM_QBIT): Likewise.
12805         (TARGET_DSP_MULTIPLY): Likewise.
12806         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12807         (arm_arch5, arm_arch5e): Delete.
12808         (arm_arch5t, arm_arch5te): Declare.
12809         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12810         (arm_arch5t): Declare.
12811         (arm_option_reconfigure_globals): Update for the above.
12812         (arm_options_perform_arch_sanity_checks): Update comment, replace
12813         use of arm_arch5 with arm_arch5t.
12814         (use_return_insn): Likewise.
12815         (arm_emit_call_insn): Likewise.
12816         (output_return_instruction): Likewise.
12817         (arm_final_prescan_insn): Likewise.
12818         (arm_coproc_builtin_available): Likewise.
12819         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12820         arm_arch5e with arm_arch5t and arm_arch5te.
12821         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12822         (arm_arch5t, arm_arch5te): Declare.
12823         * config/arm/arm-tables.opt: Regenerate.
12824         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12825         * config/arm/t-multilib: Likewise.
12826         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12827         instead of arm_arch5.
12828         (*call_reg_thumb1): Likewise.
12829         (*call_value_reg_thumb1_v5): Likewise.
12830         (*call_value_reg_thumb1): Likewise.
12831         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12832         unreachable path.
12833         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12834
12835 2018-05-18  Martin Liska  <mliska@suse.cz>
12836
12837         PR gcov-profile/84846
12838         * doc/gcov.texi: Document -t option of gcov tool.
12839
12840 2018-05-18  Martin Liska  <mliska@suse.cz>
12841
12842         PR gcov-profile/84846
12843         * gcov.c (print_usage): Add new -t option.
12844         (process_args): Handle the option.
12845         (generate_results): Use stdout as output when requested by
12846         the option.
12847
12848 2018-05-18  Martin Liska  <mliska@suse.cz>
12849
12850         PR gcov-profile/84846
12851         * coverage.c (coverage_init): Write PWD to .gcno file.
12852         * doc/gcov.texi: Document how working directory is printed.
12853         * gcov-dump.c (dump_gcov_file): Print PWD.
12854         * gcov.c (output_intermediate_file): Likewise.
12855         (read_graph_file): Read PWD string.
12856         (output_lines): Print PWD.
12857
12858 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12859
12860         PR middle-end/85817
12861         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12862         for retval and return false if all args to phi are zero.
12863
12864 2018-05-18  Richard Biener  <rguenther@suse.de>
12865
12866         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12867         method.
12868         (evrp_dom_walker::before_dom_children): Call it.
12869
12870 2018-05-18  Richard Biener  <rguenther@suse.de>
12871
12872         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12873         results when processing array refs with variable index.
12874
12875 2018-05-18  Toon Moene  <toon@moene.org>
12876
12877         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12878         directly after that of -floop-interchange. Indicate that both
12879         options are enabled by default when specifying -O3.
12880
12881 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12882
12883         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12884         iterator.  Delete separate integer-mode vec_set<mode> expander.
12885         (aarch64_simd_vec_setv2di): Delete.
12886         (vec_setv2di): Delete.
12887         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12888         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
12889         the "w, r" alternative.
12890
12891 2018-05-18  Martin Liska  <mliska@suse.cz>
12892
12893         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12894         * tree-pass.h (make_pass_lower_switch_O0): New function.
12895         * tree-switch-conversion.c (node_has_low_bound): Remove.
12896         (node_has_high_bound): Likewise.
12897         (node_is_bounded): Likewise.
12898         (class pass_lower_switch): Make it a template type and create
12899         two instances.
12900         (pass_lower_switch::execute): Add template argument.
12901         (make_pass_lower_switch): New function.
12902         (make_pass_lower_switch_O0): New function.
12903         (do_jump_if_equal): Remove.
12904         (emit_case_nodes): Simplify to just handle all 3 cases and leave
12905         all the hard work to tree optimization passes.
12906
12907 2018-05-18  Martin Liska  <mliska@suse.cz>
12908
12909         * dbgcnt.c (limit_low): Renamed from limit.
12910         (limit_high): New variable.
12911         (dbg_cnt_is_enabled): Check for upper limit.
12912         (dbg_cnt): Adjust dumping.
12913         (dbg_cnt_set_limit_by_index): Add new argument for high
12914         value.
12915         (dbg_cnt_set_limit_by_name): Likewise.
12916         (dbg_cnt_process_single_pair): Parse new format.
12917         (dbg_cnt_process_opt): Use strtok.
12918         (dbg_cnt_list_all_counters): Remove 'value' and add
12919         'limit_high'.
12920         * doc/invoke.texi: Document changes.
12921
12922 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
12923
12924         * doc/sourcebuild.texi (scalar_all_fma): Document.
12925         * tree.def (FMA_EXPR): Delete.
12926         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12927         * internal-fn.c (ternary_direct): New macro.
12928         (expand_ternary_optab_fn): Likewise.
12929         (direct_ternary_optab_supported_p): Likewise.
12930         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12931         * builtins.c (fold_builtin_fma): Delete.
12932         (fold_builtin_3): Don't call it.
12933         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12934         * expr.c (expand_expr_real_2): Likewise.
12935         * fold-const.c (operand_equal_p): Likewise.
12936         (fold_ternary_loc): Likewise.
12937         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12938         * gimple.c (DEFTREECODE): Likewise.
12939         * gimplify.c (gimplify_expr): Likewise.
12940         * optabs-tree.c (optab_for_tree_code): Likewise.
12941         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12942         * tree-eh.c (operation_could_trap_p): Likewise.
12943         (stmt_could_throw_1_p): Likewise.
12944         * tree-inline.c (estimate_operator_cost): Likewise.
12945         * tree-pretty-print.c (dump_generic_node): Likewise.
12946         (op_code_prio): Likewise.
12947         * tree-ssa-loop-im.c (stmt_cost): Likewise.
12948         * tree-ssa-operands.c (get_expr_operands): Likewise.
12949         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12950         * fold-const-call.h (fold_fma): Delete.
12951         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12952         CFN_FNMA and CFN_FNMS.
12953         (fold_fma): Delete.
12954         * genmatch.c (combined_fn): New enum.
12955         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12956         (commutative_op): New function.
12957         (commutate): Use it.  Handle more than 2 operands.
12958         (dt_operand::gen_gimple_expr): Use commutative_op.
12959         (parser::parse_expr): Allow :c to be used with non-binary
12960         operators if the commutative operand is known.
12961         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12962         CFN_FMS, CFN_FNMA and CFN_FNMS.
12963         (backprop::process_assign_use): Remove FMA_EXPR handling.
12964         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12965         (gen_hsa_fma): New function.
12966         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12967         IFN_FNMA and IFN_FNMS.
12968         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12969         * gimple-fold.h (follow_all_ssa_edges): Declare.
12970         * gimple-fold.c (follow_all_ssa_edges): New function.
12971         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12972         gimple_build interface and use follow_all_ssa_edges to fold the result.
12973         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12974         instead of checking for optabs directly.
12975         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12976         rather than FMA_EXPRs.
12977         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12978         call to IFN_FMA instead of an FMA_EXPR.
12979
12980 2018-05-17  Jim Wilson  <jimw@sifive.com>
12981
12982         * expr.c (do_tablejump): When converting index to Pmode, if we have a
12983         sign extended promoted subreg, and the range does not have the sign bit
12984         set, then do a sign extend.
12985
12986         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12987         test, check for sign extended subreg and/or constant operands, and
12988         do a sign extend in that case.
12989
12990 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
12991
12992         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12993         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12994         Add untyped.
12995         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12996         Change logics_shift_reg to logics_shift_imm.
12997         (thunderx2t99_fp_loadpair_basic): Delete.
12998         (thunderx2t99_fp_storepair_basic): Delete.
12999         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
13000         (thunderx2t99_asimd_polynomial): Delete.
13001         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13002         and neon_fp_mul_d_scalar_q.
13003         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
13004         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
13005         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
13006         (thunderx2t99_asimd_lut): Add missing tbl types.
13007         (thunderx2t99_asimd_ext): Delete.
13008         (thunderx2t99_asimd_load1_1_mult): Delete.
13009         (thunderx2t99_asimd_load1_2_mult): Delete.
13010         (thunderx2t99_asimd_load1_ldp): New.
13011         (thunderx2t99_asimd_load1): New.
13012         (thunderx2t99_asimd_load2): Add missing *load2* types.
13013         (thunderx2t99_asimd_load3): New.
13014         (thunderx2t99_asimd_load4): New.
13015         (thunderx2t99_asimd_store1_1_mult): Delete.
13016         (thunderx2t99_asimd_store1_2_mult): Delete.
13017         (thunderx2t99_asimd_store2_mult): Delete.
13018         (thunderx2t99_asimd_store2_onelane): Delete.
13019         (thunderx2t99_asimd_store_stp): New.
13020         (thunderx2t99_asimd_store1): New.
13021         (thunderx2t99_asimd_store2): New.
13022         (thunderx2t99_asimd_store3): New.
13023         (thunderx2t99_asimd_store4): New.
13024
13025 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
13026
13027         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13028         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
13029
13030 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
13031             Segher Boessenkool  <segher@kernel.crashing.org>
13032
13033         PR target/85698
13034         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13035         operand.
13036
13037 2018-05-17  Richard Biener  <rguenther@suse.de>
13038
13039         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13040         for pruning loop and prune defs feeding only already visited PHIs.
13041
13042 2018-05-17  Richard Biener  <rguenther@suse.de>
13043
13044         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13045
13046 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
13047             Richard Biener  <rguenther@suse.de>
13048
13049         PR tree-optimization/85793
13050         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13051         for VMAT_ELEMENTWISE.
13052
13053 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
13054
13055         * internal-fn.h (lookup_internal_fn): Declare
13056         * internal-fn.c (lookup_internal_fn): New function.
13057         * gimple.c (gimple_build_call_from_tree): Handle calls to
13058         internal functions.
13059         * gimple-pretty-print.c (dump_gimple_call): Print "." before
13060         internal function names.
13061         * tree-pretty-print.c (dump_generic_node): Likewise.
13062         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13063
13064 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
13065
13066         * gimple-fold.h (gimple_build): Make the function forms take
13067         combined_fn rather than built_in_function.
13068         (gimple_simplify): Likewise.
13069         * gimple-match-head.c (gimple_simplify): Likewise.
13070         * gimple-fold.c (gimple_build): Likewise.
13071         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13072         rather than gimple_build_call_internal.
13073         (get_initial_defs_for_reduction): Likewise.
13074         (vect_create_epilog_for_reduction): Likewise.
13075         (vectorizable_live_operation): Likewise.
13076
13077 2018-05-17  Martin Liska  <mliska@suse.cz>
13078
13079         * gimple-ssa-sprintf.c (format_directive): Do not use
13080         space in between 'G_' and '('.
13081
13082 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
13083
13084         PR target/85323
13085         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13086         even if the mask is not all ones.
13087
13088         PR target/85323
13089         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13090         vector.
13091         (ix86_gimple_fold_builtin): Likewise.
13092
13093         PR target/85323
13094         * config/i386/i386.c: Include tree-vector-builder.h.
13095         (ix86_vector_shift_count): New function.
13096         (ix86_fold_builtin): Fold shift builtins by scalar count.
13097         (ix86_gimple_fold_builtin): Likewise.
13098
13099         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13100         _mm512_setzero): New intrinsics.
13101
13102 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
13103             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13104
13105         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13106         code generation for cases where splatting a value is not useful.
13107         * simplify-rtx.c (simplify_ternary_operation): Simplify
13108         vec_merge across a vec_duplicate and a paradoxical subreg forming
13109         a vector mode to a vec_concat.
13110
13111 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
13112
13113         * config.gcc: Support "goldmont-plus".
13114         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13115         "goldmont-plus".
13116         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13117         PROCESSOR_GOLDMONT_PLUS.
13118         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13119         (processor_target_table): Add "goldmont-plus".
13120         (PTA_GOLDMONT_PLUS): Define.
13121         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13122         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13123         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13124         (fold_builtin_cpu): Add "goldmont-plus".
13125         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13126         (ix86_option_override_internal): Add "goldmont-plus".
13127         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13128         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13129         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13130         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13131
13132 2018-05-17  Richard Biener  <rguenther@suse.de>
13133
13134         PR tree-optimization/85757
13135         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13136         remove defs that only feed that PHI from further processing.
13137
13138 2018-05-16  Jim Wilson  <jimw@sifive.com>
13139
13140         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13141         asterisk to name.
13142         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13143
13144 2018-05-16  Mark Wielaard  <mark@klomp.org>
13145
13146         * dwarf2out.c (count_index_strings): New function.
13147         (output_indirect_strings): Call count_index_strings and generate
13148         header for dwarf_version >= 5.
13149
13150 2018-05-16  Mark Wielaard  <mark@klomp.org>
13151
13152         * dwarf2out.c (dwarf_FORM): New function.
13153         (set_indirect_string): Use dwarf_FORM.
13154         (reset_indirect_string): Likewise.
13155         (size_of_die): Likewise.
13156         (value_format): Likewise.
13157         (output_die): Likewise.
13158         (add_skeleton_AT_string): Likewise.
13159         (output_macinfo_op): Likewise.
13160         (index_string): Likewise.
13161         (output_index_string_offset): Likewise.
13162         (output_index_string): Likewise.
13163         (count_index_strings): Likewise.
13164
13165 2018-05-16  Carl Love  <cel@us.ibm.com>
13166
13167         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13168         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13169
13170 2018-05-16  Martin Jambor  <mjambor@suse.cz>
13171
13172         * ipa-prop.c (ipa_free_all_edge_args): Remove.
13173         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13174
13175 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
13176
13177         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13178         (fnma<mode>4): Likewise.
13179         (fms<mode>4): Likewise.
13180         (fnms<mode>4): Likewise.
13181         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13182         (aarch64_fnma<mode>4): Likewise.
13183         (aarch64_fms<mode>4): Likewise.
13184         (aarch64_fnms<mode>4): Likewise.
13185         (aarch64_fnmadd<mode>4): Likewise.
13186
13187 2018-05-16  Jason Merrill  <jason@redhat.com>
13188
13189         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
13190
13191 2018-05-16  Richard Biener  <rguenther@suse.de>
13192
13193         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13194         (dump_stmt_cost): Declare.
13195         (add_stmt_cost): Dump cost we add.
13196         (add_stmt_costs): New function.
13197         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13198         No longer exported.
13199         (vect_analyze_stmt): Adjust prototype.
13200         (vectorizable_condition): Likewise.
13201         (vectorizable_live_operation): Likewise.
13202         (vectorizable_reduction): Likewise.
13203         (vectorizable_induction): Likewise.
13204         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13205         cost vector to pass to vectorizable_ and record afterwards.
13206         (vect_model_reduction_cost): Take cost vector argument and adjust.
13207         (vect_model_induction_cost): Likewise.
13208         (vectorizable_reduction): Likewise.
13209         (vectorizable_induction): Likewise.
13210         (vectorizable_live_operation): Likewise.
13211         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13212         SLP_TREE_NUMBER_OF_VEC_STMTS.
13213         (vect_analyze_slp_cost_1): Remove.
13214         (vect_analyze_slp_cost): Likewise.
13215         (vect_slp_analyze_node_operations): Take visited args and
13216         a target cost vector.  Avoid processing already visited stmt sets.
13217         (vect_slp_analyze_operations): Use a local cost vector to gather
13218         costs and register those of non-discarded instances.
13219         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13220         (vect_schedule_slp_instance): Remove copying of
13221         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
13222         zero.
13223         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13224         adding cost.  Record cost entry location.
13225         (vect_prologue_cost_for_slp_op): Function to compute cost of
13226         a constant or invariant generated for SLP vect in the prologue,
13227         split out from vect_analyze_slp_cost_1.
13228         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
13229         (vect_model_promotion_demotion_cost): Likewise.
13230         (vect_model_store_cost): Likewise, make static.
13231         (vect_model_load_cost): Likewise.
13232         (vectorizable_bswap): Add cost vector arg and adjust.
13233         (vectorizable_call): Likewise.
13234         (vectorizable_simd_clone_call): Likewise.
13235         (vectorizable_conversion): Likewise.
13236         (vectorizable_assignment): Likewise.
13237         (vectorizable_shift): Likewise.
13238         (vectorizable_operation): Likewise.
13239         (vectorizable_store): Likewise.
13240         (vectorizable_load): Likewise.
13241         (vectorizable_condition): Likewise.
13242         (vectorizable_comparison): Likewise.
13243         (can_vectorize_live_stmts): Likewise.
13244         (vect_analyze_stmt): Likewise.
13245         (vect_transform_stmt): Adjust calls to vectorizable_*.
13246         * tree-vectorizer.c: Include gimple-pretty-print.h.
13247         (dump_stmt_cost): New function.
13248
13249 2018-05-16  Richard Biener  <rguenther@suse.de>
13250
13251         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
13252         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
13253         * tree-ssa-dse.c: Include tree-ssa-loop.h.
13254         (check_name): New callback.
13255         (dse_classify_store): Track cycles via a visited bitmap of PHI
13256         defs and simplify handling of in-loop and across loop dead stores
13257         and properly fail for loop-variant refs.  Handle byte-tracking with
13258         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13259         limiting the walk.
13260
13261 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
13262
13263         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
13264         (vect_get_mask_type_for_stmt): Likewise.
13265         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
13266         split out from...
13267         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
13268         to determine the statement's vector type and the vector type that
13269         should be used for calculating nunits.  Deal with cases in which
13270         the type has to be deferred.
13271         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
13272         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
13273         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
13274         (vect_determine_vf_for_stmt): New functions, split out from...
13275         (vect_determine_vectorization_factor): ...here.
13276         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
13277         (vect_get_mask_type_for_stmt): New functions, split out from
13278         vect_determine_vectorization_factor.
13279
13280 2018-05-16  Richard Biener  <rguenther@suse.de>
13281
13282         * tree-cfg.c (verify_gimple_assign_ternary): Properly
13283         verify the [VEC_]COND_EXPR embedded comparison.
13284
13285 2018-05-15  Martin Sebor  <msebor@redhat.com>
13286
13287         PR tree-optimization/85753
13288         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
13289         RECORD_TYPE in addition to ARRAY_TYPE.
13290
13291 2018-05-15  Martin Sebor  <msebor@redhat.com>
13292
13293         PR middle-end/85643
13294         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
13295
13296 2018-05-15  Richard Biener  <rguenther@suse.de>
13297
13298         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
13299         add by_clobber_p one.  Change algorithm to collect all defs
13300         representing uses we need to walk and try reducing them to
13301         a single one before failing.
13302         (dse_dom_walker::dse_optimize_stmt): Adjust.
13303
13304 2018-05-13  Mark Wielaard  <mark@klomp.org>
13305
13306         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13307         (size_of_loc_descr): Likewise.
13308         (output_loc_operands): Likewise.
13309         (output_loc_operands_raw): Likewise.
13310         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13311         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13312         (hash_loc_operands): Likewise.
13313         (compare_loc_operands): Likewise.
13314
13315 2018-05-14  Mark Wielaard  <mark@klomp.org>
13316
13317         * dwarf2out.c (count_index_addrs): New function.
13318         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13319
13320 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13321
13322         PR tree-optimization/83648
13323         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13324         return value as malloc candidate.
13325
13326 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13327
13328         PR ipa/85734
13329         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13330         param as true in call to suggest_attribute.
13331
13332 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
13333
13334         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13335         -mreadonly-in-sdata.
13336
13337 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13338
13339         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13340         New pattern.
13341         (aarch64_crypto_aesd_fused): Likewise.
13342
13343 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
13344
13345         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13346         (movsi_aarch64): Likewise.
13347         (load_pairsi): Likewise.
13348         (load_pairdi): Likewise.
13349         (store_pairsi): Likewise.
13350         (store_pairdi): Likewise.
13351         (load_pairsf): Likewise.
13352         (load_pairdf): Likewise.
13353         (store_pairsf): Likewise.
13354         (store_pairdf): Likewise.
13355         (zero_extend): Likewise.
13356         (trunc): Swap alternatives.
13357         (fcvt_target): Add '?' to prefer w over r.
13358
13359 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
13360
13361         PR target/85756
13362         * config/i386/i386.md: Disallow non-commutative arithmetics in
13363         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13364         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
13365         in the peephole2 before it.
13366
13367 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
13368
13369         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13370         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13371         (ix86_handle_option): Handle -mcldemote.
13372         * config.gcc: New header.
13373         * config/i386/cldemoteintrin.h: New file.
13374         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13375         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13376         -mcldemote.
13377         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13378         OPTION_MASK_ISA_CLDEMOTE.
13379         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13380         (ix86_valid_target_attribute_inner_p): Ditto.
13381         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13382         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13383         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13384         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13385         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13386         (cldemote): New.
13387         * config/i386/i386.opt: Add -mcldemote.
13388         * config/i386/x86intrin.h: New header.
13389         * doc/invoke.texi: Add -mcldemote.
13390
13391 2018-05-14  Richard Biener  <rguenther@suse.de>
13392
13393         * doc/match-and-simplify.texi: Adjust :s documentation.
13394
13395 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
13396
13397         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13398         intended memcpy size.
13399         (REORDER_45): Likewise.
13400
13401 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
13402
13403         * sort.cc: New file.
13404         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13405         * vec.c (qsort_chk): Use gcc_qsort.
13406         * Makefile.in (OBJS-libcommon): Add sort.o.
13407         (build/sort.o): New target.  Use it...
13408         (BUILD_RTL): ... here, and...
13409         (build/gencfn-macros): ... here, and...
13410         (build/genmatch): ... here.
13411
13412 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13413             Chung-Ju Wu  <jasonwucj@gmail.com>
13414
13415         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13416         * config/nds32/nds32-graywolf.md: New file.
13417         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13418         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13419         pipeline.
13420         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13421         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13422         * config/nds32/nds32.md (pipeline_model): Add graywolf.
13423         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13424         * config/nds32/pipelines.md: Include n15 settings.
13425
13426 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13427             Chung-Ju Wu  <jasonwucj@gmail.com>
13428
13429         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13430         * config/nds32/nds32-n13.md: New file.
13431         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13432         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13433         pipeline.
13434         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13435         * config/nds32/nds32.md (pipeline_model): Add n13.
13436         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13437         * config/nds32/pipelines.md: Include n13 settings.
13438
13439 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13440             Chung-Ju Wu  <jasonwucj@gmail.com>
13441
13442         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13443         * config/nds32/nds32-n10.md: New file.
13444         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13445         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13446         pipeline.
13447         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13448         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13449         * config/nds32/nds32.md (pipeline_model): Add n10.
13450         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13451         * config/nds32/pipelines.md: Include n10 settings.
13452
13453 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
13454             Kito Cheng  <kito.cheng@gmail.com>
13455             Chung-Ju Wu  <jasonwucj@gmail.com>
13456
13457         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13458         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13459         Add enum values for DSP extension instructions.
13460         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13461         New constraints.
13462         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13463         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13464         New code iterators.
13465         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13466         * config/nds32/nds32-dspext.md: New file for DSP implementation.
13467         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13468         * config/nds32/nds32-intrinsic.md: Likewise.
13469         * config/nds32/nds32_intrinsic.h: Likewise.
13470         * config/nds32/nds32-md-auxiliary.c: Likewise.
13471         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13472         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13473         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13474         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13475         * config/nds32/nds32-protos.h: New declarations for DSP extension.
13476         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13477         TYPE_DMAC in switch statement.
13478         * config/nds32/nds32.c: New checking and implementation for DSP
13479         extension instructions.
13480         * config/nds32/nds32.h: Likewise.
13481         * config/nds32/nds32.md: Likewise.
13482         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13483         * config/nds32/predicates.md: Implement new predicates for DSP
13484         extension.
13485
13486 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
13487
13488         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13489         Reformat alternatives and attributes so it is easier to identify
13490         which constraints/attributes go with which instruction.
13491         (mov<mode>_hardfloat32, FMOVE64): Likewise.
13492         (mov<mode>_softfloat32, FMOVE64): Likewise.
13493         (mov<mode>_hardfloat64, FMOVE64): Likewise.
13494         (mov<mode>_softfloat64, FMOVE64): Likewise.
13495
13496 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13497
13498         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13499         subsection.
13500         (Basic PowerPC Built-in Functions): The new name of the
13501         subsection previously known as "PowerPC Built-in Functions".
13502         (Basic PowerPC Built-in Functions Available on all Configurations):
13503         New subsubsection.
13504         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13505         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13506         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13507         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13508
13509 2018-05-11  Martin Jambor  <mjambor@suse.cz>
13510
13511         PR ipa/85655
13512         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13513         single const.
13514
13515 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
13516
13517         PR target/85733
13518         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13519
13520 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
13521
13522         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13523         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13524         (ix86_handle_option): Handle -mwaitpkg.
13525         * config.gcc: New header.
13526         * config/i386/cpuid.h (bit_WAITPKG): New bit.
13527         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13528         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13529         function type.
13530         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13531         OPTION_MASK_ISA_WAITPKG.
13532         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13533         (ix86_option_override_internal): Add PTA_WAITPKG.
13534         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13535         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13536         IX86_BUILTIN_TPAUSE.
13537         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13538         __builtin_ia32_umwait and __builtin_ia32_tpause.
13539         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13540         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13541         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13542         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13543         UNSPECV_TPAUSE): New.
13544         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13545         * config/i386/i386.opt: Add -mwaitpkg.
13546         * config/i386/waitpkgintrin.h: New file.
13547         * config/i386/x86intrin.h: New header.
13548         * doc/invoke.texi: Add -mwaitpkg.
13549
13550 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
13551
13552         PR target/85606
13553         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13554         equivalent.
13555         (cortex-m0): Use armv6s-m isa.
13556         (cortex-m0plus): Likewise.
13557         (cortex-m1): Likewise.
13558         (cortex-m0.small-multiply): Likewise.
13559         (cortex-m0plus.small-multiply): Likewise.
13560         (cortex-m1.small-multiply): Likewise.
13561
13562 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
13563             Jakub Jelinek  <jakub@redhat.com>
13564
13565         PR tree-optimization/85692
13566         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13567         source permute as well.
13568
13569 2018-05-11  Martin Liska  <mliska@suse.cz>
13570
13571         PR sanitizer/85556
13572         * doc/extend.texi: Document LLVM style format for no_sanitize
13573         attribute.
13574
13575 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
13576
13577         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13578         mode_supports_vsx_dform_quad to mode_supports_dq_form.
13579         (mode_supports_vsx_dform_quad): Likewise.
13580         (mode_supports_vmx_dform): Move these functions to be next to the
13581         other mode_supports functions.
13582         (mode_supports_dq_form): Likewise.
13583         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13584         mode_supports_dq_form.
13585         (reg_offset_addressing_ok_p): Likewise.
13586         (offsettable_ok_by_alignment): Likewise.
13587         (rs6000_legitimate_offset_address_p): Likewise.
13588         (legitimate_lo_sum_address_p): Likewise.
13589         (rs6000_legitimize_address): Likewise.
13590         (rs6000_legitimize_reload_address): Likewise.
13591         (rs6000_secondary_reload_inner): Likewise.
13592         (rs6000_preferred_reload_class): Likewise.
13593         (rs6000_output_move_128bit): Likewise.
13594
13595 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
13596
13597         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13598         Generate SImode target register for null target.
13599         <case IX86_BUILTIN_XGETBV>: Ditto.
13600         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13601         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13602
13603 2018-05-10  Carl Love  <cel@us.ibm.com>
13604
13605         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13606         dcbtt and dcbtstt if operands[2] is 0.
13607
13608 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
13609
13610         PR target/85693
13611         * config/i386/sse.md (usadv64qi): New expander.
13612
13613 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
13614
13615         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13616         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13617         -maltivec=be support.
13618         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13619         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13620         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13621         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13622         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13623         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13624         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13625         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13626         altivec_vsumsws): Adjust.
13627         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13628         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13629         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13630         support.
13631         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13632         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13633         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13634         (altivec_lve<VI_char>x): Delete expand.
13635         (*altivec_lve<VI_char>x_internal): Rename to...
13636         (altivec_lve<VI_char>x): ... this.
13637         (altivec_lvxl_<mode>): Delete expand.
13638         (*altivec_lvxl_<mode>_internal): Rename to ...
13639         (altivec_lvxl_<mode>): ... this.
13640         (altivec_stvxl_<mode>): Delete expand.
13641         (*altivec_stvxl_<mode>_internal): Rename to ...
13642         (altivec_stvxl_<mode>): ... this.
13643         (altivec_stve<VI_char>x): Delete expand.
13644         (*altivec_stve<VI_char>x_internal): Rename to ...
13645         (altivec_stve<VI_char>x): ... this.
13646         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13647         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13648         reduc_plus_scal_<mode>): Adjust.
13649         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13650         comment.
13651         (rs6000_cpu_cpp_builtins): Adjust.
13652         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13653         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13654         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13655         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13656         -maltivec=be support.
13657         (rs6000_split_vec_extract_var): Adjust.
13658         (rs6000_split_v4si_init): Adjust.
13659         (swap_selector_for_mode): Delete.
13660         (altivec_expand_lvx_be, altivec_expand_stvx_be,
13661         altivec_expand_stvex_be): Delete.
13662         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13663         -maltivec=be support.
13664         (rs6000_gimple_fold_builtin): Ditto.
13665         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13666         Adjust.
13667         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13668         (TARGET_DIRECT_MOVE_64BIT): Adjust.
13669         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13670         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13671         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13672         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13673         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13674         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13675         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13676         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13677         anonymous split): Adjust.
13678         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13679         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13680
13681 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
13682
13683         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13684         when --with-gxx-include-dir is also specified.
13685         * configure: Regenerate.
13686
13687 2018-05-09  Jim Wilson  <jimw@sifive.com>
13688
13689         PR target/84797
13690         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13691         * config/riscv/t-withmultilib: New.
13692         * config/riscv/withmultilib.h: New.
13693         * doc/install.texi: Document RISC-V --with-multilib-list support.
13694
13695 2018-05-09  Richard Biener  <rguenther@suse.de>
13696
13697         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13698         vector.
13699         (vect_bb_vectorization_profitable_p): Adjust.  Compute
13700         actual scalar cost using the cost vector and the add_stmt_cost
13701         machinery.
13702
13703 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13704
13705         PR rtl-optimization/85645
13706         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13707         in the REG_CFA_REGISTER note for LR, don't leave it empty.
13708
13709 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13710
13711         PR rtl-optimization/85645
13712         * shrink-wrap.c (spread_components): Return a boolean saying if
13713         anything was changed.
13714         (try_shrink_wrapping_separate): Iterate spread_components until
13715         nothing changes anymore.
13716
13717 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13718
13719         PR rtl-optimization/85645
13720         * regrename.c (build_def_use): Also kill the chains that include the
13721         destination of a REG_CFA_REGISTER note.
13722
13723 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13724
13725         PR rtl-optimization/85645
13726         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13727         insn that has a REG_CFA_REGISTER note.
13728
13729 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
13730
13731         * cfgexpand.c (expand_clobber): New function.
13732         (expand_gimple_stmt_1): Use it.
13733         * tree-vect-stmts.c (vect_clobber_variable): New function,
13734         split out from...
13735         (vectorizable_simd_clone_call): ...here.
13736         (vectorizable_store): Emit a clobber either side of an
13737         IFN_STORE_LANES sequence.
13738         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13739
13740 2018-05-09  Tom de Vries  <tom@codesourcery.com>
13741
13742         PR target/85626
13743         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13744         (define_insn "trap_if_false"): Add exit after trap.
13745
13746 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
13747
13748         PR rtl-optimization/85638
13749         * bb-reorder.c: Include common/common-target.h.
13750         (create_forwarder_block): New function extracted from...
13751         (fix_up_crossing_landing_pad): ...here.  Rename into...
13752         (dw2_fix_up_crossing_landing_pad): ...this.
13753         (sjlj_fix_up_crossing_landing_pad): New function.
13754         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13755         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13756         from both partitions and exit the loop after one iteration.
13757
13758 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13759
13760         Revert:
13761         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13762         subsection.
13763         (Basic PowerPC Built-in Functions): The new name of the
13764         subsection previously known as "PowerPC Built-in Functions".
13765         (Basic PowerPC Built-in Functions Available on all Configurations):
13766         New subsubsection.
13767         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13768         subsubsection.
13769         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13770         subsubsection.
13771         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13772         subsubsection.
13773         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13774         subsubsection.
13775
13776 2018-05-08  Jim Wilson  <jimw@sifive.com>
13777
13778         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13779         (LD_EMUL_SUFFIX): New.
13780         (LINK_SPEC): Use it.
13781
13782 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13783
13784         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13785         subsection.
13786         (Basic PowerPC Built-in Functions): The new name of the
13787         subsection previously known as "PowerPC Built-in Functions".
13788         (Basic PowerPC Built-in Functions Available on all Configurations):
13789         New subsubsection.
13790         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13791         subsubsection.
13792         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13793         subsubsection.
13794         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13795         subsubsection.
13796         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13797         subsubsection.
13798
13799 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
13800
13801         PR target/85683
13802         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13803         after cmpelim optimization.
13804
13805 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
13806
13807         * config.gcc: Support "goldmont".
13808         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13809         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13810         PROCESSOR_GOLDMONT.
13811         * config/i386/i386.c (m_GOLDMONT): Define.
13812         (processor_target_table): Add "goldmont".
13813         (PTA_GOLDMONT): Define.
13814         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13815         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13816         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13817         (fold_builtin_cpu): Add "goldmont".
13818         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13819         (ix86_option_override_internal): Add "goldmont".
13820         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13821         (processor_type): Add PROCESSOR_GOLDMONT.
13822         * config/i386/i386.md: Add CPU "glm".
13823         * config/i386/glm.md: New file.
13824         * config/i386/x86-tune.def: Add m_GOLDMONT.
13825         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13826
13827 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
13828
13829         PR target/85572
13830         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13831         E_V4DImode.
13832         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13833         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
13834         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
13835
13836         PR target/85317
13837         * config/i386/i386.c (ix86_fold_builtin): Handle
13838         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13839
13840         PR target/85480
13841         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13842         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13843
13844 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
13845
13846         PR target/85658
13847         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13848         (check_arch): Likewise.
13849         (check_fpu): Return the result rather than printing it.
13850         (end arch): Fix operator precedence.
13851         (end cpu): Likewise.
13852         (END): Print the result from check_fpu.
13853
13854 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
13855             Alan Hayward  <alan.hayward@arm.com>
13856             David Sherwood  <david.sherwood@arm.com>
13857
13858         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13859         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13860         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13861         (*fcmuo<mode>_and): New patterns.
13862
13863 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
13864
13865         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13866         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13867         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13868         (cmp_op, sve_imm_con): New code attributes.
13869         (SVE_COND_INT_CMP, imm_con): Delete.
13870         (cmp_op): Remove above unspecs from int attribute.
13871         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13872         to...
13873         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
13874         comparison-specific unspecs.
13875         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13876         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13877         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13878         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13879         (*vec_fcm<cmp_op><mode>): Rename to...
13880         (*fcm<cmp_op><mode>): ...this and adjust likewise.
13881         (*vec_fcmuo<mode>): Rename to...
13882         (*fcmuo<mode>): ...this and adjust likewise.
13883         (*pred_fcm<cmp_op><mode>): New pattern.
13884         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13885         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13886         functions.
13887         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13888         and UNORDERED.
13889         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13890         (aarch64_emit_sve_predicated_cond): New function.
13891         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13892         (aarch64_emit_unspec_cond_or): Replace with...
13893         (aarch64_emit_sve_or_conds): ...this new function.  Use
13894         aarch64_emit_sve_ptrue_op for the individual comparisons and
13895         aarch64_emit_binop to OR them together.
13896         (aarch64_emit_inverted_unspec_cond): Replace with...
13897         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
13898         aarch64_emit_sve_ptrue_op for the comparison and
13899         aarch64_emit_unop to invert the result.
13900         (aarch64_expand_sve_vec_cmp_float): Update after the above
13901         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
13902
13903 2018-05-07  Nathan Sidwell  <nathan@acm.org>
13904
13905         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13906         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13907         (Backwards Compatibility): Likewise.
13908
13909 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13910
13911         PR bootstrap/85681
13912         Revert:
13913         2018-05-07  Luis Machado  <luis.machado@linaro.org>
13914
13915         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13916         <prefetch_dynamic_strides>: New const bool field.
13917         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13918         prefetch_dynamic_strides.
13919         (exynosm1_prefetch_tune): Likewise.
13920         (thunderxt88_prefetch_tune): Likewise.
13921         (thunderx_prefetch_tune): Likewise.
13922         (thunderx2t99_prefetch_tune): Likewise.
13923         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13924         to false.
13925         (aarch64_override_options_internal): Update to set
13926         PARAM_PREFETCH_DYNAMIC_STRIDES.
13927         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13928         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13929         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13930         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13931         prefetch-dynamic-strides setting.
13932
13933         2018-05-07  Luis Machado  <luis.machado@linaro.org>
13934
13935         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13936         <minimum_stride>: New const int field.
13937         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13938         minimum_stride field.
13939         (exynosm1_prefetch_tune): Likewise.
13940         (thunderxt88_prefetch_tune): Likewise.
13941         (thunderx_prefetch_tune): Likewise.
13942         (thunderx2t99_prefetch_tune): Likewise.
13943         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13944         (aarch64_override_options_internal): Update to set
13945         PARAM_PREFETCH_MINIMUM_STRIDE.
13946         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13947         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13948         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13949         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13950         stride is constant and is below the minimum stride threshold.
13951
13952 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13953
13954         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13955         to 512.
13956
13957 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13958
13959         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13960         <prefetch_dynamic_strides>: New const bool field.
13961         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13962         prefetch_dynamic_strides.
13963         (exynosm1_prefetch_tune): Likewise.
13964         (thunderxt88_prefetch_tune): Likewise.
13965         (thunderx_prefetch_tune): Likewise.
13966         (thunderx2t99_prefetch_tune): Likewise.
13967         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13968         to false.
13969         (aarch64_override_options_internal): Update to set
13970         PARAM_PREFETCH_DYNAMIC_STRIDES.
13971         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13972         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13973         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13974         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13975         prefetch-dynamic-strides setting.
13976
13977 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13978
13979         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13980         <minimum_stride>: New const int field.
13981         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13982         minimum_stride field.
13983         (exynosm1_prefetch_tune): Likewise.
13984         (thunderxt88_prefetch_tune): Likewise.
13985         (thunderx_prefetch_tune): Likewise.
13986         (thunderx2t99_prefetch_tune): Likewise.
13987         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13988         (aarch64_override_options_internal): Update to set
13989         PARAM_PREFETCH_MINIMUM_STRIDE.
13990         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13991         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13992         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13993         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13994         stride is constant and is below the minimum stride threshold.
13995
13996 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
13997
13998         PR c++/85659
13999         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
14000         the type is addressable.  Don't force op into register if it has
14001         BLKmode.
14002
14003 2018-05-05  Roland McGrath  <mcgrathr@google.com>
14004
14005         PR other/77609
14006         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
14007         any section for which we don't know a specific type it should have,
14008         regardless of name.  Previously this was done only for the exact
14009         names ".init_array", ".fini_array", and ".preinit_array".
14010         (default_elf_asm_named_section): Add comment about
14011         relationship with default_section_type_flags and SECTION_NOTYPE.
14012         (get_section): Don't consider it a type conflict if one side has
14013         SECTION_NOTYPE and the other doesn't, as long as neither has the
14014         SECTION_BSS et al used in the default_section_type_flags logic.
14015
14016 2018-05-05  Tom de Vries  <tom@codesourcery.com>
14017
14018         PR target/85653
14019         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
14020         (workaround_barsyncs): New function.
14021         (nvptx_reorg): Use workaround_barsyncs.
14022         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14023         (define_expand "nvptx_membar_cta"): New define_expand.
14024         (define_insn "*nvptx_membar_cta"): New insn.
14025
14026 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
14027
14028         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14029         To improve optimization opportunities.
14030         * builtin-types.def: The new needed builtin types for the above.
14031
14032 2018-05-04  Richard Biener  <rguenther@suse.de>
14033
14034         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14035         * gimple-ssa-store-merging.c
14036         (imm_store_chain_info::output_merged_store): Remove redundant create,
14037         release split_store vector contents on failure.
14038         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14039         scalar stmt vector on cache hit.
14040
14041 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
14042
14043         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14044         Xilinx FP support.
14045         * config.gcc (powerpc-xilinx-eabi*): Remove.
14046         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14047         support.
14048         (fusion_addis_mem_combo_load): Ditto.
14049         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14050         FP support.
14051         (rs6000_cpu_cpp_builtins): Ditto.
14052         * config/rs6000/rs6000-linux.c
14053         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14054         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14055         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14056         support.
14057         (rs6000_setup_reg_addr_masks): Ditto.
14058         (rs6000_init_hard_regno_mode_ok): Ditto.
14059         (rs6000_option_override_internal): Ditto.
14060         (legitimate_lo_sum_address_p): Ditto.
14061         (rs6000_legitimize_address): Ditto.
14062         (rs6000_legitimize_reload_address): Ditto.
14063         (rs6000_legitimate_address_p): Ditto.
14064         (abi_v4_pass_in_fpr): Ditto.
14065         (setup_incoming_varargs): Ditto.
14066         (rs6000_gimplify_va_arg): Ditto.
14067         (rs6000_split_multireg_move): Ditto.
14068         (rs6000_savres_strategy): Ditto.
14069         (rs6000_emit_prologue_components): Ditto.
14070         (rs6000_emit_epilogue_components): Ditto.
14071         (rs6000_emit_prologue): Ditto.
14072         (rs6000_emit_epilogue): Ditto.
14073         (rs6000_elf_file_end): Ditto.
14074         (rs6000_function_value): Ditto.
14075         (rs6000_libcall_value): Ditto.
14076         * config/rs6000/rs6000.h: Ditto.
14077         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14078         (TARGET_MINMAX): ... this.  New.
14079         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14080         * config/rs6000/rs6000.md: Remove Xilinx FP support.
14081         (*movsi_internal1_single): Delete.
14082         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14083         mfpu=, mxilinx-fpu): Delete.
14084         * config/rs6000/singlefp.h: Delete.
14085         * config/rs6000/sysv4.h: Remove Xilinx FP support.
14086         * config/rs6000/t-rs6000: Ditto.
14087         * config/rs6000/t-xilinx: Delete.
14088         * config/rs6000/titan.md: Adjust for fp_type removal.
14089         * config/rs6000/vsx.md: Remove Xilinx FP support.
14090         (VStype_simple): Delete.
14091         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14092         * config/rs6000/xfpu.h: Delete.
14093         * config/rs6000/xfpu.md: Delete.
14094         * config/rs6000/xilinx.h: Delete.
14095         * config/rs6000/xilinx.opt: Delete.
14096         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14097         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14098
14099 2018-05-04  Tom de Vries  <tom@codesourcery.com>
14100
14101         PR libgomp/85639
14102         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14103         if ignore == 0.
14104
14105 2018-05-04  Richard Biener  <rguenther@suse.de>
14106
14107         PR middle-end/85627
14108         * tree-complex.c (update_complex_assignment): We are always in SSA form.
14109         (expand_complex_div_wide): Likewise.
14110         (expand_complex_operations_1): Likewise.
14111         (expand_complex_libcall): Preserve EH info of the original stmt.
14112         (tree_lower_complex): Handle removed blocks.
14113         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14114         on complex multiplication and division libcall builtins.
14115
14116 2018-05-04  Richard Biener  <rguenther@suse.de>
14117
14118         PR middle-end/85574
14119         * fold-const.c (negate_expr_p): Restrict negation of operand
14120         zero of a division to when we know that can happen without
14121         overflow.
14122         (fold_negate_expr_1): Likewise.
14123
14124 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
14125
14126         PR libstdc++/85466
14127         * real.h (real_nextafter): Declare.
14128         * real.c (real_nextafter): New function.
14129         * fold-const-call.c (fold_const_nextafter): New function.
14130         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14131         CASE_CFN_NEXTTOWARD.
14132         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14133         even when arg1_mode is different from arg0_mode.
14134
14135 2018-05-03  Nathan Sidwell  <nathan@acm.org>
14136
14137         * doc/extend.texi (Deprecated Features): Remove
14138         -ffriend-injection.
14139         (Backwards Compatibility): Likewise.
14140         * doc/invoke.texi (C++ Language Options): Likewise.
14141         (C++ Dialect Options): Likewise.
14142
14143 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
14144
14145         PR target/85530
14146         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14147         _mm512_mask_mullox_epi64): New intrinsics.
14148
14149 2018-05-03  Tom de Vries  <tom@codesourcery.com>
14150
14151         PR testsuite/85106
14152         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14153         dump files): Add offload-tree.
14154
14155 2018-05-03  Richard Biener  <rguenther@suse.de>
14156
14157         PR tree-optimization/85615
14158         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14159         to loops not nested in BBs loop father to avoid creating multi-entry
14160         loops.
14161
14162 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14163
14164         PR tree-optimization/70291
14165         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14166         arguments.  Change return type to tree.  Emit libcall as a new
14167         statement rather than replacing existing one when inplace_p is true.
14168         (expand_complex_multiplication_components): New function.
14169         (expand_complex_multiplication): Expand floating-point complex
14170         multiplication using the above.
14171         (expand_complex_division): Rename inner_type parameter to type.
14172         Update expand_complex_libcall call-site.
14173         (expand_complex_operations_1): Update expand_complex_multiplication
14174         and expand_complex_division call-sites.
14175
14176 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
14177
14178         PR target/85582
14179         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14180         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14181         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14182         the highest significant bit of the shift count mask is clear.  In
14183         check whether and[sq]i3 is needed verify that all significant bits
14184         of the shift count other than the highest are set.
14185
14186 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14187
14188         PR libgomp/82428
14189         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14190         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14191         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14192         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14193         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14194         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14195         __builtin_goacc_parlevel_size.
14196
14197 2018-05-02  Richard Biener  <rguenther@suse.de>
14198
14199         PR tree-optimization/85597
14200         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14201         do not use split vect_get_vec_defs call but call vect_get_slp_defs
14202         directly.
14203
14204 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14205
14206         PR testsuite/85106
14207         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14208         dump files): Add ltrans-tree.
14209
14210 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14211
14212         PR testsuite/85106
14213         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14214         dump files): Add wpa-ipa.
14215
14216 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
14217
14218         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
14219         powerpc*-*-linux*paired* target.
14220         * config/rs6000/750cl.h: Delete.
14221         * config/rs6000/paired.h: Delete.
14222         * config/rs6000/paired.md: Delete.
14223         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14224         float support.
14225         * config/rs6000/rs6000-builtin.def: Remove paired float support.
14226         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14227         comment.  Remove paired float support.
14228         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14229         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14230         VECTOR_PAIRED.
14231         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14232         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14233         declarations.
14234         * config/rs6000/rs6000.c: Remove paired float support.
14235         (paired_expand_vector_init, paired_expand_vector_move,
14236         paired_emit_vector_compare, paired_emit_vector_cond_expr,
14237         (paired_expand_lv_builtin, paired_expand_stv_builtin,
14238         paired_expand_builtin, paired_expand_predicate_builtin,
14239         paired_init_builtins): Delete.
14240         * config/rs6000/rs6000.h: Remove paired float support.
14241         * config/rs6000/rs6000.md: Remove paired float support.
14242         (move_from_CR_ov_bit): Delete.
14243         * config/rs6000/rs6000.opt (mpaired): Delete.
14244         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
14245         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
14246
14247 2018-05-02  Richard Biener  <rguenther@suse.de>
14248
14249         PR middle-end/85567
14250         * gimplify.c (gimplify_save_expr): When in SSA form allow
14251         SAVE_EXPRs to compute to SSA vars.
14252
14253 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
14254
14255         PR target/85582
14256         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14257         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14258         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
14259         clobber operands[2], instead use a new pseudo.  Formatting fixes.
14260
14261 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
14262
14263         PR tree-optimization/85586
14264         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
14265         exit early for statements in the same group if the accesses are
14266         not strided.
14267
14268 2018-05-02  Tom de Vries  <tom@codesourcery.com>
14269
14270         PR lto/85451
14271         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
14272         error message.
14273
14274 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
14275
14276         PR tree-optimization/85143
14277         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
14278
14279 2018-05-01  Tom de Vries  <tom@codesourcery.com>
14280
14281         PR lto/85451
14282         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
14283         not found" error message.
14284
14285 2018-05-01  Tom de Vries  <tom@codesourcery.com>
14286
14287         PR other/83786
14288         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
14289         * vec.c (test_ordered_remove_if): New function.
14290         (vec_c_tests): Call test_ordered_remove_if.
14291         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
14292         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
14293         * tree-vect-patterns.c (vect_pattern_recog_1): Use
14294         VEC_ORDERED_REMOVE_IF.
14295
14296 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14297
14298         PR tree-optimization/82665
14299         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
14300         pointer subtraction where arguments come from a memchr call.
14301
14302 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
14303
14304         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14305         --push-state --as-needed and --pop-state instead of --as-needed and
14306         --no-as-needed if ld supports it.
14307         * configure: Regenerated.
14308
14309         PR web/85578
14310         * doc/install.texi2html: Replace _002d with - and _002a with * in
14311         generated html files using sed.
14312
14313 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14314
14315         PR c++/85523
14316         * gcc-rich-location.c (blank_line_before_p): New function.
14317         (use_new_line): New function.
14318         (gcc_rich_location::add_fixit_insert_formatted): New function.
14319         * gcc-rich-location.h
14320         (gcc_rich_location::add_fixit_insert_formatted): New function.
14321
14322 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14323
14324         * selftest.c (assert_streq): Rename "expected" and "actual" to
14325         "val1" and "val2".  Extend NULL-handling to cover both inputs
14326         symmetrically, while still requiring both to be non-NULL for a pass.
14327         * selftest.h (assert_streq): Rename "expected" and "actual" to
14328         "val1" and "val2".
14329         (ASSERT_EQ): Likewise.
14330         (ASSERT_EQ_AT): Likewise.
14331         (ASSERT_KNOWN_EQ): Likewise.
14332         (ASSERT_KNOWN_EQ_AT): Likewise.
14333         (ASSERT_NE): Likewise.
14334         (ASSERT_MAYBE_NE): Likewise.
14335         (ASSERT_MAYBE_NE_AT): Likewise.
14336         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
14337         the assertion to pass.
14338         (ASSERT_STREQ_AT): Likewise.
14339
14340 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
14341
14342         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14343         interaction with -pie.
14344
14345 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14346
14347         * selftest.h: Fix alphabetization of per-source-file selftest
14348         declarations.
14349
14350 2018-04-30  Jason Merrill  <jason@redhat.com>
14351
14352         PR c++/61982 - dead stores to destroyed objects.
14353         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14354         of clobber.
14355
14356 2018-04-30  Jason Merrill  <jason@redhat.com>
14357
14358         * tree.c (build_clobber): New.
14359         * tree.h: Declare it.
14360         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14361
14362 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14363
14364         * diagnostic-show-locus.c (layout::layout): Update for
14365         location_get_source_line returning a char_span.
14366         (struct char_span): Move to input.h.
14367         (struct correction): Update for fields in char_span becoming
14368         private.
14369         (struct source_line): Update for location_get_source_line
14370         returning a char_span.
14371         (layout::print_line): Likewise.
14372         * edit-context.c (edited_file::print_content): Likewise.
14373         (edited_file::print_diff_hunk): Likewise.
14374         (edited_file::print_run_of_changed_lines): Likewise.
14375         (edited_file::get_num_lines): Likewise.
14376         (edited_line::edited_line): Likewise.
14377         * final.c (asm_show_source): Likewise.
14378         * input.c (location_get_source_line): Convert return type
14379         from const char * to char_span, losing the final "line_len"
14380         param.
14381         (dump_location_info): Update for the above.
14382         (get_substring_ranges_for_loc): Likewise.  Use a char_span
14383         when handling the literal within the line.
14384         (test_reading_source_line): Update for location_get_source_line
14385         returning a char_span.
14386         * input.h (class char_span): Move here from
14387         diagnostic-show-locus.c, converting from a struct to a class.
14388         Make data members private.
14389         (char_span::operator bool): New.
14390         (char_span::length): New.
14391         (char_span::get_buffer): New.
14392         (char_span::operator[]): New.
14393         (char_span::subspan): Make const.
14394         (char_span::xstrdup): New.
14395         (location_get_source_line): Convert return type from const char *
14396         to char_span, losing the final "line_size" param.
14397
14398 2018-04-30  Jan Hubicka  <jh@suse.cz>
14399
14400         * lto-wrapper.c (ltrans_priorities): New static var.
14401         (cmp_priority): New.
14402         (run_gcc): Read priorities and if doing parallel build order
14403         the Makefile by them.
14404
14405 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14406
14407         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14408
14409 2018-04-30  Richard Biener  <rguenther@suse.de>
14410
14411         * tree-cfg.c (verify_address): Remove base argument, add
14412         flag whether to check TREE_ADDRESSABLE and do that.
14413         (verify_expr): Remove.
14414         (verify_types_in_gimple_reference): Add pieces from verify_expr.
14415         (verify_gimple_assign_single): Likewise.
14416         (verify_gimple_switch): Likewise.
14417         (verify_expr_location_1): Dereference tp once.  Add (disabled)
14418         piece from verify_expr.
14419         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14420
14421 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14422
14423         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14424
14425 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14426
14427         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14428         (small_data_pattern): Likewise.
14429         (arc_rewrite_small_data): Likewise.
14430         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14431         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14432         (get_symbol_alignment): New function.
14433         (legitimate_small_data_address_p): Likewise.
14434         (legitimate_scaled_address): Update, call
14435         legitimate_small_data_address_p.
14436         (output_sdata): New static variable.
14437         (arc_print_operand): Update how we handle small data operands.
14438         (arc_print_operand_address): Likewise.
14439         (arc_legitimate_address_p): Update, use
14440         legitimate_small_data_address_p.
14441         (arc_rewrite_small_data_p): Remove.
14442         (arc_rewrite_small_data_1): Likewise.
14443         (arc_rewrite_small_data): Likewise.
14444         (small_data_pattern): Likewise.
14445         (compact_sda_memory_operand): Update to use
14446         legitimate_small_data_address_p and get_symbol_alignment.
14447         (prepare_move_operands): Don't rewite sdata pattern.
14448         (prepare_extend_operands): Remove.
14449         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14450         pattern.
14451         (zero_extendqisi2): Likewise.
14452         (zero_extendhisi2): Likewise.
14453         (extendqihi2): Likewise.
14454         (extendqisi2): Likewise.
14455         (extendhisi2): Likewise.
14456         (addsi3): Likewise.
14457         (subsi3): Likewise.
14458         (andsi3): Likewise.
14459         * config/arc/constraints.md (Usd): Change it to memory constraint.
14460
14461 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14462
14463         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14464         as source of std instructions.
14465         * config/arc/arc.md (movsi_insn): Update pattern predicate to
14466         allow 6-bit constants as source for store instructions.
14467         (movdi_insn): Update instruction pattern to allow 6-bit constants
14468         as source for store instructions.
14469
14470 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
14471
14472         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14473
14474 2018-04-30  Nathan Sidwell  <nathan@acm.org>
14475             Sandra Loosemore <sandra@codesourcery.com>
14476
14477         * dumpfile.c (dump_open): Allow '-' for stdout.
14478         * doc/invoke.texi (Developer Options): Document dump filename
14479         determination early.  Document stdin/stdout selection.
14480
14481 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
14482
14483         Microblaze Target: PIC data text relative
14484
14485         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14486         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14487         Add declaration.
14488         * config/microblaze/microblaze.h (microblaze_constant_address_p):
14489         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14490         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14491         New addressing mode for data-text relative position indepenedent code.
14492         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14493         'ADDRESS_SYMBOLIC_TXT_REL'.
14494         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14495         (microblaze_legitimate_pic_operand): Exclude function calls from
14496         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14497         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14498         addresses cases.
14499         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14500         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14501         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14502         for 'address + offset'.
14503         (microblaze_expand_prologue): Add new function prologue call for
14504         'r20' assignation.
14505         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14506         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14507         table in case of TARGET_PIC_DATA_TEXT_REL.
14508         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14509         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14510         Add new macros 'UNSPEC_TEXT',
14511         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14512         + exclude function calls from 'UNSPEC_PLT' in case of data text
14513         relative mode.
14514         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14515         new target hook for generating address diff vector tables in case of
14516         flag_pic.
14517         * doc/tm.texi : Regenerate.
14518         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14519         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14520         of addr diff vector generation.
14521         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14522         target hook definition.
14523         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14524         Add default function for generate_pic_addr_diff_vec -> flag_pic.
14525         * doc/invoke.texi (Add new pic option): Add new microblaze pic
14526         option for data text relative.
14527
14528 2018-04-30  Richard Biener  <rguenther@suse.de>
14529
14530         * tree-chrec.h (evolution_function_is_constant_p): Remove
14531         redundant check.
14532         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14533
14534 2018-04-30  Richard Biener  <rguenther@suse.de>
14535
14536         PR bootstrap/85571
14537         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14538
14539 2018-04-30  Richard Biener  <rguenther@suse.de>
14540
14541         PR tree-optimization/28364
14542         PR tree-optimization/85275
14543         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14544         copying first exit test.
14545
14546 2018-04-28  Mark Wielaard  <mark@klomp.org>
14547
14548         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14549         dwarf_version >= 5.
14550         (dwarf_AT): Handle DW_AT_addr_base.
14551         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14552
14553 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
14554
14555         PR target/84431
14556         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14557         (*ashl<dwi>3_doubleword_mask_1): Ditto.
14558         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14559         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14560
14561 2018-04-28  Richard Biener  <rguenther@suse.de>
14562
14563         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14564         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14565         to reflect use.  Only add interesting stmts.
14566
14567 2018-04-27  Martin Jambor  <mjambor@suse.cz>
14568
14569         PR ipa/85549
14570         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14571         the jump function allows for passing through aggregate values.
14572
14573 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
14574
14575         * input.h (in_system_header_at): Convert from macro to inline
14576         function.
14577         (from_macro_expansion_at): Likewise.
14578         (from_macro_definition_at): Likewise.
14579
14580 2018-04-27  Jeff Law  <law@redhat.com>
14581
14582         * config.gcc: Mark tile* targets as deprecated/obsolete.
14583
14584 2018-04-27  Richard Biener  <rguenther@suse.de>
14585
14586         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14587         fix for ILP32.
14588
14589 2018-04-27  Richard Biener  <rguenther@suse.de>
14590
14591         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14592
14593 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
14594
14595         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14596         with Yd constraint. Set "preferred_for_speed" attribute from
14597         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14598         with Yd constraint.
14599         (*movdi_internal): Ditto.
14600         (movti_interunit splitters): Remove
14601         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14602         (movdi_interunit splitters): Ditto.
14603         * config/i386/constraints.md (Ye): Remove.
14604         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14605
14606 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14607
14608         PR target/85512
14609         * config/aarch64/constraints.md (Usg): Limit to 31.
14610         (Usj): Limit to 63.
14611
14612 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
14613
14614         PR tree-optimization/85529
14615         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14616         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14617         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14618         zero extension or masking of the MSB bit.
14619         (optimize_range_tests): Add FIRST_BB argument, pass it through
14620         to optimize_range_tests_var_bound.
14621         (maybe_optimize_range_tests, reassociate_bb): Adjust
14622         optimize_range_tests callers.
14623
14624 2018-04-26  Richard Biener  <rguenther@suse.de>
14625             Jakub Jelinek  <jakub@redhat.com>
14626
14627         * cgraph.h (symbol_table): Just declare debug method here.
14628         * symtab.c (symbol_table::debug): Define.
14629
14630 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
14631
14632         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14633
14634 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
14635
14636         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14637         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14638         (*movdi_internal): Substitute Yi and Yj constraint with x
14639         and Ym and Yn constraint with y constraint.  Update "isa"
14640         attribute and set "preferred_for_speed" attribute from
14641         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14642         (*movsi_internal): Ditto.
14643         (*movdf_internal): Ditto.
14644         (*movsf_internal): Ditto.
14645         (*zero_extendsidi2): Ditto.
14646         * config/i386/sse.md (vec_set<mode>_0): Ditto.
14647         (sse2_loadld): Ditto.
14648         (*vec_extract<ssevecmodelower>_0): Ditto.
14649         (*vec_extractv4si_0_zext_sse4): Ditto.
14650         (vec_concatv2di): Ditto.
14651         (*vec_dup<mode>): Ditto.
14652         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14653         * config/i386/constraints.md (Yi): Remove.
14654         (Yj): Remove.
14655         (Ym): Remove.
14656         (Yn): Remove.
14657
14658 2018-04-26  Nathan Sidwell  <nathan@acm.org>
14659
14660         * dumpfile.c (dump_open): New.
14661         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14662         (dump_finish): Detect stdio/stderr by value not name.
14663
14664 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
14665
14666         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14667
14668 2018-04-26  Tom de Vries  <tom@codesourcery.com>
14669
14670         PR target/84952
14671         * config/nvptx/nvptx.c (verify_neutering_jumps)
14672         (verify_neutering_labels): New function
14673         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14674
14675 2018-04-26  Tom de Vries  <tom@codesourcery.com>
14676
14677         PR target/84025
14678         * config/nvptx/nvptx.c (needs_neutering_p): New function.
14679         (nvptx_single): Use needs_neutering_p to skip over insns that do not
14680         need neutering.
14681
14682 2018-04-26  Richard Biener <rguenther@suse.de>
14683             Tom de Vries  <tom@codesourcery.com>
14684
14685         PR lto/85422
14686         * lto-streamer-out.c (output_function): Fixup loops if required to match
14687         discovery done in the reader.
14688
14689 2018-04-26  Richard Biener  <rguenther@suse.de>
14690
14691         PR tree-optimization/85116
14692         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14693         have a loop exit from the single latch predecessor.  Remove
14694         case of header with just condition.
14695         (ch_base::copy_headers): Exclude infinite loops from any
14696         processing.
14697         (pass_ch::execute): Record exits.
14698
14699 2018-04-26  Richard Biener  <rguenther@suse.de>
14700
14701         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14702         prologue cost vector and pass it to vect_get_load_cost.
14703         (vect_get_peeling_costs_all_drs): Likewise.
14704         (vect_peeling_hash_get_lowest_cost): Likewise.
14705         (vect_enhance_data_refs_alignment): Likewise.
14706
14707 2018-04-26  Richard Biener  <rguenther@suse.de>
14708
14709         PR middle-end/85450
14710         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14711         checking of integer<->pointer conversions.
14712         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14713         sign-/zero-extending pointer types.
14714         (expand_omp_for_static_chunk): Likewise.
14715
14716 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
14717             Jean Lee  <xiaoyur347@gmail.com>
14718
14719         * config/mips/mips.c (mips_asan_shadow_offset): New function.
14720         (TARGET_ASAN_SHADOW_OFFSET): Define.
14721         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14722         true for -fsanitize=address.
14723
14724 2018-04-25  Mark Wielaard  <mark@klomp.org>
14725
14726         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14727         shorter ones.
14728
14729 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
14730
14731         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14732         than "alu", remove explicit "memory" and "imm_disp" attributes.
14733         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14734
14735         PR middle-end/85414
14736         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14737         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14738         gen_lowpart_no_emit.
14739
14740 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
14741
14742         PR target/85473
14743         * config/i386/i386.c (ix86_expand_builtin): Change memory
14744         operand to XI, extend p0 to Pmode.
14745         * config/i386/i386.md: Change unspec volatile and operand
14746         1 mode to XI, change operand 0 mode to P.
14747
14748 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14749
14750         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14751         GET_MODE_MASK before any checking.
14752         (nds32_can_use_bset_p): Likewise.
14753         (nds32_can_use_btgl_p): Likewise.
14754
14755 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14756
14757         * config/nds32/nds32-doubleword.md: New define_split pattern for
14758         illegal register number.
14759
14760 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14761
14762         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14763
14764 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14765
14766         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14767
14768 2018-04-25  Richard Biener  <rguenther@suse.de>
14769
14770         * lto-streamer.h (LTO_major_version): Bump to 8.
14771
14772 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
14773
14774         * BASE-VER: Set to 9.0.0.
14775
14776 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
14777
14778         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14779         in __abskf2 and __powikf2.
14780
14781 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14782
14783         PR target/85512
14784         * config/aarch64/constraints.md (Usg, Usj): New constraints.
14785         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14786         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14787         Use the above on operand 2.  Reindent.
14788         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14789
14790 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
14791
14792         PR target/85485
14793         * common/config/i386/i386-common.c (ix86_handle_option): Don't
14794         handle OPT_mcet.
14795         * config/i386/i386.opt (mcet): Removed.
14796         * doc/install.texi: Remove -mcet documentation.
14797         * doc/invoke.texi: Likewise.
14798
14799 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
14800
14801         PR target/85485
14802         * doc/install.texi: Remove -mcet from bootstrap-cet.
14803
14804 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14805
14806         PR target/85511
14807         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14808         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14809         if TARGET_64BIT.
14810
14811         PR target/85503
14812         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14813         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14814         containing a CONST_VECTOR.
14815
14816 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
14817
14818         * doc/install.texi: Update newlib dependency for nvptx.
14819
14820 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14821
14822         PR target/85508
14823         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14824         instead of INTVAL when shifting x left.
14825
14826 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
14827
14828         PR tree-optimization/85478
14829         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14830         vect_grouped_store_supported for single element vectors.
14831
14832 2018-04-24  Richard Biener  <rguenther@suse.de>
14833
14834         PR target/85491
14835         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14836         load cost increase to the case of non-constant step.
14837
14838 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14839
14840         PR target/84828
14841         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14842         destination if any_malformed_asm.
14843
14844 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
14845
14846         PR middle-end/85496
14847         * expr.c (store_field): In the bitfield case, if the value comes from
14848         a function call and is returned in registers by means of a PARALLEL,
14849         do not change the mode of the temporary unless BLKmode and VOIDmode.
14850
14851 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
14852
14853         PR rtl-optimization/85423
14854         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14855         dependencies to debug insns when the previous insn is non-debug.
14856
14857 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
14858
14859         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14860         enums into a single definition.
14861         (fls): Fix predicates and printing.
14862         (seti): Likewise.
14863
14864 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
14865
14866         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14867         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14868         and short u6 immediate.
14869         (check_if_valid_sleep_operand): Remove.
14870         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14871
14872 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14873
14874         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14875         flag_always_save_lp condition.
14876         * config/nds32/nds32.opt (malways-save-lp): New option.
14877
14878 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
14879
14880         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14881         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14882         * config/nds32/nds32.h
14883         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14884         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14885
14886 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
14887
14888         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14889         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14890
14891 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
14892             Chung-Ju Wu  <jasonwucj@gmail.com>
14893
14894         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14895         Declare.
14896         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14897         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14898
14899 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14900
14901         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14902
14903 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14904
14905         * config/nds32/nds32-protos.h (nds32_data_alignment,
14906         nds32_local_alignment): Declare.
14907         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14908         nds32_local_alignment): New functions.
14909         (TARGET_CONSTANT_ALIGNMENT): Define.
14910         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14911
14912 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14913
14914         * config/nds32/nds32.c
14915         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14916         (TARGET_MODES_TIEABLE_P): Likewise.
14917
14918 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14919
14920         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14921         level Ofast and Og.
14922
14923 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
14924             Chung-Ju Wu  <jasonwucj@gmail.com>
14925
14926         * config/nds32/constants.md (unspec_volatile_element): Add enum values
14927         for unaligned access.
14928         * config/nds32/nds32-intrinsic.c: Implementation of expanding
14929         unaligned access.
14930         * config/nds32/nds32-intrinsic.md: Likewise.
14931         * config/nds32/nds32_intrinsic.h: Likewise.
14932         * config/nds32/nds32.h (nds32_builtins): Likewise.
14933         * config/nds32/nds32.opt (munaligned-access): New option.
14934         * config/nds32/nds32.c (nds32_asm_file_start): Display
14935         flag_unaligned_access status.
14936
14937 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
14938
14939         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14940         -mno-relax is present.
14941         * config/riscv/linux.h (LINK_SPEC): Ditto.
14942
14943 2018-04-20  Martin Sebor  <msebor@redhat.com>
14944
14945         PR c/85365
14946         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14947         for null pointers.
14948         (gimple_fold_builtin_stxcpy_chk): Same.
14949         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14950
14951 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
14952
14953         PR target/85456
14954         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14955         __powikf2 when long double is IEEE 128-bit.
14956
14957 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
14958
14959         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14960         step to make sure stack always aligned.
14961
14962 2018-04-20  Carl Love  <cel@us.ibm.com>
14963
14964         PR target/83402
14965         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14966         size check for arg0.
14967
14968 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
14969             Tom de Vries  <tom@codesourcery.com>
14970
14971         PR target/85445
14972         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14973         Emit insns for calls too.
14974         (nvptx_find_par): Always look for worker-level predecessor insn.
14975         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
14976         calls.
14977         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14978         (nvptx_process_pars): Propagate frames for calls.
14979
14980 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
14981
14982         PR target/85469
14983         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14984         Removed.
14985         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14986         (ix86_handle_option): Don't handle OPT_mibt.
14987         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14988         __SHSTK__.
14989         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14990         has_ibt and ibt.
14991         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14992         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14993         (ix86_target_macros): Define __CET__ with flag_cf_protection
14994         for -fcf-protection.
14995         * config/i386/i386.c (isa2_opts): Remove -mibt.
14996         * config/i386/i386.h (TARGET_IBT): Removed.
14997         (TARGET_IBT_P): Likewise.
14998         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14999         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
15000         * config/i386/i386.opt (mcet): Update help message.
15001         (mshstk): Likewise.
15002         (mibt): Removed.
15003         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
15004         -mcet as an alias for -mshstk.
15005
15006 2018-04-20  Richard Biener <rguenther@suse.de>
15007
15008         PR middle-end/85475
15009         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
15010         complexity by forcing a single use of the multiply operand.
15011
15012 2018-04-20  Martin Jambor  <mjambor@suse.cz>
15013
15014         ipa/85449
15015         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
15016         recursion dependency to only apply to non-clones.
15017
15018 2018-04-20  Martin Jambor  <mjambor@suse.cz>
15019
15020         ipa/85447
15021         * ipa-cp.c (create_specialized_node): Check that clones of
15022         self-recursive edges exist during IPA-CP.
15023
15024 2018-04-19  Toon Moene  <toon@moene.org>
15025
15026         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15027         by -O3.
15028
15029 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
15030
15031         PR tree-optimization/85467
15032         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
15033         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
15034         VECTOR_CST element to type.
15035
15036 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
15037
15038         PR target/85397
15039         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15040         * config/i386/i386.md (builtin_setjmp_setup): Removed.
15041         (builtin_longjmp): Likewise.
15042         (save_stack_nonlocal): New pattern.
15043         (restore_stack_nonlocal): Likewise.
15044
15045 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
15046
15047         PR target/85404
15048         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15049         Replace ASM_OUTPUT_LABEL with fprintf.
15050
15051 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
15052
15053         PR target/85417
15054         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15055         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15056         * config/i386/i386-c.c (ix86_target_macros_internal): Also
15057         define __IBT__ and __SHSTK__ for -fcf-protection.
15058         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15059         TARGET_IBT.
15060         (ix86_trampoline_init): Likewise.
15061         (x86_output_mi_thunk): Likewise.
15062         (ix86_notrack_prefixed_insn_p): Likewise.
15063         (ix86_option_override_internal): Don't disallow -fcf-protection.
15064         * config/i386/i386.md (rdssp<mode>): Also enable for
15065         -fcf-protection.
15066         (incssp<mode>): Likewise.
15067         (nop_endbr): Likewise.
15068         * config/i386/i386.opt (mcet): Change help message to built-in
15069         functions only.
15070         (mibt): Likewise.
15071         (mshstk): Likewise.
15072         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15073         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
15074         enable CET built-in functions.
15075
15076 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
15077
15078         * common/config/i386/i386-common.c
15079         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15080         OPTION_MASK_ISA_MOVDIRI_UNSET,
15081         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15082         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15083         * config.gcc (movdirintrin.h): New header.
15084         * config/i386/cpuid.h (bit_MOVDIRI,
15085         bit_MOVDIR64B): New bits.
15086         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15087         and -mmvodir64b.
15088         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15089         (VOID, PVOID, PCVOID)): New function types.
15090         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15091         __builtin_ia32_directstoreu_u64,
15092         __builtin_ia32_movdir64b): New builtins.
15093         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15094         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15095         and -mmovdiri.
15096         (ix86_valid_target_attribute_inner_p): Ditto.
15097         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15098         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15099         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15100         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15101         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15102         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15103         (movdiri<mode>, movdir64b_<mode>): New.
15104         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15105         * config/i386/immintrin.h: Include movdirintrin.h.
15106         * config/i386/movdirintrin.h: New file.
15107         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15108
15109 2018-04-19  Richard Biener  <rguenther@suse.de>
15110
15111         PR middle-end/85455
15112         * cfg.c (clear_bb_flags): When loop state says we have
15113         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15114
15115 2018-04-19  Richard Biener  <rguenther@suse.de>
15116
15117         PR tree-optimization/84737
15118         * tree-vect-data-refs.c (vect_copy_ref_info): New function
15119         copying restrict info.
15120         (vect_setup_realignment): Use it.
15121         * tree-vectorizer.h (vect_copy_ref_info): Declare.
15122         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15123         the first DR to all generated stores.
15124         (vectorizable_load): Likewise for loads.
15125
15126 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
15127
15128         PR tree-optimization/85446
15129         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15130         the integral and pointer types to have the same precision.
15131
15132         * doc/install.texi: Document --disable-cet being the default and
15133         --enable-cet=auto.
15134
15135 2018-04-18  Martin Liska  <mliska@suse.cz>
15136
15137         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15138         style.
15139
15140 2018-04-18  Martin Liska  <mliska@suse.cz>
15141
15142         Revert
15143         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
15144
15145         PR ipa/83983
15146         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15147         arguments if they are comparable.
15148
15149 2018-04-18  Martin Liska  <mliska@suse.cz>
15150
15151         Revert
15152         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
15153
15154         PR lto/84805
15155         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15156         incomplete types.
15157
15158 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
15159
15160         PR target/85388
15161         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15162         ENDBR after calling __morestack.
15163
15164 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
15165
15166         PR jit/85384
15167         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15168         by using gcc_base_ver to generate a gcc_driver_version, and use
15169         it when generating GCC_DRIVER_NAME.
15170         * configure: Regenerate.
15171
15172 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
15173
15174         PR target/81084
15175         * config.gcc: Obsolete powerpc*-*-*spe*.
15176
15177 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15178
15179         PR debug/84637
15180         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15181         (stabstr_D): Change type of unum from unsigned int to
15182         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
15183         type.
15184
15185 2018-04-17  Jim Wilson  <jimw@sifive.com>
15186
15187         PR 84856
15188         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15189         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15190         Set arg_pointer_offset after using pretend_args_size.
15191
15192 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15193
15194         PR rtl-optimization/85431
15195         * dse.c (record_store): Ignore zero width stores.
15196
15197         PR sanitizer/85230
15198         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
15199         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15200         __builtin_stack_restore rather than after it.
15201         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15202         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15203         argument instead of virtual_dynamic_stack_rtx.
15204
15205 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15206
15207         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15208         New prototype.
15209         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15210         Add note to error message to explain internal mapping of overloaded
15211         built-in function name to non-overloaded built-in function name.
15212         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15213         function.
15214
15215 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
15216
15217         PR target/85424
15218         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15219         where the inputs overlap with the output.
15220
15221 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15222
15223         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15224         (=v, v) alternative and explicit "memory" attribute.
15225         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
15226         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15227         attributes.
15228         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15229         "sselog1" type instead of "sselog".
15230         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15231         "sselog".  Remove explicit "memory" attribute.
15232         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15233         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15234         attributes.
15235         (vec_extract_hi_v32hi): Merge all alternatives into one, use
15236         "sselog1" type instead of "sselog".  Remove explicit "memory"
15237         attribute.
15238         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
15239         use "sselog1" type instead of "sselog".  Remove explicit "memory"
15240         attribute.
15241         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
15242         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15243         attributes.
15244         (vec_extract_hi_v64qi): Merge all alternatives into one, use
15245         "sselog1" type instead of "sselog".  Remove explicit "memory"
15246         attribute.
15247         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
15248         use "sselog1" type instead of "sselog".  Remove explicit "memory"
15249         attribute.
15250
15251         PR target/85430
15252         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
15253
15254         PR middle-end/85414
15255         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
15256         on a SUBREG.
15257
15258 2018-04-17  Martin Jambor  <mjambor@suse.cz>
15259
15260         PR ipa/85421
15261         * ipa-cp.c (create_specialized_node): Call
15262         expand_all_artificial_thunks if necessary.
15263
15264 2018-04-17  Martin Liska  <mliska@suse.cz>
15265
15266         PR lto/85405
15267         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
15268         in message, remote space in between '_G' and '('.
15269
15270 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
15271
15272         PR target/85281
15273         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
15274         avx512f_vmcmp<mode>3<round_saeonly_name>,
15275         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
15276         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
15277         avx512f_rndscale<mode><round_saeonly_name>,
15278         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
15279         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
15280         Use %<iptr>2 instead of %2 for -masm=intel.
15281         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
15282         avx512f_vcvttss2usi<round_saeonly_name>,
15283         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
15284         -masm=intel.
15285         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
15286         avx512f_vcvttsd2usi<round_saeonly_name>,
15287         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
15288         Use %q1 instead of %1 for -masm=intel.
15289         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
15290         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
15291         of %3 for -masm=intel.
15292         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
15293         -masm=intel.
15294         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
15295         -masm=intel.
15296         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
15297         -masm=intel.
15298         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
15299         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
15300         %g1.
15301         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
15302         -masm=intel.
15303         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15304         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15305         %g1 and one with %0 and %1.
15306         (avx512er_vmrcp28<mode><round_saeonly_name>,
15307         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15308         %1 for -masm=intel.
15309         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15310         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15311         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15312         of %0 and %{%4%} for -masm=intel.
15313         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15314         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15315         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15316         order of %0 and %{%5%}%{z%} for -masm=intel.
15317
15318 2018-04-17  Jan Hubicka  <jh@suse.cz>
15319
15320         PR lto/85405
15321         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15322
15323 2018-04-17  Martin Liska  <mliska@suse.cz>
15324
15325         PR ipa/85329
15326         * multiple_target.c (create_dispatcher_calls): Set apostrophes
15327         for target_clone error message.  Make default implementation
15328         clone to be a local declaration.
15329         (separate_attrs): Add new argument and check for an empty
15330         string.
15331         (expand_target_clones): Handle it.
15332         (ipa_target_clone): Make redirection just for target_clones
15333         functions.
15334
15335 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
15336             Tom de Vries  <tom@codesourcery.com>
15337
15338         PR middle-end/84955
15339         * omp-expand.c (expand_oacc_for): Add dummy false branch for
15340         tiled basic blocks without omp continue statements.
15341
15342 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
15343
15344         PR target/83660
15345         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15346         vec_extract expression as having side effects to make sure it gets
15347         a cleanup point.
15348
15349 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
15350
15351         PR target/85403
15352         * config/i386/i386.c (get_builtin_code_for_version): Check
15353         error_mark_node.
15354
15355 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
15356
15357         PR target/84331
15358         * config.gcc: Support "skylake".
15359         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15360         PROCESSOR_SKYLAKE.
15361         * config/i386/i386.c (m_SKYLAKE): Define.
15362         (processor_target_table): Add "skylake".
15363         (ix86_option_override_internal): Add "skylake".
15364         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15365         PROCESSOR_CANNONLAKE.
15366         (get_builtin_code_for_version): Fix priority for
15367         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15368         PROCESSOR_SKYLAKE-AVX512.
15369         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15370         (processor_type): Add PROCESSOR_SKYLAKE.
15371
15372 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
15373             Jason Merrill  <jason@redhat.com>
15374
15375         PR c++/85112
15376         * convert.c (convert_to_integer_1): Use direct recursion for
15377         enumeral types and types with a precision less than the number
15378         of bits in their mode.
15379
15380 2018-04-16  Julia Koval  <julia.koval@intel.com>
15381
15382         PR target/84413
15383         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15384         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15385
15386 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
15387
15388         PR target/85293
15389         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15390         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15391         and -mno-direct-move.
15392
15393 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
15394
15395         PR target/83402
15396         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15397         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15398         Ensure negative shifts result in {0}.
15399
15400 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
15401
15402         PR rtl-optimization/79916
15403         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15404         regs (if any) to define how to gnerate SD moves when LRA is in
15405         progress.
15406
15407 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
15408
15409         PR rtl-optimization/85393
15410         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15411         * except.c (expand_dw2_landing_pad_for_region): Make static.
15412         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15413         a label and unconditional jump to old_bb, rather than
15414         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15415         basic block.
15416
15417         PR rtl-optimization/85376
15418         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15419         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15420         instead of a specific value.
15421
15422 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
15423             Bin Cheng  <bin.cheng@arm.com>
15424
15425         PR tree-optimization/82965
15426         PR tree-optimization/83991
15427         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15428         by_profile_only parameter.
15429         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15430         information if the loop was predicted to iterate too many times.
15431         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15432
15433 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
15434
15435         PR lto/71991
15436         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15437         always inline.
15438
15439 2018-04-13  Martin Liska  <mliska@suse.cz>
15440             Jakub Jelinek  <jakub@redhat.com>
15441
15442         PR middle-end/81657
15443         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15444         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15445         * builtins.c (expand_builtin_memory_copy_args): Use
15446         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15447         handle dest_addr == pc_rtx.
15448
15449 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
15450
15451         PR target/85291
15452         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15453         asked to not generate direct moves.
15454         (fix_trunc<mode>si2_stfiwx): Similar.
15455         (fix_trunc<mode>si2_internal): Similar.
15456
15457 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15458
15459         PR debug/83157
15460         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15461         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15462         lookup if dest in some wider mode is known to be const0_rtx and
15463         if so, record permanent equivalence for it to be ZERO_EXTEND of
15464         the narrower mode destination.
15465
15466 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
15467
15468         * lto-streamer-out.c (output_function): Revert 259346.
15469         * omp-expand.c (expand_oacc_for): Likewise.
15470
15471 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
15472
15473         PR rtl-optimization/85354
15474         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15475         * sel-sched.c (sel_global_init): ... here.
15476
15477 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
15478
15479         PR target/85238
15480         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15481         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15482         mode for PE-COFF targets.
15483         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15484         (i386_pe_asm_lto_end): Likewise.
15485         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15486         (TARGET_ASM_LTO_END): Likewise.
15487         * config/i386/winnt.c (saved_debug_info_level): New static variable.
15488         (i386_pe_asm_lto_start): New function.
15489         (i386_pe_asm_lto_end): Likewise.
15490
15491 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
15492             Richard Biener  <rguenther@suse.de>
15493
15494         PR middle-end/84955
15495         * lto-streamer-out.c (output_function): Fix CFG loop state before
15496         streaming out.
15497         * omp-expand.c (expand_oacc_for): Handle calls to internal
15498         functions like regular functions.
15499
15500 2018-04-12  Richard Biener  <rguenther@suse.de>
15501
15502         PR lto/85371
15503         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15504         for the early LTO debug to properly generate references to it
15505         during DIE emission.  Do not re-use that for the skeleton for
15506         split-dwarf.
15507         (dwarf2out_early_finish): Likewise.
15508
15509 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15510
15511         PR target/85328
15512         * config/i386/sse.md
15513         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15514         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15515         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15516         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15517         and output is a reg, avoid creating invalid lowpart subreg, but
15518         instead split into a 512-bit move.  Don't split if not AVX512VL,
15519         input is xmm16+ reg and output is a mem.
15520         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15521         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15522         xmm16+ reg and output is a mem.
15523
15524 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15525
15526         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15527         also for flag_dwarf2_cfi_asm.
15528
15529 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15530
15531         PR rtl-optimization/85342
15532         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15533         a bool scalar var inside of the loop instead.  Don't try to update
15534         recog_data.operand after failed apply_change_group.
15535
15536 2018-04-12  Tom de Vries  <tom@codesourcery.com>
15537
15538         PR target/85296
15539         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15540         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
15541         array with flexible array member as array without given dimension.
15542         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15543         argument for undefined param to true.
15544
15545 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
15546
15547         PR target/85321
15548         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15549         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15550         from PowerPC section.
15551         * config/rs6000/sysv4.opt (mcall-): Improve help text.
15552         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15553         help text that is too long.
15554         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15555         help text that is too long.
15556         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15557         help text that is too long.
15558
15559 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
15560
15561         * config/alpha/alpha.md (stack_probe_internal): Rename
15562         from "probe_stack".  Update all callers.
15563
15564 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15565
15566         PR rtl-optimization/84566
15567         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15568         sched_macro_fuse_insns.
15569
15570 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15571
15572         PR target/84301
15573         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15574         (compute_block_dependences): ... from here.
15575
15576 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15577
15578         PR tree-optimization/85331
15579         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15580         from int to HOST_WIDE_INT.
15581
15582 2018-04-11  Martin Jambor  <mjambor@suse.cz>
15583
15584         PR ipa/84149
15585         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15586         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15587         not the same as the source val.
15588         (cgraph_edge_brings_value_p): New parameter.
15589         (gather_edges_for_value): Pass destination value to
15590         cgraph_edge_brings_value_p.
15591         (perhaps_add_new_callers): Likewise.
15592         (get_info_about_necessary_edges): Likewise and exclude values brought
15593         only by self-recursive edges.
15594         (create_specialized_node): Redirect only clones of self-calling edges.
15595         (+self_recursive_pass_through_p): New function.
15596         (find_more_scalar_values_for_callers_subset): Use it.
15597         (find_aggregate_values_for_callers_subset): Likewise.
15598         (known_aggs_to_agg_replacement_list): Removed.
15599         (decide_whether_version_node): Re-calculate known constants for all
15600         remaining context clones.
15601
15602 2018-04-11  Richard Biener  <rguenther@suse.de>
15603
15604         PR lto/85339
15605         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15606         from early DWARF output.
15607         (dwarf2out_early_finish): Output line info unconditionally into
15608         early DWARF and add reference to it.
15609
15610 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15611
15612         PR target/85281
15613         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15614         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15615         other than V2DFmode using iptr mode attribute.
15616         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15617
15618 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15619
15620         PR rtl-optimization/84659
15621         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15622
15623 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15624
15625         PR debug/85302
15626         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15627         SIZEP is NULL.
15628         (output_loc_list): Pass address of a dummy size variable even in the
15629         locview handling loop.
15630         (index_location_lists): Add comment on why skip_loc_list_entry can't
15631         call size_of_locs.
15632
15633 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
15634
15635         PR target/85261
15636         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15637         into register.
15638
15639 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
15640
15641         PR target/85321
15642         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15643         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15644         and -mstring-compare-inline-limit.
15645
15646 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15647
15648         PR target/85287
15649         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15650         for stack clash protection in a register whenever we need it to be in
15651         a register.
15652
15653 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15654
15655         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15656         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15657
15658 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15659
15660         PR target/85321
15661         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15662         the help text.
15663         (mlong-double-): Ditto.
15664         * config/rs6000/sysv4.opt (msdata=): Ditto.
15665         (mtls-size=): Ditto.
15666
15667 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15668
15669         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15670         erroneous entries for
15671         "vector int vec_ldl (int, long int *)", and
15672         "vector unsigned int vec_ldl (int, unsigned long int *)".
15673         Add comments and entries for
15674         "vector bool char vec_ldl (int, bool char *)",
15675         "vector bool short vec_ldl (int, bool short *)",
15676         "vector bool int vec_ldl (int, bool int *)",
15677         "vector bool long long vec_ldl (int, bool long long *)",
15678         "vector pixel vec_ldl (int, pixel *)",
15679         "vector long long vec_ldl (int, long long *)",
15680         "vector unsigned long long vec_ldl (int, unsigned long long *)".
15681         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15682         type tree bool_long_long_type_node and correct definition of
15683         bool_V2DI_type_node to make reference to this new type tree.
15684         (rs6000_mangle_type): Replace erroneous reference to
15685         bool_long_type_node with bool_long_long_type_node.
15686         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15687         comments to emphasize sign distinctions for char and int types and
15688         replace RS6000_BTI_bool_long constant with
15689         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
15690         use of RS6000_BTI_pixel.
15691         (bool_long_type_node): Remove this macro definition.
15692         (bool_long_long_type_node): New macro definition
15693
15694 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
15695
15696         PR rtl-optimization/85300
15697         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15698         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15699         simplify_unary_operation fails.
15700
15701 2018-04-10  Martin Liska  <mliska@suse.cz>
15702
15703         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15704         cgraph_edge and ipa_ref.
15705
15706 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
15707
15708         PR target/85177
15709         PR target/85255
15710         * config/i386/sse.md
15711         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15712         computation of the VEC_MERGE selector from mask.
15713         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15714         Fix decoding of the VEC_MERGE selector into mask.
15715
15716 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
15717
15718         PR tree-optimization/85286
15719         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15720
15721 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
15722
15723         * final.c (final_1): Set insn_last_address as well as
15724         insn_current_address.
15725
15726 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15727
15728         PR target/85173
15729         * explow.c (emit_stack_probe): Call validize_mem on memory location
15730         before passing it to gen_probe_stack.  Create address operand and
15731         legitimize it for the probe_stack_address case.
15732
15733 2018-04-09  Jan Hubicka  <jh@suse.cz>
15734
15735         PR lto/85078
15736         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15737         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15738         * tree.c (free_lang_data_in_type): Fix handling of binfos;
15739         walk basetypes.
15740         (free_lang_data): Rebuild type inheritance graph.
15741
15742 2018-04-09  Martin Sebor  <msebor@redhat.com>
15743
15744         * invoke.texi (-finline-small-functions): Mention other optimization
15745         options.
15746         (-findirect-inlining, -fpartial-inlining): Same.
15747         (-finline-functions-called-once): Same.
15748         (-freorder-blocks-and-partition): Same.
15749
15750 2018-04-09  Jan Hubicka  <jh@suse.cz>
15751
15752         PR rtl/84058
15753         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15754         jumps; choose last target that matches the criteria (i.e.
15755         no partition changes for non-crossing jumps).
15756         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15757         support for redirecting crossing jumps to non-crossing.
15758
15759 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
15760
15761         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15762         also for naked functions.
15763
15764 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
15765
15766         * config/arc/arc.md (add_shift): New pattern.
15767         (add_shift2): Likewise.
15768         (sub_shift): Likewise.
15769         (sub_shift_cmp0_noout): Likewise.
15770         (compare_si_ashiftsi): Likewise.
15771         (xbfu_cmp0_noout): New combine pattern.
15772         (xbfu_cmp0"): Likewise.
15773         (movsi_set_cc_insn): Place the predicable variant first.
15774         (commutative_binary_cmp0_noout): Remove clobber.
15775         (commutative_binary_cmp0): New pattern.
15776         (noncommutative_binary_cmp0): Likewise.
15777         (noncommutative_binary_cmp0_noout): Likewise.
15778         (noncommutative_binary_comparison_result_used): Removed.
15779         (rsub_cmp0): New pattern.
15780         (rsub_cmp0_noout): Likewise.
15781         (extzvsi): Changed, keep only meaningful variants.
15782         (SQH, SEZ): New iterators.
15783         (SQH_postfix): New mode attribute.
15784         (SEZ_prefix): New code attribute.
15785         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15786         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15787         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15788         of numerical value.
15789         (noncommutative_operator): Check the availability of barrel
15790         shifter option.
15791
15792 2018-04-09  Richard Biener  <rguenther@suse.de>
15793
15794         PR tree-optimization/85284
15795         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15796         Only use the niter constraining form of simple_iv when the exit
15797         is always executed.
15798
15799 2018-04-09  Tom de Vries  <tom@codesourcery.com>
15800
15801         PR target/84041
15802         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15803         (define_expand "*memory_barrier"): New define_expand.
15804         (define_insn "memory_barrier"): New insn.
15805
15806 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15807
15808         PR rtl-optimization/80463
15809         PR rtl-optimization/83972
15810         PR rtl-optimization/83480
15811
15812         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15813         correct producer for the insn.
15814         (tidy_control_flow): Fixup seqnos in case of debug insns.
15815
15816 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15817
15818         PR rtl-optimization/83913
15819
15820         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15821         different sched-times when merging exprs.
15822
15823 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15824
15825         PR rtl-optimization/83962
15826
15827         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15828         tidy_fallthru_edge and tidy_control_flow.
15829
15830 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15831
15832         PR rtl-optimization/83530
15833
15834         * sel-sched.c (force_next_insn): New global variable.
15835         (remove_insn_for_debug): When force_next_insn is true, also leave only
15836         next insn in the ready list.
15837         (sel_sched_region): When the region wasn't scheduled, make another pass
15838         over it with force_next_insn set to 1.
15839
15840 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
15841
15842         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15843         into tm_file.
15844         * config/nds32/constants.md (unspec_volatile_element): Add enum values
15845         for interrupt control.
15846         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15847         functions for interrupt control.
15848         * config/nds32/nds32-intrinsic.md: Likewise.
15849         * config/nds32/nds32_intrinsic.h: Likewise.
15850         * config/nds32/nds32.h (nds32_builtins): Likewise.
15851
15852 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
15853
15854         * config/nds32/nds32.c (nds32_init_machine_status,
15855         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15856         strict_aligned_p field.
15857         (nds32_expand_to_rtl_hook): New function.
15858         (TARGET_EXPAND_TO_RTL_HOOK): Define.
15859         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15860
15861 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15862             Chung-Ju Wu  <jasonwucj@gmail.com>
15863
15864         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15865         * config/nds32/nds32-n7.md: New file.
15866         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15867         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15868         pipeline.
15869         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15870         * config/nds32/nds32.md (pipeline_model): Add n7.
15871         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15872         * config/nds32/pipelines.md: Include n7 settings.
15873
15874 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15875             Chung-Ju Wu  <jasonwucj@gmail.com>
15876
15877         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15878         * config/nds32/nds32-e8.md: New file.
15879         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15880         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15881         pipeline.
15882         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15883         * config/nds32/nds32.md (pipeline_model): Add e8.
15884         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15885         * config/nds32/pipelines.md: Include e8 settings.
15886
15887 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15888             Chung-Ju Wu  <jasonwucj@gmail.com>
15889
15890         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15891         * config/nds32/nds32-n8.md: New file.
15892         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15893         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15894         pipeline.
15895         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15896         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15897         * config/nds32/nds32.md (pipeline_model): Add n8.
15898         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15899         * config/nds32/pipelines.md: Include n8 settings.
15900
15901 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15902             Chung-Ju Wu  <jasonwucj@gmail.com>
15903
15904         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15905         * config/nds32/nds32-n9-2r1w.md: New file.
15906         * config/nds32/nds32-n9-3r2w.md: New file.
15907         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15908         nds32_register_ports): New or modify for cpu n9.
15909         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15910         pipeline.
15911         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15912         * config/nds32/nds32-utils.c: New file.
15913         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15914         TARGET_MUL_SLOW): Define.
15915         * config/nds32/nds32.md (pipeline_model): New attribute.
15916         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15917         New options that support cpu n9.
15918         * config/nds32/pipelines.md: Include n9 settings.
15919         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15920
15921 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
15922
15923         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15924         information if necessary.
15925         (output_cond_branch_compare_zero): Likewise.
15926         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15927         (nds32_target_alignment): Refine for alignment.
15928         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15929         (FUNCTION_BOUNDARY): Modify.
15930         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15931         align case.
15932         * config/nds32/nds32.opt (malways-align, malign-functions): New.
15933
15934 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
15935
15936         * config/nds32/constants.md (unspec_volatile_element): Add values for
15937         TLB operation and data prefetch.
15938         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15939         functions for TLB operation and data prefetch.
15940         * config/nds32/nds32-intrinsic.md: Likewise.
15941         * config/nds32/nds32_intrinsic.h: Likewise.
15942         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15943         (nds32_print_operand): Likewise.
15944         * config/nds32/nds32.h (nds32_builtins): Likewise.
15945
15946 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
15947         Andrew Pinski <pinsika@gcc.gnu.org>
15948
15949         PR middle-end/82976
15950         * match.pd: Use constant_boolean_node of correct type instead of
15951         boolean_true_node or boolean_false_node for simplifying
15952         pointer comparisons to zero.
15953
15954 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
15955
15956         PR tree-optimization/80021
15957         * tree.c (verify_type_variant): Make error call in verify_variant_match
15958         translatable and remove final full stop.
15959
15960 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
15961
15962         * config/nds32/constants.md (unspec_volatile_element): Add
15963         UNSPEC_VOLATILE_EH_RETURN.
15964         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15965         nds32_output_stack_pop): Support dwarf exception handling process.
15966         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15967         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15968         exception handling process.
15969         (nds32_compute_stack_frame): Likewise.
15970         (nds32_return_addr_rtx): Likewise.
15971         (nds32_initial_elimination_offset): Likewise.
15972         (nds32_expand_prologue): Likewise.
15973         (nds32_expand_epilogue): Likewise.
15974         (nds32_dynamic_chain_address): New function.
15975         * config/nds32/nds32.h (machine_function): Add fields for dwarf
15976         exception handling.
15977         (DYNAMIC_CHAIN_ADDRESS): Define.
15978         (EH_RETURN_DATA_REGNO): Define.
15979         (EH_RETURN_STACKADJ_RTX): Define.
15980         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15981         patterns for dwarf exception handling.
15982
15983 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
15984
15985         * config/nds32/nds32.h: Clean up obsolete macros.
15986
15987 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
15988
15989         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15990         Add enum values for particular instructions.
15991         * config/nds32/nds32-intrinsic.c: Implementation of expanding
15992         particular intrinsic functions.
15993         * config/nds32/nds32-intrinsic.md: Likewise.
15994         * config/nds32/nds32_intrinsic.h: Likewise.
15995         * config/nds32/nds32.h (nds32_builtins): Likewise.
15996         * config/nds32/nds32.md (type): Add pbsad and pbsada.
15997         (btst, ave): New patterns for particular instructions.
15998
15999 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16000
16001         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16002         Add enum values for atomic load/store and memory sync.
16003         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
16004         and memory sync.
16005         * config/nds32/nds32-intrinsic.md: Likewise.
16006         * config/nds32/nds32_intrinsic.h: Likewise.
16007         * config/nds32/nds32.h (nds32_builtins): Likewise.
16008
16009 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
16010
16011         PR tree-optimization/85257
16012         * fold-const.c (native_encode_vector): If not all elts could fit
16013         and off is -1, return 0 rather than offset.
16014         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
16015         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
16016         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
16017         adjust buffer in native_interpret_expr call.
16018
16019 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16020
16021         * config/nds32/constants.md (unspec_volatile_element): Add cache
16022         control enum values.
16023         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16024         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
16025         * config/nds32/nds32.c (nds32_cctl_names): New.
16026         (nds32_print_operand): Handle cache control register names.
16027         * config/nds32/nds32.h (nds32_builtins): New enum values.
16028         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16029         macros.
16030         * config/nds32/nds32.md (type): Add mmu.
16031         * config/nds32/pipelines.md (simple_insn): Add mmu.
16032
16033 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16034
16035         * config/nds32/nds32.md (type): Remove call.
16036         * config/nds32/pipelines.md (simple_insn): Likewise.
16037
16038 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16039
16040         * config/nds32/constants.md (unspec_volatile_element): Add
16041         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16042         UNSPEC_VOLATILE_FMFCFG.
16043         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16044         description for fmfcfg and fmfcsr.
16045         (bdesc_1arg): Add fmtcsr.
16046         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16047         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16048         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16049         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16050         unspec_fmfcfg): New patterns.
16051         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16052         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16053         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16054         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16055         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16056         __nds32__fmfcfg): Define.
16057
16058 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16059
16060         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16061         intrinsic register names.
16062         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16063         intrinsic register enum values and macros.
16064
16065 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16066
16067         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16068         for load/store addressing form.
16069         (nds32_print_operand_address): Likewise.
16070
16071 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
16072
16073         PR target/85196
16074         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16075         based on LABEL_REF.  Remove useless assertion.
16076         (pic_address_needs_scratch): Fix formatting.
16077         (sparc_legitimize_pic_address): Minor tweaks.
16078         (sparc_delegitimize_address): Adjust assertion accordingly.
16079         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16080         into symbolic_operand.
16081         (movsi_high_pic_label_ref): Likewise.
16082         (movsi_lo_sum_pic_label_ref): Likewise.
16083         (movdi_pic_label_ref): Likewise.
16084         (movdi_high_pic_label_ref): Likewise.
16085         (movdi_lo_sum_pic_label_ref): Likewise.
16086
16087 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
16088
16089         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16090         custom LIB_SPEC setup.
16091
16092 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
16093             Kito Cheng  <kito.cheng@gmail.com>
16094
16095         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16096         * config/riscv/freebsd.h: New.
16097
16098 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
16099
16100         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16101         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16102         file.
16103
16104 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
16105             Kito Cheng  <kito.cheng@gmail.com>
16106
16107         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16108         nds32_output_call, nds32_symbol_binds_local_p): New functions.
16109         * config/nds32/nds32-protos.h (nds32_output_call,
16110         nds32_output_return): Declare.
16111         * config/nds32/nds32.md: Refine all the call and return patterns.
16112
16113 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
16114
16115         PR debug/85252
16116         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16117         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16118
16119         PR rtl-optimization/84872
16120         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16121         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16122         EDGE_CROSSING edge.
16123
16124 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
16125
16126         * expr.c (copy_blkmode_to_reg): Revert 254862.
16127         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16128
16129 2018-04-06  Richard Biener  <rguenther@suse.de>
16130
16131         PR middle-end/85244
16132         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16133         after seeing a component reference with an adjacent field.  Treat
16134         refs to arrays at struct end of external decls similar to
16135         refs to unconstrained commons.
16136
16137 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
16138
16139         PR sanitizer/85213
16140         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16141         look through SAVE_EXPRs with non-side-effects argument.  Adjust
16142         recursive calls.
16143         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16144         save_p here.
16145
16146 2018-04-06  Richard Biener  <rguenther@suse.de>
16147
16148         PR middle-end/85180
16149         * alias.c (find_base_term): New wrapper around find_base_term
16150         unwinding CSELIB_VAL_PTR changes.
16151         (find_base_term): Do not restore CSELIB_VAL_PTR during the
16152         recursion.
16153
16154 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
16155
16156         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16157         instructions.
16158         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16159         constant definitions.
16160         ("nop"): lr 0,0 -> nopr r0
16161         ("nop_lr0", "nop_lr1"): New insn definitions.
16162
16163 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
16164
16165         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16166         NDS32_V3PUSH_AVAILABLE_P macro.
16167
16168 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
16169             Chung-Ju Wu  <jasonwucj@gmail.com>
16170
16171         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16172         (nds32*-*-*): Add float and fpu_config into supported_defaults.
16173         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16174         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16175         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16176         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16177         * config/nds32/constraints.md: New constraints and checking for hard
16178         float configuration.
16179         * config/nds32/iterators.md: New mode iterator and attribute for hard
16180         float configuration.
16181         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16182         patterns.
16183         * config/nds32/nds32-fpu.md: New file.
16184         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16185         deal with hard float code generation.
16186         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16187         ARCH_V3S.
16188         (abi_type, float_reg_number): New enum type.
16189         * config/nds32/nds32-predicates.c: New predicates for hard float.
16190         * config/nds32/nds32-protos.h: Declare functions for hard float.
16191         * config/nds32/nds32.c: Implementation for hard float configuration.
16192         * config/nds32/nds32.h: Definitions for hard float configuration.
16193         * config/nds32/nds32.md: Include hard float machine description and
16194         modify patterns for hard float configuration.
16195         * config/nds32/nds32.opt: New options for hard float configuration.
16196         * config/nds32/predicates.md: New predicates for hard float
16197         configuration.
16198
16199 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
16200
16201         * common/config/nds32/nds32-common.c
16202         (nds32_option_optimization_table): Enable -mreleax-hint by default.
16203
16204 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
16205
16206         PR middle-end/85195
16207         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16208         CONSTRUCTOR_ELT (ctor, ...)->value.
16209
16210 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
16211
16212         PR target/85193
16213         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16214
16215 2018-04-05  Tom de Vries  <tom@codesourcery.com>
16216
16217         PR target/85204
16218         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16219         cond jump.
16220
16221 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
16222             Kito Cheng  <kito.cheng@gmail.com>
16223
16224         * config/nds32/constraints.md (U33): Fine-tune checking condition.
16225         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16226         * config/nds32/nds32.h (nds32_16bit_address_type): Add
16227         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16228
16229 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
16230             Kito Cheng  <kito.cheng@gmail.com>
16231
16232         * config/nds32/constraints.md (Ufe): New memory constraint.
16233         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16234         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16235         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
16236         operands.
16237         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
16238         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
16239
16240 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16241
16242         * config/nds32/nds32.md: Use optimize_size in the condition for
16243         alu-shift instructions.
16244
16245 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16246
16247         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
16248
16249 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16250
16251         * config/nds32/nds32.md (negsi2): Refine pattern.
16252
16253 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
16254             Chung-Ju Wu  <jasonwucj@gmail.com>
16255
16256         * config/nds32/iterators.md (shift_rotate): New code iterator.
16257         (shift): New code attribute.
16258         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
16259         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
16260         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
16261         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
16262         bit-wise operations.
16263         (andsi3, *andsi3): Ditto.
16264         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
16265         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
16266         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
16267         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
16268         nds32_ior_operand, nds32_xor_operand): New predicates.
16269
16270 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16271
16272         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
16273         (addsi3, subsi3): ... this.
16274
16275 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16276
16277         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
16278
16279 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16280
16281         * config/nds32/nds32.md: Adjust indention.
16282
16283 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
16284
16285         * config/nds32/nds32.md (feature): New attribute.
16286
16287 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
16288
16289         * config/nds32/nds32.md (subtype): New attribute.
16290
16291 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
16292
16293         PR target/85203
16294         * config/arm/arm-builtins.c (arm_expand_builtin): Change
16295         expansion to perform a bitwise AND of the argument followed by a
16296         boolean negation of the result.
16297
16298 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
16299
16300         PR rtl-optimization/84878
16301         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
16302         the basic block.  Assert the use reference is not artificial and that
16303         it has an associated insn.
16304
16305 2018-04-04  Michael Matz  <matz@suse.de>
16306
16307         * builtins.c (compute_objsize): Pass correct operand
16308         to array_at_struct_end_p.
16309
16310 2018-04-04  Richard Biener  <rguenther@suse.de>
16311
16312         PR lto/85176
16313         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16314         from contexts for DINFO_LEVEL_TERSE and below.
16315
16316 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16317
16318         * config/nds32/nds32-doubleword.md (move_<mode>): Require
16319         resiter_operand condition.
16320         * config/nds32/nds32.md (*move<mode>): Ditto.
16321
16322 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16323             Monk Chiang  <sh.chiang04@gmail.com>
16324
16325         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16326
16327 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16328
16329         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16330
16331 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16332             Kito Cheng  <kito.cheng@gmail.com>
16333
16334         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16335         nds32_cond_code_str, output_cond_branch,
16336         output_cond_branch_compare_zero, nds32_expand_cbranch,
16337         nds32_expand_cstore, nds32_expand_movcc,
16338         nds32_output_cbranchsi4_equality_zero,
16339         nds32_output_cbranchsi4_equality_reg,
16340         nds32_output_cbranchsi4_equality_reg_or_const_int,
16341         nds32_output_cbranchsi4_greater_less_zero: New functions.
16342         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16343         nds32_expand_cstore, nds32_expand_movcc,
16344         nds32_output_cbranchsi4_equality_zero,
16345         nds32_output_cbranchsi4_equality_reg,
16346         nds32_output_cbranchsi4_equality_reg_or_const_int,
16347         nds32_output_cbranchsi4_greater_less_zero): Declare.
16348         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16349         nds32_rimm11s_operand): New predicates.
16350         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16351         * config/nds32/nds32.md: Rewrite all the branch and conditional move
16352         patterns.
16353
16354 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16355
16356         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16357         * config/nds32/nds32.md: Ditto.
16358         * config/nds32/pipelines.md: Ditto.
16359
16360 2018-04-04  Richard Biener  <rguenther@suse.de>
16361
16362         PR tree-optimization/85168
16363         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16364         propagating abnormals.
16365
16366 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16367
16368         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16369
16370 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16371             Kito Cheng  <kito.cheng@gmail.com>
16372
16373         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16374         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16375         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16376         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16377         * config/nds32/nds32.md (sibcall_internal): New.
16378         (sibcall_register): Remove.
16379         (sibcall_immediate): Remove.
16380         (sibcall_value_internal): New.
16381         (sibcall_value_register): Remove.
16382         (sibcall_value_immediate): Remove.
16383         * config/nds32/predicates.md (nds32_general_register_operand): New.
16384         (nds32_call_address_operand): New.
16385
16386 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
16387
16388         PR rtl-optimization/85167
16389         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16390         bb_defs if *split_p, instead preinitialize it to NULL.
16391
16392         PR tree-optimization/85156
16393         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16394         evaluating the argument multiple times.
16395
16396 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
16397
16398         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16399         than vector.
16400         (_mm_cvtpd_ps): Likewise.
16401         (_mm_cvttpd_epi32): Likewise.
16402         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16403         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16404         vector, pixel, and bool following altivec.h include.
16405
16406 2018-04-03  Martin Sebor  <msebor@redhat.com>
16407
16408         * doc/extend.texi (Common Function Attributes): Clarify.
16409         (const attribute): Likewise.
16410         (pure attribute): Likewise.
16411
16412 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
16413
16414         PR target/85169
16415         * config/i386/i386.c (ix86_expand_vector_set): Use
16416         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
16417
16418 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
16419
16420         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16421         instructions when changing rounding bits to preserve precision bits
16422         in the x87 control word.
16423
16424 2018-04-03  Martin Liska  <mliska@suse.cz>
16425
16426         PR tree-optimization/82491
16427         * rtl.h (strip_offset_and_add): Replace += suboffset with
16428         poly_uint64 () + suboffset.
16429
16430 2018-03-29  Martin Liska  <mliska@suse.cz>
16431             Martin Jambor  <mjambor@suse.cz>
16432
16433         PR ipa/84947
16434         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16435         param_type is not an integral or pointer type.
16436
16437 2018-04-03  Richard Biener  <rguenther@suse.de>
16438
16439         * sese.h (recompute_all_dominators): Remove.
16440
16441 2018-04-02  Martin Sebor  <msebor@redhat.com>
16442
16443         * doc/invoke.texi (-Wrestrict): Fix typos.
16444
16445 2018-04-02  Jim Wilson  <jimw@sifive.com>
16446
16447         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16448         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16449         (<optab>di3, <optab>si3_extend): Likewise.
16450         (<optab>si3_mask, <optab>si3_mask_1): New.
16451         (<optab>di3_mask, <optab>di3_mask_1): New.
16452         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16453         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16454         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16455
16456 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
16457
16458         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16459         example.
16460
16461 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16462
16463         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16464         (nds32_canonicalize_comparison): New function.
16465
16466 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16467             Kito Cheng  <kito.cheng@gmail.com>
16468             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
16469
16470         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16471         * config/nds32/constants.md (unspec_volatile_element): Add
16472         UNSPEC_VOLATILE_RELAX_GROUP.
16473         * config/nds32/nds32-relax-opt.c: New file.
16474         * config/nds32/nds32-predicates.c
16475         (nds32_symbol_load_store_p): New function.
16476         * config/nds32/nds32-protos.h
16477         (nds32_symbol_load_store_p): Declare function.
16478         (make_pass_nds32_relax_opt): Declare new rtl pass function.
16479         * config/nds32/nds32.c
16480         (nds32_register_pass): New function to register pass.
16481         (nds32_register_passes): New function to register passes.
16482         * config/nds32/nds32.md (relax_group): New pattern.
16483         * config/nds32/nds32.opt (mrelax-hint): New option.
16484         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16485
16486 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
16487
16488         * config/nds32/t-nds32: Modify files dependency.
16489
16490 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16491
16492         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16493         (PROFILE_HOOK): Define its implementation.
16494
16495 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16496
16497         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16498         type and 32-bit size.
16499
16500 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
16501
16502         PR middle-end/85090
16503         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16504         (V_128_256): New mode iterator.
16505         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16506         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16507         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16508         of V.
16509         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16510         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16511
16512 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
16513
16514         PR target/83315
16515         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16516         NaN inputs correctly.
16517
16518 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
16519
16520         PR target/80546
16521         * config/rs6000/vsx.md (??r): New mode attribute.
16522         (*vsx_mov<mode>_64bit): Use it.
16523         (*vsx_mov<mode>_32bit): Likewise.
16524
16525 2018-03-30  Martin Sebor  <msebor@redhat.com>
16526
16527         PR tree-optimization/84818
16528         * builtins.c (check_access): Use warning_n.
16529
16530 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
16531
16532         PR target/83822
16533         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16534         condition.
16535         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16536         condition.
16537
16538 2018-03-30  Julia Koval  <julia.koval@intel.com>
16539
16540         PR target/84413
16541         * x86-tune.def (movx, partial_reg_dependency): Enable for
16542         m_SKYLAKE_AVX512.
16543
16544 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
16545
16546         PR inline-asm/84985
16547         * lra-constraints.c (process_alt_operands): Move setting
16548         this_alternative_matches below.
16549
16550 2018-03-29  Martin Liska  <mliska@suse.cz>
16551
16552         PR lto/84995.
16553         * doc/invoke.texi: Document how LTO works with debug info.
16554         Describe auto-load support of binutils.  Mention 'x86-64'
16555         as valid option value of -march option.
16556
16557 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
16558
16559         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16560
16561         PR c/85094
16562         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16563         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16564         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16565         checking.
16566
16567 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
16568
16569         PR target/84912
16570         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16571         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16572         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16573         for RS6000_BTM_POWERPC64.
16574         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16575         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16576         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16577         definition.
16578         (DIVDE): Use it.
16579         (DIVDEU): Likewise.
16580
16581 2018-03-28  Carl Love  <cel@us.ibm.com>
16582
16583         Revert
16584         2017-09-27  Carl Love  <cel@us.ibm.com>
16585
16586         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16587         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16588         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16589         fctiw instruction.
16590
16591 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
16592
16593         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16594         instead of __vector bool.
16595         (_mm_max_pu8): Likewise.
16596         (_mm_min_pi16): Likewise.
16597
16598 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
16599
16600         PR target/84912
16601         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16602         (DIVWEUO): Likewise.
16603         (DIVDEO): Likewise.
16604         (DIVDEUO): Likewise.
16605         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16606         DIVWEUO and DIVDEUO.
16607         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16608         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16609         (div_extend): Likewise.
16610         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16611         builtin function.
16612         (__builtin_divweuo): Likewise.
16613         (__builtin_divdeo): Likewise.
16614         (__builtin_divdeuo): Likewise.
16615
16616 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
16617
16618         PR target/85095
16619         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16620         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16621
16622         PR tree-optimization/82004
16623         * gimple-match-head.c (optimize_pow_to_exp): New function.
16624         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16625         Don't fold to exp if optimize_pow_to_exp is false.
16626
16627 2018-03-28  Martin Liska  <mliska@suse.cz>
16628
16629         PR other/84819
16630         * calls.c (initialize_argument_information): Fix trailing space.
16631         * common.opt: Fix typo and provide better explanation for
16632         -fsanitize-coverage option.
16633         * config/i386/i386.opt: Fix typo.
16634
16635 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
16636             Martin Liska  <mliska@suse.cz>
16637
16638         PR sanitizer/85081
16639         * gimplify.c (asan_poison_variable): Don't do the check for
16640         gimplify_omp_ctxp here.
16641         (gimplify_decl_expr): Do it here.
16642         (gimplify_target_expr): Likewise.
16643
16644 2018-03-28  Martin Liska  <mliska@suse.cz>
16645
16646         PR target/84988
16647         * config/i386/i386.c (ix86_function_arg_advance): Do not call
16648         chkp_type_bounds_count if MPX is not enabled.
16649
16650 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
16651
16652         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16653
16654 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
16655
16656         PR target/84914
16657         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16658         function to create the function decl for complex long double
16659         multiply and divide for -mabi=ieeelongdouble.
16660         (init_float128_ieee): Call it.
16661
16662 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
16663
16664         PR target/85044
16665         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16666         -fcf-protection=branch -mibt.
16667         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16668
16669 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16670
16671         PR target/81863
16672         * config/arm/arm.c (arm_valid_symbolic_address): Handle
16673         arm_word_relocations.
16674
16675 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
16676
16677         PR target/85056
16678         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16679         extern array declarations.
16680
16681 2018-03-27  Richard Biener  <rguenther@suse.de>
16682
16683         PR middle-end/84067
16684         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16685         explicit single_use checks.
16686
16687 2018-03-27  Richard Biener  <rguenther@suse.de>
16688
16689         PR tree-optimization/85082
16690         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16691         Valueize the VUSE.
16692
16693 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16694
16695         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16696         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16697         Turn on fasynchronous-unwind-tables and funwind-tables.
16698
16699 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
16700
16701         PR target/85073
16702         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16703         (*bmi_blsr_<mode>_ccz): Ditto.
16704
16705 2018-03-26  Tom de Vries  <tom@codesourcery.com>
16706
16707         PR tree-optimization/85063
16708         * omp-general.c (offloading_function_p): New function.  Factor out
16709         of ...
16710         * omp-offload.c (pass_omp_target_link::gate): ... here.
16711         * omp-general.h (offloading_function_p): Declare.
16712         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16713         with attribute omp declare target for offloading functions.
16714
16715 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
16716
16717         PR tree-optimization/84005
16718         * tree-data-ref.h (get_base_for_alignment): Declare.
16719         * tree-data-ref.c (get_base_for_alignment_1): New function.
16720         (get_base_for_alignment): Likewise.
16721         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16722         get_base_for_alignment to find a suitable base object, instead
16723         of always using drb->base_address.
16724
16725 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
16726
16727         PR inline-asm/85022
16728         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16729         known size by default.
16730
16731 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
16732
16733         PR inline-asm/85030
16734         * lra-constraints.c (process_alt_operands): Don't match BLKmode
16735         and non BLKmode operands.
16736
16737 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16738
16739         PR target/85026
16740         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16741         Clean up attributes.
16742
16743 2018-03-23  Richard Biener  <rguenther@suse.de>
16744
16745         PR debug/85020
16746         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16747         we are going to emit early debug for LTO.
16748
16749 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
16750
16751         PR inline-asm/85034
16752         * function.c (match_asm_constraints_1): Don't optimize if input
16753         doesn't satisfy general_operand predicate for output's mode.
16754
16755         PR inline-asm/85022
16756         * alias.c (write_dependence_p): Don't require for x_canonicalized
16757         non-VOIDmode if x has VOIDmode.
16758
16759         PR sanitizer/85029
16760         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16761         just don't try to optimize it rather than assert it never happens.
16762
16763 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16764
16765         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16766         macro expansions for definition of ST_INTERNAL_<mode> and
16767         LD_INTERNAL_<mode> builtins.
16768         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16769         Remove prototype.
16770         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16771         function.
16772         (altivec_expand_st_builtin): Likewise.
16773         (altivec_expand_builtin): Remove calls to deleted functions.
16774         (rs6000_address_for_altivec): Delete this function.
16775         * config/rs6000/vector.md: Remove expands for
16776         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16777
16778 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
16779
16780         PR target/84826
16781         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16782         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16783         re-computing once computed.
16784         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16785         (arm_init_machine_status): Initialize
16786         machine->static_chain_stack_bytes.
16787
16788 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16789
16790         PR target/84760
16791         * doc/extend.texi: Add four new prototypes for vec_ld.
16792         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16793         definitions for more logical presentation.
16794         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16795         entries for V1TI variants of __builtin_altivec_ld builtin.
16796         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16797         handling of V1TI variant of LVX icode pattern.
16798         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16799         (rs6000_gimple_fold_builtin): Likewise.
16800         (altivec_init_builtins): Add code to define
16801         __builtin_altivec_lvx_v1ti function.
16802
16803 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
16804
16805         PR inline-asm/84941
16806         * function.c (match_asm_constraints_1): Don't do the optimization
16807         if input isn't a REG, SUBREG, MEM or constant.
16808
16809 2018-03-22  Tom de Vries  <tom@codesourcery.com>
16810
16811         PR tree-optimization/84956
16812         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16813         bb_has_abnormal_pred.
16814
16815 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
16816
16817         PR sanitizer/85018
16818         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16819         DECL_INITIAL (decl) to decl at the end.
16820         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16821         adjust the comment.
16822
16823 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
16824
16825         * doc/extend.texi (__builtin_tgmath): Document when complex
16826         integer types are treated as _Complex _Float64.
16827
16828 2018-03-21  Tom de Vries  <tom@codesourcery.com>
16829
16830         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16831
16832 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16833
16834         PR tree-optimization/84960
16835         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16836         if it is ENTRY block, move them into single succ of ENTRY in that case.
16837
16838 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
16839
16840         PR tree-optimization/84811
16841         * poly-int.h (poly_span_traits): Remove the T3 parameter and
16842         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16843         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16844         (known_subrange_p): Update accordingly.  Cast each value involved
16845         in the size comparison, rather than casting the result of the
16846         subtraction.
16847
16848 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16849
16850         PR tree-optimization/84982
16851         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16852         by flipping the least significant bit rather than all bits from
16853         bitpos to bitpos + bitsize - 1.
16854
16855 2018-03-21  Nathan Sidwell  <nathan@acm.org>
16856
16857         * doc/extend.texi (Deprecated Features): Remove mention of
16858         long-deleted deprecations.
16859
16860 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16861
16862         PR jit/84288
16863         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16864         * configure: Regenerate.
16865
16866 2018-03-21  Tom de Vries  <tom@codesourcery.com>
16867
16868         PR tree-optimization/83126
16869         * tree-parloops.c (num_phis): New function.
16870         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16871
16872 2018-03-21  Nathan Sidwell  <nathan@acm.org>
16873
16874         * doc/extend.texi (Deprecated Features): Update deprecated flags,
16875         mention anon-struct/union members and trailing attributes.
16876
16877 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
16878
16879         PR tree-optimization/84969
16880         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16881         builtin memset partitions if they set different rhs values.
16882
16883 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16884
16885         PR rtl-optimization/84989
16886         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16887         VEC_DUPLICATE with scalar result mode.
16888
16889 2018-03-21  Martin Liska  <mliska@suse.cz>
16890
16891         PR ipa/84963
16892         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16893         not intended return statement.
16894
16895 2018-03-21  Martin Liska  <mliska@suse.cz>
16896
16897         PR target/84988
16898         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16899         (chkp_find_bound_slots_1): Limit number of iterations.
16900
16901 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
16902
16903         PR target/84838
16904         * Minor grammar fixes for x86 options.
16905
16906 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
16907
16908         PR debug/84875
16909         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16910         holding REG_CFA_RESTORE notes, instead turn them into a USE.
16911
16912 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
16913
16914         PR target/83789
16915         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16916         (altivec_lvx_<mode>_1op): Likewise.
16917         (altivec_stvx_<mode>_2op): Likewise.
16918         (altivec_stvx_<mode>_1op): Likewise.
16919         (altivec_lvx_<VM2:mode>): New define_expand.
16920         (altivec_stvx_<VM2:mode>): Likewise.
16921         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16922         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16923         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16924         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16925         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16926         (rs6000_gen_lvx): Likewise.
16927         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16928         (altivec_expand_stv_builtin): Likewise.
16929         (altivec_expand_builtin): Likewise.
16930         * config/rs6000/vector.md: Likewise.
16931
16932 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16933
16934         PR target/82518
16935         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16936         BYTES_BIG_ENDIAN.
16937
16938 2018-03-20  Richard Biener  <rguenther@suse.de>
16939
16940         PR target/84986
16941         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16942         sign-conversions as zero, fall back to standard scalar_stmt
16943         cost for the rest.
16944
16945 2018-03-20  Martin Liska  <mliska@suse.cz>
16946
16947         PR ipa/84825
16948         * predict.c (rebuild_frequencies): Handle case when we have
16949         PROFILE_ABSENT, but flag_guess_branch_prob is false.
16950
16951 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
16952
16953         PR target/84990
16954         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16955         flag_section_anchors.
16956         * varasm.c (use_blocks_for_decl_p): Remove hack for
16957         dw2_force_const_mem.
16958
16959         PR target/84845
16960         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16961         to ...
16962         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
16963         be created, use lowpart_subreg of operands[0] rather than operands[0]
16964         itself.
16965         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16966         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16967         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16968         and n constraint instead of aarch64_shift_imm_di and Usd.
16969         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16970         (*aarch64_<optab>_reg_minus<mode>3): ... this.
16971
16972 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
16973
16974         PR target/82989
16975         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16976         to favor GPR over NEON registers.
16977         (<shift>di3_neon): Likewise.
16978
16979 2018-03-20  Tom de Vries  <tom@codesourcery.com>
16980
16981         PR target/84952
16982         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16983         (nvptx_process_pars): Emit bar.sync asap and alap.
16984
16985 2018-03-20  Tom de Vries  <tom@codesourcery.com>
16986
16987         PR target/84954
16988         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16989         seen_label if seen_label is already set.
16990
16991 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
16992
16993         PR target/84945
16994         * config/i386/i386.c (fold_builtin_cpu): For features above 31
16995         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16996         Use 1U instead of 1.  Formatting fixes.
16997
16998         PR c/84953
16999         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
17000         instead of TREE_TYPE (s1) for the return value.
17001
17002 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
17003
17004         PR tree-optimization/84946
17005         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
17006         bitsize + bitsize in poly_uint64 rather than poly_int64.
17007
17008         PR sanitizer/78651
17009         * dwarf2asm.c: Include fold-const.c.
17010         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
17011         of decl rather than decl itself.
17012
17013         PR rtl-optimization/84643
17014         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
17015
17016 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
17017
17018         PR sanitizer/78651
17019         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
17020         calling assemble_variable.
17021
17022 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
17023
17024         PR target/81647
17025         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17026         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17027
17028 2018-03-19  Jim Wilson  <jimw@sifive.com>
17029
17030         PR bootstrap/84856
17031         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
17032         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
17033         (riscv_first_stack_step): Likewise.
17034         (riscv_option_override): Use STACK_BOUNDARY instead of
17035         MIN_STACK_BOUNDARY.
17036         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17037         MIN_STACK_BOUNDARY.
17038         (BIGGEST_ALIGNMENT): Set to 128.
17039         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17040         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17041         STACK_BOUNDARY.
17042
17043 2018-03-19  Richard Biener  <rguenther@suse.de>
17044
17045         PR tree-optimization/84933
17046         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17047         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17048
17049 2018-03-19  Richard Biener  <rguenther@suse.de>
17050
17051         PR tree-optimization/84859
17052         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17053         (cond_if_else_store_replacement): Perform sinking operation on
17054         single-store BBs regardless of MAX_STORES_TO_SINK setting.
17055         Generalize what a BB with a single eligible store is.
17056
17057 2018-03-19  Richard Biener  <rguenther@suse.de>
17058
17059         PR tree-optimization/84929
17060         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17061         chrec_is_positive against non-chrec arg.
17062
17063 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
17064
17065         PR target/84711
17066         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17067
17068 2018-03-18  Martin Liska  <mliska@suse.cz>
17069
17070         PR rtl-optimization/84635
17071         * regrename.c (build_def_use): Use matches_mode only when
17072         matches >= 0.
17073
17074 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
17075
17076         PR tree-optimization/84913
17077         * tree-vect-loop.c (vectorizable_reduction): Don't try to
17078         vectorize chains of COND_EXPRs.
17079
17080 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
17081
17082         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17083
17084 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
17085
17086         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17087
17088 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
17089
17090         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17091
17092 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
17093             Kito Cheng  <kito.cheng@gmail.com>
17094
17095         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17096         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17097         (nds32_adjust_reg_alloc_order): New function.
17098         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17099
17100 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
17101
17102         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17103         nds32_print_operand, nds32_print_operand_address): Use
17104         HOST_WIDE_INT_PRINT_DEC instead.
17105
17106 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
17107
17108         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17109
17110 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
17111
17112         PR target/84902
17113         * config/i386/i386.c (initial_ix86_tune_features,
17114         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17115         unsigned long long.
17116         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17117         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17118         rather than 1u << ix86_tune.  Formatting fix.
17119         (ix86_option_override_internal): Change ix86_arch_mask from
17120         unsigned int to unsigned HOST_WIDE_INT, initialize to
17121         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17122         (ix86_function_specific_restore): Likewise.
17123
17124 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
17125
17126         PR target/84899
17127         * postreload.c (reload_combine_recognize_pattern): Perform
17128         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17129         truncate_int_for_mode the result for the destination's mode.
17130
17131         PR c/84909
17132         * hsa-gen.c (mem_type_for_type): Fix comment typo.
17133         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17134         Likewise.
17135         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17136         Likewise.
17137
17138 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
17139
17140         PR target/84876
17141         * lra-assigns.c (lra_split_hard_reg_for): Don't use
17142         regno_allocno_class_array and sorted_pseudos.
17143         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17144         insns where regno is used.
17145
17146 2018-03-16  Martin Liska  <mliska@suse.cz>
17147
17148         PR ipa/84833
17149         * multiple_target.c (create_dispatcher_calls): Redirect
17150         reference in the symbol table.
17151
17152 2018-03-16  Martin Liska  <mliska@suse.cz>
17153
17154         PR ipa/84722
17155         * multiple_target.c (create_dispatcher_calls): Redirect also
17156         an alias.
17157
17158 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
17159
17160         PR c++/79937
17161         PR c++/82410
17162         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17163         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17164         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17165
17166 2018-03-16  Julia Koval  <julia.koval@intel.com>
17167
17168         * doc/invoke.texi (Skylake Server): Add CLWB.
17169         Cannonlake): Remove CLWB.
17170
17171 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
17172
17173         PR tree-optimization/84841
17174         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17175         1 << 3.
17176         (FLOAT_ONE_CONST_TYPE): Define.
17177         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17178         (sort_by_operand_rank): Put entries with higher constant_type last
17179         rather than first to match comments.
17180
17181 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
17182
17183         * config/nios2/nios2.md (movsi_internal): Fix thinko in
17184         split predicate.
17185
17186 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
17187
17188         PR c++/79085
17189         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17190         check and use address of target always.
17191
17192 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
17193
17194         PR target/84574
17195         * config/i386/i386.c (indirect_thunk_needed): Update comments.
17196         (indirect_thunk_bnd_needed): Likewise.
17197         (indirect_thunks_used): Likewise.
17198         (indirect_thunks_bnd_used): Likewise.
17199         (indirect_return_needed): New.
17200         (indirect_return_bnd_needed): Likewise.
17201         (output_indirect_thunk_function): Add a bool argument for
17202         function return.
17203         (output_indirect_thunk_function): Don't generate alias for
17204         function return thunk.
17205         (ix86_code_end): Call output_indirect_thunk_function to generate
17206         function return thunks.
17207         (ix86_output_function_return): Set indirect_return_bnd_needed
17208         and indirect_return_needed instead of indirect_thunk_bnd_needed
17209         and indirect_thunk_needed.
17210
17211 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
17212
17213         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17214         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17215         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17216
17217 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
17218             Paul Hua <paul.hua.gm@gmail.com>
17219
17220         PR c/84852
17221         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17222
17223 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
17224
17225         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17226         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17227         resp. SFmode cases.
17228
17229 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
17230
17231         PR target/84711
17232         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17233         instead of GET_MODE_SIZE when comparing Units.
17234
17235 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
17236
17237         PR target/68256
17238         * varasm.c (hash_section): Return an unchangeble hash value
17239         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
17240         Return !aarch64_can_use_per_function_literal_pools_p ().
17241
17242 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
17243
17244         PR target/84860
17245         * optabs.c (emit_conditional_move): Pass address of cmode's copy
17246         rather than address of cmode as last argument to prepare_cmp_insn.
17247
17248 2018-03-15  Julia Koval  <julia.koval@intel.com>
17249
17250         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17251         F_AVX512VNNI, F_AVX512BITALG): New.
17252
17253 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
17254
17255         PR target/83451
17256         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
17257         insn for floating-point loads and stores.
17258
17259 2018-03-14  Carl Love  <cel@us.ibm.com>
17260
17261         * config/rs6000/rs6000-c.c: Add macro definitions for
17262         ALTIVEC_BUILTIN_VEC_PERMXOR.
17263         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
17264         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
17265         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
17266         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
17267         UNSPEC_VPERMXOR.
17268         * config/doc/extend.texi: Add prototypes for vec_permxor.
17269
17270 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
17271
17272         PR c/84852
17273         * diagnostic-show-locus.c (class layout_point): Convert m_line
17274         from int to linenum_type.
17275         (line_span::comparator): Use linenum "compare" function when
17276         comparing line numbers.
17277         (test_line_span): New function.
17278         (layout_range::contains_point): Convert param "row" from int to
17279         linenum_type.
17280         (layout_range::intersects_line_p): Likewise.
17281         (layout::will_show_line_p): Likewise.
17282         (layout::print_source_line): Likewise.
17283         (layout::should_print_annotation_line_p): Likewise.
17284         (layout::print_annotation_line): Likewise.
17285         (layout::print_leading_fixits): Likewise.
17286         (layout::annotation_line_showed_range_p): Likewise.
17287         (struct line_corrections): Likewise for field m_row.
17288         (line_corrections::line_corrections): Likewise for param "row".
17289         (layout::print_trailing_fixits): Likewise.
17290         (layout::get_state_at_point): Likewise.
17291         (layout::get_x_bound_for_row): Likewise.
17292         (layout::print_line): Likewise.
17293         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
17294         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
17295         * input.c (selftest::test_linenum_comparisons): New function.
17296         (selftest::input_c_tests): Call it.
17297         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
17298         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
17299         * selftest.h (ASSERT_GT): New macro.
17300         (ASSERT_GT_AT): New macro.
17301         (ASSERT_LT): New macro.
17302         (ASSERT_LT_AT): New macro.
17303
17304 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
17305
17306         PR rtl-optimization/84780
17307         * combine.c (distribute_links): Don't make a link based on pc_rtx.
17308
17309 2018-03-14  Martin Liska  <mliska@suse.cz>
17310
17311         * tree.c (record_node_allocation_statistics): Use
17312         get_stats_node_kind.
17313         (get_stats_node_kind): New function extracted from
17314         record_node_allocation_statistics.
17315         (free_node): Use get_stats_node_kind.
17316
17317 2018-03-14  Richard Biener  <rguenther@suse.de>
17318
17319         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17320         that the value-set of ANTIC_IN doesn't grow.
17321
17322         Revert
17323         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17324         member.
17325         (BB_VISITED_WITH_VISITED_SUCCS): New define.
17326         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17327
17328 2018-03-14  Julia Koval  <julia.koval@intel.com>
17329
17330         * config.gcc (icelake-client, icelake-server): New.
17331         (icelake): Remove.
17332         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17333         (initial_ix86_arch_features): Ditto.
17334         (PTA_SKYLAKE): Add SGX.
17335         (PTA_ICELAKE): Remove.
17336         (PTA_ICELAKE_CLIENT): New.
17337         (PTA_ICELAKE_SERVER): New.
17338         (ix86_option_override_internal): Split up icelake on icelake client and
17339         icelake server.
17340         (get_builtin_code_for_version): Ditto.
17341         (fold_builtin_cpu): Ditto.
17342         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17343         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17344         * config/i386/i386.h (processor_type): Ditto.
17345         * doc/invoke.texi: Ditto.
17346
17347 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
17348
17349         PR sanitizer/83392
17350         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17351         INTEGER_CST offset, add it together with bitpos / 8 and
17352         sign extend based on POINTER_SIZE.
17353
17354         PR target/84844
17355         Revert
17356         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
17357
17358         PR target/78090
17359         * config/i386/constraints.md (Yc): New register constraint.
17360         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17361         Use Yc constraint for alternative 2 of operand 0.  Remove
17362         preferred_for_speed attribute.
17363
17364 2018-03-14  Richard Biener  <rguenther@suse.de>
17365
17366         PR tree-optimization/84830
17367         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17368         with the old one to avoid oscillations.
17369
17370 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
17371
17372         PR target/83712
17373         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17374         pseudos.
17375         (assign_by_spills): Return a flag of reload assignment failure.
17376         Do not process the reload assignment failures.  Do not spill other
17377         reload pseudos if they has the same reg class.  Update n if
17378         necessary.
17379         (lra_assign): Add a return arg.  Set up from the result of
17380         assign_by_spills call.
17381         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17382         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17383         usage_insns if it is not NULL.
17384         (spill_hard_reg_in_range): New function.
17385         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17386         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17387         function prototypes.
17388         (lra_assign): Change prototype.
17389         * lra.c (lra): Add code to deal with fails by splitting hard reg
17390         live ranges.
17391
17392 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
17393
17394         * config/riscv/riscv.opt (mrelax): New option.
17395         * config/riscv/riscv.c (riscv_file_start): Emit ".option
17396         "norelax" when riscv_mrelax is disabled.
17397         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17398
17399 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
17400
17401         PR target/84743
17402         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17403         reassociation for int modes.
17404
17405 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17406
17407         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17408         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17409         for big-endian.
17410         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17411         * config/aarch64/aarch64-sve.md
17412         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17413         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17414         (*extend<mode><Vwide>2): Rename to...
17415         (aarch64_sve_extend<mode><Vwide>2): ...this.
17416         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17417         renaming the old pattern to...
17418         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
17419         unsigned packs.
17420         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17421         define_expand, renaming the old pattern to...
17422         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17423         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17424         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17425         account when deciding which SVE instruction the optab should use.
17426         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17427
17428 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17429
17430         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17431         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17432         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17433         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17434         (tlsdesc_small_<mode>): Turn a define_expand and use
17435         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
17436         (tlsdesc_small_advsimd_<mode>): ...this.
17437         (tlsdesc_small_sve_<mode>): New pattern.
17438
17439 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17440
17441         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17442         (UNSPEC_UMUL_HIGHPART): New constants.
17443         (MUL_HIGHPART): New int iteraor.
17444         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17445         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17446         define_expand.
17447         (*<su>mul<mode>3_highpart): New define_insn.
17448
17449 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
17450
17451         PR lto/84805
17452         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17453         incomplete types.
17454
17455 2018-03-13  Martin Liska  <mliska@suse.cz>
17456
17457         PR ipa/84658.
17458         * (sem_item_optimizer::sem_item_optimizer): Initialize new
17459         vector.
17460         (sem_item_optimizer::~sem_item_optimizer): Release it.
17461         (sem_item_optimizer::merge_classes): Register variable aliases.
17462         (sem_item_optimizer::fixup_pt_set): New function.
17463         (sem_item_optimizer::fixup_points_to_sets): Likewise.
17464         * ipa-icf.h: Declare new variables and functions.
17465
17466 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
17467
17468         PR middle-end/84834
17469         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17470         integer_pow2p@2 and test integer_pow2p in condition.
17471         (A < 0 ? C : 0): Similarly for @1.
17472
17473         PR middle-end/84831
17474         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17475         characters starting at p contain '\0' character, don't look beyond
17476         that.
17477
17478         PR target/84827
17479         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17480         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17481
17482         PR target/84828
17483         * reg-stack.c (change_stack): Change update_end var from int to
17484         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17485         also call set_block_for_insn on the newly added insns and rescan.
17486
17487         PR target/84786
17488         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17489         on the last operand.
17490
17491         PR c++/84704
17492         * tree.c (stabilize_reference_1): Return save_expr (e) for
17493         STATEMENT_LIST even if it doesn't have side-effects.
17494
17495 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
17496
17497         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17498
17499 2018-03-12  Renlin Li  <renlin.li@arm.com>
17500
17501         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17502         aarch64_output_scalar_simd_mov_immediate.
17503
17504 2018-03-12  Martin Sebor  <msebor@redhat.com>
17505
17506         PR tree-optimization/83456
17507         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17508         for perfectly overlapping calls to memcpy.
17509         (gimple_fold_builtin_memory_chk): Same.
17510         (gimple_fold_builtin_strcpy): Handle no-warning.
17511         (gimple_fold_builtin_stxcpy_chk): Same.
17512         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17513
17514 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
17515
17516         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17517         parameter.  Use it for SFmode.
17518         (rs6000_function_arg_advance_1): Adjust.
17519         (rs6000_function_arg): Adjust.
17520         (rs6000_gimplify_va_arg): Pass false for that new parameter.
17521
17522 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
17523
17524         PR rtl-optimization/84169
17525         PR rtl-optimization/84780
17526         * combine.c (can_combine_p): Check for a 2-insn combination whether
17527         the destination register is used between the two insns, too.
17528
17529 2018-03-12  Richard Biener  <rguenther@suse.de>
17530
17531         PR tree-optimization/84803
17532         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17533         for refs DR analysis didn't process.
17534
17535 2018-03-12  Richard Biener  <rguenther@suse.de>
17536
17537         PR tree-optimization/84777
17538         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17539         force-vectorize loops ignore whether we are optimizing for size.
17540
17541 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
17542
17543         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17544         (TARGET_MD_ASM_ADJUST): Define.
17545
17546 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
17547             Kito Cheng  <kito.cheng@gmail.com>
17548             Chung-Ju Wu  <jasonwucj@gmail.com>
17549
17550         * config/nds32/nds32.c (nds32_compute_stack_frame,
17551         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17552         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17553         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17554         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17555         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17556         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17557         * config/nds32/nds32.md (prologue, epilogue): Use macro
17558         NDS32_V3PUSH_AVAILABLE_P to do checking.
17559
17560 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
17561
17562         PR debug/58150
17563         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17564         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17565         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
17566         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
17567         addition of most attributes on !orig_type_die or the attribute not
17568         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17569
17570 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17571             Chung-Ju Wu  <jasonwucj@gmail.com>
17572
17573         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17574         __NDS32_VH__ macro.
17575         * config/nds32/nds32.opt (mvh): New option.
17576
17577 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17578             Chung-Ju Wu  <jasonwucj@gmail.com>
17579
17580         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17581         function.
17582         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17583         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17584         definition.
17585
17586 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17587             Chung-Ju Wu  <jasonwucj@gmail.com>
17588
17589         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17590         function.
17591         * config/nds32/nds32-multiple.md (strlensi): New pattern.
17592         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17593
17594 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
17595             Kito Cheng  <kito.cheng@gmail.com>
17596             Chung-Ju Wu  <jasonwucj@gmail.com>
17597
17598         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17599         UNSPEC_FFMISM and UNSPEC_FLMISM.
17600         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17601         for ffb, ffmism and flmism.
17602         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17603         (unspec_ffmism): Ditto.
17604         (unspec_flmism): Ditto.
17605         (nds32_expand_builtin_impl): Check if string extension is available.
17606         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17607         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17608
17609 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
17610
17611         Reverting patch:
17612         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
17613
17614         PR target/83712
17615         * lra-assigns.c (assign_by_spills): Return a flag of reload
17616         assignment failure.  Do not process the reload assignment
17617         failures.  Do not spill other reload pseudos if they has the same
17618         reg class.
17619         (lra_assign): Add a return arg.  Set up from the result of
17620         assign_by_spills call.
17621         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17622         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17623         usage_insns if it is not NULL.
17624         (spill_hard_reg_in_range): New function.
17625         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17626         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17627         function prototypes.
17628         (lra_assign): Change prototype.
17629         * lra.c (lra): Add code to deal with fails by splitting hard reg
17630         live ranges.
17631
17632 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
17633
17634         PR target/84807
17635         * config/i386/i386.opt: Replace Enforcment with Enforcement.
17636
17637 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
17638
17639         PR debug/84620
17640         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17641         (dw_val_node): Add val_symbolic_view.
17642         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17643         (symview_upper_bound): New.
17644         (new_line_info_table): Initialize symviews_since_reset.
17645         (dwarf2out_source_line): Count symviews_since_reset and set
17646         symview_upper_bound.
17647         (dw_val_equal_p): Handle symview.
17648         (add_AT_symview): New.
17649         (print_dw_val): Handle symview.
17650         (attr_checksum, attr_checksum_ordered): Likewise.
17651         (same_dw_val_p, size_of_die): Likewise.
17652         (value_format, output_die): Likewise.
17653         (add_high_low_attributes): Use add_AT_symview for entry_view.
17654         (dwarf2out_finish): Reset symview_upper_bound, clear
17655         zero_view_p.
17656
17657 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
17658
17659         PR target/83969
17660         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17661         Add strict argument and use it.
17662         (rs6000_split_multireg_move): Update for new strict argument.
17663         (mem_operand_gpr): Disallow all non-offsettable addresses.
17664         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17665
17666 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
17667
17668         PR target/84772
17669         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17670         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17671         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17672
17673         PR c++/84767
17674         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17675         decl, use remap_type if we want to use the type.
17676
17677 2018-03-09  Martin Sebor  <msebor@redhat.com>
17678
17679         PR tree-optimization/84526
17680         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17681         Remove dead code.
17682         (builtin_access::generic_overlap): Be prepared to handle non-array
17683         base objects.
17684
17685 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
17686
17687         PR rtl-optimization/84682
17688         * lra-constraints.c (process_address_1): Check is_address flag
17689         for address constraints.
17690         (process_alt_operands): Likewise.
17691         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17692         preprocess_constraints.
17693         * recog.h (preprocess_constraints): Add oploc parameter.
17694         Adjust callers.
17695         * recog.c (preprocess_constraints): Test address_operand for
17696         CT_ADDRESS constraints.
17697
17698 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
17699
17700         PR target/83712
17701         * lra-assigns.c (assign_by_spills): Return a flag of reload
17702         assignment failure.  Do not process the reload assignment
17703         failures.  Do not spill other reload pseudos if they has the same
17704         reg class.
17705         (lra_assign): Add a return arg.  Set up from the result of
17706         assign_by_spills call.
17707         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17708         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17709         usage_insns if it is not NULL.
17710         (spill_hard_reg_in_range): New function.
17711         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17712         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17713         function prototypes.
17714         (lra_assign): Change prototype.
17715         * lra.c (lra): Add code to deal with fails by splitting hard reg
17716         live ranges.
17717
17718 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17719
17720         PR target/83193
17721         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17722         Accept complain bool parameter.  Only emit errors if it is true.
17723         (arm_parse_cpu_option_name): Likewise.
17724         (arm_target_thumb_only): Adjust callers of the above.
17725         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17726         prototype to take a default true bool parameter.
17727         (arm_parse_arch_option_name): Likewise.
17728
17729 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
17730             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
17731
17732         PR jit/64089
17733         PR jit/84288
17734         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17735         * configure: Regenerate.
17736         * configure.ac ("linker --version-script option"): New.
17737         ("linker soname option"): New.
17738
17739 2018-03-09  Richard Biener  <rguenther@suse.de>
17740
17741         PR tree-optimization/84775
17742         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17743         immediate uses of predicate stmts and mark them modified.
17744
17745         Revert
17746         PR tree-optimization/84178
17747         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17748         to caller.
17749         (version_loop_for_if_conversion): Delay update_ssa call.
17750         (tree_if_conversion): Delay update_ssa until after predicate
17751         insertion.
17752
17753 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
17754
17755         PR target/84763
17756         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17757         when the function accesses prior frames.
17758
17759 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
17760
17761         PR debug/84456
17762         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17763         gen_llsym, otherwise call maybe_gen_llsym.
17764
17765         PR inline-asm/84742
17766         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17767         has ',' character inside of it.
17768
17769 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17770
17771         PR target/84748
17772         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17773         as clobbering CC_REGNUM.
17774
17775 2018-03-08  Richard Biener  <rguenther@suse.de>
17776
17777         PR middle-end/84552
17778         * tree-scalar-evolution.c: Include tree-into-ssa.h.
17779         (follow_copies_to_constant): Do not follow SSA names registered
17780         for update.
17781
17782 2018-03-08  Richard Biener  <rguenther@suse.de>
17783
17784         PR tree-optimization/84178
17785         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17786         to caller.
17787         (version_loop_for_if_conversion): Delay update_ssa call.
17788         (tree_if_conversion): Delay update_ssa until after predicate
17789         insertion.
17790
17791 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
17792
17793         PR tree-optimization/84178
17794         * tree-if-conv.c (release_bb_predicate): Remove the
17795         the assertion that the stmts have NULL use_ops.
17796         Discard the statements, asserting that they haven't
17797         yet been added to a BB.
17798
17799 2018-03-08  Richard Biener  <rguenther@suse.de>
17800
17801         PR tree-optimization/84746
17802         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17803         (phi_translate): Pass in destination ANTIC_OUT set.
17804         (phi_translate_1): Likewise.  For a simplified result lookup
17805         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17806         (phi_translate_set): Adjust.
17807         (do_pre_regular_insertion): Likewise.
17808         (do_pre_partial_partial_insertion): Likewise.
17809
17810 2018-03-08  Martin Liska  <mliska@suse.cz>
17811
17812         PR gcov-profile/84735
17813         * doc/gcov.texi: Document usage of profile files.
17814         * gcov-io.h: Document changes in the format.
17815
17816 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
17817
17818         PR debug/84404
17819         PR debug/84408
17820         * dwarf2out.c (struct dw_line_info_table): Update comments for
17821         view == -1.
17822         (FORCE_RESET_NEXT_VIEW): New.
17823         (FORCE_RESETTING_VIEW_P): New.
17824         (RESETTING_VIEW_P): Check for -1 too.
17825         (ZERO_VIEW_P): Likewise.
17826         (new_line_info_table): Force-reset next view.
17827         (dwarf2out_begin_function): Likewise.
17828         (dwarf2out_source_line): Simplify zero_view_p initialization.
17829         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17830         view directly.  Omit view when omitting .loc at line 0.
17831
17832 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
17833
17834         PR tree-optimization/84740
17835         * tree-switch-conversion.c (process_switch): Call build_constructors
17836         only if info.phi_count is non-zero.
17837
17838         PR tree-optimization/84739
17839         * tree-tailcall.c (find_tail_calls): Check call arguments against
17840         DECL_ARGUMENTS (current_function_decl) rather than
17841         DECL_ARGUMENTS (func) when checking for tail recursion.
17842
17843 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
17844
17845         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17846         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17847         Volker Reichelt's entry and add entries for people that perform
17848         GCC fuzzy testing and report numerous bugs.
17849
17850 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
17851
17852         PR target/82411
17853         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17854         readonly data in sdata, if that is disabled.
17855         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17856         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17857         -mreadonly-in-sdata option.
17858
17859 2018-03-07  Martin Sebor  <msebor@redhat.com>
17860
17861         PR tree-optimization/84468
17862         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17863         basic block when looking for nul assignment.
17864
17865 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
17866
17867         PR target/84277
17868         * except.h (output_function_exception_table): Adjust prototype.
17869         * except.c (output_function_exception_table): Remove FNNAME parameter
17870         and add SECTION parameter.  Ouput one part of the table at a time.
17871         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17872         the first part of the exception table and emit unwind directives.
17873         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17874         (i386_pe_seh_cold_init): Likewise.
17875         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17876         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17877         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17878         (ix86_output_call_insn): Emit a nop in one more case for SEH.
17879         * config/i386/winnt.c: Include except.h.
17880         (struct seh_frame_state): Add reg_offset, after_prologue and
17881         in_cold_section fields.
17882         (i386_pe_seh_end_prologue): Set seh->after_prologue.
17883         (i386_pe_seh_cold_init): New function.
17884         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17885         to seh->in_cold_section.
17886         (seh_emit_push): Record the offset of the push.
17887         (seh_emit_save): Record the offet of the save.
17888         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17889         Test seh->after_prologue to disregard the epilogue.
17890         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17891         (i386_pe_end_cold_function): New function.
17892
17893 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
17894
17895         PR fortran/84565
17896         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17897         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17898
17899         PR c++/84704
17900         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17901         on tmp_var.
17902         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17903         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17904
17905         PR middle-end/84723
17906         * multiple_target.c: Include tree-inline.h and intl.h.
17907         (expand_target_clones): Diagnose and fail if node->definition and
17908         !tree_versionable_function_p (node->decl).
17909
17910 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
17911
17912         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17913         sprint_ul.
17914         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17915         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17916         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17917
17918 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
17919
17920         PR target/84710
17921         * combine.c (try_combine): Use reg_or_subregno instead of handling
17922         just paradoxical SUBREGs and REGs.
17923
17924 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
17925
17926         * config/arc/arc.c (arc_finalize_pic): Remove function.
17927         (arc_must_save_register): We use single base PIC register, remove
17928         checks to save/restore the PIC register.
17929         (arc_expand_prologue): Likewise.
17930         * config/arc/arc-protos.h (arc_set_default_type_attributes):
17931         Remove.
17932         (arc_verify_short): Likewise.
17933         (arc_attr_type): Likewise.
17934         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17935         (walk_stores): Likewise.
17936         (arc_address_cost): Make it static.
17937         (arc_verify_short): Likewise.
17938         (branch_dest): Likewise.
17939         (arc_attr_type): Likewise.
17940         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17941         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17942         (arc_final_prescan_insn): Remove inserting the nops due to
17943         hardware hazards.  It is done in reorg step.
17944         (insn_length_variant_t): Remove.
17945         (insn_length_parameters_t): Likewise.
17946         (arc_insn_length_parameters): Likewise.
17947         (arc_get_insn_variants): Likewise.
17948         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17949
17950 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
17951
17952         PR inline-asm/84683
17953         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17954         assertion failure.
17955
17956         PR tree-optimization/84687
17957         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17958         on new_node->decl.
17959         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17960
17961 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17962
17963         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17964         Rename to ppc_speculation_barrier.
17965         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17966         __builtin_ppc_speculation_barrier.
17967
17968 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
17969
17970         PR target/84700
17971         * combine.c (combine_simplify_rtx): Don't try to simplify if
17972         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17973         are equal to x.
17974
17975 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
17976
17977         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17978         to 32 bytes when compiling for POWER9.
17979
17980 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
17981
17982         PR target/84564
17983         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17984         regparm >= 3 with no arg reg available also for calls with
17985         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
17986
17987         PR target/84524
17988         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17989         orig,vex.
17990         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
17991
17992 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
17993
17994         PR target/84264
17995         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17996
17997 2018-03-05  Richard Biener  <rguenther@suse.de>
17998
17999         PR tree-optimization/84486
18000         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
18001         When inserting a __builtin_assume_aligned call set the LHS
18002         SSA name alignment info accordingly.
18003
18004 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
18005
18006         PR tree-optimization/84114
18007         * config/aarch64/aarch64.c (aarch64_reassociation_width)
18008         Avoid reassociation of FLOAT_MODE addition.
18009
18010 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
18011
18012         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
18013         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
18014         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
18015         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
18016         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
18017         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
18018         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
18019         and -mwbnoinvd.
18020         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
18021         __builtin_ia32_wbinvd): New builtins.
18022         (SPECIAL_ARGS2): New.
18023         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
18024         (SPECIAL_ARGS2): New.
18025         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18026         (ix86_valid_target_attribute_inner_p): Ditto.
18027         (ix86_init_mmx_sse_builtins): Add special_args2.
18028         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18029         TARGET_WBNOINVD_P): New.
18030         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18031         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
18032         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
18033         * config/i386/immintrin.h (_wbinvd): New intrinsic.
18034         * config/i386/pconfigintrin.h: New file.
18035         * config/i386/wbnoinvdintrin.h: Ditto.
18036         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18037         wbnoinvdintrin.h.
18038         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18039
18040 2018-03-05  Richard Biener  <rguenther@suse.de>
18041
18042         PR tree-optimization/84670
18043         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18044         member.
18045         (BB_VISITED_WITH_VISITED_SUCCS): New define.
18046         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18047         (compute_antic_aux): Only assert the number of values in ANTIC_IN
18048         doesn't grow if all successors (recursively) were visited at least
18049         once.
18050
18051 2018-03-05  Richard Biener  <rguenther@suse.de>
18052
18053         PR tree-optimization/84650
18054         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18055         if executed in the loop pipeline.
18056
18057 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
18058
18059         * doc/configfiles.texi (Configuration Files): Move info about
18060         conditionalizing $target-protos.h to...
18061         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
18062         differs from $target-protos.h.
18063
18064 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
18065             Chung-Ju Wu  <jasonwucj@gmail.com>
18066
18067         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18068         * config/nds32/nds32-multiple.md (setmemsi): Define.
18069         * config/nds32/nds32-memory-manipulation.c
18070         (nds32_gen_dup_4_byte_to_word_value): New.
18071         (emit_setmem_word_loop): New.
18072         (emit_setmem_byte_loop): New.
18073         (nds32_expand_setmem_loop): New.
18074         (nds32_expand_setmem_loop_v3m): New.
18075         (nds32_expand_setmem_unroll): New.
18076         (nds32_expand_setmem): New.
18077
18078 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
18079             Chung-Ju Wu  <jasonwucj@gmail.com>
18080
18081         * config/nds32/nds32-memory-manipulation.c
18082         (nds32_emit_load_store): New.
18083         (nds32_emit_post_inc_load_store): New.
18084         (nds32_emit_mem_move): New.
18085         (nds32_emit_mem_move_block): New.
18086         (nds32_expand_movmemsi_loop_unknown_size): New.
18087         (nds32_expand_movmemsi_loop_known_size): New.
18088         (nds32_expand_movmemsi_loop): New.
18089         (nds32_expand_movmemsi_unroll): New.
18090         (nds32_expand_movmemqi): Rename ...
18091         (nds32_expand_movmemsi): ... to this.
18092         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18093         (movmemsi): ... to this.
18094         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18095         (nds32_expand_movmemsi): ... to this.
18096
18097 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
18098             Monk Chiang  <sh.chiang04@gmail.com>
18099             Chung-Ju Wu  <jasonwucj@gmail.com>
18100
18101         * config/nds32/nds32-protos.h
18102         (nds32_expand_load_multiple): New arguments.
18103         (nds32_expand_store_multiple): Ditto.
18104         (nds32_valid_multiple_load_store): Rename ...
18105         (nds32_valid_multiple_load_store_p): ... to this.
18106         * config/nds32/nds32-memory-manipulation.c
18107         (nds32_expand_load_multiple): Refine implementation.
18108         (nds32_expand_store_multiple): Ditto.
18109         * config/nds32/nds32-multiple.md
18110         (load_multiple): Update nds32_expand_load_multiple interface.
18111         (store_multiple): Update nds32_expand_store_multiple interface.
18112         * config/nds32/nds32-predicates.c
18113         (nds32_valid_multiple_load_store): Rename ...
18114         (nds32_valid_multiple_load_store_p): ... to this and refine
18115         implementation.
18116         * config/nds32/predicates.md
18117         (nds32_load_multiple_and_update_address_operation): New predicate.
18118         (nds32_store_multiple_and_update_address_operation): New predicate.
18119
18120 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
18121             Chung-Ju Wu  <jasonwucj@gmail.com>
18122
18123         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18124         (combo): New attribute.
18125         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18126
18127 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
18128
18129         * config/nds32/nds32.opt: Change -mcmodel= default value.
18130
18131 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
18132             Monk Chiang  <sh.chiang04@gmail.com>
18133             Chung-Ju Wu  <jasonwucj@gmail.com>
18134
18135         * config/nds32/constants.md (unspec_element): New enum.
18136         * config/nds32/constraints.md (Umw): New constraint.
18137         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18138         * config/nds32/nds32-intrinsic.md: Likewise.
18139         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18140         (nds32_valid_smw_lwm_base_p): New.
18141         (nds32_output_smw_single_word): New.
18142         (nds32_output_lmw_single_word): New.
18143         (nds32_expand_unaligned_load): New.
18144         (nds32_expand_unaligned_store): New.
18145         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18146         (nds32_output_smw_single_word): Declare.
18147         (nds32_output_lmw_single_word): Declare.
18148         (nds32_expand_unaligned_load): Declare.
18149         (nds32_expand_unaligned_store): Declare.
18150         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18151         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18152         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18153         NDS32_BUILTIN_UASTORE_DW.
18154         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18155         predicate.
18156
18157 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
18158             Kito Cheng  <kito.cheng@gmail.com>
18159             Chung-Ju Wu  <jasonwucj@gmail.com>
18160
18161         * config/nds32/nds32-intrinsic.c
18162         (nds32_expand_builtin_null_ftype_reg): Delete.
18163         (nds32_expand_builtin_reg_ftype_imm): Ditto.
18164         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18165         (nds32_read_argument): New.
18166         (nds32_legitimize_target): Ditto.
18167         (nds32_legitimize_argument): Ditto.
18168         (nds32_check_constant_argument): Ditto.
18169         (nds32_expand_unop_builtin): Ditto.
18170         (nds32_expand_unopimm_builtin): Ditto.
18171         (nds32_expand_binop_builtin): Ditto.
18172         (nds32_builtin_decl_impl): Ditto.
18173         (builtin_description): Ditto.
18174         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18175         (nds32_init_builtins_impl): Ditto.
18176         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18177         (nds32_builtin_decl): New.
18178         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18179         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18180
18181 2018-03-02  Jeff Law  <law@redhat.com>
18182
18183         * reorg.c (stop_search_p): Handle DEBUG_INSN.
18184         (redundant_insn, fill_simple_delay_slots): Likewise.
18185         (fill_slots_from_thread): Likewise.
18186         * resource.c (mark_referenced_resources): Likewise.
18187         (mark_set_resources, find_dead_or_set_registers): Likewise.
18188
18189 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18190
18191         * substring-locations.h (format_warning_va): Formatting fix for
18192         ATTRIBUTE_GCC_DIAG.
18193         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18194         argument.
18195         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18196         * substring-locations.c: Include intl.h.
18197         (format_warning_va): Turned into small wrapper around
18198         format_warning_n_va, renamed to ...
18199         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18200         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18201         use ngettext.
18202         (format_warning_at_substring_n): New function.
18203         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18204         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
18205         format_warning_at_substring with just a shorter name instead of
18206         const function pointer.
18207         (fmtwarn_n): New function.
18208         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18209         appropriate, get rid of all the fmtstr temporaries, move conditionals
18210         with G_() wrapped string literals directly into fmtwarn arguments,
18211         cast dir.len to (int), formatting fixes.
18212
18213 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
18214
18215         * doc/invoke.texi: Remove "Cilk Plus" references.
18216
18217 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18218             Richard Biener  <rguenther@suse.de>
18219
18220         PR ipa/84628
18221         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18222         for error or warning attributes if CALL_FROM_THUNK_P is set.
18223         Formatting fixes.
18224
18225 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18226
18227         PR target/56540
18228         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18229         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18230
18231         PR target/56540
18232         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18233         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18234
18235         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
18236         instead of -1U in last predictors element's probability member.
18237
18238 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
18239
18240         PR ipa/83983
18241         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18242         arguments if they are comparable.
18243
18244 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
18245
18246         PR tree-optimization/84634
18247         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
18248         masks and masked_loop_p with a single loop_masks, making sure it's
18249         null for bb vectorization.
18250
18251 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
18252
18253         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
18254         (vect_analyze_data_ref_access): Use loop->safe_len rather than
18255         loop->force_vectorize to check whether there is no alias.
18256
18257 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
18258
18259         PR target/84614
18260         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
18261         prototypes.
18262         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
18263         comments.
18264         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
18265         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
18266         instead of a loop around prev_real_insn.
18267         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
18268         prev_real_insn.
18269
18270         PR inline-asm/84625
18271         * config/i386/i386.c (ix86_print_operand): Use conditional
18272         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
18273         zero vector.
18274
18275 2018-03-02  Richard Biener  <rguenther@suse.de>
18276
18277         PR tree-optimization/84427
18278         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
18279         (bitmap_set_subtract_values): Rewrite to handle multiple
18280         exprs per value.
18281         (clean): Likewise.
18282         (prune_clobbered_mems): Likewise.
18283         (phi_translate): Take edge instead of pred/phiblock.
18284         (phi_translate_1): Likewise.
18285         (phi_translate_set): Likewise.  Insert all translated
18286         exprs for a value into the set, keeping possibly multiple
18287         expressions per value.
18288         (compute_antic_aux): Adjust for phi_translate changes.
18289         When intersecting union the expressions and prune those
18290         not in the final value set, keeping possibly multiple
18291         expressions per value.  Do not use value-insertion
18292         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
18293         all expressions.  Add verification that the value-sets
18294         only shrink during iteration.
18295         (compute_partial_antic_aux): Adjust for the phi_translate changes.
18296         (do_pre_regular_insertion): Likewise.
18297         (do_pre_partial_partial_insertion): Likewise.
18298
18299 2018-03-02  Richard Biener  <rguenther@suse.de>
18300
18301         PR target/82005
18302         * config/darwin.c (saved_debug_info_level): New static global.
18303         (darwin_asm_lto_start): Disable debug info generation for LTO out.
18304         (darwin_asm_lto_end): Restore debug info generation settings.
18305
18306 2018-03-01  Martin Liska  <mliska@suse.cz>
18307
18308         PR sanitizer/82484
18309         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18310         volatile arguments.
18311
18312 2018-03-01  Richard Biener  <rguenther@suse.de>
18313
18314         PR debug/84645
18315         * dwarf2out.c (gen_variable_die): Properly handle late VLA
18316         type annotation with LTO when debug was disabled at compile-time.
18317
18318 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
18319
18320         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18321         XINT with INTVAL.
18322         (mips_final_postscan_insn): Likewise.
18323
18324 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
18325
18326         PR rtl-optimization/84528
18327         * alias.c (init_alias_target): Add commentary.
18328         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18329         a unique base value if the frame pointer is not eliminated
18330         to the stack pointer.
18331
18332 2018-03-01  Tom de Vries  <tom@codesourcery.com>
18333
18334         PR rtl-optimization/83327
18335         * lra-int.h (hard_regs_spilled_into): Declare.
18336         * lra.c (hard_regs_spilled_into): Define.
18337         (init_reg_info): Init hard_regs_spilled_into.
18338         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18339         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18340         (process_bb_lives): Handle hard_regs_spilled_into.
18341         (lra_create_live_ranges_1): Before doing liveness propagation, clear
18342         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18343
18344 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
18345
18346         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18347         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18348         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18349         * config/rs6000/aix72.h: New file.
18350
18351 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
18352
18353         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18354         instead of warning_at with conditional singular and plural messages
18355         where possible.
18356
18357         PR target/52991
18358         * stor-layout.c (update_alignment_for_field): For
18359         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18360         && !DECL_PACKED (field), do the alignment update, just use
18361         only desired_align instead of MAX (type_align, desired_align)
18362         as the alignment.
18363         (place_field): Don't do known_align < desired_align handling
18364         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18365         is non-NULL, instead do it after rli->prev_field handling and
18366         only if not within a bitfield word.  For DECL_PACKED (field)
18367         use type_align of BITS_PER_UNIT.
18368
18369 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
18370
18371         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18372         superfluous parentheses and trailing spaces.
18373
18374 2018-02-28  Richard Biener  <rguenther@suse.de>
18375
18376         PR tree-optimization/84584
18377         * graphite-scop-detection.c (scop_detection::add_scop): Discard
18378         SCoPs with fake exit edge.
18379
18380 2018-02-28  Martin Liska  <mliska@suse.cz>
18381
18382         PR testsuite/84597
18383         * timevar.c (timer::print): Fix format to properly print 100%
18384         values.
18385
18386 2018-02-28  Richard Biener  <rguenther@suse.de>
18387
18388         PR middle-end/84607
18389         * genmatch.c (capture_info::walk_match): Do not mark
18390         captured expressions without operands as expr_p given
18391         they act more like predicates and should be subject to
18392         "lost tail" side-effect preserving.
18393
18394 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
18395
18396         PR rtl-optimization/81611
18397         * auto-inc-dec.c (attempt_change): Move dead note from
18398         mem_insn if it's the next use of regno
18399         (find_address): Take address use of reg holding
18400         non-incremented value.  Add parm to limit search to the named
18401         reg only.
18402         (merge_in_block): Attempt to use a mem insn that is the next
18403         use of the original regno.
18404
18405 2018-02-27  Martin Sebor  <msebor@redhat.com>
18406
18407         PR c++/83871
18408         * doc/invoke.texi (-Wmissing-attributes): New option.
18409         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18410
18411 2018-02-27  Martin Sebor  <msebor@redhat.com>
18412
18413         PR translation/84207
18414         * diagnostic-core.h (warning_n, error_n, inform_n): Change
18415         n argument to unsigned HOST_WIDE_INT.
18416         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18417         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
18418         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18419         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18420
18421 2018-02-27  Richard Biener  <rguenther@suse.de>
18422
18423         PR tree-optimization/84512
18424         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18425         Do not use the estimate returned from record_stmt_cost for
18426         the scalar iteration cost but sum properly using add_stmt_cost.
18427
18428 2018-02-27  Richard Biener  <rguenther@suse.de>
18429
18430         PR tree-optimization/84466
18431         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18432         Adjust last change to less strictly validate use operands.
18433
18434 2018-02-27  Martin Liska  <mliska@suse.cz>
18435
18436         PR gcov-profile/84548
18437         * gcov.c (process_file): Allow partial overlap and consider it
18438         also as group functions.
18439         (output_lines): Properly calculate range of lines for a group.
18440
18441 2018-02-27  Martin Liska  <mliska@suse.cz>
18442
18443         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18444         'ggc' suffixes.  Change first column width.
18445         (timer::print): Fix formatting of the column.
18446
18447 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
18448
18449         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18450         preserve inline entry blocks for the sake of debug inline
18451         entry point markers alone.
18452         (remove_unused_locals): Suggest in comments a better place to
18453         force the preservation of inline entry blocks that are
18454         otherwise unused, but do not preserve them.
18455
18456 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18457
18458         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18459
18460 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18461
18462         PR target/84039
18463         * config/i386/constraints.md (Bs): Replace
18464         ix86_indirect_branch_register with
18465         TARGET_INDIRECT_BRANCH_REGISTER.
18466         (Bw): Likewise.
18467         * config/i386/i386.md (indirect_jump): Likewise.
18468         (tablejump): Likewise.
18469         (*sibcall_memory): Likewise.
18470         (*sibcall_value_memory): Likewise.
18471         Peepholes of indirect call and jump via memory: Likewise.
18472         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18473         (*sibcall_value_GOT_32): Likewise.
18474         * config/i386/predicates.md (indirect_branch_operand): Likewise.
18475         (GOT_memory_operand): Likewise.
18476         (call_insn_operand): Likewise.
18477         (sibcall_insn_operand): Likewise.
18478         (GOT32_symbol_operand): Likewise.
18479         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18480
18481 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
18482
18483         PR rtl-optimization/83496
18484         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18485         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
18486         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18487         redundant insn, if any.
18488         (relax_delay_slots): Likewise.
18489         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18490
18491 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
18492
18493         PR tree-optimization/83965
18494         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18495         that grouped statements are part of a reduction chain.  Return
18496         true if the statement is not marked as a reduction itself but
18497         is part of a group.
18498         (vect_recog_dot_prod_pattern): Don't check whether the statement
18499         is part of a group here.
18500         (vect_recog_sad_pattern): Likewise.
18501         (vect_recog_widen_sum_pattern): Likewise.
18502
18503 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
18504
18505         PR debug/84545
18506         * final.c (rest_of_clean_state): Also look for calls inside sequences.
18507
18508 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18509
18510         PR target/84530
18511         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18512         the bool argument.
18513         (ix86_output_indirect_function_return): New prototype.
18514         (ix86_split_simple_return_pop_internal): Likewise.
18515         * config/i386/i386.c (indirect_return_via_cx): New.
18516         (indirect_return_via_cx_bnd): Likewise.
18517         (indirect_thunk_name): Handle return va CX_REG.
18518         (output_indirect_thunk_function): Create alias for
18519         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18520         (ix86_output_indirect_jmp): Remove the bool argument.
18521         (ix86_output_indirect_function_return): New function.
18522         (ix86_split_simple_return_pop_internal): Likewise.
18523         * config/i386/i386.md (*indirect_jump): Don't pass false
18524         to ix86_output_indirect_jmp.
18525         (*tablejump_1): Likewise.
18526         (simple_return_pop_internal): Change it to define_insn_and_split.
18527         Call ix86_split_simple_return_pop_internal to split it for
18528         -mfunction-return=.
18529         (simple_return_indirect_internal): Call
18530         ix86_output_indirect_function_return instead of
18531         ix86_output_indirect_jmp.
18532
18533 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
18534
18535         PR bootstrap/84405
18536         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18537         memset and value initialization afterwards.
18538
18539 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
18540
18541         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18542
18543 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18544
18545         PR target/84521
18546         * common/config/aarch64/aarch64-common.c
18547         (aarch_option_optimization_table[]): Switch
18548         off fomit-frame-pointer
18549
18550 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
18551             Chung-Ju Wu  <jasonwucj@gmail.com>
18552
18553         * config/nds32/nds32-multiple.md (load_multiple): Disallow
18554         volatile memory.
18555         (store_multiple): Ditto.
18556
18557 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
18558
18559         * config.gcc: Add --with-cpu support for nds32 target.
18560         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18561         * config/nds32/nds32.opt: Add -mcpu= option.
18562
18563 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
18564
18565         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18566         isel=yes): Warn for these deprecated options.
18567
18568 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
18569
18570         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18571         ISA_2_5_MASKS_EMBEDDED.
18572
18573 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
18574
18575         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18576         p->max as pointers rather than using iterative_hash_expr.
18577
18578 2018-02-23  Carl Love  <cel@us.ibm.com>
18579
18580         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18581         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18582         BU_P8V_OVERLOAD_2.
18583         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18584         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18585         P8V_BUILTIN_VEC_VUNSIGNED2.
18586
18587 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
18588
18589         PR target/81572
18590         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18591         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18592         LRA_UNKNOWN_ALT.
18593         * lra-constraints.c (curr_insn_transform): Set up
18594         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
18595         LRA_UNKNOWN_ALT.
18596         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18597         * lra-eliminations.c (spill_pseudos): Ditto.
18598         (process_insn_for_elimination): Ditto.
18599         * lra-lives.c (reg_early_clobber_p): Use the new macros.
18600         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18601         LRA_NON_CLOBBERED_ALT.
18602
18603 2018-02-22  Martin Sebor  <msebor@redhat.com>
18604
18605         PR tree-optimization/84480
18606         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18607         to maybe_diag_stxncpy_trunc.  Call it.
18608         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18609         from gimple_fold_builtin_strcpy.  Print inlining stack.
18610         (handle_builtin_stxncpy): Print inlining stack.
18611         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18612
18613 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
18614
18615         PR target/84176
18616         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18617         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18618         and -fcheck-pointer-bounds are used together.
18619         (indirect_thunk_prefix): New enum.
18620         (indirect_thunk_need_prefix): New function.
18621         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
18622         "_nt" instead of "_bnd" for NOTRACK prefix.
18623         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18624         (output_indirect_thunk_function): Likewise.
18625         (): Likewise.
18626         (ix86_code_end): Update output_indirect_thunk_function calls.
18627         (ix86_output_indirect_branch_via_reg): Replace
18628         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18629         (ix86_output_indirect_branch_via_push): Likewise.
18630         (ix86_output_function_return): Likewise.
18631         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18632         incompatible with -fcf-protection=branch and
18633         -fcheck-pointer-bounds.
18634
18635 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
18636
18637         PR target/83335
18638         * config/aarch64/aarch64.c (aarch64_print_address_internal):
18639         Change gcc_assert call to output_operand_lossage.
18640
18641 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
18642
18643         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18644
18645 2018-02-22  DJ Delorie  <dj@redhat.com>
18646             Sebastian Perta  <sebastian.perta@renesas.com>
18647             Oleg Endo  <olegendo@gcc.gnu.org>
18648
18649         * config/rx/rx.c (rx_rtx_costs): New function.
18650         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18651
18652 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18653
18654         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18655
18656 2018-02-22  Martin Liska  <mliska@suse.cz>
18657
18658         PR driver/83193
18659         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18660         Add "native" as a possible value.
18661
18662 2018-02-22  Martin Liska  <mliska@suse.cz>
18663
18664         PR driver/83193
18665         * config/i386/i386.c (ix86_option_override_internal):
18666         Add "native" as a possible value for -march and -mtune.
18667
18668 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
18669
18670         PR target/84502
18671         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18672         to all type variants.
18673
18674         PR tree-optimization/84503
18675         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18676         width as info->bitpos + info->bitsize - start.
18677         (merged_store_group::merge_overlapping): Simplify width computation.
18678         (check_no_overlap): New function.
18679         (imm_store_chain_info::try_coalesce_bswap): Compute expected
18680         start + width and last_order of the group, fail if check_no_overlap
18681         fails.
18682         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18683         to group if check_no_overlap fails.
18684
18685 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
18686
18687         * config/rs6000/altivec.md: Delete contraint arguments to
18688         define_expand, define_split, and define_peephole2, and in
18689         define_insn_and_split if always unused.
18690         * config/rs6000/darwin.md: Ditto.
18691         * config/rs6000/dfp.md: Ditto.
18692         * config/rs6000/rs6000.md: Ditto.
18693         * config/rs6000/sync.md: Ditto.
18694         * config/rs6000/vector.md: Ditto.
18695         * config/rs6000/vsx.md: Ditto.
18696
18697 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
18698
18699         * config/rs6000/altivec.md: Write output control strings as braced
18700         blocks instead of double-quoted strings.
18701         * config/rs6000/darwin.md: Ditto.
18702         * config/rs6000/rs6000.md: Ditto.
18703         * config/rs6000/vector.md: Ditto.
18704         * config/rs6000/vsx.md: Ditto.
18705
18706 2018-02-21  Jason Merrill  <jason@redhat.com>
18707
18708         PR c++/84314 - ICE with templates and fastcall attribute.
18709         * attribs.c (build_type_attribute_qual_variant): Remove assert.
18710
18711 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
18712
18713         * ipa-cp.c (determine_versionability): Fix comment typos.
18714
18715 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
18716
18717         PR c/84229
18718         * ipa-cp.c (determine_versionability): Do not version functions caling
18719         va_arg_pack.
18720
18721 2018-02-21  Martin Liska  <mliska@suse.cz>
18722
18723         PR driver/83193
18724         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18725         Add "native" as a possible value.
18726         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
18727         the macro when native cpu detection is available.
18728
18729 2018-02-21  Martin Liska  <mliska@suse.cz>
18730
18731         PR driver/83193
18732         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18733         Add "native" as a possible value.
18734         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18735         when native cpu detection is available.
18736
18737 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
18738             Martin Sebor  <msebor@redhat.com>
18739
18740         PR tree-optimization/84478
18741         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18742         false.
18743         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18744         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18745         support which is conservatively correct, for 2 only stay conservative
18746         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
18747         argument to the 2 argument get_range_strlen, adjust 6 arg
18748         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18749         false.
18750         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18751         (gimple_fold_builtin_strlen): Pass true as last argument to
18752         get_range_strlen.
18753
18754 2018-02-20  Martin Sebor  <msebor@redhat.com>
18755
18756         PR middle-end/84095
18757         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18758         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
18759         (builtin_memref::builtin_memref): Factor out parts into
18760         set_base_and_offset and call it.
18761
18762 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
18763
18764         PR middle-end/84406
18765         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18766         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18767         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
18768         search at the associated MODE_INT.
18769
18770 2018-02-20  Jeff Law  <law@redhat.com>
18771
18772         PR middle-end/82123
18773         PR tree-optimization/81592
18774         PR middle-end/79257
18775         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18776         for range data rather than using global data.
18777         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18778         range data rather than using global data.
18779         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18780         pass it to children as needed.
18781         (struct directive::fmtresult): Similarly.
18782         (struct directive::set_width): Similarly.
18783         (struct directive::set_precision): Similarly.
18784         (format_integer, format_directive, parse_directive): Similarly.
18785         (format_none): Accept unnamed vr_values parameter.
18786         (format_percent, format_floating, format_character): Similarly.
18787         (format_string, format_plain): Similarly.
18788         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18789         the EVRP range analyzer for range data rather than using global data.
18790         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18791         gimple-ssa-evrp-analyze.h
18792         (class sprintf_dom_walker): Add after_dom_children member function.
18793         Add evrp_range_analyzer member.
18794         (sprintf_dom_walker::before_dom_children): Call into the EVRP
18795         range analyzer as needed.
18796         (sprintf_dom_walker::after_dom_children): New member function.
18797         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18798         if not optimizing.
18799         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18800         (evrp_range_analyzer::pop_to_marker): Likewise.
18801
18802 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
18803
18804         PR tree-optimization/84419
18805         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18806         with the required type if its current type is compatible but
18807         different.
18808
18809 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
18810
18811         PR middle-end/82004
18812         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18813         after vectorization.
18814
18815 2018-02-20  Martin Liska  <mliska@suse.cz>
18816
18817         PR driver/83193
18818         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18819         possible values if we don't have a hint.
18820
18821 2018-02-20  Martin Liska  <mliska@suse.cz>
18822
18823         PR c/84310
18824         PR target/79747
18825         * final.c (shorten_branches): Build align_tab array with one
18826         more element.
18827         * opts.c (finish_options): Add alignment option limit check.
18828         (MAX_CODE_ALIGN): Likewise.
18829         (MAX_CODE_ALIGN_VALUE): Likewise.
18830         * doc/invoke.texi: Document maximum allowed option value for
18831         all -falign-* options.
18832
18833 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
18834
18835         PR target/84146
18836         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18837         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18838         * var-tracking.c (emit_note_insn_var_location): Remove all references
18839         to NOTE_INSN_CALL_ARG_LOCATION.
18840         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18841         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18842         Use copy_rtx_if_shared.
18843         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18844         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18845         (dwarf2out_var_location): Remove handling of
18846         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18847         on call_insn.
18848         * final.c (final_scan_insn): Remove all references to
18849         NOTE_INSN_CALL_ARG_LOCATION.
18850         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
18851         before dumping final insns.
18852         * except.c (emit_note_eh_region_end): Remove all references to
18853         NOTE_INSN_CALL_ARG_LOCATION.
18854         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18855         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18856         * config/arc/arc.c (hwloop_optimize): Likewise.
18857         * config/arm/arm.c (create_fix_barrier): Likewise.
18858         * config/s390/s390.c (s390_chunkify_start): Likewise.
18859         * config/sh/sh.c (find_barrier): Likewise.
18860         * config/i386/i386.c (rest_of_insert_endbranch,
18861         ix86_seh_fixup_eh_fallthru): Likewise.
18862         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18863         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18864         * config/frv/frv.c (frv_function_prologue): Likewise.
18865         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
18866         reg note.
18867         (note_outside_basic_block_p): Remove all references to
18868         NOTE_INSN_CALL_ARG_LOCATION.
18869         * gengtype.c (adjust_field_rtx_def): Likewise.
18870         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18871         Likewise.
18872         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18873         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18874
18875         PR c++/84444
18876         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18877         is ADDR_EXPR.
18878
18879         PR tree-optimization/84452
18880         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18881         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18882         is NULL.
18883
18884 2018-02-19  Martin Liska  <mliska@suse.cz>
18885
18886         PR sanitizer/82183
18887         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18888
18889 2018-02-19  Martin Liska  <mliska@suse.cz>
18890             Richard Sandiford  <richard.sandiford@linaro.org>
18891
18892         PR tree-optimization/82491
18893         * gimple-fold.c (get_base_constructor): Make earlier bail out
18894         to prevent ubsan.
18895
18896 2018-02-19  Carl Love  <cel@us.ibm.com>
18897
18898         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18899         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18900         BU_P8V_OVERLOAD_1.
18901         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18902         P8V_BUILTIN_VEC_NEG.
18903
18904 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
18905
18906         * config/rl78/rl78.md (movdf): New define expand.
18907
18908 2018-02-19  Martin Liska  <mliska@suse.cz>
18909
18910         PR other/80589
18911         * doc/invoke.texi: Fix typo.
18912         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18913
18914 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
18915
18916         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18917         handle rs6000_single_float and rs6000_double_float specially for
18918         e500 family CPUs.
18919
18920 2018-02-16  Jeff Law  <law@redhat.com>
18921
18922         * config/rx/rx.c (add_pop_cfi_notes): New function.;
18923         (pop_regs): Use it.
18924
18925 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
18926
18927         PR ipa/84425
18928         * ipa-inline.c (inline_small_functions): Fix a typo.
18929
18930 2018-02-16  Nathan Sidwell  <nathan@acm.org>
18931
18932         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18933
18934 2018-02-16  Carl Love  <cel@us.ibm.com>
18935
18936         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18937         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18938         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18939         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18940         expansion to P8V_BUILTIN_VEC_FLOAT2.
18941
18942 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
18943
18944         PR rtl-optimization/70023
18945         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18946         src_regno into account.
18947
18948 2018-02-16  Carl Love  <cel@us.ibm.com>
18949
18950         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18951         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18952         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18953         * config/rs6000/rs6000.c: Remove case statements for
18954         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18955         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18956         and P9V_BUILTIN_VEC_VINSERT4B.
18957         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18958         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18959         * config/rs6000/vsx.md:
18960         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18961         vec_insert4b.
18962
18963 2018-02-16  Carl Love  <cel@us.ibm.com>
18964
18965         * config/rs6000/altivec.h: Add builtin names vec_extract4b
18966         vec_insert4b.
18967         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18968         definitions.
18969         * config/rs6000/rs6000-c.c: Add the definitions for
18970         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18971         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18972         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18973         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
18974         definition for insert4b and define insn *insert3b_internal.
18975         * doc/extend.texi: Add documentation for vec_extract4b.
18976
18977 2018-02-16  Nathan Sidwell  <nathan@acm.org>
18978
18979         * doc/extend.texi (Backwards Compatibility): Mention friend
18980         injection.  Note for-scope is deprecated.
18981         * doc/invoke.texi (-ffriend-injection): Deprecate.
18982
18983 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
18984
18985         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18986         that moved to I2, also allow destinations that are a paradoxical
18987         subreg (instead of a normal reg).
18988
18989 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
18990
18991         PR target/83831
18992         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18993         to QImode.
18994
18995 2018-02-16  Richard Biener  <rguenther@suse.de>
18996
18997         PR tree-optimization/84037
18998         PR tree-optimization/84016
18999         PR target/82862
19000         * config/i386/i386.c (ix86_builtin_vectorization_cost):
19001         Adjust vec_construct for the fact we need additional higher latency
19002         128bit inserts for AVX256 and AVX512 vector builds.
19003         (ix86_add_stmt_cost): Scale vector construction cost for
19004         elementwise loads.
19005
19006 2018-02-16  Richard Biener  <rguenther@suse.de>
19007
19008         PR tree-optimization/84417
19009         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
19010         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
19011         (non_rewritable_lvalue_p): Likewise, use poly-ints.
19012
19013 2018-02-16  Martin Liska  <mliska@suse.cz>
19014
19015         PR sanitizer/84307
19016         * internal-fn.def (ASAN_CHECK): Set proper flags.
19017         (ASAN_MARK): Likewise.
19018
19019 2018-02-16  Julia Koval  <julia.koval@intel.com>
19020
19021         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
19022         from PTA_CANNONLAKE.
19023
19024 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
19025
19026         PR target/84272
19027         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19028         Use ++iter rather than iter++ for std::list iterators.
19029         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
19030         defer deleting them until all nodes in the forest are processed.  Do
19031         free even leaf nodes.  Change to_process into auto_vec.
19032
19033         PR bootstrap/84405
19034         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19035         * vec.h (vec_default_construct): Use memset instead of placement new
19036         if BROKEN_VALUE_INITIALIZATION is defined.
19037         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19038         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19039         is defined.
19040
19041         PR rtl-optimization/83723
19042         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19043         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
19044         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
19045         recursive calls.
19046         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19047         callers.
19048         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19049
19050 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
19051
19052         PR rtl-optimization/81443
19053         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19054         from inner REGs to paradoxical SUBREGs.
19055
19056 2018-02-16  Richard Biener  <rguenther@suse.de>
19057
19058         PR tree-optimization/84399
19059         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19060         For operands we can analyze at their definition make sure we can
19061         analyze them at each use as well.
19062
19063 2018-02-16  Richard Biener  <rguenther@suse.de>
19064
19065         PR tree-optimization/84190
19066         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19067         volatile accesses if the decl isn't volatile.
19068
19069 2018-02-15  Jason Merrill  <jason@redhat.com>
19070
19071         PR c++/84314 - ICE with templates and fastcall attribute.
19072         * attribs.c (build_type_attribute_qual_variant): Don't clobber
19073         TYPE_CANONICAL on an existing type.
19074
19075 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
19076
19077         PR tree-optimization/84383
19078         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19079         dstoff nor call operand_equal_p if dstbase is NULL.
19080
19081         PR tree-optimization/84334
19082         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19083         also a CONSTANT_CLASS_P, punt.
19084
19085 2018-02-14  Jim Wilson  <jimw@sifive.com>
19086
19087         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19088         first SMALL_OPERAND check.  New local min_second_step.  Move assert
19089         to where locals are set.  Add TARGET_RVC support.
19090         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19091
19092 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
19093
19094         * doc/invoke.texi: Correct -Wformat-overflow code sample.
19095
19096 2018-02-14  Martin Sebor  <msebor@redhat.com>
19097
19098         PR tree-optimization/83698
19099         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19100         arrays constrain the offset range to their bounds.
19101         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19102         (builtin_access::overlap): Avoid setting the size of overlap if it's
19103         already been set.
19104         (maybe_diag_overlap): Also consider arrays when deciding what values
19105         of offsets to include in diagnostics.
19106
19107 2018-02-14  Martin Sebor  <msebor@redhat.com>
19108
19109         PR c/84108
19110         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19111         that correspond to the kind of a declaration.
19112
19113 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
19114
19115         PR target/83984
19116         * config/pa/pa.md: Load address of PIC label using the linkage table
19117         if the label is nonlocal.
19118
19119 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19120
19121         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19122         warning message if user requests -maltivec=be.
19123         * doc/invoke.texi: Document deprecation of -maltivec=be.
19124
19125 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
19126
19127         PR target/84220
19128         * config/rs6000/rs6000-c.c: Update definitions for
19129         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19130         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19131
19132 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
19133
19134         PR target/84239
19135         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19136         add _get_ssp intrinsics. Remove argument from
19137         __builtin_ia32_rdssp[d|q].
19138         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19139         * config/i386/i386-builtin.def: Remove argument from
19140         __builtin_ia32_rdssp[d|q].
19141         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19142         ix86_expand_special_args_builtin for _rdssp[d|q].
19143         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19144         Clear register before usage.
19145         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19146         Add documentation for new _get_ssp and _inc_ssp intrinsics.
19147
19148 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
19149
19150         PR tree-optimization/84357
19151         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19152         operand 1 of an ARRAY_REF too.
19153
19154 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
19155
19156         PR target/83831
19157         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19158         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19159         declarations.
19160         (set_of_reg): New struct.
19161         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19162         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19163         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19164         functions.
19165         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19166         Split into bitclr, bitset, bitinvert patterns if appropriate.
19167         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19168         use rx_fuse_in_memory_bitop.
19169         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19170         to named insn, correct maximum insn length.
19171
19172 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
19173
19174         PR target/79242
19175         * machmode.def: Define a complex mode for PARTIAL_INT.
19176         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19177         MODE_PARTIAL_INT.
19178         * doc/rtl.texi: Document CSPImode.
19179         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19180         handling.
19181         (msp430_hard_regno_nregs_with_padding): Likewise.
19182
19183 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
19184
19185         PR target/84279
19186         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19187
19188 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
19189
19190         PR rtl-optimization/84169
19191         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
19192         we generated a parallel as new i3 and we split that to new i2 and i3
19193         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
19194         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
19195         those to i2, not i1.  Partially rewrite this scan code.
19196
19197 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
19198
19199         PR c/82210
19200         * stor-layout.c (place_field): For variable length fields, adjust
19201         offset_align afterwards not just based on the field's alignment,
19202         but also on the size.
19203
19204         PR middle-end/84309
19205         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19206         of exps and logs in the use_exp2 case.
19207
19208 2018-02-13  Jeff Law  <law@redhat.com>
19209
19210         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19211         entry for "vector".
19212
19213         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19214         ARGS as unused.
19215
19216 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
19217
19218         PR debug/84342
19219         PR debug/84319
19220         * common.opt (gas-loc-support, gas-locview-support): New.
19221         (ginline-points, ginternal-reset-location-views): New.
19222         * doc/invoke.texi: Document them.  Use @itemx where intended.
19223         (gvariable-location-views): Adjust.
19224         * target.def (reset_location_view): New.
19225         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19226         (TARGET_RESET_LOCATION_VIEW): New.
19227         * doc/tm.texi: Rebuilt.
19228         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19229         (dwarf2out_default_as_locview_support): New.
19230         (output_asm_line_debug_info): Use option variables.
19231         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19232         (output_loc_list): Likewise.
19233         (add_high_low_attributes): Check option variables.
19234         Don't output entry view attribute in strict mode.
19235         (gen_inlined_subroutine_die): Check option variables.
19236         (dwarf2out_inline_entry): Likewise.
19237         (init_sections_and_labels): Likewise.
19238         (dwarf2out_early_finish): Likewise.
19239         (maybe_reset_location_view): New, from...
19240         (dwarf2out_var_location): ... here.  Call it.
19241         * debug.h (dwarf2out_default_as_loc_support): Declare.
19242         (dwarf2out_default_as_locview_support): Declare.
19243         * hooks.c (hook_int_rtx_insn_0): New.
19244         * hooks.h (hook_int_rtx_insn_0): Declare.
19245         * toplev.c (process_options): Take -gas-loc-support and
19246         -gas-locview-support from dwarf2out.  Enable
19247         -gvariable-location-views by default only with locview
19248         assembler support.  Enable -ginternal-reset-location-views by
19249         default only if the target defines the corresponding hook.
19250         Enable -ginline-points by default if location views are
19251         enabled; force it disabled if statement frontiers are
19252         disabled.
19253         * tree-inline.c (expand_call_inline): Check option variables.
19254         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
19255
19256 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
19257
19258         PR tree-optimization/84321
19259         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
19260         handling.  Also check whether the anti-range contains any values
19261         that satisfy the mask; switch to a VR_RANGE if not.
19262
19263 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
19264
19265         PR sanitizer/84340
19266         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
19267
19268 2018-02-13  Martin Jambor  <mjambor@suse.cz>
19269
19270         PR c++/83990
19271         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
19272         of call statements, also set location of a load to a temporary.
19273
19274 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
19275
19276         * config/rl78/rl78.c (add_vector_labels): New function.
19277         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
19278         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
19279         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
19280         which checks that no arguments are passed.
19281         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
19282         * doc/extend.texi: Documentation for the new attribute.
19283
19284 2018-02-13  Andreas Schwab  <schwab@suse.de>
19285
19286         * config/riscv/linux.h (CPP_SPEC): Define.
19287
19288 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
19289
19290         PR target/84335
19291         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
19292         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
19293         OPTION_MASK_ISA_AES as first argument to def_builtin_const
19294         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
19295         instead of OPTION_MASK_ISA_PCLMUL as first argument to
19296         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
19297         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
19298         temporarily for AES and PCLMUL builtins.
19299
19300         PR tree-optimization/84339
19301         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
19302         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19303         Formatting fixes.
19304
19305         PR middle-end/84309
19306         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19307         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19308         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19309         inline function.
19310         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19311         inline function.
19312         * omp-simd-clone.h: New file.
19313         * omp-simd-clone.c: Include omp-simd-clone.h.
19314         (expand_simd_clones): No longer static.
19315         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19316         cgraph.h and omp-simd-clone.h.
19317         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19318         (vect_recog_widen_shift_pattern): Formatting fix.
19319         (vect_pattern_recog_1): Don't check optab for calls.
19320
19321         PR target/84336
19322         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19323         operands[2] into a REG before using gen_lowpart on it.
19324
19325 2018-02-12  Jeff Law  <law@redhat.com>
19326
19327         PR target/83760
19328         * config/sh/sh.c (find_barrier): Consider a sibling call
19329         a barrier as well.
19330
19331         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19332         successfully back substituting a reg.
19333
19334 2018-02-12  Richard Biener  <rguenther@suse.de>
19335
19336         PR tree-optimization/84037
19337         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19338         parameter, move visited init to caller.
19339         (vect_slp_analyze_operations): Separate cost from validity
19340         check, initialize visited once for all instances.
19341         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19342         for all instances.
19343         * tree-vect-stmts.c (vect_model_simple_cost): Make early
19344         out an assert.
19345         (vect_model_promotion_demotion_cost): Likewise.
19346         (vectorizable_bswap): Guard cost modeling with !slp_node
19347         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19348         SLP stmts.
19349         (vectorizable_call): Likewise.
19350         (vectorizable_conversion): Likewise.
19351         (vectorizable_assignment): Likewise.
19352         (vectorizable_shift): Likewise.
19353         (vectorizable_operation): Likewise.
19354         (vectorizable_store): Likewise.
19355         (vectorizable_load): Likewise.
19356         (vectorizable_condition): Likewise.
19357         (vectorizable_comparison): Likewise.
19358
19359 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
19360
19361         PR sanitizer/84307
19362         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19363         (ASAN_MARK): Fix fnspec to account for return value, change pointer
19364         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19365
19366 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
19367
19368         PR middle-end/83665
19369         * params.def (inline-min-speedup): Increase from 8 to 15.
19370         (max-inline-insns-auto): Decrease from 40 to 30.
19371         * ipa-split.c (consider_split): Add some buffer for function to
19372         be considered inlining candidate.
19373         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19374         default values.
19375
19376 2018-02-12  Richard Biener  <rguenther@suse.de>
19377
19378         PR tree-optimization/84037
19379         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19380         matched stmts if we cannot swap the non-matched ones.
19381
19382 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
19383
19384         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19385         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19386         _mm_maskz_scalef_round_ss): New intrinsics.
19387         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19388         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19389         __builtin_ia32_scalefss_round): Remove.
19390         (__builtin_ia32_scalefsd_mask_round,
19391         __builtin_ia32_scalefss_mask_round): New intrinsics.
19392         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19393         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19394         ((match_operand:VF_128 2 "<round_nimm_predicate>"
19395         "<round_constraint>")): Changed to ...
19396         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19397         "<round_scalar_constraint>")): ... this.
19398         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19399         %0, %1, %2<round_op3>}"): Changed to ...
19400         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19401         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19402         %2<round_scalar_mask_op3>}"): ... this.
19403         * config/i386/subst.md (round_scalar_nimm_predicate): New.
19404
19405 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
19406
19407         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19408         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19409         (_mm_maskz_sqrt_round_ss): New intrinsics.
19410         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19411         (__builtin_ia32_sqrtsd_mask_round)
19412         (__builtin_ia32_sqrtss_mask_round): New builtins.
19413         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19414         (__builtin_ia32_sqrtss_round): Remove.
19415         (__builtin_ia32_sqrtsd_mask_round)
19416         (__builtin_ia32_sqrtss_mask_round): New builtins.
19417         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19418         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19419         ((match_operand:VF_128 1 "vector_operand"
19420         "xBm,<round_constraint>")): Changed to ...
19421         ((match_operand:VF_128 1 "vector_operand"
19422         "xBm,<round_scalar_constraint>")): ... this.
19423         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19424         %0, %2, %<iptr>1<round_op3>}): Changed to ...
19425         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19426         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19427         %<iptr>1<round_scalar_mask_op3>}): ... this.
19428         ((set_attr "prefix" "<round_prefix>")): Changed to ...
19429         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19430
19431 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
19432
19433         PR target/84266
19434         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19435         Cast vec_cmpeq result to correct type.
19436         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19437         Cast vec_cmpgt result to correct type.
19438
19439 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
19440
19441         * final.c (final_scan_insn_1): Renamed from...
19442         (final_scan_insn): ... this.  New wrapper, to recover
19443         seen from the outermost call in recursive ones.
19444         * config/sparc/sparc.c (output_return): Drop seen from call.
19445         (output_sibcall): Likewise.
19446         * config/visium/visium.c (output_branch): Likewise.
19447
19448 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
19449
19450         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19451         function label.
19452
19453 2018-02-10  Alan Modra  <amodra@gmail.com>
19454
19455         PR target/84300
19456         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19457         Specify LR as an input.
19458
19459 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
19460
19461         PR sanitizer/83987
19462         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19463         remove_member_access_dummy_vars): New functions.
19464         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19465         lower_omp_1, execute_lower_omp): Use them.
19466
19467         PR rtl-optimization/84308
19468         * shrink-wrap.c (spread_components): Release todo vector.
19469
19470 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
19471
19472         PR rtl-optimization/57193
19473         * ira-color.c (struct allocno_color_data): Add member
19474         conflict_allocno_hard_prefs.
19475         (update_conflict_allocno_hard_prefs): New.
19476         (bucket_allocno_compare_func): Add a preference based on
19477         conflict_allocno_hard_prefs.
19478         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19479         (color_allocnos): Remove a dead code.  Initiate
19480         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
19481
19482 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
19483
19484         PR target/84226
19485         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19486         constraint from =wa to wa.  Avoid a subreg on the output operand,
19487         instead use a pseudo and subreg it in a move.
19488         (p9_xxbrd_<mode>): Changed to ...
19489         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19490         (p9_xxbrd_v2df): New expander.
19491         (p9_xxbrw_<mode>): Changed to ...
19492         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19493         (p9_xxbrw_v4sf): New expander.
19494
19495 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
19496
19497         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19498
19499 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
19500
19501         PR target/83926
19502         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19503         multiply in 32-bit mode.
19504         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19505         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19506         mode.
19507
19508 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
19509
19510         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19511         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19512         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19513         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19514
19515 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
19516
19517         PR lto/84213
19518         * dwarf2out.c (is_trivial_indirect_ref): New function.
19519         (dwarf2out_late_global_decl): Do not generate a location
19520         attribute for variables that have a non-trivial DECL_VALUE_EXPR
19521         and that are not defined in the current unit.
19522
19523 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
19524
19525         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19526         instead of a libcall for UNORDERED.
19527
19528 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
19529
19530         PR target/82641
19531         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19532         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19533
19534 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19535
19536         PR target/PR84295
19537         * config/s390/s390.c (s390_set_current_function): Invoke
19538         s390_indirect_branch_settings also if fndecl didn't change.
19539
19540 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
19541
19542         * config/rs6000/rs6000.md (blockage): Set length to zero.
19543
19544 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
19545
19546         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19547
19548 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
19549
19550         PR sanitizer/84285
19551         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19552         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19553         -static-lib*san.
19554
19555         PR debug/84252
19556         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19557         PARALLEL incoming that failed vt_get_decl_and_offset check.
19558
19559         PR middle-end/84237
19560         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19561         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19562         TREE_READONLY bit.
19563         (get_variable_section): For decls in named .bss* sections pass true as
19564         second argument to bss_initializer_p.
19565
19566 2018-02-09  Marek Polacek  <polacek@redhat.com>
19567             Jakub Jelinek  <jakub@redhat.com>
19568
19569         PR c++/83659
19570         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19571         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
19572         Sync some changes from cxx_fold_indirect_ref.
19573
19574 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
19575
19576         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19577         markers.
19578         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19579         (BLOCK_INLINE_ENTRY_LABEL): New.
19580         (dwarf2out_var_location): Disregard inline entry markers.
19581         (inline_entry_data): New struct.
19582         (inline_entry_data_hasher): New hashtable type.
19583         (inline_entry_data_hasher::hash): New.
19584         (inline_entry_data_hasher::equal): New.
19585         (inline_entry_data_table): New variable.
19586         (add_high_low_attributes): Add DW_AT_entry_pc and
19587         DW_AT_GNU_entry_view attributes if a pending entry is found
19588         in inline_entry_data_table.  Add old entry_pc attribute only
19589         if debug nonbinding markers are disabled.
19590         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19591         markers are enabled.
19592         (block_within_block_p, dwarf2out_inline_entry): New.
19593         (dwarf2out_finish): Check that no entries remained in
19594         inline_entry_data_table.
19595         * final.c (reemit_insn_block_notes): Handle inline entry notes.
19596         (final_scan_insn, notice_source_line): Likewise.
19597         (rest_of_clean_state): Skip inline entry markers.
19598         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19599         markers.
19600         * gimple.c (gimple_build_debug_inline_entry): New.
19601         * gimple.h (enum gimple_debug_subcode): Add
19602         GIMPLE_DEBUG_INLINE_ENTRY.
19603         (gimple_build_debug_inline_entry): Declare.
19604         (gimple_debug_inline_entry_p): New.
19605         (gimple_debug_nonbind_marker_p): Adjust.
19606         * insn-notes.def (INLINE_ENTRY): New.
19607         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19608         inline entry marker notes.
19609         (print_insn): Likewise.
19610         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19611         (INSN_DEBUG_MARKER_KIND): Likewise.
19612         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19613         * tree-inline.c (expand_call_inline): Build and insert
19614         debug_inline_entry stmt.
19615         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19616         inline entry blocks early, if nonbind markers are enabled.
19617         (dump_scope_block): Dump fragment info.
19618         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19619         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19620         (gimple_build_debug_inline_entry): New.
19621         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19622         Enable/disable inline entry points too.
19623         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19624         (DEBUG_INSN): Describe inline entry markers.
19625
19626         * common.opt (gvariable-location-views): New.
19627         (gvariable-location-views=incompat5): New.
19628         * config.in: Rebuilt.
19629         * configure: Rebuilt.
19630         * configure.ac: Test assembler for view support.
19631         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19632         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19633         * dwarf2out.c (var_loc_view): New typedef.
19634         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19635         (dwarf2out_locviews_in_attribute): New.
19636         (dwarf2out_locviews_in_loclist): New.
19637         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19638         (enum dw_line_info_opcode): Add LI_adv_address.
19639         (struct dw_line_info_table): Add view.
19640         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19641         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19642         (zero_view_p): New variable.
19643         (ZERO_VIEW_P): New macro.
19644         (output_asm_line_debug_info): New.
19645         (struct var_loc_node): Add view.
19646         (add_AT_view_list, AT_loc_list): New.
19647         (add_var_loc_to_decl): Add view param.  Test it against last.
19648         (new_loc_list): Add view params.  Record them.
19649         (AT_loc_list_ptr): Handle loc and view lists.
19650         (view_list_to_loc_list_val_node): New.
19651         (print_dw_val): Handle dw_val_class_view_list.
19652         (size_of_die): Likewise.
19653         (value_format): Likewise.
19654         (loc_list_has_views): New.
19655         (gen_llsym): Set vl_symbol too.
19656         (maybe_gen_llsym, skip_loc_list_entry): New.
19657         (dwarf2out_maybe_output_loclist_view_pair): New.
19658         (output_loc_list): Output view list or entries too.
19659         (output_view_list_offset): New.
19660         (output_die): Handle dw_val_class_view_list.
19661         (output_dwarf_version): New.
19662         (output_compilation_unit_header): Use it.
19663         (output_skeleton_debug_sections): Likewise.
19664         (output_rnglists, output_line_info): Likewise.
19665         (output_pubnames, output_aranges): Update version comments.
19666         (output_one_line_info_table): Output view numbers in asm comments.
19667         (dw_loc_list): Determine current endview, pass it to new_loc_list.
19668         Call maybe_gen_llsym.
19669         (loc_list_from_tree_1): Adjust.
19670         (add_AT_location_description): Create view list attribute if
19671         needed, check it's absent otherwise.
19672         (convert_cfa_to_fb_loc_list): Adjust.
19673         (maybe_emit_file): Call output_asm_line_debug_info for test.
19674         (dwarf2out_var_location): Reset views as needed.  Precompute
19675         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
19676         attribute.  Set view.
19677         (new_line_info_table): Reset next view.
19678         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19679         (dwarf2out_source_line): Likewise.  Output view resets and labels to
19680         the assembler, or select appropriate line info opcodes.
19681         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19682         (optimize_string_length): Catch it.  Adjust.
19683         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
19684         dw_val_class_view_list, and remove it if no longer needed.
19685         (hash_loc_list): Hash view numbers.
19686         (loc_list_hasher::equal): Compare them.
19687         (optimize_location_lists): Check whether a view list symbol is
19688         needed, and whether the locview attribute is present, and
19689         whether they match.  Remove the locview attribute if no longer
19690         needed.
19691         (index_location_lists): Call skip_loc_list_entry for test.
19692         (dwarf2out_finish): Call output_asm_line_debug_info for test.
19693         Use output_dwarf_version.
19694         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19695         (struct dw_val_node): Add val_view_list.
19696         * final.c (SEEN_NEXT_VIEW): New.
19697         (set_next_view_needed): New.
19698         (clear_next_view_needed): New.
19699         (maybe_output_next_view): New.
19700         (final_start_function): Rename to...
19701         (final_start_function_1): ... this.  Take pointer to FIRST,
19702         add SEEN parameter.  Emit param bindings in the initial view.
19703         (final_start_function): Reintroduce SEEN-less interface.
19704         (final): Rename to...
19705         (final_1): ... this.  Take SEEN parameter.  Output final pending
19706         next view at the end.
19707         (final): Reintroduce seen-less interface.
19708         (final_scan_insn): Output pending next view before switching
19709         sections or ending a block.  Mark the next view as needed when
19710         outputting variable locations.  Notify debug backend of section
19711         changes, and of location view changes.
19712         (rest_of_handle_final): Adjust.
19713         * toplev.c (process_options): Autodetect value for debug variable
19714         location views option.  Warn on incompat5 without -gdwarf-5.
19715         * doc/invoke.texi (gvariable-location-views): New.
19716         (gvariable-location-views=incompat5): New.
19717         (gno-variable-location-views): New.
19718
19719 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
19720
19721         PR tree-optimization/84136
19722         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19723         that the result of find_edge is non-NULL.
19724
19725 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
19726
19727         PR target/83008
19728         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19729         storing integer register in SImode.  Fix cost of 256 and 512
19730         byte aligned SSE register store.
19731
19732 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
19733
19734         * config/i386/i386.c (ix86_multiplication_cost): Fix
19735         multiplication cost for TARGET_AVX512DQ.
19736
19737 2018-02-08  Marek Polacek  <polacek@redhat.com>
19738
19739         PR tree-optimization/84238
19740         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19741         get_range_strlen.
19742
19743 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19744
19745         PR tree-optimization/84265
19746         * tree-vect-stmts.c (vectorizable_store): Don't treat
19747         VMAT_CONTIGUOUS accesses as grouped.
19748         (vectorizable_load): Likewise.
19749
19750 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19751
19752         PR tree-optimization/81635
19753         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19754         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19755         (test_round_for_mask): New functions.
19756         (wide_int_cc_tests): Call test_round_for_mask.
19757         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19758         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19759         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19760         range returned by get_range_info.
19761
19762 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
19763
19764         PR ipa/81360
19765         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19766         * symtab.c: Include builtins.h
19767         (symtab_node::output_to_lto_symbol_table_p): Move here
19768         from lto-streamer-out.c:output_symbol_p.
19769         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19770         (output_symbol_p): Move all logic to symtab.c
19771         (produce_symtab): Update.
19772
19773 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19774
19775         * config/s390/s390-opts.h (enum indirect_branch): Define.
19776         * config/s390/s390-protos.h (s390_return_addr_from_memory)
19777         (s390_indirect_branch_via_thunk)
19778         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19779         (enum s390_indirect_branch_type): Define.
19780         * config/s390/s390.c (struct s390_frame_layout, struct
19781         machine_function): Remove.
19782         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19783         (indirect_branch_table_label_no, indirect_branch_table_name):
19784         Define variables.
19785         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19786         (enum s390_indirect_branch_option): Define.
19787         (s390_return_addr_from_memory): New function.
19788         (s390_handle_string_attribute): New function.
19789         (s390_attribute_table): Add new attribute handler.
19790         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19791         (s390_indirect_branch_via_thunk): New function.
19792         (s390_indirect_branch_via_inline_thunk): New function.
19793         (s390_function_ok_for_sibcall): When jumping via thunk disallow
19794         sibling call optimization for non z10 compiles.
19795         (s390_emit_call): Force indirect branch target to be a single
19796         register.  Add r1 clobber for non-z10 compiles.
19797         (s390_emit_epilogue): Emit return jump via return_use expander.
19798         (s390_reorg): Handle JUMP_INSNs as execute targets.
19799         (s390_option_override_internal): Perform validity checks for the
19800         new command line options.
19801         (s390_indirect_branch_attrvalue): New function.
19802         (s390_indirect_branch_settings): New function.
19803         (s390_set_current_function): Invoke s390_indirect_branch_settings.
19804         (s390_output_indirect_thunk_function):  New function.
19805         (s390_code_end): Implement target hook.
19806         (s390_case_values_threshold): Implement target hook.
19807         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19808         macros.
19809         * config/s390/s390.h (struct s390_frame_layout)
19810         (struct machine_function): Move here from s390.c.
19811         (TARGET_INDIRECT_BRANCH_NOBP_RET)
19812         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19813         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19814         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19815         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19816         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19817         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19818         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19819         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19820         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19821         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19822         (mnemonic attribute): Add values which aren't recognized
19823         automatically.
19824         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19825         pattern for branch conversion.  Fix mnemonic attribute.
19826         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19827         indirect branch via thunk if requested.
19828         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19829         ("*indirect_jump"): Disable for branch conversion using out of
19830         line thunks.
19831         ("indirect_jump_via_thunk<mode>_z10")
19832         ("indirect_jump_via_thunk<mode>")
19833         ("indirect_jump_via_inlinethunk<mode>_z10")
19834         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19835         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19836         ("casesi_jump_via_inlinethunk<mode>_z10")
19837         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19838         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19839         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19840         ("*indirect2_jump"): Disable for branch conversion.
19841         ("casesi_jump"): Turn into expander and expand patterns for branch
19842         conversion.
19843         ("return_use"): New expander.
19844         ("*return"): Emit return via thunk and rename it to ...
19845         ("*return<mode>"): ... this one.
19846         * config/s390/s390.opt: Add new options and and enum for the
19847         option values.
19848
19849 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19850
19851         * lra-constraints.c (match_reload): Unconditionally use
19852         gen_lowpart_SUBREG, rather than selecting between that
19853         and equivalent gen_rtx_SUBREG code.
19854
19855 2018-02-08  Richard Biener  <rguenther@suse.de>
19856
19857         PR tree-optimization/84233
19858         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19859         changed flag instead of boguously re-using phi_inserted.
19860
19861 2018-02-08  Martin Jambor  <mjambor@suse.cz>
19862
19863         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19864         static local variables.
19865
19866 2018-02-08  Richard Biener  <rguenther@suse.de>
19867
19868         PR tree-optimization/84278
19869         * tree-vect-stmts.c (vectorizable_store): When looking for
19870         smaller vector types to perform grouped strided loads/stores
19871         make sure the mode is supported by the target.
19872         (vectorizable_load): Likewise.
19873
19874 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
19875
19876         * config/aarch64/aarch64.c (aarch64_components_for_bb):
19877         Increase LDP/STP opportunities by adding adjacent callee-saves.
19878
19879 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
19880
19881         PR rtl-optimization/84068
19882         PR rtl-optimization/83459
19883         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19884
19885 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
19886
19887         PR tree-optimization/84224
19888         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19889         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19890         non-zero arguments.
19891
19892 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
19893
19894         PR target/84113
19895         * config/rs6000/altivec.md (*restore_world): Remove LR use.
19896         * config/rs6000/predicates.md (restore_world_operation): Adjust op
19897         count, remove one USE.
19898
19899 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
19900
19901         * doc/install.texi (Configuration): Document the
19902         --with-long-double-format={ibm,ieee} PowerPC configuration
19903         options.
19904
19905         PR target/84154
19906         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19907         Convert from define_expand to be define_insn_and_split.  Rework
19908         float/double/_Float128 conversions to QI/HI/SImode to work with
19909         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
19910         conversions to QI/HImode types did a store and then a load to
19911         truncate the value.  For conversions to VSX registers, don't split
19912         the insn, instead emit the code directly.  Use the code iterator
19913         any_fix to combine signed and unsigned conversions.
19914         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19915         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19916         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19917         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19918         (fix_<mode>di2_hw): Likewise.
19919         (fixuns_<mode>di2_hw): Likewise.
19920         (fix_<mode>si2_hw): Likewise.
19921         (fixuns_<mode>si2_hw): Likewise.
19922         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19923         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19924         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19925         fix<uns>_trunc<SFDF:mode>si2_p8.
19926         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19927         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19928         (fix<uns>_<mode>_mem): Likewise.
19929         (fctiw<u>z_<mode>_mem): Likewise.
19930         (fix<uns>_<mode>_mem): Likewise.
19931         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19932         the register allocator from doing a direct move to the GPRs to do
19933         a store, and instead use the ISA 3.0 store byte/half-word from
19934         vector register instruction.  For IEEE 128-bit floating point,
19935         also optimize stores of 32-bit ints.
19936         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19937
19938 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
19939
19940         * genextract.c (push_pathstr_operand): New function to support
19941         [a-zA-Z].
19942         (walk_rtx): Call push_pathstr_operand.
19943         (print_path): Support [a-zA-Z].
19944
19945 2018-02-07  Richard Biener  <rguenther@suse.de>
19946
19947         PR tree-optimization/84037
19948         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19949         (cse_and_gimplify_to_preheader): Declare.
19950         (vect_get_place_in_interleaving_chain): Likewise.
19951         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19952         ivexpr_map.
19953         (_loop_vec_info::~_loop_vec_info): Delete it.
19954         (cse_and_gimplify_to_preheader): New function.
19955         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19956         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19957         (vectorizable_load): Likewise.  For grouped stores always base
19958         the IV on the first element.
19959         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19960         condition before gimplifying.
19961
19962 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
19963
19964         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19965         *DIV_EXPR and *MOD_EXPR.
19966
19967 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
19968
19969         PR target/84248
19970         * config/i386/i386.c (ix86_option_override_internal): Mask out
19971         the CF_SET bit when checking -fcf-protection.
19972
19973 2018-02-07  Tom de Vries  <tom@codesourcery.com>
19974
19975         PR libgomp/84217
19976         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19977         enough.
19978
19979 2018-02-07  Richard Biener  <rguenther@suse.de>
19980
19981         PR tree-optimization/84204
19982         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19983         this place.
19984
19985         PR tree-optimization/84205
19986         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19987         special-case isl_ast_op_zdiv_r.
19988
19989         PR tree-optimization/84223
19990         * graphite-scop-detection.c (gather_bbs::before_dom_children):
19991         Only add conditions from within the region.
19992         (gather_bbs::after_dom_children): Adjust.
19993
19994 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
19995
19996         PR target/84209
19997         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19998         * config/avr/avr.md: Only post-reload split REG-REG moves if
19999         either register is GENERAL_REG_P.
20000
20001 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
20002
20003         PR tree-optimization/84235
20004         * tree-ssa-scopedtables.c
20005         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
20006         if the subtraction is performed in floating point type where NaNs are
20007         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
20008         build 1.  Formatting fix.
20009
20010 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
20011
20012         PR target/84146
20013         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
20014         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
20015         and skip it regardless of bb boundaries.  Use CALL_P macro,
20016         don't test INSN_P (insn) together with CALL_P or JUMP_P check
20017         unnecessarily, formatting fix.
20018
20019 2018-02-06  Michael Collison  <michael.collison@arm.com>
20020
20021         * config/arm/thumb2.md:
20022         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
20023         (*thumb_mov_notscc): Ditto.
20024
20025 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
20026
20027         PR target/84154
20028         * config/rs6000/rs6000.md (su code attribute): Use "u" for
20029         unsigned_fix, not "s".
20030
20031 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20032
20033         * configure.ac (gcc_fn_eh_frame_ro): New function.
20034         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20035         correct .eh_frame permissions.
20036         * configure: Regenerate.
20037
20038 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
20039
20040         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20041         irrelevant options.
20042
20043 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20044
20045         * config/rs6000/rs6000.c (rs6000_option_override_internal):
20046         Display warning message for -mno-speculate-indirect-jumps.
20047
20048 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
20049
20050         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20051         Undocumented.
20052         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20053
20054 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
20055
20056         PR tree-optimization/84225
20057         * tree-eh.c (find_trapping_overflow): Only call
20058         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20059
20060 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
20061
20062         PR target/84145
20063         * config/i386/i386.c: Reimplement the check of possible options
20064         -mibt/-mshstk conbination. Change error messages.
20065         * doc/invoke.texi: Fix a typo: remove extra '='.
20066
20067 2018-02-06  Marek Polacek  <polacek@redhat.com>
20068
20069         PR tree-optimization/84228
20070         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20071
20072 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
20073
20074         PR target/82641
20075         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20076         emitted arch directives.
20077         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20078         __ARM_FEATURE_COPROC before changing architectures.
20079
20080 2018-02-06  Richard Biener  <rguenther@suse.de>
20081
20082         * config/i386/i386.c (print_reg): Fix typo.
20083         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20084
20085 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
20086
20087         * configure: Regenerate.
20088
20089 2018-02-05  Martin Sebor  <msebor@redhat.com>
20090
20091         PR tree-optimization/83369
20092         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20093         inlining context.
20094
20095 2018-02-05  Martin Liska  <mliska@suse.cz>
20096
20097         * doc/invoke.texi: Cherry-pick upstream r323995.
20098
20099 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
20100
20101         * ira.c (ira_init_register_move_cost): Adjust comment.
20102
20103 2018-02-05  Martin Liska  <mliska@suse.cz>
20104
20105         PR gcov-profile/84137
20106         * doc/gcov.texi: Fix typo in documentation.
20107
20108 2018-02-05  Martin Liska  <mliska@suse.cz>
20109
20110         PR gcov-profile/83879
20111         * doc/gcov.texi: Document necessity of --dynamic-list-data when
20112         using dlopen functionality.
20113
20114 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
20115
20116         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20117         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20118         _mm_maskz_range_ss, _mm_mask_range_round_ss,
20119         _mm_maskz_range_round_ss): New intrinsics.
20120         (__builtin_ia32_rangesd128_round)
20121         (__builtin_ia32_rangess128_round): Remove.
20122         (__builtin_ia32_rangesd128_mask_round,
20123         __builtin_ia32_rangess128_mask_round): New builtins.
20124         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20125         __builtin_ia32_rangess128_round): Remove.
20126         (__builtin_ia32_rangesd128_mask_round,
20127         __builtin_ia32_rangess128_mask_round): New builtins.
20128         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20129         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20130         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20131         "<round_saeonly_constraint>")): Changed to ...
20132         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20133         "<round_saeonly_scalar_constraint>")): ... this.
20134         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20135         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20136         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20137         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20138         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20139
20140 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
20141
20142         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20143         options.
20144         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20145         Remove all values except native, 8540 and 8548.
20146
20147 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
20148
20149         * config/i386/i386.c (ix86_output_function_return): Pass
20150         INVALID_REGNUM, instead of -1, as invalid register number to
20151         indirect_thunk_name and output_indirect_thunk.
20152
20153 2018-02-02  Julia Koval  <julia.koval@intel.com>
20154
20155         * config.gcc: Add -march=icelake.
20156         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20157         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20158         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20159         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20160         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20161         (processor_target_table): Add icelake.
20162         (ix86_option_override_internal): Handle new PTAs.
20163         (get_builtin_code_for_version): Handle icelake.
20164         (M_INTEL_COREI7_ICELAKE): New.
20165         (fold_builtin_cpu): Handle icelake.
20166         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20167         * doc/invoke.texi: Add -march=icelake.
20168
20169 2018-02-02  Julia Koval  <julia.koval@intel.com>
20170
20171         * config/i386/i386.c (ix86_option_override_internal): Change flags type
20172         to wide_int_bitmask.
20173         * wide-int-bitmask.h: New.
20174
20175 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
20176
20177         PR target/84066
20178         * config/i386/i386.md: Replace Pmode with word_mode in
20179         builtin_setjmp_setup and builtin_longjmp to support x32.
20180
20181 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
20182
20183         PR target/56010
20184         PR target/83743
20185         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20186         #include "opts.h".
20187         (rs6000_supported_cpu_names): New static variable.
20188         (linux_cpu_translation_table): Likewise.
20189         (elf_platform) <cpu>: Define new static variable and use it.
20190         Translate kernel AT_PLATFORM name to canonical name if needed.
20191         Error if platform name is unknown.
20192
20193 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
20194
20195         PR target/84089
20196         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20197
20198 2018-02-01  Jeff Law  <law@redhat.com>
20199
20200         PR target/84128
20201         * config/i386/i386.c (release_scratch_register_on_entry): Add new
20202         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
20203         the scratch if RELEASE_VIA_POP is false.
20204         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20205         If we have to save a temporary register, decrement SIZE appropriately.
20206         Pass new arguments to release_scratch_register_on_entry.
20207         (ix86_adjust_stack_and_probe): Likewise.
20208         (ix86_emit_probe_stack_range): Pass new arguments to
20209         release_scratch_register_on_entry.
20210
20211 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
20212
20213         PR rtl-optimization/84157
20214         * combine.c (change_zero_ext): Use REG_P predicate in
20215         front of HARD_REGISTER_P predicate.
20216
20217 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
20218
20219         * config/avr/avr.c (avr_option_override): Move disabling of
20220         -fdelete-null-pointer-checks to...
20221         * common/config/avr/avr-common.c (avr_option_optimization_table):
20222         ...here.
20223
20224 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20225
20226         PR tree-optimization/81635
20227         * tree-data-ref.c (split_constant_offset_1): For types that
20228         wrap on overflow, try to use range info to prove that wrapping
20229         cannot occur.
20230
20231 2018-02-01  Renlin Li  <renlin.li@arm.com>
20232
20233         PR target/83370
20234         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20235         TAILCALL_ADDR_REGS.
20236         (aarch64_register_move_cost): Likewise.
20237         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
20238         TAILCALL_ADDR_REGS.
20239         (REG_CLASS_NAMES): Likewise.
20240         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
20241         TAILCALL_ADDR_REGS. Remove IP registers.
20242         * config/aarch64/aarch64.md (Ucs): Update register constraint.
20243
20244 2018-02-01  Richard Biener  <rguenther@suse.de>
20245
20246         * domwalk.h (dom_walker::dom_walker): Add additional constructor
20247         for specifying RPO order and allow NULL for that.
20248         * domwalk.c (dom_walker::dom_walker): Likewise.
20249         (dom_walker::walk): Handle NULL RPO order.
20250         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
20251         in RPO order.
20252         (rewrite_update_dom_walker): Likewise.
20253         (mark_def_dom_walker): Likewise.
20254
20255 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20256
20257         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
20258         (aarch64_maybe_expand_sve_subreg_move): Declare.
20259         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
20260         * config/aarch64/predicates.md (aarch64_any_register_operand): New
20261         predicate.
20262         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
20263         that are semantically a reverse operation.
20264         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
20265         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
20266         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
20267         functions.
20268         (aarch64_can_change_mode_class): For big-endian, forbid changes
20269         between two SVE modes if they have different element sizes.
20270
20271 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20272
20273         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
20274         the TImode handling for big-endian targets.
20275
20276 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20277
20278         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
20279         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
20280         not just bytes.
20281         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
20282         Remove BSWAP handing for big-endian targets and use the form of
20283         LD1RQ appropariate for the mode.
20284
20285 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20286
20287         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
20288         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
20289         duplicated element.
20290
20291 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
20292
20293         PR tearget/83845
20294         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
20295         check for operands that need to go through aarch64_sve_reload_be.
20296
20297 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
20298
20299         PR tree-optimization/81661
20300         PR tree-optimization/84117
20301         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
20302         * tree-eh.c: Include gimplify.h.
20303         (find_trapping_overflow, replace_trapping_overflow,
20304         rewrite_to_non_trapping_overflow): New functions.
20305         * tree-vect-loop.c: Include tree-eh.h.
20306         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20307         * tree-data-ref.c: Include tree-eh.h.
20308         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20309
20310 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
20311
20312         PR rtl-optimization/84123
20313         * combine.c (change_zero_ext): Check if hard register satisfies
20314         can_change_dest_mode before calling gen_lowpart_SUBREG.
20315
20316 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
20317
20318         PR target/82444
20319         * ira.c (ira_init_register_move_cost): Remove assert.
20320
20321 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20322
20323         PR rtl-optimization/84071
20324         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20325         * doc/tm.texi: Regenerate.
20326
20327 2018-01-31  Richard Biener  <rguenther@suse.de>
20328
20329         PR tree-optimization/84132
20330         * tree-data-ref.c (analyze_miv_subscript): Properly
20331         check whether evolution_function_is_affine_multivariate_p
20332         before calling gcd_of_steps_may_divide_p.
20333
20334 2018-01-31  Julia Koval  <julia.koval@intel.com>
20335
20336         PR target/83618
20337         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20338         * config/i386/i386.md (rdpid_rex64) New.
20339         (rdpid): Make 32bit only.
20340
20341 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
20342
20343         PR lto/84105
20344         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20345         an IDENTIFIER_NODE for FUNCTION_TYPE's.
20346
20347 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20348
20349         Revert
20350         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
20351
20352         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20353
20354 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20355
20356         PR rtl-optimization/84071
20357         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20358         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20359
20360 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
20361
20362         * config/arc/arc.c (arc_handle_aux_attribute): New function.
20363         (arc_attribute_table): Add 'aux' attribute.
20364         (arc_in_small_data_p): Consider aux like variables.
20365         (arc_is_aux_reg_p): New function.
20366         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20367         (arc_get_aux_arg): New function.
20368         (prepare_move_operands): Handle aux-register access.
20369         (arc_handle_aux_attribute): New function.
20370         * doc/extend.texi (ARC Variable attributes): Add subsection.
20371
20372 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
20373
20374         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20375         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20376         (arc_attribute_table): Add 'uncached' attribute.
20377         (arc_print_operand): Print '.di' flag for uncached memory
20378         accesses.
20379         (arc_in_small_data_p): Do not consider for small data the uncached
20380         types.
20381         (arc_is_uncached_mem_p): New function.
20382         * config/arc/predicates.md (compact_store_memory_operand): Check
20383         for uncached memory accesses.
20384         (nonvol_nonimm_operand): Likewise.
20385         * doc/extend.texi (ARC Type Attribute): New subsection.
20386
20387 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
20388
20389         PR c/84100
20390         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20391         falign-loops=): Add Optimization flag.
20392
20393 2018-01-30  Jeff Law  <law@redhat.com>
20394
20395         PR target/84064
20396         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20397         INT_REGISTERS_SAVED.  Check it prior to calling
20398         get_scratch_register_on_entry.
20399         (ix86_adjust_stack_and_probe): Similarly.
20400         (ix86_emit_probe_stack_range): Similarly.
20401         (ix86_expand_prologue): Corresponding changes.
20402
20403 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20404
20405         PR target/40411
20406         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20407         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20408
20409 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
20410
20411         PR target/84112
20412         * lra-constraints.c (curr_insn_transform): Process AND in the
20413         address.
20414
20415 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
20416
20417         PR rtl-optimization/83986
20418         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20419         dependence against last_pending_memory_flush in addition to
20420         pending_jump_insns.
20421
20422 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
20423
20424         PR tree-optimization/81611
20425         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20426         copies.
20427
20428 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20429
20430         PR target/83758
20431         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20432         a reg rtx.
20433
20434 2018-01-30  Richard Biener  <rguenther@suse.de>
20435             Jakub Jelinek  <jakub@redhat.com>
20436
20437         PR tree-optimization/84111
20438         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20439         inner loops added during recursion, as they don't have up-to-date
20440         SSA form.
20441
20442 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
20443
20444         PR ipa/81360
20445         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20446         (can_inline_edge_by_limits_p): ... here.
20447         (can_early_inline_edge_p, check_callers,
20448         update_caller_keys, update_callee_keys, recursive_inlining,
20449         add_new_edges_to_heap, speculation_useful_p,
20450         inline_small_functions,
20451         inline_small_functions, flatten_function,
20452         inline_to_all_callers_1): Update.
20453
20454 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
20455
20456         * profile-count.c (profile_count::combine_with_ipa_count): Handle
20457         zeros correctly.
20458
20459 2018-01-30  Richard Biener  <rguenther@suse.de>
20460
20461         PR tree-optimization/83008
20462         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20463         invariant and constant vector uses in stmts when they need
20464         more than one stmt.
20465
20466 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20467
20468         PR bootstrap/84017
20469         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20470         * configure: Regenerate.
20471
20472 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20473
20474         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20475         pattern.
20476         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20477         Use gen_rtx_REG rather than gen_lowpart.
20478
20479 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20480
20481         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20482         rather than 0 when creating partial subregs.
20483
20484 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20485
20486         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20487         of usage.
20488
20489 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20490
20491         PR target/81550
20492         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20493         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20494         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20495         flags.  This restores the settings used before the 2017-07-24.
20496         Turning off pre increment/decrement/modify allows IVOPTS to
20497         optimize DF/SF loops where the index is an int.
20498
20499 2018-01-29  Richard Biener  <rguenther@suse.de>
20500             Kelvin Nilsen  <kelvin@gcc.gnu.org>
20501
20502         PR bootstrap/80867
20503         * tree-vect-stmts.c (vectorizable_call): Don't call
20504         targetm.vectorize_builtin_md_vectorized_function if callee is
20505         NULL.
20506
20507 2018-01-22  Carl Love  <cel@us.ibm.com>
20508
20509         * doc/extend.tex: Fix typo in second arg in
20510         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20511
20512 2018-01-29  Richard Biener  <rguenther@suse.de>
20513
20514         PR tree-optimization/84086
20515         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20516         (flush_ssaname_freelist): When SSA names were released reset
20517         the SCEV hash table.
20518
20519 2018-01-29  Richard Biener  <rguenther@suse.de>
20520
20521         PR tree-optimization/84057
20522         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20523         removed paths when removing edges.
20524
20525 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
20526
20527         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20528         -mfunction-return=@var{choice}.
20529
20530 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
20531
20532         PR diagnostic/84034
20533         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20534         Handle CR like TAB.
20535         (layout::print_source_line): Likewise.
20536         (test_get_line_width_without_trailing_whitespace): Add test cases.
20537
20538 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
20539
20540         PR middle-end/84040
20541         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20542         debug insns.
20543
20544 2018-01-26  Jim Wilson  <jimw@sifive.com>
20545
20546         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20547
20548         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20549         specified.
20550
20551 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20552
20553         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20554         and CMP + SUB-immediate -> SUBS.
20555
20556 2018-01-26  Martin Sebor  <msebor@redhat.com>
20557
20558         PR tree-optimization/83896
20559         * tree-ssa-strlen.c (get_string_len): Rename...
20560         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
20561         Avoid assuming length is constant.
20562         (handle_char_store): Use HOST_WIDE_INT for string length.
20563
20564 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
20565
20566         PR target/81763
20567         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20568         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20569
20570 2018-01-26  Richard Biener  <rguenther@suse.de>
20571
20572         PR rtl-optimization/84003
20573         * dse.c (record_store): Only record redundant stores when
20574         the earlier store aliases at least all accesses the later one does.
20575
20576 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
20577
20578         PR rtl-optimization/83985
20579         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20580         REG_CFA_RESTORE insns.
20581         (delete_unmarked_insns): Don't ignore separate shrink wrapping
20582         REG_CFA_RESTORE insns here.
20583
20584         PR c/83989
20585         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20586         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20587
20588 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20589
20590         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20591         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20592         (arc_init): Likewise.
20593         (arc_override_options): Likewise.
20594         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20595         value.
20596         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20597         support.
20598         * config/arc/arc.h (TARGET_DBNZ): Define.
20599         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20600         properly set the tune attribute.
20601         (dbnz): Use TARGET_DBNZ guard.
20602         * config/arc/arc.opt (mtune): Add core3 option.
20603
20604 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20605
20606         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20607         recognize new pic like addresses.
20608         (arc_delegitimize_address): Clean up.
20609
20610 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20611
20612         * config/arc/arc-arches.def: Option mrf16 valid for all
20613         architectures.
20614         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20615         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20616         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20617         * config/arc/arc-tables.opt: Regenerate.
20618         * config/arc/arc.c (arc_conditional_register_usage): Handle
20619         reduced register file case.
20620         (arc_file_start): Set must have build attributes.
20621         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20622         mrf16 option value.
20623         * config/arc/arc.opt (mrf16): Add new option.
20624         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20625         * config/arc/genmultilib.awk: Handle new mrf16 option.
20626         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20627         * config/arc/t-multilib: Regenerate.
20628         * doc/invoke.texi (ARC Options): Document mrf16 option.
20629
20630 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20631
20632         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20633         * config/arc/arc.c (arc_handle_secure_attribute): New function.
20634         (arc_attribute_table): Add 'secure_call' attribute.
20635         (arc_print_operand): Print secure call operand.
20636         (arc_function_ok_for_sibcall): Don't optimize tail calls when
20637         secure.
20638         (arc_is_secure_call_p): New function.  * config/arc/arc.md
20639         (call_i): Add support for sjli instruction.
20640         (call_value_i): Likewise.
20641         * config/arc/constraints.md (Csc): New constraint.
20642
20643 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20644             John Eric Martin  <John.Martin@emmicro-us.com>
20645
20646         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20647         * config/arc/arc.c (_arc_jli_section): New struct.
20648         (arc_jli_section): New type.
20649         (rc_jli_sections): New static variable.
20650         (arc_handle_jli_attribute): New function.
20651         (arc_attribute_table): Add jli_always and jli_fixed attribute.
20652         (arc_file_end): New function.
20653         (TARGET_ASM_FILE_END): Define.
20654         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20655         (arc_add_jli_section): New function.
20656         (jli_call_scan): Likewise.
20657         (arc_reorg): Call jli_call_scan.
20658         (arc_output_addsi): Remove 'S' from printing asm operand.
20659         (arc_is_jli_call_p): New function.
20660         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20661         operand.
20662         (movhi_insn): Likewise.
20663         (movsi_insn): Likewise.
20664         (movsi_set_cc_insn): Likewise.
20665         (loadqi_update): Likewise.
20666         (load_zeroextendqisi_update): Likewise.
20667         (load_signextendqisi_update): Likewise.
20668         (loadhi_update): Likewise.
20669         (load_zeroextendhisi_update): Likewise.
20670         (load_signextendhisi_update): Likewise.
20671         (loadsi_update): Likewise.
20672         (loadsf_update): Likewise.
20673         (movsicc_insn): Likewise.
20674         (bset_insn): Likewise.
20675         (bxor_insn): Likewise.
20676         (bclr_insn): Likewise.
20677         (bmsk_insn): Likewise.
20678         (bicsi3_insn): Likewise.
20679         (cmpsi_cc_c_insn): Likewise.
20680         (movsi_ne): Likewise.
20681         (movsi_cond_exec): Likewise.
20682         (clrsbsi2): Likewise.
20683         (norm_f): Likewise.
20684         (normw): Likewise.
20685         (swap): Likewise.
20686         (divaw): Likewise.
20687         (flag): Likewise.
20688         (sr): Likewise.
20689         (kflag): Likewise.
20690         (ffs): Likewise.
20691         (ffs_f): Likewise.
20692         (fls): Likewise.
20693         (call_i): Remove 'S' asm letter, add jli instruction.
20694         (call_value_i): Likewise.
20695         * config/arc/arc.op (mjli-always): New option.
20696         * config/arc/constraints.md (Cji): New constraint.
20697         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20698         operand.
20699         (subsf3_fpx): Likewise.
20700         (mulsf3_fpx): Likewise.
20701         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20702         asm operand.
20703         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20704         function attrbutes.
20705         * doc/invoke.texi (ARC): Document mjli-always option.
20706
20707 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
20708
20709         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20710         avoid addition with 0 and use incw and decw where possible.
20711
20712 2018-01-26  Richard Biener  <rguenther@suse.de>
20713
20714         PR tree-optimization/81082
20715         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20716         association if it requires casting to unsigned.
20717         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20718         from fold_plusminus_mult_expr to catch important cases late when
20719         range info is available.
20720
20721 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20722
20723         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20724         * configure.ac (hidden_linkonce): New test.
20725         * configure: Regenerate.
20726         * config.in: Regenerate.
20727
20728 2018-01-26  Julia Koval  <julia.koval@intel.com>
20729
20730         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20731         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20732         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20733         _mm_mask_bitshuffle_epi64_mask): Fix type.
20734         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20735         USI_FTYPE_V4DI_V4DI_USI): Remove.
20736         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20737         __builtin_ia32_vpshufbitqmb256_mask,
20738         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20739         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20740         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20741
20742 2018-01-26  Alan Modra  <amodra@gmail.com>
20743
20744         PR target/84033
20745         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20746         UNSPEC_VBPERMQ.  Sort other unspecs.
20747
20748 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
20749
20750         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20751
20752 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
20753
20754         PR middle-end/83055
20755         * predict.c (drop_profile): Do not push/pop cfun; update also
20756         node->count.
20757         (handle_missing_profiles): Fix logic looking for zero profiles.
20758
20759 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
20760
20761         PR middle-end/83977
20762         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20763         on functions with #pragma omp declare simd or functions with simd
20764         attribute.
20765         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20766         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20767         Remove trailing \n from warning_at calls.
20768
20769 2018-01-25  Tom de Vries  <tom@codesourcery.com>
20770
20771         PR target/84028
20772         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20773         for neutered workers.
20774
20775 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
20776
20777         PR target/68467
20778         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20779         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20780
20781 2018-01-24  Jeff Law  <law@redhat.com>
20782
20783         PR target/83994
20784         * i386.c (get_probe_interval): Move to earlier point.
20785         (ix86_compute_frame_layout): If -fstack-clash-protection and
20786         the frame is larger than the probe interval, then use pushes
20787         to save registers rather than reg->mem moves.
20788         (ix86_expand_prologue): Remove conditional for int_registers_saved
20789         assertion.
20790
20791 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
20792
20793         PR target/84014
20794         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20795         min/max for never referenced object.
20796
20797 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
20798
20799         PR middle-end/83977
20800         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20801         here.
20802         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20803         attributes from DECL_ATTRIBUTES (decl) without affecting
20804         DECL_ATTRIBUTES (current_function_decl).
20805         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20806         functions with non-NULL DECL_ABSTRACT_ORIGIN.
20807
20808 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
20809
20810         PR tree-optimization/83979
20811         * fold-const.c (fold_comparison): Use constant_boolean_node
20812         instead of boolean_{true,false}_node.
20813
20814 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
20815
20816         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20817         with zero counts.
20818
20819 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20820
20821         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20822         Simplify the clause that sets the length attribute.
20823         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20824         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20825         clause that sets the length attribute.
20826         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20827
20828 2018-01-24  Tom de Vries  <tom@codesourcery.com>
20829
20830         PR target/83589
20831         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20832         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20833         Add strict parameter.
20834         (prevent_branch_around_nothing): Insert dummy insn between branch to
20835         label and label with no ptx insn inbetween.
20836         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20837
20838 2018-01-24  Tom de Vries  <tom@codesourcery.com>
20839
20840         PR target/81352
20841         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20842         for neutered threads in warp.
20843         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20844
20845 2018-01-24  Richard Biener  <rguenther@suse.de>
20846
20847         PR tree-optimization/83176
20848         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20849         operands.
20850
20851 2018-01-24  Richard Biener  <rguenther@suse.de>
20852
20853         PR tree-optimization/82819
20854         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20855         code generating pluses that are no-ops in the target precision.
20856
20857 2018-01-24  Richard Biener  <rguenther@suse.de>
20858
20859         PR middle-end/84000
20860         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20861
20862 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
20863
20864         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20865         to merge probabilities.
20866         * predict.c (probably_never_executed): Also mark as cold functions
20867         with global 0 profile and guessed local profile.
20868         * profile-count.c (profile_probability::combine_with_count): New
20869         member function.
20870         * profile-count.h (profile_probability::operator*,
20871         profile_probability::operator*=, profile_probability::operator/,
20872         profile_probability::operator/=): Reduce precision to adjusted
20873         and set value to guessed on contradictory divisions.
20874         (profile_probability::combine_with_freq): Remove.
20875         (profile_probability::combine_wiht_count): Declare.
20876         (profile_count::force_nonzero):: Set to adjusted.
20877         (profile_count::probability_in):: Set quality to adjusted.
20878         * tree-ssa-tail-merge.c (replace_block_by): Use
20879         combine_with_count.
20880
20881 2018-01-23  Andrew Waterman  <andrew@sifive.com>
20882             Jim Wilson  <jimw@sifive.com>
20883
20884         * config/riscv/riscv.c (riscv_stack_boundary): New.
20885         (riscv_option_override): Set riscv_stack_boundary.  Handle
20886         riscv_preferred_stack_boundary_arg.
20887         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20888         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20889         (STACK_BOUNDARY): Set to riscv_stack_boundary.
20890         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20891         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20892         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20893
20894 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
20895
20896         PR target/83905
20897         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20898         of struct ix86_frame.
20899         (ix86_expand_epilogue): Likewise.  Add a local variable for
20900         the reg_save_offset field in struct ix86_frame.
20901
20902 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
20903
20904         PR tree-optimization/82604
20905         * tree-loop-distribution.c (enum partition_kind): New enum item
20906         PKIND_PARTIAL_MEMSET.
20907         (partition_builtin_p): Support above new enum item.
20908         (generate_code_for_partition): Ditto.
20909         (compute_access_range): Differentiate cases that equality can be
20910         proven at all loops, the innermost loops or no loops.
20911         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20912         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
20913         (finalize_partitions, distribute_loop): Don't fuse partition of
20914         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20915         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20916         parloop is enabled.
20917
20918 2018-01-23  Martin Liska  <mliska@suse.cz>
20919
20920         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20921         order to ignore the predictor.
20922         (PRED_POLYMORPHIC_CALL): Likewise.
20923         (PRED_RECURSIVE_CALL): Likewise.
20924
20925 2018-01-23  Martin Liska  <mliska@suse.cz>
20926
20927         * tree-profile.c (tree_profiling): Print function header to
20928         aware reader which function we are working on.
20929         * value-prof.c (gimple_find_values_to_profile): Do not print
20930         not interesting value histograms.
20931
20932 2018-01-23  Martin Liska  <mliska@suse.cz>
20933
20934         * profile-count.h (enum profile_quality): Add
20935         profile_uninitialized as the first value. Do not number values
20936         as they are zero based.
20937         (profile_count::verify): Update sanity check.
20938         (profile_probability::verify): Likewise.
20939
20940 2018-01-23  Nathan Sidwell  <nathan@acm.org>
20941
20942         * doc/invoke.texi (ffor-scope): Deprecate.
20943
20944 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
20945
20946         PR tree-optimization/83510
20947         * domwalk.c (set_all_edges_as_executable): New function.
20948         (dom_walker::dom_walker): Convert bool param
20949         "skip_unreachable_blocks" to enum reachability.  Move setup of
20950         edge flags to set_all_edges_as_executable and only do it when
20951         reachability is REACHABLE_BLOCKS.
20952         * domwalk.h (enum dom_walker::reachability): New enum.
20953         (dom_walker::dom_walker): Convert bool param
20954         "skip_unreachable_blocks" to enum reachability.
20955         (set_all_edges_as_executable): New decl.
20956         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
20957         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20958         "reachability".
20959         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20960         but converting true to REACHABLE_BLOCKS.
20961         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20962         * tree-vrp.c
20963         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20964         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20965         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20966         REACHABLE_BLOCKS.
20967         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20968         if check_all_array_refs will be called.
20969
20970 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
20971
20972         * tree.c (selftest::test_location_wrappers): Add more test
20973         coverage.
20974
20975 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
20976
20977         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20978         (selftest::test_bit_in_range): Likewise.
20979
20980 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
20981
20982         PR testsuite/83888
20983         * doc/sourcebuild.texi (vect_float): Say that the selector
20984         only describes the situation when -funsafe-math-optimizations is on.
20985         (vect_float_strict): Document.
20986
20987 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
20988
20989         PR tree-optimization/83965
20990         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20991         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20992         instead of checking only for a reduction.
20993         (vect_recog_widen_sum_pattern): Likewise.
20994
20995 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
20996
20997         * predict.c (probably_never_executed): Only use precise profile info.
20998         (compute_function_frequency): Skip after inlining hack since we now
20999         have quality checking.
21000
21001 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
21002
21003         * profile-count.h (profile_probability::very_unlikely,
21004         profile_probability::unlikely, profile_probability::even): Set
21005         precision to guessed.
21006
21007 2018-01-23  Richard Biener  <rguenther@suse.de>
21008
21009         PR tree-optimization/83963
21010         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
21011         Properly terminate dominator walk when crossing the exit edge not
21012         when visiting its source block.
21013
21014 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
21015
21016         PR c++/83918
21017         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
21018         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
21019
21020 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
21021
21022         PR tree-optimization/83957
21023         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
21024         semicolon after for body surrounded by braces.
21025
21026         PR tree-optimization/83081
21027         * profile-count.h (profile_probability::split): New method.
21028         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21029         Use profile_probability::split.
21030         (do_compare_rtx_and_jump): Fix adjustment of probabilities
21031         when splitting a single conditional jump into 2.
21032
21033 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
21034
21035         PR tree-optimization/69452
21036         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21037         decl.
21038
21039 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21040
21041         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21042         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21043         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21044
21045 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21046
21047         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21048         declaration.
21049         * config/rl78/rl78.md (movdi): New define_expand.
21050         * config/rl78/rl78.c (rl78_split_movdi): New function.
21051
21052 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
21053
21054         PR target/83862
21055         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21056         no longer used.
21057         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21058         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21059         128-bit to produce an UNSPEC move to get the double word with the
21060         signbit and then a shift directly to do signbit.
21061         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21062         implementation with a new version that just does either a direct
21063         move or a regular move.  Move memory interface to separate insns.
21064         Move insns so they are next to the expander.
21065         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21066         with signbit move.  Split big and little endian case.
21067         (signbit<mode>2_dm_mem_le): Likewise.
21068         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21069         (signbit<mode>2_dm2): Likewise.
21070
21071 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21072
21073         * config/rl78/rl78.md (anddi3): New define_expand.
21074
21075 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21076
21077         * config/rl78/rl78.md (umindi3): New define_expand.
21078
21079 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21080
21081         * config/rl78/rl78.md (smindi3): New define_expand.
21082
21083 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21084
21085         * config/rl78/rl78.md (smaxdi3): New define_expand.
21086
21087 2018-01-22  Carl Love  <cel@us.ibm.com>
21088
21089         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21090         LVX_V1TI): Add macro expansion.
21091         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21092         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21093         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21094         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21095         Change check to determine if the instruction is a byte reversing
21096         entry.  Fix typo in comment.
21097         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21098         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21099         Add def_builtin calls for new builtins.
21100         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21101         Add define_insn expansion.
21102
21103 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21104
21105         * config/rl78/rl78.md (umaxdi3): New define_expand.
21106
21107 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
21108
21109         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21110         for non-QImode registers.
21111
21112 2018-01-22  Richard Biener  <rguenther@suse.de>
21113
21114         PR tree-optimization/83963
21115         * graphite-scop-detection.c (scop_detection::get_sese): Delay
21116         including the loop exit block.
21117         (scop_detection::merge_sese): Likewise.
21118         (scop_detection::add_scop): Do it here instead.
21119
21120 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21121
21122         * doc/sourcebuild.texi (arm_softfloat): Document.
21123
21124 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
21125
21126         PR gcc/77734
21127         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21128         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21129         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21130
21131 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21132             David Edelsohn  <dje.gcc@gmail.com>
21133
21134         PR target/83946
21135         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21136         Change "crset eq" to "crset 2".
21137         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21138         (*call_indirect_aix<mode>_nospec): Likewise.
21139         (*call_value_indirect_aix<mode>_nospec): Likewise.
21140         (*call_indirect_elfv2<mode>_nospec): Likewise.
21141         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21142         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21143         change assembly output from . to $.
21144         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21145         (indirect_jump<mode>_nospec): Change assembly output from . to $.
21146         (*tablejump<mode>_internal1_nospec): Likewise.
21147
21148 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
21149
21150         PR target/80870
21151         * config/sh/sh_optimize_sett_clrt.cc:
21152         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21153
21154 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
21155
21156         PR tree-optimization/83940
21157         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21158         offset_dt to vect_constant_def rather than vect_unknown_def_type.
21159         (vect_check_load_store_mask): Add a mask_dt_out parameter and
21160         use it to pass back the definition type.
21161         (vect_check_store_rhs): Likewise rhs_dt_out.
21162         (vect_build_gather_load_calls): Add a mask_dt argument and use
21163         it instead of a call to vect_is_simple_use.
21164         (vectorizable_store): Update calls to vect_check_load_store_mask
21165         and vect_check_store_rhs.  Use the dt returned by the latter instead
21166         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
21167         instead of calls to vect_is_simple_use.  Pass the scalar rather
21168         than the vector operand to vect_is_simple_use when handling
21169         second and subsequent copies of an rhs value.
21170         (vectorizable_load): Update calls to vect_check_load_store_mask
21171         and vect_build_gather_load_calls.  Use the cached mask_dt and
21172         gs_info.offset_dt instead of calls to vect_is_simple_use.
21173
21174 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
21175
21176         PR middle-end/83945
21177         * tree-emutls.c: Include gimplify.h.
21178         (lower_emutls_2): New function.
21179         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21180         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21181         it before further processing.
21182
21183         PR target/83930
21184         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21185         UINTVAL (trueop1) instead of INTVAL (op1).
21186
21187 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
21188
21189         PR debug/81570
21190         PR debug/83728
21191         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21192         INCOMING_FRAME_SP_OFFSET if not defined.
21193         (scan_trace): Add ENTRY argument.  If true and
21194         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21195         emit a note to adjust the CFA offset.
21196         (create_cfi_notes): Adjust scan_trace callers.
21197         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21198         INCOMING_FRAME_SP_OFFSET in the CIE.
21199         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21200         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21201         Likewise.
21202         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21203         * doc/tm.texi: Regenerated.
21204
21205 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21206
21207         PR rtl-optimization/83147
21208         * lra-constraints.c (remove_inheritance_pseudos): Use
21209         lra_substitute_pseudo_within_insn.
21210
21211 2018-01-19  Tom de Vries  <tom@codesourcery.com>
21212             Cesar Philippidis  <cesar@codesourcery.com>
21213
21214         PR target/83920
21215         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21216
21217 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
21218
21219         PR target/83790
21220         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21221         spaces for function labels.
21222
21223 2018-01-19  Martin Liska  <mliska@suse.cz>
21224
21225         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21226         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21227         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21228         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21229         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21230         (PRED_CONST_RETURN): Change from 69 to 65.
21231         (PRED_NULL_RETURN): Change from 91 to 71.
21232         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21233         (PRED_LOOP_GUARD): Change from 66 to 73.
21234
21235 2018-01-19  Martin Liska  <mliska@suse.cz>
21236
21237         * predict.c (predict_insn_def): Add new assert.
21238         (struct branch_predictor): Change type to signed integer.
21239         (test_prediction_value_range): Amend test to cover
21240         PROB_UNINITIALIZED.
21241         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
21242         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
21243         (PRED_LOOP_ITERATIONS_MAX): Likewise.
21244         (PRED_LOOP_IV_COMPARE): Likewise.
21245         * predict.h (PROB_UNINITIALIZED): Define new constant.
21246
21247 2018-01-19  Martin Liska  <mliska@suse.cz>
21248
21249         * predict.c (dump_prediction): Add new format for
21250         analyze_brprob.py script which is enabled with -details
21251         suboption.
21252         * profile-count.h (precise_p): New function.
21253
21254 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
21255
21256         PR tree-optimization/83922
21257         * tree-vect-loop.c (vect_verify_full_masking): Return false if
21258         there are no statements that need masking.
21259         (vect_active_double_reduction_p): New function.
21260         (vect_analyze_loop_operations): Use it when handling phis that
21261         are not in the loop header.
21262
21263 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
21264
21265         PR tree-optimization/83914
21266         * tree-vect-loop.c (vectorizable_induction): Don't convert
21267         init_expr or apply the peeling adjustment for inductions
21268         that are nested within the vectorized loop.
21269
21270 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21271
21272         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
21273         instead of NEG.
21274
21275 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
21276
21277         PR sanitizer/81715
21278         PR testsuite/83882
21279         * function.h (gimplify_parameters): Add gimple_seq * argument.
21280         * function.c: Include gimple.h and options.h.
21281         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
21282         for the added local temporaries if needed.
21283         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
21284         if there are any parameter cleanups, wrap whole body into a
21285         try/finally with the cleanups.
21286
21287 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
21288
21289         PR target/82964
21290         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
21291         Use GET_MODE_CLASS for scalar floating point.
21292
21293 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
21294
21295         PR ipa/82256
21296         patch by PaX Team
21297         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
21298         Fix call of call_cgraph_insertion_hooks.
21299
21300 2018-01-18  Martin Sebor  <msebor@redhat.com>
21301
21302         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21303
21304 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
21305
21306         PR ipa/83619
21307         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21308         frequencies.
21309
21310 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
21311
21312         PR other/70268
21313         * common.opt: (-ffile-prefix-map): New option.
21314         * opts.c (common_handle_option): Defer it.
21315         * opts-global.c (handle_common_deferred_options): Handle it.
21316         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21317         * file-prefix-map.h: New file.
21318         (remap_debug_filename, add_debug_prefix_map): ...here.
21319         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21320         * final.c (debug_prefix_map, add_debug_prefix_map
21321         remap_debug_filename): Move to...
21322         * file-prefix-map.c: New file.
21323         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21324         generalize, get rid of alloca(), use strrchr() instead of strchr().
21325         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21326         Implement in terms of add_prefix_map().
21327         (remap_macro_filename, remap_debug_filename): Implement in term of
21328         remap_filename().
21329         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21330         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21331         * dbxout.c: Include file-prefix-map.h.
21332         * varasm.c: Likewise.
21333         * vmsdbgout.c: Likewise.
21334         * xcoffout.c: Likewise.
21335         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21336         * doc/cppopts.texi (-fmacro-prefix-map): Document.
21337         * doc/invoke.texi (-ffile-prefix-map): Document.
21338         (-fdebug-prefix-map): Update description.
21339
21340 2018-01-18  Martin Liska  <mliska@suse.cz>
21341
21342         * config/i386/i386.c (indirect_thunk_name): Document that also
21343         lfence is emitted.
21344         (output_indirect_thunk): Document why both instructions
21345         (pause and lfence) are generated.
21346
21347 2018-01-18  Richard Biener  <rguenther@suse.de>
21348
21349         PR tree-optimization/83887
21350         * graphite-scop-detection.c
21351         (scop_detection::get_nearest_dom_with_single_entry): Remove.
21352         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21353         (scop_detection::merge_sese): Re-implement with a flood-fill
21354         algorithm that properly finds a SESE region if it exists.
21355
21356 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
21357
21358         PR c/61240
21359         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21360         pointer_diff optimizations use view_convert instead of convert.
21361
21362 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21363
21364         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21365         Generate different code for -mno-speculate-indirect-jumps.
21366         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21367         (*call_indirect_aix<mode>): Disable for
21368         -mno-speculate-indirect-jumps.
21369         (*call_indirect_aix<mode>_nospec): New define_insn.
21370         (*call_value_indirect_aix<mode>): Disable for
21371         -mno-speculate-indirect-jumps.
21372         (*call_value_indirect_aix<mode>_nospec): New define_insn.
21373         (*sibcall_nonlocal_sysv<mode>): Generate different code for
21374         -mno-speculate-indirect-jumps.
21375         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21376
21377 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21378
21379         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21380         long double type, set the flags for noting the default long double
21381         type, even if we don't pass or return a long double type.
21382
21383 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
21384
21385         PR ipa/83051
21386         * ipa-inline.c (flatten_function): Do not overwrite final inlining
21387         failure.
21388
21389 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
21390
21391         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21392         support for merge[hl].
21393         (fold_mergehl_helper): New helper function.
21394         (tree-vector-builder.h): New #include for tree_vector_builder usage.
21395         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21396         (altivec_vmrglw_direct): Add xxmrglw insn.
21397
21398 2018-01-17  Andrew Waterman  <andrew@sifive.com>
21399
21400         * config/riscv/riscv.c (riscv_conditional_register_usage): If
21401         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21402
21403 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
21404
21405         PR lto/83121
21406         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21407         call the lto_location_cache before reading the
21408         DECL_SOURCE_LOCATION of the types.
21409
21410 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
21411             Richard Sandiford  <richard.sandiford@linaro.org>
21412
21413         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21414         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21415         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21416         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21417         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21418         Add declaration.
21419         * config/aarch64/constraints.md (aarch64_movti_operand):
21420         Limit immediates.
21421         * config/aarch64/predicates.md (Uti): Add new constraint.
21422
21423 2018-01-17  Carl Love  <cel@us.ibm.com>
21424
21425         * config/rs6000/vsx.md (define_expand xl_len_r,
21426         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21427         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21428         lxvll.
21429         (define_expand, define_insn): Move the shift left from  the
21430         define_insn to the define_expand for lxvl and stxvl instructions.
21431         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21432         and XL_LEN_R definitions to PURE.
21433
21434 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
21435
21436         * config/i386/i386.c (indirect_thunk_name): Declare regno
21437         as unsigned int.  Compare regno with INVALID_REGNUM.
21438         (output_indirect_thunk): Ditto.
21439         (output_indirect_thunk_function): Ditto.
21440         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
21441         in the call to output_indirect_thunk_function.
21442
21443 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
21444
21445         PR middle-end/83884
21446         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21447         rather than the size of inner_type to determine the stack slot size
21448         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21449
21450 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
21451
21452         PR target/83546
21453         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21454         to PTA_SILVERMONT.
21455
21456 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
21457
21458         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21459         endian Linux systems to optionally enable multilibs for selecting
21460         the long double type if the user configured an explicit type.
21461         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21462         have no long double multilibs if not defined.
21463         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21464         warn if the user used -mabi={ieee,ibm}longdouble and we built
21465         multilibs for long double.
21466         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21467         appropriate multilib option.
21468         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21469         multilib options.
21470         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21471         for building long double multilibs.
21472         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21473
21474 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
21475
21476         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21477         copies.
21478
21479         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21480         64 bits.
21481         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21482         128 bits.
21483
21484         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21485         variables.
21486
21487         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21488         return value.
21489
21490 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
21491
21492         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21493         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21494
21495 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21496
21497         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21498         different rtl trees depending on TARGET_64BIT.
21499         (rs6000_gen_lvx): Likewise.
21500
21501 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
21502
21503         * config/visium/visium.md (nop): Tweak comment.
21504         (hazard_nop): Likewise.
21505
21506 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21507
21508         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21509         -mspeculate-indirect-jumps.
21510         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21511         for -mno-speculate-indirect-jumps.
21512         (*call_indirect_elfv2<mode>_nospec): New define_insn.
21513         (*call_value_indirect_elfv2<mode>): Disable for
21514         -mno-speculate-indirect-jumps.
21515         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21516         (indirect_jump): Emit different RTL for
21517         -mno-speculate-indirect-jumps.
21518         (*indirect_jump<mode>): Disable for
21519         -mno-speculate-indirect-jumps.
21520         (*indirect_jump<mode>_nospec): New define_insn.
21521         (tablejump): Emit different RTL for
21522         -mno-speculate-indirect-jumps.
21523         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21524         (tablejumpsi_nospec): New define_expand.
21525         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21526         (tablejumpdi_nospec): New define_expand.
21527         (*tablejump<mode>_internal1): Disable for
21528         -mno-speculate-indirect-jumps.
21529         (*tablejump<mode>_internal1_nospec): New define_insn.
21530         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21531         option.
21532
21533 2018-01-16  Artyom Skrobov tyomitch@gmail.com
21534
21535         * caller-save.c (insert_save): Drop unnecessary parameter.  All
21536         callers updated.
21537
21538 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21539             Richard Biener  <rguenth@suse.de>
21540
21541         PR libgomp/83590
21542         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21543         return early, inline manually is_gimple_sizepos.  Make sure if we
21544         call gimplify_expr we don't end up with a gimple constant.
21545         * tree.c (variably_modified_type_p): Don't return true for
21546         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
21547         * gimplify.h (is_gimple_sizepos): Remove.
21548
21549 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21550
21551         PR tree-optimization/83857
21552         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21553         vectorizable_live_operation for pure SLP statements.
21554         (vectorizable_live_operation): Handle PHIs.
21555
21556 2018-01-16  Richard Biener  <rguenther@suse.de>
21557
21558         PR tree-optimization/83867
21559         * tree-vect-stmts.c (vect_transform_stmt): Precompute
21560         nested_in_vect_loop_p since the scalar stmt may get invalidated.
21561
21562 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21563
21564         PR c/83844
21565         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21566         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21567         If off is not INTEGER_CST, issue a may not be aligned warning
21568         rather than isn't aligned.  Use isn%'t rather than isn't.
21569         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21570         into MULT_EXPR.
21571         <case MULT_EXPR>: Improve the case when bottom and one of the
21572         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21573         operand, in that case check if the other operand is multiple of
21574         bottom divided by the INTEGER_CST operand.
21575
21576 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21577
21578         PR target/83858
21579         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21580         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21581         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21582         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21583         * config/pa/pa.c (pa_function_arg_advance): Likewise.
21584         (pa_function_arg, pa_arg_partial_bytes): Likewise.
21585         (pa_function_arg_size): New function.
21586
21587 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21588
21589         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21590         in a separate statement.
21591
21592 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21593
21594         PR tree-optimization/83847
21595         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21596         group gathers and scatters.
21597
21598 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21599
21600         PR rtl-optimization/86620
21601         * params.def (max-sched-ready-insns): Bump minimum value to 1.
21602
21603         PR rtl-optimization/83213
21604         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21605         to last if both are JUMP_INSNs.
21606
21607         PR tree-optimization/83843
21608         * gimple-ssa-store-merging.c
21609         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21610         store_immediate_info for bswap/nop orig_stores.
21611
21612 2018-01-15  Andrew Waterman  <andrew@sifive.com>
21613
21614         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21615         !TARGET_MUL.
21616         <UDIV>: Increase cost if !TARGET_DIV.
21617
21618 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
21619
21620         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21621         (define_attr "cr_logical_3op"): New.
21622         (cceq_ior_compare): Adjust.
21623         (cceq_ior_compare_complement): Adjust.
21624         (*cceq_rev_compare): Adjust.
21625         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21626         (is_cracked_insn): Adjust.
21627         (insn_must_be_first_in_group): Adjust.
21628         * config/rs6000/40x.md: Adjust.
21629         * config/rs6000/440.md: Adjust.
21630         * config/rs6000/476.md: Adjust.
21631         * config/rs6000/601.md: Adjust.
21632         * config/rs6000/603.md: Adjust.
21633         * config/rs6000/6xx.md: Adjust.
21634         * config/rs6000/7450.md: Adjust.
21635         * config/rs6000/7xx.md: Adjust.
21636         * config/rs6000/8540.md: Adjust.
21637         * config/rs6000/cell.md: Adjust.
21638         * config/rs6000/e300c2c3.md: Adjust.
21639         * config/rs6000/e500mc.md: Adjust.
21640         * config/rs6000/e500mc64.md: Adjust.
21641         * config/rs6000/e5500.md: Adjust.
21642         * config/rs6000/e6500.md: Adjust.
21643         * config/rs6000/mpc.md: Adjust.
21644         * config/rs6000/power4.md: Adjust.
21645         * config/rs6000/power5.md: Adjust.
21646         * config/rs6000/power6.md: Adjust.
21647         * config/rs6000/power7.md: Adjust.
21648         * config/rs6000/power8.md: Adjust.
21649         * config/rs6000/power9.md: Adjust.
21650         * config/rs6000/rs64.md: Adjust.
21651         * config/rs6000/titan.md: Adjust.
21652
21653 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21654
21655         * config/i386/predicates.md (indirect_branch_operand): Rewrite
21656         ix86_indirect_branch_register logic.
21657
21658 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21659
21660         * config/i386/constraints.md (Bs): Update
21661         ix86_indirect_branch_register check.  Don't check
21662         ix86_indirect_branch_register with GOT_memory_operand.
21663         (Bw): Likewise.
21664         * config/i386/predicates.md (GOT_memory_operand): Don't check
21665         ix86_indirect_branch_register here.
21666         (GOT32_symbol_operand): Likewise.
21667
21668 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21669
21670         * config/i386/predicates.md (constant_call_address_operand):
21671         Rewrite ix86_indirect_branch_register logic.
21672         (sibcall_insn_operand): Likewise.
21673
21674 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21675
21676         * config/i386/constraints.md (Bs): Replace
21677         ix86_indirect_branch_thunk_register with
21678         ix86_indirect_branch_register.
21679         (Bw): Likewise.
21680         * config/i386/i386.md (indirect_jump): Likewise.
21681         (tablejump): Likewise.
21682         (*sibcall_memory): Likewise.
21683         (*sibcall_value_memory): Likewise.
21684         Peepholes of indirect call and jump via memory: Likewise.
21685         * config/i386/i386.opt: Likewise.
21686         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21687         (GOT_memory_operand): Likewise.
21688         (call_insn_operand): Likewise.
21689         (sibcall_insn_operand): Likewise.
21690         (GOT32_symbol_operand): Likewise.
21691
21692 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
21693
21694         PR middle-end/83837
21695         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21696         type rather than type addr's type points to.
21697         (expand_omp_atomic_mutex): Likewise.
21698         (expand_omp_atomic): Likewise.
21699
21700 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21701
21702         PR target/83839
21703         * config/i386/i386.c (output_indirect_thunk_function): Use
21704         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21705         for  __x86_return_thunk.
21706
21707 2018-01-15  Richard Biener  <rguenther@suse.de>
21708
21709         PR middle-end/83850
21710         * expmed.c (extract_bit_field_1): Fix typo.
21711
21712 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21713
21714         PR target/83687
21715         * config/arm/iterators.md (VF): New mode iterator.
21716         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21717         Remove integer-related logic from pattern.
21718         (neon_vabd<mode>_3): Likewise.
21719
21720 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
21721
21722         PR middle-end/82694
21723         * common.opt (fstrict-overflow): No longer an alias.
21724         (fwrapv-pointer): New option.
21725         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21726         also for pointer types based on flag_wrapv_pointer.
21727         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21728         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21729         opts->x_flag_wrapv got set.
21730         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21731         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21732         POINTER_TYPE_OVERFLOW_UNDEFINED.
21733         * match.pd: Likewise in address comparison pattern.
21734         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21735
21736 2018-01-15  Richard Biener  <rguenther@suse.de>
21737
21738         PR lto/83804
21739         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21740         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
21741         Reset type names to their identifier if their TYPE_DECL doesn't
21742         have linkage (and thus is used for ODR and devirt).
21743         (save_debug_info_for_decl): Remove.
21744         (save_debug_info_for_type): Likewise.
21745         (add_tree_to_fld_list): Adjust.
21746         * tree-pretty-print.c (dump_generic_node): Make dumping of
21747         type names more robust.
21748
21749 2018-01-15  Richard Biener  <rguenther@suse.de>
21750
21751         * BASE-VER: Bump to 8.0.1.
21752
21753 2018-01-14  Martin Sebor  <msebor@redhat.com>
21754
21755         PR other/83508
21756         * builtins.c (check_access): Avoid warning when the no-warning bit
21757         is set.
21758
21759 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
21760
21761         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21762         * ira-color (allocno_hard_regs_compare): Likewise.
21763
21764 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
21765
21766         PR target/83013
21767         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21768         Use .pushsection/.popsection.
21769
21770 2018-01-14  Martin Sebor  <msebor@redhat.com>
21771
21772         PR c++/81327
21773         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21774
21775 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
21776
21777         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21778         entry from extra_headers.
21779         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21780         extra_headers, make the list bitwise identical to the i?86-*-* one.
21781
21782 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21783
21784         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21785         -mcmodel=large with -mindirect-branch=thunk,
21786         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21787         -mfunction-return=thunk-extern.
21788         * doc/invoke.texi: Document -mcmodel=large is incompatible with
21789         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21790         -mfunction-return=thunk and -mfunction-return=thunk-extern.
21791
21792 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21793
21794         * config/i386/i386.c (print_reg): Print the name of the full
21795         integer register without '%'.
21796         (ix86_print_operand): Handle 'V'.
21797         * doc/extend.texi: Document 'V' modifier.
21798
21799 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21800
21801         * config/i386/constraints.md (Bs): Disallow memory operand for
21802         -mindirect-branch-register.
21803         (Bw): Likewise.
21804         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21805         (GOT_memory_operand): Likewise.
21806         (call_insn_operand): Likewise.
21807         (sibcall_insn_operand): Likewise.
21808         (GOT32_symbol_operand): Likewise.
21809         * config/i386/i386.md (indirect_jump): Call convert_memory_address
21810         for -mindirect-branch-register.
21811         (tablejump): Likewise.
21812         (*sibcall_memory): Likewise.
21813         (*sibcall_value_memory): Likewise.
21814         Disallow peepholes of indirect call and jump via memory for
21815         -mindirect-branch-register.
21816         (*call_pop): Replace m with Bw.
21817         (*call_value_pop): Likewise.
21818         (*sibcall_pop_memory): Replace m with Bs.
21819         * config/i386/i386.opt (mindirect-branch-register): New option.
21820         * doc/invoke.texi: Document -mindirect-branch-register option.
21821
21822 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21823
21824         * config/i386/i386-protos.h (ix86_output_function_return): New.
21825         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21826         set function_return_type.
21827         (indirect_thunk_name): Add ret_p to indicate thunk for function
21828         return.
21829         (output_indirect_thunk_function): Pass false to
21830         indirect_thunk_name.
21831         (ix86_output_indirect_branch_via_reg): Likewise.
21832         (ix86_output_indirect_branch_via_push): Likewise.
21833         (output_indirect_thunk_function): Create alias for function
21834         return thunk if regno < 0.
21835         (ix86_output_function_return): New function.
21836         (ix86_handle_fndecl_attribute): Handle function_return.
21837         (ix86_attribute_table): Add function_return.
21838         * config/i386/i386.h (machine_function): Add
21839         function_return_type.
21840         * config/i386/i386.md (simple_return_internal): Use
21841         ix86_output_function_return.
21842         (simple_return_internal_long): Likewise.
21843         * config/i386/i386.opt (mfunction-return=): New option.
21844         (indirect_branch): Mention -mfunction-return=.
21845         * doc/extend.texi: Document function_return function attribute.
21846         * doc/invoke.texi: Document -mfunction-return= option.
21847
21848 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21849
21850         * config/i386/i386-opts.h (indirect_branch): New.
21851         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21852         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21853         with local indirect jump when converting indirect call and jump.
21854         (ix86_set_indirect_branch_type): New.
21855         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21856         (indirectlabelno): New.
21857         (indirect_thunk_needed): Likewise.
21858         (indirect_thunk_bnd_needed): Likewise.
21859         (indirect_thunks_used): Likewise.
21860         (indirect_thunks_bnd_used): Likewise.
21861         (INDIRECT_LABEL): Likewise.
21862         (indirect_thunk_name): Likewise.
21863         (output_indirect_thunk): Likewise.
21864         (output_indirect_thunk_function): Likewise.
21865         (ix86_output_indirect_branch_via_reg): Likewise.
21866         (ix86_output_indirect_branch_via_push): Likewise.
21867         (ix86_output_indirect_branch): Likewise.
21868         (ix86_output_indirect_jmp): Likewise.
21869         (ix86_code_end): Call output_indirect_thunk_function if needed.
21870         (ix86_output_call_insn): Call ix86_output_indirect_branch if
21871         needed.
21872         (ix86_handle_fndecl_attribute): Handle indirect_branch.
21873         (ix86_attribute_table): Add indirect_branch.
21874         * config/i386/i386.h (machine_function): Add indirect_branch_type
21875         and has_local_indirect_jump.
21876         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21877         to true.
21878         (tablejump): Likewise.
21879         (*indirect_jump): Use ix86_output_indirect_jmp.
21880         (*tablejump_1): Likewise.
21881         (simple_return_indirect_internal): Likewise.
21882         * config/i386/i386.opt (mindirect-branch=): New option.
21883         (indirect_branch): New.
21884         (keep): Likewise.
21885         (thunk): Likewise.
21886         (thunk-inline): Likewise.
21887         (thunk-extern): Likewise.
21888         * doc/extend.texi: Document indirect_branch function attribute.
21889         * doc/invoke.texi: Document -mindirect-branch= option.
21890
21891 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
21892
21893         PR ipa/83051
21894         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21895
21896 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
21897
21898         * ipa-inline.c (want_inline_small_function_p): Return false if
21899         inlining has already failed with CIF_FINAL_ERROR.
21900         (update_caller_keys): Call want_inline_small_function_p before
21901         can_inline_edge_p.
21902         (update_callee_keys): Likewise.
21903
21904 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21905
21906         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21907         New function.
21908         (rs6000_quadword_masked_address_p): Likewise.
21909         (quad_aligned_load_p): Likewise.
21910         (quad_aligned_store_p): Likewise.
21911         (const_load_sequence_p): Add comment to describe the outer-most loop.
21912         (mimic_memory_attributes_and_flags): New function.
21913         (rs6000_gen_stvx): Likewise.
21914         (replace_swapped_aligned_store): Likewise.
21915         (rs6000_gen_lvx): Likewise.
21916         (replace_swapped_aligned_load): Likewise.
21917         (replace_swapped_load_constant): Capitalize argument name in
21918         comment describing this function.
21919         (rs6000_analyze_swaps): Add a third pass to search for vector loads
21920         and stores that access quad-word aligned addresses and replace
21921         with stvx or lvx instructions when appropriate.
21922         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21923         New function prototype.
21924         (rs6000_quadword_masked_address_p): Likewise.
21925         (rs6000_gen_lvx): Likewise.
21926         (rs6000_gen_stvx): Likewise.
21927         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21928         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21929         when memory address is aligned.
21930         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21931         this split to select lvx instruction when memory address is aligned.
21932         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21933         instruction when memory address is aligned.
21934         (*vsx_le_perm_load_v16qi): Likewise.
21935         (four unnamed splitters): Modify to select the stvx instruction
21936         when memory is aligned.
21937
21938 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
21939
21940         * predict.c (determine_unlikely_bbs): Handle correctly BBs
21941         which appears in the queue multiple times.
21942
21943 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21944             Alan Hayward  <alan.hayward@arm.com>
21945             David Sherwood  <david.sherwood@arm.com>
21946
21947         * tree-vectorizer.h (vec_lower_bound): New structure.
21948         (_loop_vec_info): Add check_nonzero and lower_bounds.
21949         (LOOP_VINFO_CHECK_NONZERO): New macro.
21950         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21951         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21952         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21953         fields.  Make seg_len the distance travelled, not including the
21954         access size.
21955         (dr_direction_indicator): Declare.
21956         (dr_zero_step_indicator): Likewise.
21957         (dr_known_forward_stride_p): Likewise.
21958         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21959         tree-ssanames.h.
21960         (runtime_alias_check_p): Allow runtime alias checks with
21961         variable strides.
21962         (operator ==): Compare access_size and align.
21963         (prune_runtime_alias_test_list): Rework for new distinction between
21964         the access_size and seg_len.
21965         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
21966         segment lengths.
21967         (get_segment_min_max): New function.
21968         (create_intersect_range_checks): Use it.
21969         (dr_step_indicator): New function.
21970         (dr_direction_indicator): Likewise.
21971         (dr_zero_step_indicator): Likewise.
21972         (dr_known_forward_stride_p): Likewise.
21973         * tree-loop-distribution.c (data_ref_segment_size): Return
21974         DR_STEP * (niters - 1).
21975         (compute_alias_check_pairs): Update call to the dr_with_seg_len
21976         constructor.
21977         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21978         (vect_preserves_scalar_order_p): New function, split out from...
21979         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
21980         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21981         (vect_vfa_access_size): New function.
21982         (vect_vfa_align): Likewise.
21983         (vect_compile_time_alias): Take access_size_a and access_b arguments.
21984         (dump_lower_bound): New function.
21985         (vect_check_lower_bound): Likewise.
21986         (vect_small_gap_p): Likewise.
21987         (vectorizable_with_step_bound_p): Likewise.
21988         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21989         depencies if the vectorization factor is 1.  Convert the checks
21990         for nonzero steps into checks on the bounds of DR_STEP.  Try using
21991         a bunds check for variable steps if the minimum required step is
21992         relatively small. Update calls to the dr_with_seg_len
21993         constructor and to vect_compile_time_alias.
21994         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21995         function.
21996         (vect_loop_versioning): Call it.
21997         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21998         when retrying.
21999         (vect_estimate_min_profitable_iters): Account for any bounds checks.
22000
22001 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22002             Alan Hayward  <alan.hayward@arm.com>
22003             David Sherwood  <david.sherwood@arm.com>
22004
22005         * doc/sourcebuild.texi (vect_scatter_store): Document.
22006         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
22007         optabs.
22008         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
22009         Document.
22010         * genopinit.c (main): Add supports_vec_scatter_store and
22011         supports_vec_scatter_store_cached to target_optabs.
22012         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
22013         IFN_MASK_SCATTER_STORE.
22014         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
22015         functions.
22016         * internal-fn.h (internal_store_fn_p): Declare.
22017         (internal_fn_stored_value_index): Likewise.
22018         * internal-fn.c (scatter_store_direct): New macro.
22019         (expand_scatter_store_optab_fn): New function.
22020         (direct_scatter_store_optab_supported_p): New macro.
22021         (internal_store_fn_p): New function.
22022         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
22023         IFN_MASK_SCATTER_STORE.
22024         (internal_fn_mask_index): Likewise.
22025         (internal_fn_stored_value_index): New function.
22026         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22027         for scatter stores.
22028         * optabs-query.h (supports_vec_scatter_store_p): Declare.
22029         * optabs-query.c (supports_vec_scatter_store_p): New function.
22030         * tree-vectorizer.h (vect_get_store_rhs): Declare.
22031         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
22032         true for scatter stores.
22033         (vect_gather_scatter_fn_p): Handle scatter stores too.
22034         (vect_check_gather_scatter): Consider using scatter stores if
22035         supports_vec_scatter_store_p.
22036         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22037         scatter stores too.
22038         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22039         internal_fn_stored_value_index.
22040         (check_load_store_masking): Handle scatter stores too.
22041         (vect_get_store_rhs): Make public.
22042         (vectorizable_call): Use internal_store_fn_p.
22043         (vectorizable_store): Handle scatter store internal functions.
22044         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22045         when deciding whether the end of the group has been reached.
22046         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22047         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22048         (mask_scatter_store<mode>): New insns.
22049
22050 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22051             Alan Hayward  <alan.hayward@arm.com>
22052             David Sherwood  <david.sherwood@arm.com>
22053
22054         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22055         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22056         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22057         function.
22058         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22059         Use vect_truncate_gather_scatter_offset if we can't treat the
22060         operation as a normal gather load or scatter store.
22061         (get_group_load_store_type): Take the gather_scatter_info
22062         as argument.  Try using a gather load or scatter store for
22063         single-element groups.
22064         (get_load_store_type): Update calls to get_group_load_store_type
22065         and vect_use_strided_gather_scatters_p.
22066
22067 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22068             Alan Hayward  <alan.hayward@arm.com>
22069             David Sherwood  <david.sherwood@arm.com>
22070
22071         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22072         optional tree argument.
22073         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22074         null target hooks.
22075         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22076         but continue to use the current value as a fallback.
22077         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22078         to compare the updates.
22079         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22080         (get_load_store_type): Use it when handling a strided access.
22081         (vect_get_strided_load_store_ops): New function.
22082         (vect_get_data_ptr_increment): Likewise.
22083         (vectorizable_load): Handle strided gather loads.  Always pass
22084         a step to vect_create_data_ref_ptr and bump_vector_ptr.
22085
22086 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22087             Alan Hayward  <alan.hayward@arm.com>
22088             David Sherwood  <david.sherwood@arm.com>
22089
22090         * doc/md.texi (gather_load@var{m}): Document.
22091         (mask_gather_load@var{m}): Likewise.
22092         * genopinit.c (main): Add supports_vec_gather_load and
22093         supports_vec_gather_load_cached to target_optabs.
22094         * optabs-tree.c (init_tree_optimization_optabs): Use
22095         ggc_cleared_alloc to allocate target_optabs.
22096         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22097         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22098         functions.
22099         * internal-fn.h (internal_load_fn_p): Declare.
22100         (internal_gather_scatter_fn_p): Likewise.
22101         (internal_fn_mask_index): Likewise.
22102         (internal_gather_scatter_fn_supported_p): Likewise.
22103         * internal-fn.c (gather_load_direct): New macro.
22104         (expand_gather_load_optab_fn): New function.
22105         (direct_gather_load_optab_supported_p): New macro.
22106         (direct_internal_fn_optab): New function.
22107         (internal_load_fn_p): Likewise.
22108         (internal_gather_scatter_fn_p): Likewise.
22109         (internal_fn_mask_index): Likewise.
22110         (internal_gather_scatter_fn_supported_p): Likewise.
22111         * optabs-query.c (supports_at_least_one_mode_p): New function.
22112         (supports_vec_gather_load_p): Likewise.
22113         * optabs-query.h (supports_vec_gather_load_p): Declare.
22114         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22115         and memory_type field.
22116         (NUM_PATTERNS): Bump to 15.
22117         * tree-vect-data-refs.c: Include internal-fn.h.
22118         (vect_gather_scatter_fn_p): New function.
22119         (vect_describe_gather_scatter_call): Likewise.
22120         (vect_check_gather_scatter): Try using internal functions for
22121         gather loads.  Recognize existing calls to a gather load function.
22122         (vect_analyze_data_refs): Consider using gather loads if
22123         supports_vec_gather_load_p.
22124         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22125         (vect_get_gather_scatter_offset_type): Likewise.
22126         (vect_convert_mask_for_vectype): Likewise.
22127         (vect_add_conversion_to_patterm): Likewise.
22128         (vect_try_gather_scatter_pattern): Likewise.
22129         (vect_recog_gather_scatter_pattern): New pattern recognizer.
22130         (vect_vect_recog_func_ptrs): Add it.
22131         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22132         internal_fn_mask_index and internal_gather_scatter_fn_p.
22133         (check_load_store_masking): Take the gather_scatter_info as an
22134         argument and handle gather loads.
22135         (vect_get_gather_scatter_ops): New function.
22136         (vectorizable_call): Check internal_load_fn_p.
22137         (vectorizable_load): Likewise.  Handle gather load internal
22138         functions.
22139         (vectorizable_store): Update call to check_load_store_masking.
22140         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22141         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22142         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22143         (aarch64_gather_scale_operand_d): New predicates.
22144         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22145         (mask_gather_load<mode>): New insns.
22146
22147 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22148             Alan Hayward  <alan.hayward@arm.com>
22149             David Sherwood  <david.sherwood@arm.com>
22150
22151         * optabs.def (fold_left_plus_optab): New optab.
22152         * doc/md.texi (fold_left_plus_@var{m}): Document.
22153         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22154         * internal-fn.c (fold_left_direct): Define.
22155         (expand_fold_left_optab_fn): Likewise.
22156         (direct_fold_left_optab_supported_p): Likewise.
22157         * fold-const-call.c (fold_const_fold_left): New function.
22158         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22159         * tree-parloops.c (valid_reduction_p): New function.
22160         (gather_scalar_reductions): Use it.
22161         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22162         (vect_finish_replace_stmt): Declare.
22163         * tree-vect-loop.c (fold_left_reduction_fn): New function.
22164         (needs_fold_left_reduction_p): New function, split out from...
22165         (vect_is_simple_reduction): ...here.  Accept reductions that
22166         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22167         (vect_force_simple_reduction): Also store the reduction type in
22168         the assignment's STMT_VINFO_REDUC_TYPE.
22169         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22170         (merge_with_identity): New function.
22171         (vect_expand_fold_left): Likewise.
22172         (vectorize_fold_left_reduction): Likewise.
22173         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
22174         scalar phi in place for it.  Check for target support and reject
22175         cases that would reassociate the operation.  Defer the transform
22176         phase to vectorize_fold_left_reduction.
22177         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22178         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22179         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22180
22181 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22182
22183         * tree-if-conv.c (predicate_mem_writes): Remove redundant
22184         call to ifc_temp_var.
22185
22186 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22187             Alan Hayward  <alan.hayward@arm.com>
22188             David Sherwood  <david.sherwood@arm.com>
22189
22190         * target.def (legitimize_address_displacement): Take the original
22191         offset as a poly_int.
22192         * targhooks.h (default_legitimize_address_displacement): Update
22193         accordingly.
22194         * targhooks.c (default_legitimize_address_displacement): Likewise.
22195         * doc/tm.texi: Regenerate.
22196         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22197         as an argument, moving assert of ad->disp == ad->disp_term to...
22198         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
22199         Try calling targetm.legitimize_address_displacement before expanding
22200         the address rather than afterwards, and adjust for the new interface.
22201         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22202         Match the new hook interface.  Handle SVE addresses.
22203         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22204         new hook interface.
22205
22206 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22207
22208         * Makefile.in (OBJS): Add early-remat.o.
22209         * target.def (select_early_remat_modes): New hook.
22210         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22211         * doc/tm.texi: Regenerate.
22212         * targhooks.h (default_select_early_remat_modes): Declare.
22213         * targhooks.c (default_select_early_remat_modes): New function.
22214         * timevar.def (TV_EARLY_REMAT): New timevar.
22215         * passes.def (pass_early_remat): New pass.
22216         * tree-pass.h (make_pass_early_remat): Declare.
22217         * early-remat.c: New file.
22218         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22219         function.
22220         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22221
22222 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22223             Alan Hayward  <alan.hayward@arm.com>
22224             David Sherwood  <david.sherwood@arm.com>
22225
22226         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22227         vfm1 with a bound_epilog parameter.
22228         (vect_do_peeling): Update calls accordingly, and move the prologue
22229         call earlier in the function.  Treat the base bound_epilog as 0 for
22230         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
22231         this base when peeling for gaps.
22232         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22233         with fully-masked loops.
22234         (vect_estimate_min_profitable_iters): Handle the single peeled
22235         iteration in that case.
22236
22237 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22238             Alan Hayward  <alan.hayward@arm.com>
22239             David Sherwood  <david.sherwood@arm.com>
22240
22241         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
22242         single-element interleaving even if the size is not a power of 2.
22243         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
22244         accesses for single-element interleaving if the group size is
22245         not a power of 2.
22246
22247 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22248             Alan Hayward  <alan.hayward@arm.com>
22249             David Sherwood  <david.sherwood@arm.com>
22250
22251         * doc/md.texi (fold_extract_last_@var{m}): Document.
22252         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
22253         * optabs.def (fold_extract_last_optab): New optab.
22254         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
22255         * internal-fn.c (fold_extract_direct): New macro.
22256         (expand_fold_extract_optab_fn): Likewise.
22257         (direct_fold_extract_optab_supported_p): Likewise.
22258         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
22259         * tree-vect-loop.c (vect_model_reduction_cost): Handle
22260         EXTRACT_LAST_REDUCTION.
22261         (get_initial_def_for_reduction): Do not create an initial vector
22262         for EXTRACT_LAST_REDUCTION reductions.
22263         (vectorizable_reduction): Leave the scalar phi in place for
22264         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
22265         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
22266         epilogue code for EXTRACT_LAST_REDUCTION and defer the
22267         transform phase to vectorizable_condition.
22268         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
22269         split out from...
22270         (vect_finish_stmt_generation): ...here.
22271         (vect_finish_replace_stmt): New function.
22272         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
22273         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
22274         pattern.
22275         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
22276
22277 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22278             Alan Hayward  <alan.hayward@arm.com>
22279             David Sherwood  <david.sherwood@arm.com>
22280
22281         * doc/md.texi (extract_last_@var{m}): Document.
22282         * optabs.def (extract_last_optab): New optab.
22283         * internal-fn.def (EXTRACT_LAST): New internal function.
22284         * internal-fn.c (cond_unary_direct): New macro.
22285         (expand_cond_unary_optab_fn): Likewise.
22286         (direct_cond_unary_optab_supported_p): Likewise.
22287         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
22288         loops using EXTRACT_LAST.
22289         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
22290         (extract_last_<mode>): ...this optab.
22291         (vec_extract<mode><Vel>): Update accordingly.
22292
22293 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22294             Alan Hayward  <alan.hayward@arm.com>
22295             David Sherwood  <david.sherwood@arm.com>
22296
22297         * target.def (empty_mask_is_expensive): New hook.
22298         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
22299         * doc/tm.texi: Regenerate.
22300         * targhooks.h (default_empty_mask_is_expensive): Declare.
22301         * targhooks.c (default_empty_mask_is_expensive): New function.
22302         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22303         if the target says that empty masks are expensive.
22304         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22305         New function.
22306         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22307
22308 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22309             Alan Hayward  <alan.hayward@arm.com>
22310             David Sherwood  <david.sherwood@arm.com>
22311
22312         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22313         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22314         (vect_use_loop_mask_for_alignment_p): New function.
22315         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22316         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22317         niters_skip argument.  Make sure that the first niters_skip elements
22318         of the first iteration are inactive.
22319         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22320         Update call to vect_set_loop_masks_directly.
22321         (get_misalign_in_elems): New function, split out from...
22322         (vect_gen_prolog_loop_niters): ...here.
22323         (vect_update_init_of_dr): Take a code argument that specifies whether
22324         the adjustment should be added or subtracted.
22325         (vect_update_init_of_drs): Likewise.
22326         (vect_prepare_for_masked_peels): New function.
22327         (vect_do_peeling): Skip prologue peeling if we're using a mask
22328         instead.  Update call to vect_update_inits_of_drs.
22329         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22330         mask_skip_niters.
22331         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22332         alignment.  Do not include the number of peeled iterations in
22333         the minimum threshold in that case.
22334         (vectorizable_induction): Adjust the start value down by
22335         LOOP_VINFO_MASK_SKIP_NITERS iterations.
22336         (vect_transform_loop): Call vect_prepare_for_masked_peels.
22337         Take the number of skipped iterations into account when calculating
22338         the loop bounds.
22339         * tree-vect-stmts.c (vect_gen_while_not): New function.
22340
22341 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22342             Alan Hayward  <alan.hayward@arm.com>
22343             David Sherwood  <david.sherwood@arm.com>
22344
22345         * doc/sourcebuild.texi (vect_fully_masked): Document.
22346         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22347         default value to 0.
22348         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22349         split out from...
22350         (vect_analyze_loop_2): ...here. Don't check the vectorization
22351         factor against the number of loop iterations if the loop is
22352         fully-masked.
22353
22354 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22355             Alan Hayward  <alan.hayward@arm.com>
22356             David Sherwood  <david.sherwood@arm.com>
22357
22358         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22359         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22360         (dump_groups): Update accordingly.
22361         (iv_use::mem_type): New member variable.
22362         (address_p): New function.
22363         (record_use): Add a mem_type argument and initialize the new
22364         mem_type field.
22365         (record_group_use): Add a mem_type argument.  Use address_p.
22366         Remove obsolete null checks of base_object.  Update call to record_use.
22367         (find_interesting_uses_op): Update call to record_group_use.
22368         (find_interesting_uses_cond): Likewise.
22369         (find_interesting_uses_address): Likewise.
22370         (get_mem_type_for_internal_fn): New function.
22371         (find_address_like_use): Likewise.
22372         (find_interesting_uses_stmt): Try find_address_like_use before
22373         calling find_interesting_uses_op.
22374         (addr_offset_valid_p): Use the iv mem_type field as the type
22375         of the addressed memory.
22376         (add_autoinc_candidates): Likewise.
22377         (get_address_cost): Likewise.
22378         (split_small_address_groups_p): Use address_p.
22379         (split_address_groups): Likewise.
22380         (add_iv_candidate_for_use): Likewise.
22381         (autoinc_possible_for_pair): Likewise.
22382         (rewrite_groups): Likewise.
22383         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22384         (determine_group_iv_cost): Update after split of USE_ADDRESS.
22385         (get_alias_ptr_type_for_ptr_address): New function.
22386         (rewrite_use_address): Rewrite address uses in calls that were
22387         identified by find_address_like_use.
22388
22389 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22390             Alan Hayward  <alan.hayward@arm.com>
22391             David Sherwood  <david.sherwood@arm.com>
22392
22393         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22394         TARGET_MEM_REFs.
22395         * gimple-expr.h (is_gimple_addressable: Likewise.
22396         * gimple-expr.c (is_gimple_address): Likewise.
22397         * internal-fn.c (expand_call_mem_ref): New function.
22398         (expand_mask_load_optab_fn): Use it.
22399         (expand_mask_store_optab_fn): Likewise.
22400
22401 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22402             Alan Hayward  <alan.hayward@arm.com>
22403             David Sherwood  <david.sherwood@arm.com>
22404
22405         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22406         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22407         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22408         (cond_umax@var{mode}): Document.
22409         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22410         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22411         (cond_umin_optab, cond_umax_optab): New optabs.
22412         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22413         (COND_IOR, COND_XOR): New internal functions.
22414         * internal-fn.h (get_conditional_internal_fn): Declare.
22415         * internal-fn.c (cond_binary_direct): New macro.
22416         (expand_cond_binary_optab_fn): Likewise.
22417         (direct_cond_binary_optab_supported_p): Likewise.
22418         (get_conditional_internal_fn): New function.
22419         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22420         Cope with reduction statements that are vectorized as calls rather
22421         than assignments.
22422         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22423         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22424         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22425         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22426         (UNSPEC_COND_EOR): New unspecs.
22427         (optab): Add mappings for them.
22428         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22429         (sve_int_op, sve_fp_op): New int attributes.
22430
22431 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22432             Alan Hayward  <alan.hayward@arm.com>
22433             David Sherwood  <david.sherwood@arm.com>
22434
22435         * optabs.def (while_ult_optab): New optab.
22436         * doc/md.texi (while_ult@var{m}@var{n}): Document.
22437         * internal-fn.def (WHILE_ULT): New internal function.
22438         * internal-fn.h (direct_internal_fn_supported_p): New override
22439         that takes two types as argument.
22440         * internal-fn.c (while_direct): New macro.
22441         (expand_while_optab_fn): New function.
22442         (convert_optab_supported_p): Likewise.
22443         (direct_while_optab_supported_p): New macro.
22444         * wide-int.h (wi::udiv_ceil): New function.
22445         * tree-vectorizer.h (rgroup_masks): New structure.
22446         (vec_loop_masks): New typedef.
22447         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22448         and fully_masked_p.
22449         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22450         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22451         (vect_max_vf): New function.
22452         (slpeel_make_loop_iterate_ntimes): Delete.
22453         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22454         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22455         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22456         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22457         internal-fn.h, stor-layout.h and optabs-query.h.
22458         (vect_set_loop_mask): New function.
22459         (add_preheader_seq): Likewise.
22460         (add_header_seq): Likewise.
22461         (interleave_supported_p): Likewise.
22462         (vect_maybe_permute_loop_masks): Likewise.
22463         (vect_set_loop_masks_directly): Likewise.
22464         (vect_set_loop_condition_masked): Likewise.
22465         (vect_set_loop_condition_unmasked): New function, split out from
22466         slpeel_make_loop_iterate_ntimes.
22467         (slpeel_make_loop_iterate_ntimes): Rename to..
22468         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
22469         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22470         (vect_do_peeling): Update call accordingly.
22471         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22472         loops.
22473         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22474         mask_compare_type, can_fully_mask_p and fully_masked_p.
22475         (release_vec_loop_masks): New function.
22476         (_loop_vec_info): Use it to free the loop masks.
22477         (can_produce_all_loop_masks_p): New function.
22478         (vect_get_max_nscalars_per_iter): Likewise.
22479         (vect_verify_full_masking): Likewise.
22480         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22481         retries, and free the mask rgroups before retrying.  Check loop-wide
22482         reasons for disallowing fully-masked loops.  Make the final decision
22483         about whether use a fully-masked loop or not.
22484         (vect_estimate_min_profitable_iters): Do not assume that peeling
22485         for the number of iterations will be needed for fully-masked loops.
22486         (vectorizable_reduction): Disable fully-masked loops.
22487         (vectorizable_live_operation): Likewise.
22488         (vect_halve_mask_nunits): New function.
22489         (vect_double_mask_nunits): Likewise.
22490         (vect_record_loop_mask): Likewise.
22491         (vect_get_loop_mask): Likewise.
22492         (vect_transform_loop): Handle the case in which the final loop
22493         iteration might handle a partial vector.  Call vect_set_loop_condition
22494         instead of slpeel_make_loop_iterate_ntimes.
22495         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22496         (check_load_store_masking): New function.
22497         (prepare_load_store_mask): Likewise.
22498         (vectorizable_store): Handle fully-masked loops.
22499         (vectorizable_load): Likewise.
22500         (supportable_widening_operation): Use vect_halve_mask_nunits for
22501         booleans.
22502         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22503         (vect_gen_while): New function.
22504         * config/aarch64/aarch64.md (umax<mode>3): New expander.
22505         (aarch64_uqdec<mode>): New insn.
22506
22507 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22508             Alan Hayward  <alan.hayward@arm.com>
22509             David Sherwood  <david.sherwood@arm.com>
22510
22511         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22512         (reduc_xor_scal_optab): New optabs.
22513         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22514         (reduc_xor_scal_@var{m}): Document.
22515         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22516         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22517         internal functions.
22518         * fold-const-call.c (fold_const_call): Handle them.
22519         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22520         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22521         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22522         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22523         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22524         (UNSPEC_XORV): New unspecs.
22525         (optab): Add entries for them.
22526         (BITWISEV): New int iterator.
22527         (bit_reduc_op): New int attributes.
22528
22529 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22530             Alan Hayward  <alan.hayward@arm.com>
22531             David Sherwood  <david.sherwood@arm.com>
22532
22533         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22534         * internal-fn.def (VEC_SHL_INSERT): New internal function.
22535         * optabs.def (vec_shl_insert_optab): New optab.
22536         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22537         (duplicate_and_interleave): Likewise.
22538         * tree-vect-loop.c: Include internal-fn.h.
22539         (neutral_op_for_slp_reduction): New function, split out from
22540         get_initial_defs_for_reduction.
22541         (get_initial_def_for_reduction): Handle option 2 for variable-length
22542         vectors by loading the neutral value into a vector and then shifting
22543         the initial value into element 0.
22544         (get_initial_defs_for_reduction): Replace the code argument with
22545         the neutral value calculated by neutral_op_for_slp_reduction.
22546         Use gimple_build_vector for constant-length vectors.
22547         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22548         but the first group_size elements have a neutral value.
22549         Use duplicate_and_interleave otherwise.
22550         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22551         Update call to get_initial_defs_for_reduction.  Handle SLP
22552         reductions for variable-length vectors by creating one vector
22553         result for each scalar result, with the elements associated
22554         with other scalar results stubbed out with the neutral value.
22555         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22556         Require IFN_VEC_SHL_INSERT for double reductions on
22557         variable-length vectors, or SLP reductions that have
22558         a neutral value.  Require can_duplicate_and_interleave_p
22559         support for variable-length unchained SLP reductions if there
22560         is no neutral value, such as for MIN/MAX reductions.  Also require
22561         the number of vector elements to be a multiple of the number of
22562         SLP statements when doing variable-length unchained SLP reductions.
22563         Update call to vect_create_epilog_for_reduction.
22564         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22565         and remove initial values.
22566         (duplicate_and_interleave): Make public.
22567         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22568         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22569
22570 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22571             Alan Hayward  <alan.hayward@arm.com>
22572             David Sherwood  <david.sherwood@arm.com>
22573
22574         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22575         (can_duplicate_and_interleave_p): New function.
22576         (vect_get_and_check_slp_defs): Take the vector of statements
22577         rather than just the current one.  Remove excess parentheses.
22578         Restriction rejectinon of vect_constant_def and vect_external_def
22579         for variable-length vectors to boolean types, or types for which
22580         can_duplicate_and_interleave_p is false.
22581         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22582         (duplicate_and_interleave): New function.
22583         (vect_get_constant_vectors): Use gimple_build_vector for
22584         constant-length vectors and suitable variable-length constant
22585         vectors.  Use duplicate_and_interleave for other variable-length
22586         vectors.  Don't defer the update when inserting new statements.
22587
22588 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22589             Alan Hayward  <alan.hayward@arm.com>
22590             David Sherwood  <david.sherwood@arm.com>
22591
22592         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22593         min_profitable_iters doesn't go negative.
22594
22595 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22596             Alan Hayward  <alan.hayward@arm.com>
22597             David Sherwood  <david.sherwood@arm.com>
22598
22599         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22600         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22601         * optabs.def (vec_mask_load_lanes_optab): New optab.
22602         (vec_mask_store_lanes_optab): Likewise.
22603         * internal-fn.def (MASK_LOAD_LANES): New internal function.
22604         (MASK_STORE_LANES): Likewise.
22605         * internal-fn.c (mask_load_lanes_direct): New macro.
22606         (mask_store_lanes_direct): Likewise.
22607         (expand_mask_load_optab_fn): Handle masked operations.
22608         (expand_mask_load_lanes_optab_fn): New macro.
22609         (expand_mask_store_optab_fn): Handle masked operations.
22610         (expand_mask_store_lanes_optab_fn): New macro.
22611         (direct_mask_load_lanes_optab_supported_p): Likewise.
22612         (direct_mask_store_lanes_optab_supported_p): Likewise.
22613         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22614         parameter.
22615         (vect_load_lanes_supported): Likewise.
22616         * tree-vect-data-refs.c (strip_conversion): New function.
22617         (can_group_stmts_p): Likewise.
22618         (vect_analyze_data_ref_accesses): Use it instead of checking
22619         for a pair of assignments.
22620         (vect_store_lanes_supported): Take a masked_p parameter.
22621         (vect_load_lanes_supported): Likewise.
22622         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22623         vect_store_lanes_supported and vect_load_lanes_supported.
22624         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22625         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22626         parameter.  Don't allow gaps for masked accesses.
22627         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
22628         and vect_load_lanes_supported.
22629         (get_load_store_type): Take a masked_p parameter and update
22630         call to get_group_load_store_type.
22631         (vectorizable_store): Update call to get_load_store_type.
22632         Handle IFN_MASK_STORE_LANES.
22633         (vectorizable_load): Update call to get_load_store_type.
22634         Handle IFN_MASK_LOAD_LANES.
22635
22636 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22637             Alan Hayward  <alan.hayward@arm.com>
22638             David Sherwood  <david.sherwood@arm.com>
22639
22640         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22641         modes for SVE.
22642         * config/aarch64/aarch64-protos.h
22643         (aarch64_sve_struct_memory_operand_p): Declare.
22644         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22645         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22646         (VPRED, vpred): Handle SVE structure modes.
22647         * config/aarch64/constraints.md (Utx): New constraint.
22648         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22649         (aarch64_sve_struct_nonimmediate_operand): New predicates.
22650         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22651         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22652         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22653         structure modes.  Split into pieces after RA.
22654         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22655         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22656         New patterns.
22657         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22658         SVE structure modes.
22659         (aarch64_classify_address): Likewise.
22660         (sizetochar): Move earlier in file.
22661         (aarch64_print_operand): Handle SVE register lists.
22662         (aarch64_array_mode): New function.
22663         (aarch64_sve_struct_memory_operand_p): Likewise.
22664         (TARGET_ARRAY_MODE): Redefine.
22665
22666 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22667             Alan Hayward  <alan.hayward@arm.com>
22668             David Sherwood  <david.sherwood@arm.com>
22669
22670         * target.def (array_mode): New target hook.
22671         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22672         * doc/tm.texi: Regenerate.
22673         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22674         * hooks.c (hook_optmode_mode_uhwi_none): New function.
22675         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22676         targetm.array_mode.
22677         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
22678         type sizes.
22679
22680 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22681             Alan Hayward  <alan.hayward@arm.com>
22682             David Sherwood  <david.sherwood@arm.com>
22683
22684         * fold-const.c (fold_binary_loc): Check the argument types
22685         rather than the result type when testing for a vector operation.
22686
22687 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22688
22689         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22690         * doc/tm.texi: Regenerate.
22691
22692 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22693             Alan Hayward  <alan.hayward@arm.com>
22694             David Sherwood  <david.sherwood@arm.com>
22695
22696         * doc/invoke.texi (-msve-vector-bits=): Document new option.
22697         (sve): Document new AArch64 extension.
22698         * doc/md.texi (w): Extend the description of the AArch64
22699         constraint to include SVE vectors.
22700         (Upl, Upa): Document new AArch64 predicate constraints.
22701         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22702         enum.
22703         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22704         (msve-vector-bits=): New option.
22705         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22706         SVE when these are disabled.
22707         (sve): New extension.
22708         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22709         modes.  Adjust their number of units based on aarch64_sve_vg.
22710         (MAX_BITSIZE_MODE_ANY_MODE): Define.
22711         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22712         aarch64_addr_query_type.
22713         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22714         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22715         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22716         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22717         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22718         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22719         (aarch64_simd_imm_zero_p): Delete.
22720         (aarch64_check_zero_based_sve_index_immediate): Declare.
22721         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22722         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22723         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22724         (aarch64_sve_float_mul_immediate_p): Likewise.
22725         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22726         rather than an rtx.
22727         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22728         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22729         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22730         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22731         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22732         (aarch64_regmode_natural_size): Likewise.
22733         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22734         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22735         left one place.
22736         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22737         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22738         for VG and the SVE predicate registers.
22739         (V_ALIASES): Add a "z"-prefixed alias.
22740         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22741         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22742         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22743         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22744         (REG_CLASS_NAMES): Add entries for them.
22745         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
22746         and the predicate registers.
22747         (aarch64_sve_vg): Declare.
22748         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22749         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22750         (REGMODE_NATURAL_SIZE): Define.
22751         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22752         SVE macros.
22753         * config/aarch64/aarch64.c: Include cfgrtl.h.
22754         (simd_immediate_info): Add a constructor for series vectors,
22755         and an associated step field.
22756         (aarch64_sve_vg): New variable.
22757         (aarch64_dbx_register_number): Handle VG and the predicate registers.
22758         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22759         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22760         (VEC_ANY_DATA, VEC_STRUCT): New constants.
22761         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22762         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22763         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22764         (aarch64_get_mask_mode): New functions.
22765         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22766         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22767         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
22768         predicate modes and predicate registers.  Explicitly restrict
22769         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
22770         to store a vector mode if it is recognized by
22771         aarch64_classify_vector_mode.
22772         (aarch64_regmode_natural_size): New function.
22773         (aarch64_hard_regno_caller_save_mode): Return the original mode
22774         for predicates.
22775         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22776         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22777         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22778         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22779         functions.
22780         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
22781         does not overlap dest if the function is frame-related.  Handle
22782         SVE constants.
22783         (aarch64_split_add_offset): New function.
22784         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22785         them aarch64_add_offset.
22786         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22787         and update call to aarch64_sub_sp.
22788         (aarch64_add_cfa_expression): New function.
22789         (aarch64_expand_prologue): Pass extra temporary registers to the
22790         functions above.  Handle the case in which we need to emit new
22791         DW_CFA_expressions for registers that were originally saved
22792         relative to the stack pointer, but now have to be expressed
22793         relative to the frame pointer.
22794         (aarch64_output_mi_thunk): Pass extra temporary registers to the
22795         functions above.
22796         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
22797         IP0 and IP1 values for SVE frames.
22798         (aarch64_expand_vec_series): New function.
22799         (aarch64_expand_sve_widened_duplicate): Likewise.
22800         (aarch64_expand_sve_const_vector): Likewise.
22801         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22802         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
22803         into the register, rather than emitting a SET directly.
22804         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22805         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22806         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22807         (offset_9bit_signed_scaled_p): New functions.
22808         (aarch64_replicate_bitmask_imm): New function.
22809         (aarch64_bitmask_imm): Use it.
22810         (aarch64_cannot_force_const_mem): Reject expressions involving
22811         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
22812         (aarch64_classify_index): Handle SVE indices, by requiring
22813         a plain register index with a scale that matches the element size.
22814         (aarch64_classify_address): Handle SVE addresses.  Assert that
22815         the mode of the address is VOIDmode or an integer mode.
22816         Update call to aarch64_classify_symbol.
22817         (aarch64_classify_symbolic_expression): Update call to
22818         aarch64_classify_symbol.
22819         (aarch64_const_vec_all_in_range_p): New function.
22820         (aarch64_print_vector_float_operand): Likewise.
22821         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
22822         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
22823         and the FP immediates 1.0 and 0.5.
22824         (aarch64_print_address_internal): Handle SVE addresses.
22825         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22826         (aarch64_regno_regclass): Handle predicate registers.
22827         (aarch64_secondary_reload): Handle big-endian reloads of SVE
22828         data modes.
22829         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22830         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22831         (aarch64_convert_sve_vector_bits): New function.
22832         (aarch64_override_options): Use it to handle -msve-vector-bits=.
22833         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22834         rather than an rtx.
22835         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22836         Handle SVE vector and predicate modes.  Accept VL-based constants
22837         that need only one temporary register, and VL offsets that require
22838         no temporary registers.
22839         (aarch64_conditional_register_usage): Mark the predicate registers
22840         as fixed if SVE isn't available.
22841         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22842         Return true for SVE vector and predicate modes.
22843         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22844         rather than an unsigned int.  Handle SVE modes.
22845         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
22846         SVE modes.
22847         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22848         if SVE is enabled.
22849         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22850         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22851         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22852         (aarch64_sve_float_mul_immediate_p): New functions.
22853         (aarch64_sve_valid_immediate): New function.
22854         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22855         Explicitly reject structure modes.  Check for INDEX constants.
22856         Handle PTRUE and PFALSE constants.
22857         (aarch64_check_zero_based_sve_index_immediate): New function.
22858         (aarch64_simd_imm_zero_p): Delete.
22859         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22860         vector modes.  Accept constants in the range of CNT[BHWD].
22861         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22862         ask for an Advanced SIMD mode.
22863         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22864         (aarch64_simd_vector_alignment): Handle SVE predicates.
22865         (aarch64_vectorize_preferred_vector_alignment): New function.
22866         (aarch64_simd_vector_alignment_reachable): Use it instead of
22867         the vector size.
22868         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22869         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22870         functions.
22871         (MAX_VECT_LEN): Delete.
22872         (expand_vec_perm_d): Add a vec_flags field.
22873         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22874         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22875         (aarch64_evpc_ext): Don't apply a big-endian lane correction
22876         for SVE modes.
22877         (aarch64_evpc_rev): Rename to...
22878         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
22879         (aarch64_evpc_rev_global): New function.
22880         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22881         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22882         MAX_VECT_LEN.
22883         (aarch64_evpc_sve_tbl): New function.
22884         (aarch64_expand_vec_perm_const_1): Update after rename of
22885         aarch64_evpc_rev.  Handle SVE permutes too, trying
22886         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22887         than aarch64_evpc_tbl.
22888         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22889         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22890         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22891         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22892         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22893         (aarch64_expand_sve_vcond): New functions.
22894         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22895         of aarch64_vector_mode_p.
22896         (aarch64_dwarf_poly_indeterminate_value): New function.
22897         (aarch64_compute_pressure_classes): Likewise.
22898         (aarch64_can_change_mode_class): Likewise.
22899         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22900         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22901         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22902         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22903         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22904         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22905         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22906         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22907         constraints.
22908         (Dn, Dl, Dr): Accept const as well as const_vector.
22909         (Dz): Likewise.  Compare against CONST0_RTX.
22910         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22911         of "vector" where appropriate.
22912         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22913         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22914         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22915         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22916         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22917         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22918         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22919         (v_int_equiv): Extend to SVE modes.
22920         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22921         mode attributes.
22922         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22923         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22924         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22925         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22926         (SVE_COND_FP_CMP): New int iterators.
22927         (perm_hilo): Handle the new unpack unspecs.
22928         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22929         attributes.
22930         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22931         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22932         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22933         (aarch64_equality_operator, aarch64_constant_vector_operand)
22934         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22935         (aarch64_sve_nonimmediate_operand): Likewise.
22936         (aarch64_sve_general_operand): Likewise.
22937         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22938         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22939         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22940         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22941         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22942         (aarch64_sve_float_arith_immediate): Likewise.
22943         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22944         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22945         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22946         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22947         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22948         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22949         (aarch64_sve_float_arith_operand): Likewise.
22950         (aarch64_sve_float_arith_with_sub_operand): Likewise.
22951         (aarch64_sve_float_mul_operand): Likewise.
22952         (aarch64_sve_vec_perm_operand): Likewise.
22953         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22954         (aarch64_mov_operand): Accept const_poly_int and const_vector.
22955         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22956         as well as const_vector.
22957         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22958         in file.  Use CONST0_RTX and CONSTM1_RTX.
22959         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
22960         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22961         Use aarch64_simd_imm_zero.
22962         * config/aarch64/aarch64-sve.md: New file.
22963         * config/aarch64/aarch64.md: Include it.
22964         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22965         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22966         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22967         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22968         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22969         (sve): New attribute.
22970         (enabled): Disable instructions with the sve attribute unless
22971         TARGET_SVE.
22972         (movqi, movhi): Pass CONST_POLY_INT operaneds through
22973         aarch64_expand_mov_immediate.
22974         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22975         CNT[BHSD] immediates.
22976         (movti): Split CONST_POLY_INT moves into two halves.
22977         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22978         Split additions that need a temporary here if the destination
22979         is the stack pointer.
22980         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22981         (*add<mode>3_poly_1): New instruction.
22982         (set_clobber_cc): New expander.
22983
22984 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22985
22986         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22987         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
22988         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22989         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22990         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
22991         Change innermode from fixed_mode_size to machine_mode.
22992         (simplify_subreg): Update call accordingly.  Handle a constant-sized
22993         subreg of a variable-length CONST_VECTOR.
22994
22995 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22996             Alan Hayward  <alan.hayward@arm.com>
22997             David Sherwood  <david.sherwood@arm.com>
22998
22999         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
23000         (add_offset_to_base): New function, split out from...
23001         (create_mem_ref): ...here.  When handling a scale other than 1,
23002         check first whether the address is valid without the offset.
23003         Add it into the base if so, leaving the index and scale as-is.
23004
23005 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
23006
23007         PR c++/83778
23008         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
23009         fold_for_warn before checking if arg2 is INTEGER_CST.
23010
23011 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
23012
23013         * config/rs6000/predicates.md (load_multiple_operation): Delete.
23014         (store_multiple_operation): Delete.
23015         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
23016         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
23017         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
23018         guarded by TARGET_STRING.
23019         (rs6000_output_load_multiple): Delete.
23020         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
23021         OPTION_MASK_STRING / TARGET_STRING handling.
23022         (print_operand) <'N', 'O'>: Add comment that these are unused now.
23023         (const rs6000_opt_masks) <"string">: Change mask to 0.
23024         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
23025         (MASK_STRING): Delete.
23026         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23027         parts.  Simplify.
23028         (load_multiple): Delete.
23029         (*ldmsi8): Delete.
23030         (*ldmsi7): Delete.
23031         (*ldmsi6): Delete.
23032         (*ldmsi5): Delete.
23033         (*ldmsi4): Delete.
23034         (*ldmsi3): Delete.
23035         (store_multiple): Delete.
23036         (*stmsi8): Delete.
23037         (*stmsi7): Delete.
23038         (*stmsi6): Delete.
23039         (*stmsi5): Delete.
23040         (*stmsi4): Delete.
23041         (*stmsi3): Delete.
23042         (movmemsi_8reg): Delete.
23043         (corresponding unnamed define_insn): Delete.
23044         (movmemsi_6reg): Delete.
23045         (corresponding unnamed define_insn): Delete.
23046         (movmemsi_4reg): Delete.
23047         (corresponding unnamed define_insn): Delete.
23048         (movmemsi_2reg): Delete.
23049         (corresponding unnamed define_insn): Delete.
23050         (movmemsi_1reg): Delete.
23051         (corresponding unnamed define_insn): Delete.
23052         * config/rs6000/rs6000.opt (mno-string): New.
23053         (mstring): Replace by deprecation warning stub.
23054         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23055
23056 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
23057
23058         * regrename.c (regrename_do_replace): If replacing the same
23059         reg multiple times, try to reuse last created gen_raw_REG.
23060
23061         PR debug/81155
23062         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23063         main to workaround a bug in GDB.
23064
23065 2018-01-12  Tom de Vries  <tom@codesourcery.com>
23066
23067         PR target/83737
23068         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23069
23070 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
23071
23072         PR rtl-optimization/80481
23073         * ira-color.c (get_cap_member): New function.
23074         (allocnos_conflict_by_live_ranges_p): Use it.
23075         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23076         (setup_slot_coalesced_allocno_live_ranges): Ditto.
23077
23078 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
23079
23080         PR target/83628
23081         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23082         (*saddl_se_1): Ditto.
23083         (*ssubsi_1): Ditto.
23084         (*ssubl_se_1): Ditto.
23085
23086 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
23087
23088         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23089         rather than wi::to_widest for DR_INITs.
23090         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23091         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23092         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23093         INTEGER_CSTs.
23094         (vect_analyze_group_access_1): Note that here.
23095
23096 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
23097
23098         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23099         polynomial type sizes.
23100
23101 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
23102
23103         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23104         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23105         (gimple_add_tmp_var): Likewise.
23106
23107 2018-01-12  Martin Liska  <mliska@suse.cz>
23108
23109         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23110         (gimple_alloc_sizes): Likewise.
23111         (dump_gimple_statistics): Use PRIu64 in printf format.
23112         * gimple.h: Change uint64_t to int.
23113
23114 2018-01-12  Martin Liska  <mliska@suse.cz>
23115
23116         * tree-core.h: Use uint64_t instead of int.
23117         * tree.c (tree_node_counts): Likewise.
23118         (tree_node_sizes): Likewise.
23119         (dump_tree_statistics): Use PRIu64 in printf format.
23120
23121 2018-01-12  Martin Liska  <mliska@suse.cz>
23122
23123         * Makefile.in: As qsort_chk is implemented in vec.c, add
23124         vec.o to linkage of gencfn-macros.
23125         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23126         passing the info to record_node_allocation_statistics.
23127         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23128         and pass the info.
23129         * ggc-common.c (struct ggc_usage): Add operator== and use
23130         it in operator< and compare function.
23131         * mem-stats.h (struct mem_usage): Likewise.
23132         * vec.c (struct vec_usage): Remove operator< and compare
23133         function. Can be simply inherited.
23134
23135 2018-01-12  Martin Jambor  <mjambor@suse.cz>
23136
23137         PR target/81616
23138         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23139         * tree-ssa-math-opts.c: Include domwalk.h.
23140         (convert_mult_to_fma_1): New function.
23141         (fma_transformation_info): New type.
23142         (fma_deferring_state): Likewise.
23143         (cancel_fma_deferring): New function.
23144         (result_of_phi): Likewise.
23145         (last_fma_candidate_feeds_initial_phi): Likewise.
23146         (convert_mult_to_fma): Added deferring logic, split actual
23147         transformation to convert_mult_to_fma_1.
23148         (math_opts_dom_walker): New type.
23149         (math_opts_dom_walker::after_dom_children): New method, body moved
23150         here from pass_optimize_widening_mul::execute, added deferring logic
23151         bits.
23152         (pass_optimize_widening_mul::execute): Moved most of code to
23153         math_opts_dom_walker::after_dom_children.
23154         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23155         * config/i386/i386.c (ix86_option_override_internal): Added
23156         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23157
23158 2018-01-12  Richard Biener  <rguenther@suse.de>
23159
23160         PR debug/83157
23161         * dwarf2out.c (gen_variable_die): Do not reset old_die for
23162         inline instance vars.
23163
23164 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
23165
23166         PR target/81819
23167         * config/rx/rx.c (rx_is_restricted_memory_address):
23168         Handle SUBREG case.
23169
23170 2018-01-12  Richard Biener  <rguenther@suse.de>
23171
23172         PR tree-optimization/80846
23173         * target.def (split_reduction): New target hook.
23174         * targhooks.c (default_split_reduction): New function.
23175         * targhooks.h (default_split_reduction): Declare.
23176         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23177         target requests first reduce vectors by combining low and high
23178         parts.
23179         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23180         (get_vectype_for_scalar_type_and_size): Export.
23181         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23182         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23183         * doc/tm.texi: Regenerate.
23184         * config/i386/i386.c (ix86_split_reduction): Implement
23185         TARGET_VECTORIZE_SPLIT_REDUCTION.
23186
23187 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23188
23189         PR target/83368
23190         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23191         in PIC mode except for TARGET_VXWORKS_RTP.
23192         * config/sparc/sparc.c: Include cfgrtl.h.
23193         (TARGET_INIT_PIC_REG): Define.
23194         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23195         (sparc_pic_register_p): New predicate.
23196         (sparc_legitimate_address_p): Use it.
23197         (sparc_legitimize_pic_address): Likewise.
23198         (sparc_delegitimize_address): Likewise.
23199         (sparc_mode_dependent_address_p): Likewise.
23200         (gen_load_pcrel_sym): Remove 4th parameter.
23201         (load_got_register): Adjust call to above.  Remove obsolete stuff.
23202         (sparc_expand_prologue): Do not call load_got_register here.
23203         (sparc_flat_expand_prologue): Likewise.
23204         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23205         (sparc_use_pseudo_pic_reg): New function.
23206         (sparc_init_pic_reg): Likewise.
23207         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23208         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23209
23210 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
23211
23212         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23213         Add item for branch_cost.
23214
23215 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23216
23217         PR rtl-optimization/83565
23218         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23219         not extend the result to a larger mode for rotate operations.
23220         (num_sign_bit_copies1): Likewise.
23221
23222 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23223
23224         PR target/40411
23225         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23226         -symbolic.
23227         Use values-Xc.o for -pedantic.
23228         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23229
23230 2018-01-12  Martin Liska  <mliska@suse.cz>
23231
23232         PR ipa/83054
23233         * ipa-devirt.c (final_warning_record::grow_type_warnings):
23234         New function.
23235         (possible_polymorphic_call_targets): Use it.
23236         (ipa_devirt): Likewise.
23237
23238 2018-01-12  Martin Liska  <mliska@suse.cz>
23239
23240         * profile-count.h (enum profile_quality): Use 0 as invalid
23241         enum value of profile_quality.
23242
23243 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23244
23245         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
23246         -mext-string options.
23247
23248 2018-01-12  Richard Biener  <rguenther@suse.de>
23249
23250         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
23251         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
23252         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
23253         Likewise.
23254         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
23255
23256 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
23257
23258         * configure.ac (--with-long-double-format): Add support for the
23259         configuration option to change the default long double format on
23260         PowerPC systems.
23261         * config.gcc (powerpc*-linux*-*): Likewise.
23262         * configure: Regenerate.
23263         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
23264         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
23265         used without modification.
23266
23267 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23268
23269         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
23270         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
23271         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
23272         MISC_BUILTIN_SPEC_BARRIER.
23273         (rs6000_init_builtins): Likewise.
23274         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
23275         enum value.
23276         (speculation_barrier): New define_insn.
23277         * doc/extend.texi: Document __builtin_speculation_barrier.
23278
23279 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
23280
23281         PR target/83203
23282         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
23283         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
23284         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
23285         iterators.
23286         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
23287         integral modes instead of "ss" and "sd".
23288         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
23289         vectors with 32-bit and 64-bit elements.
23290         (vecdupssescalarmodesuffix): New mode attribute.
23291         (vec_dup<mode>): Use it.
23292
23293 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
23294
23295         PR target/83330
23296         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
23297         frame if argument is passed on stack.
23298
23299 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
23300
23301         PR target/82682
23302         * ree.c (combine_reaching_defs): Optimize also
23303         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23304         reg2=any_extend(exp); reg1=reg2;, formatting fix.
23305
23306 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
23307
23308         PR middle-end/83189
23309         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23310
23311 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
23312
23313         PR middle-end/83718
23314         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23315         after they are computed.
23316
23317 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
23318
23319         PR tree-optimization/83695
23320         * gimple-loop-linterchange.cc
23321         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23322         reset cached scev information after interchange.
23323         (pass_linterchange::execute): Remove call to scev_reset_htab.
23324
23325 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23326
23327         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23328         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23329         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23330         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23331         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23332         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23333         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23334         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23335         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23336         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23337         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23338         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23339         (V_lane_reg): Likewise.
23340         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23341         New define_expand.
23342         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23343         (vfmal_lane_low<mode>_intrinsic,
23344         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23345         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23346         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23347         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23348         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23349         vfmsl_lane_high<mode>_intrinsic): New define_insns.
23350
23351 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23352
23353         * config/arm/arm-cpus.in (fp16fml): New feature.
23354         (ALL_SIMD): Add fp16fml.
23355         (armv8.2-a): Add fp16fml as an option.
23356         (armv8.3-a): Likewise.
23357         (armv8.4-a): Add fp16fml as part of fp16.
23358         * config/arm/arm.h (TARGET_FP16FML): Define.
23359         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23360         when appropriate.
23361         * config/arm/arm-modes.def (V2HF): Define.
23362         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23363         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23364         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23365         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23366         vfmsl_low, vfmsl_high): New set of builtins.
23367         * config/arm/iterators.md (PLUSMINUS): New code iterator.
23368         (vfml_op): New code attribute.
23369         (VFMLHALVES): New int iterator.
23370         (VFML, VFMLSEL): New mode attributes.
23371         (V_reg): Define mapping for V2HF.
23372         (V_hi, V_lo): New mode attributes.
23373         (VF_constraint): Likewise.
23374         (vfml_half, vfml_half_selector): New int attributes.
23375         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23376         define_expand.
23377         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23378         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23379         New define_insn.
23380         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23381         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23382         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23383         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
23384         documentation.
23385         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23386         Document new effective target and option set.
23387
23388 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23389
23390         * config/arm/arm-cpus.in (armv8_4): New feature.
23391         (ARMv8_4a): New fgroup.
23392         (armv8.4-a): New arch.
23393         * config/arm/arm-tables.opt: Regenerate.
23394         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23395         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23396         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23397         Add matching rules for -march=armv8.4-a and extensions.
23398         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23399
23400 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
23401
23402         PR target/81821
23403         * config/rx/rx.md (BW): New mode attribute.
23404         (sync_lock_test_and_setsi): Add mode suffix to insn output.
23405
23406 2018-01-11  Richard Biener  <rguenther@suse.de>
23407
23408         PR tree-optimization/83435
23409         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23410         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23411         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23412
23413 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23414             Alan Hayward  <alan.hayward@arm.com>
23415             David Sherwood  <david.sherwood@arm.com>
23416
23417         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23418         field.
23419         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
23420         (aarch64_print_address_internal): Use it to check for a zero offset.
23421
23422 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23423             Alan Hayward  <alan.hayward@arm.com>
23424             David Sherwood  <david.sherwood@arm.com>
23425
23426         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23427         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23428         Return a poly_int64 rather than a HOST_WIDE_INT.
23429         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23430         rather than a HOST_WIDE_INT.
23431         * config/aarch64/aarch64.h (aarch64_frame): Protect with
23432         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
23433         hard_fp_offset, frame_size, initial_adjust, callee_offset and
23434         final_offset from HOST_WIDE_INT to poly_int64.
23435         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23436         to_constant when getting the number of units in an Advanced SIMD
23437         mode.
23438         (aarch64_builtin_vectorized_function): Check for a constant number
23439         of units.
23440         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23441         GET_MODE_SIZE.
23442         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23443         attribute instead of GET_MODE_NUNITS.
23444         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23445         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23446         GET_MODE_SIZE for fixed-size registers.
23447         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23448         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23449         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23450         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23451         (aarch64_print_operand, aarch64_print_address_internal)
23452         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23453         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23454         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23455         Handle polynomial GET_MODE_SIZE.
23456         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
23457         wider than SImode without modification.
23458         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23459         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23460         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23461         passing and returning SVE modes.
23462         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23463         rather than GEN_INT.
23464         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23465         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23466         (aarch64_allocate_and_probe_stack_space): Likewise.
23467         (aarch64_layout_frame): Cope with polynomial offsets.
23468         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23469         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
23470         polynomial offsets.
23471         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23472         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23473         poly_int64 rather than a HOST_WIDE_INT.
23474         (aarch64_get_separate_components, aarch64_process_components)
23475         (aarch64_expand_prologue, aarch64_expand_epilogue)
23476         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23477         (aarch64_anchor_offset): New function, split out from...
23478         (aarch64_legitimize_address): ...here.
23479         (aarch64_builtin_vectorization_cost): Handle polynomial
23480         TYPE_VECTOR_SUBPARTS.
23481         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23482         GET_MODE_NUNITS.
23483         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23484         number of elements from the PARALLEL rather than the mode.
23485         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23486         rather than GET_MODE_BITSIZE.
23487         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23488         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23489         (aarch64_expand_vec_perm_const_1): Handle polynomial
23490         d->perm.length () and d->perm elements.
23491         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
23492         Apply to_constant to d->perm elements.
23493         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23494         polynomial CONST_VECTOR_NUNITS.
23495         (aarch64_move_pointer): Take amount as a poly_int64 rather
23496         than an int.
23497         (aarch64_progress_pointer): Avoid temporary variable.
23498         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23499         the mode attribute instead of GET_MODE.
23500
23501 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23502             Alan Hayward  <alan.hayward@arm.com>
23503             David Sherwood  <david.sherwood@arm.com>
23504
23505         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23506         x exists before using it.
23507         (aarch64_add_constant_internal): Rename to...
23508         (aarch64_add_offset_1): ...this.  Replace regnum with separate
23509         src and dest rtxes.  Handle the case in which they're different,
23510         including when the offset is zero.  Replace scratchreg with an rtx.
23511         Use 2 additions if there is no spare register into which we can
23512         move a 16-bit constant.
23513         (aarch64_add_constant): Delete.
23514         (aarch64_add_offset): Replace reg with separate src and dest
23515         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
23516         Use aarch64_add_offset_1.
23517         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23518         an rtx rather than an int.  Take the delta as a poly_int64
23519         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
23520         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23521         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23522         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23523         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23524         and aarch64_add_sp.
23525         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23526         aarch64_add_constant.
23527
23528 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23529
23530         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23531         Use scalar_float_mode.
23532
23533 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23534
23535         * config/aarch64/aarch64-simd.md
23536         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23537         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23538         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23539         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23540         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23541         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23542         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23543         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23544         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23545         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23546
23547 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23548
23549         PR target/83514
23550         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23551         targ_options->x_arm_arch_string is non NULL.
23552
23553 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
23554
23555         * config/aarch64/aarch64.h
23556         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
23557
23558 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
23559
23560         PR target/82096
23561         * expmed.c (emit_store_flag_force): Swap if const op0
23562         and change VOIDmode to mode of op0.
23563
23564 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23565
23566         PR rtl-optimization/83761
23567         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23568         than bytes to mode_for_size.
23569
23570 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23571
23572         PR middle-end/83189
23573         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23574         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23575         profile.
23576
23577 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23578
23579         PR middle-end/83575
23580         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23581         when in layout mode.
23582         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23583         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23584         partition fixup.
23585
23586 2018-01-10  Michael Collison  <michael.collison@arm.com>
23587
23588         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23589         * config/aarch64/aarch64-option-extension.def: Add
23590         AARCH64_OPT_EXTENSION of 'fp16fml'.
23591         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23592         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23593         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23594         * config/aarch64/constraints.md (Ui7): New constraint.
23595         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23596         (VFMLA_SEL_W): Ditto.
23597         (f16quad): Ditto.
23598         (f16mac1): Ditto.
23599         (VFMLA16_LOW): New int iterator.
23600         (VFMLA16_HIGH): Ditto.
23601         (UNSPEC_FMLAL): New unspec.
23602         (UNSPEC_FMLSL): Ditto.
23603         (UNSPEC_FMLAL2): Ditto.
23604         (UNSPEC_FMLSL2): Ditto.
23605         (f16mac): New code attribute.
23606         * config/aarch64/aarch64-simd-builtins.def
23607         (aarch64_fmlal_lowv2sf): Ditto.
23608         (aarch64_fmlsl_lowv2sf): Ditto.
23609         (aarch64_fmlalq_lowv4sf): Ditto.
23610         (aarch64_fmlslq_lowv4sf): Ditto.
23611         (aarch64_fmlal_highv2sf): Ditto.
23612         (aarch64_fmlsl_highv2sf): Ditto.
23613         (aarch64_fmlalq_highv4sf): Ditto.
23614         (aarch64_fmlslq_highv4sf): Ditto.
23615         (aarch64_fmlal_lane_lowv2sf): Ditto.
23616         (aarch64_fmlsl_lane_lowv2sf): Ditto.
23617         (aarch64_fmlal_laneq_lowv2sf): Ditto.
23618         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23619         (aarch64_fmlalq_lane_lowv4sf): Ditto.
23620         (aarch64_fmlsl_lane_lowv4sf): Ditto.
23621         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23622         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23623         (aarch64_fmlal_lane_highv2sf): Ditto.
23624         (aarch64_fmlsl_lane_highv2sf): Ditto.
23625         (aarch64_fmlal_laneq_highv2sf): Ditto.
23626         (aarch64_fmlsl_laneq_highv2sf): Ditto.
23627         (aarch64_fmlalq_lane_highv4sf): Ditto.
23628         (aarch64_fmlsl_lane_highv4sf): Ditto.
23629         (aarch64_fmlalq_laneq_highv4sf): Ditto.
23630         (aarch64_fmlsl_laneq_highv4sf): Ditto.
23631         * config/aarch64/aarch64-simd.md:
23632         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23633         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23634         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23635         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23636         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23637         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23638         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23639         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23640         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23641         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23642         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23643         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23644         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23645         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23646         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23647         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23648         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23649         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23650         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23651         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23652         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23653         (vfmlsl_low_u32): Ditto.
23654         (vfmlalq_low_u32): Ditto.
23655         (vfmlslq_low_u32): Ditto.
23656         (vfmlal_high_u32): Ditto.
23657         (vfmlsl_high_u32): Ditto.
23658         (vfmlalq_high_u32): Ditto.
23659         (vfmlslq_high_u32): Ditto.
23660         (vfmlal_lane_low_u32): Ditto.
23661         (vfmlsl_lane_low_u32): Ditto.
23662         (vfmlal_laneq_low_u32): Ditto.
23663         (vfmlsl_laneq_low_u32): Ditto.
23664         (vfmlalq_lane_low_u32): Ditto.
23665         (vfmlslq_lane_low_u32): Ditto.
23666         (vfmlalq_laneq_low_u32): Ditto.
23667         (vfmlslq_laneq_low_u32): Ditto.
23668         (vfmlal_lane_high_u32): Ditto.
23669         (vfmlsl_lane_high_u32): Ditto.
23670         (vfmlal_laneq_high_u32): Ditto.
23671         (vfmlsl_laneq_high_u32): Ditto.
23672         (vfmlalq_lane_high_u32): Ditto.
23673         (vfmlslq_lane_high_u32): Ditto.
23674         (vfmlalq_laneq_high_u32): Ditto.
23675         (vfmlslq_laneq_high_u32): Ditto.
23676         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23677         (AARCH64_FL_FOR_ARCH8_4): New.
23678         (AARCH64_ISA_F16FML): New ISA flag.
23679         (TARGET_F16FML): New feature flag for fp16fml.
23680         (doc/invoke.texi): Document new fp16fml option.
23681
23682 2018-01-10  Michael Collison  <michael.collison@arm.com>
23683
23684         * config/aarch64/aarch64-builtins.c:
23685         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23686         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23687         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23688         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23689         (AARCH64_ISA_SHA3): New ISA flag.
23690         (TARGET_SHA3): New feature flag for sha3.
23691         * config/aarch64/iterators.md (sha512_op): New int attribute.
23692         (CRYPTO_SHA512): New int iterator.
23693         (UNSPEC_SHA512H): New unspec.
23694         (UNSPEC_SHA512H2): Ditto.
23695         (UNSPEC_SHA512SU0): Ditto.
23696         (UNSPEC_SHA512SU1): Ditto.
23697         * config/aarch64/aarch64-simd-builtins.def
23698         (aarch64_crypto_sha512hqv2di): New builtin.
23699         (aarch64_crypto_sha512h2qv2di): Ditto.
23700         (aarch64_crypto_sha512su0qv2di): Ditto.
23701         (aarch64_crypto_sha512su1qv2di): Ditto.
23702         (aarch64_eor3qv8hi): Ditto.
23703         (aarch64_rax1qv2di): Ditto.
23704         (aarch64_xarqv2di): Ditto.
23705         (aarch64_bcaxqv8hi): Ditto.
23706         * config/aarch64/aarch64-simd.md:
23707         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23708         (aarch64_crypto_sha512su0qv2di): Ditto.
23709         (aarch64_crypto_sha512su1qv2di): Ditto.
23710         (aarch64_eor3qv8hi): Ditto.
23711         (aarch64_rax1qv2di): Ditto.
23712         (aarch64_xarqv2di): Ditto.
23713         (aarch64_bcaxqv8hi): Ditto.
23714         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23715         (vsha512h2q_u64): Ditto.
23716         (vsha512su0q_u64): Ditto.
23717         (vsha512su1q_u64): Ditto.
23718         (veor3q_u16): Ditto.
23719         (vrax1q_u64): Ditto.
23720         (vxarq_u64): Ditto.
23721         (vbcaxq_u16): Ditto.
23722         * config/arm/types.md (crypto_sha512): New type attribute.
23723         (crypto_sha3): Ditto.
23724         (doc/invoke.texi): Document new sha3 option.
23725
23726 2018-01-10  Michael Collison  <michael.collison@arm.com>
23727
23728         * config/aarch64/aarch64-builtins.c:
23729         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23730         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23731         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23732         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23733         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23734         (AARCH64_ISA_SM4): New ISA flag.
23735         (TARGET_SM4): New feature flag for sm4.
23736         * config/aarch64/aarch64-simd-builtins.def
23737         (aarch64_sm3ss1qv4si): Ditto.
23738         (aarch64_sm3tt1aq4si): Ditto.
23739         (aarch64_sm3tt1bq4si): Ditto.
23740         (aarch64_sm3tt2aq4si): Ditto.
23741         (aarch64_sm3tt2bq4si): Ditto.
23742         (aarch64_sm3partw1qv4si): Ditto.
23743         (aarch64_sm3partw2qv4si): Ditto.
23744         (aarch64_sm4eqv4si): Ditto.
23745         (aarch64_sm4ekeyqv4si): Ditto.
23746         * config/aarch64/aarch64-simd.md:
23747         (aarch64_sm3ss1qv4si): Ditto.
23748         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23749         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23750         (aarch64_sm4eqv4si): Ditto.
23751         (aarch64_sm4ekeyqv4si): Ditto.
23752         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23753         (sm3part_op): Ditto.
23754         (CRYPTO_SM3TT): Ditto.
23755         (CRYPTO_SM3PART): Ditto.
23756         (UNSPEC_SM3SS1): New unspec.
23757         (UNSPEC_SM3TT1A): Ditto.
23758         (UNSPEC_SM3TT1B): Ditto.
23759         (UNSPEC_SM3TT2A): Ditto.
23760         (UNSPEC_SM3TT2B): Ditto.
23761         (UNSPEC_SM3PARTW1): Ditto.
23762         (UNSPEC_SM3PARTW2): Ditto.
23763         (UNSPEC_SM4E): Ditto.
23764         (UNSPEC_SM4EKEY): Ditto.
23765         * config/aarch64/constraints.md (Ui2): New constraint.
23766         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23767         * config/arm/types.md (crypto_sm3): New type attribute.
23768         (crypto_sm4): Ditto.
23769         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23770         (vsm3tt1aq_u32): Ditto.
23771         (vsm3tt1bq_u32): Ditto.
23772         (vsm3tt2aq_u32): Ditto.
23773         (vsm3tt2bq_u32): Ditto.
23774         (vsm3partw1q_u32): Ditto.
23775         (vsm3partw2q_u32): Ditto.
23776         (vsm4eq_u32): Ditto.
23777         (vsm4ekeyq_u32): Ditto.
23778         (doc/invoke.texi): Document new sm4 option.
23779
23780 2018-01-10  Michael Collison  <michael.collison@arm.com>
23781
23782         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23783         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23784         (AARCH64_FL_FOR_ARCH8_4): New.
23785         (AARCH64_FL_V8_4): New flag.
23786         (doc/invoke.texi): Document new armv8.4-a option.
23787
23788 2018-01-10  Michael Collison  <michael.collison@arm.com>
23789
23790         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23791         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23792         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23793         * config/aarch64/aarch64-option-extension.def: Add
23794         AARCH64_OPT_EXTENSION of 'sha2'.
23795         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23796         (crypto): Disable sha2 and aes if crypto disabled.
23797         (crypto): Enable aes and sha2 if enabled.
23798         (simd): Disable sha2 and aes if simd disabled.
23799         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23800         New flags.
23801         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23802         (TARGET_SHA2): New feature flag for sha2.
23803         (TARGET_AES): New feature flag for aes.
23804         * config/aarch64/aarch64-simd.md:
23805         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23806         conditional on TARGET_AES.
23807         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23808         (aarch64_crypto_sha1hsi): Make pattern conditional
23809         on TARGET_SHA2.
23810         (aarch64_crypto_sha1hv4si): Ditto.
23811         (aarch64_be_crypto_sha1hv4si): Ditto.
23812         (aarch64_crypto_sha1su1v4si): Ditto.
23813         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23814         (aarch64_crypto_sha1su0v4si): Ditto.
23815         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23816         (aarch64_crypto_sha256su0v4si): Ditto.
23817         (aarch64_crypto_sha256su1v4si): Ditto.
23818         (doc/invoke.texi): Document new aes and sha2 options.
23819
23820 2018-01-10  Martin Sebor  <msebor@redhat.com>
23821
23822         PR tree-optimization/83781
23823         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23824         as string arrays.
23825
23826 2018-01-11  Martin Sebor  <msebor@gmail.com>
23827             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23828
23829         PR tree-optimization/83501
23830         PR tree-optimization/81703
23831
23832         * tree-ssa-strlen.c (get_string_cst): Rename...
23833         (get_string_len): ...to this.  Handle global constants.
23834         (handle_char_store): Adjust.
23835
23836 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
23837             Jim Wilson  <jimw@sifive.com>
23838
23839         * config/riscv/riscv-protos.h (riscv_output_return): New.
23840         * config/riscv/riscv.c (struct machine_function): New naked_p field.
23841         (riscv_attribute_table, riscv_output_return),
23842         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23843         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23844         (riscv_compute_frame_info): Only compute frame->mask if not a naked
23845         function.
23846         (riscv_expand_prologue): Add early return for naked function.
23847         (riscv_expand_epilogue): Likewise.
23848         (riscv_function_ok_for_sibcall): Return false for naked function.
23849         (riscv_set_current_function): New.
23850         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23851         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23852         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23853         * doc/extend.texi (RISC-V Function Attributes): New.
23854
23855 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
23856
23857         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23858         check for 128-bit long double before checking TCmode.
23859         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23860         128-bit long doubles before checking TFmode or TCmode.
23861         (FLOAT128_IBM_P): Likewise.
23862
23863 2018-01-10  Martin Sebor  <msebor@redhat.com>
23864
23865         PR tree-optimization/83671
23866         * builtins.c (c_strlen): Unconditionally return zero for the empty
23867         string.
23868         Use -Warray-bounds for warnings.
23869         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23870         for non-constant array indices with COMPONENT_REF, arrays of
23871         arrays, and pointers to arrays.
23872         (gimple_fold_builtin_strlen): Determine and set length range for
23873         non-constant character arrays.
23874
23875 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
23876
23877         PR middle-end/81897
23878         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23879         empty blocks.
23880
23881 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
23882
23883         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23884
23885 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
23886
23887         PR target/83399
23888         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23889         VECTOR_MEM_ALTIVEC_OR_VSX_P.
23890         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23891         indexed_or_indirect_operand predicate.
23892         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23893         (*vsx_le_perm_load_v8hi): Likewise.
23894         (*vsx_le_perm_load_v16qi): Likewise.
23895         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23896         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23897         (*vsx_le_perm_store_v8hi): Likewise.
23898         (*vsx_le_perm_store_v16qi): Likewise.
23899         (eight unnamed splitters): Likewise.
23900
23901 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
23902
23903         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23904         * config/rs6000/emmintrin.h: Likewise.
23905         * config/rs6000/mmintrin.h: Likewise.
23906         * config/rs6000/xmmintrin.h: Likewise.
23907
23908 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
23909
23910         PR c++/43486
23911         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23912         "public_flag".
23913         * tree.c (tree_nop_conversion): Return true for location wrapper
23914         nodes.
23915         (maybe_wrap_with_location): New function.
23916         (selftest::check_strip_nops): New function.
23917         (selftest::test_location_wrappers): New function.
23918         (selftest::tree_c_tests): Call it.
23919         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23920         (maybe_wrap_with_location): New decl.
23921         (EXPR_LOCATION_WRAPPER_P): New macro.
23922         (location_wrapper_p): New inline function.
23923         (tree_strip_any_location_wrapper): New inline function.
23924
23925 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
23926
23927         PR target/83735
23928         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23929         stack_realign_offset for the largest alignment of stack slot
23930         actually used.
23931         (ix86_find_max_used_stack_alignment): New function.
23932         (ix86_finalize_stack_frame_flags): Use it.  Set
23933         max_used_stack_alignment if we don't realign stack.
23934         * config/i386/i386.h (machine_function): Add
23935         max_used_stack_alignment.
23936
23937 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
23938
23939         * config/arm/arm.opt (-mbranch-cost): New option.
23940         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23941         account.
23942
23943 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
23944
23945         PR target/83629
23946         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23947         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23948
23949 2018-01-10  Richard Biener  <rguenther@suse.de>
23950
23951         PR debug/83765
23952         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23953         early out so it also covers the case where we have a non-NULL
23954         origin.
23955
23956 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
23957
23958         PR tree-optimization/83753
23959         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23960         for non-strided grouped accesses if the number of elements is 1.
23961
23962 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23963
23964         PR target/81616
23965         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23966         * i386.h (TARGET_USE_GATHER): Define.
23967         * x86-tune.def (X86_TUNE_USE_GATHER): New.
23968
23969 2018-01-10  Martin Liska  <mliska@suse.cz>
23970
23971         PR bootstrap/82831
23972         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23973         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23974         partitioning.
23975         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23976         CLEANUP_NO_PARTITIONING is not set.
23977
23978 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
23979
23980         * doc/rtl.texi: Remove documentation of (const ...) wrappers
23981         for vectors, as a partial revert of r254296.
23982         * rtl.h (const_vec_p): Delete.
23983         (const_vec_duplicate_p): Don't test for vector CONSTs.
23984         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23985         * expmed.c (make_tree): Likewise.
23986
23987         Revert:
23988         * common.md (E, F): Use CONSTANT_P instead of checking for
23989         CONST_VECTOR.
23990         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23991         checking for CONST_VECTOR.
23992
23993 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
23994
23995         PR middle-end/83575
23996         * predict.c (force_edge_cold): Handle in more sane way edges
23997         with no prediction.
23998
23999 2018-01-09  Carl Love  <cel@us.ibm.com>
24000
24001         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
24002         V4SI, V4SF types.
24003         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
24004         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
24005         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
24006         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
24007         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
24008         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
24009         * config/rs6000/rs6000-protos.h: Add extern defition for
24010         rs6000_generate_float2_double_code.
24011         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
24012         function.
24013         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
24014         (float2_v2df): Add define_expand.
24015
24016 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
24017
24018         PR target/83628
24019         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
24020         op_mode in the force_to_mode call.
24021
24022 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
24023
24024         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
24025         instead of checking each element individually.
24026         (aarch64_evpc_uzp): Likewise.
24027         (aarch64_evpc_zip): Likewise.
24028         (aarch64_evpc_ext): Likewise.
24029         (aarch64_evpc_rev): Likewise.
24030         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24031         instead of checking each element individually.  Return true without
24032         generating rtl if
24033         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
24034         whether all selected elements come from the same input, instead of
24035         checking each element individually.  Remove calls to gen_rtx_REG,
24036         start_sequence and end_sequence and instead assert that no rtl is
24037         generated.
24038
24039 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
24040
24041         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24042         order of HIGH and CONST checks.
24043
24044 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
24045
24046         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24047         if the destination isn't an SSA_NAME.
24048
24049 2018-01-09  Richard Biener  <rguenther@suse.de>
24050
24051         PR tree-optimization/83668
24052         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24053         move prologue...
24054         (canonicalize_loop_form): ... here, renamed from ...
24055         (canonicalize_loop_closed_ssa_form): ... this and amended to
24056         swap successor edges for loop exit blocks to make us use
24057         the RPO order we need for initial schedule generation.
24058
24059 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
24060
24061         PR tree-optimization/64811
24062         * match.pd: When optimizing comparisons with Inf, avoid
24063         introducing or losing exceptions from comparisons with NaN.
24064
24065 2018-01-09  Martin Liska  <mliska@suse.cz>
24066
24067         PR sanitizer/82517
24068         * asan.c (shadow_mem_size): Add gcc_assert.
24069
24070 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
24071
24072         Don't save registers in main().
24073
24074         PR target/83738
24075         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24076         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24077         * config/avr/avr.c (avr_set_current_function): Don't error if
24078         naked, OS_task or OS_main are specified at the same time.
24079         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24080         OS_main.
24081         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24082         attribute.
24083         * common/config/avr/avr-common.c (avr_option_optimization_table):
24084         Switch on -mmain-is-OS_task for optimizing compilations.
24085
24086 2018-01-09  Richard Biener  <rguenther@suse.de>
24087
24088         PR tree-optimization/83572
24089         * graphite.c: Include cfganal.h.
24090         (graphite_transform_loops): Connect infinite loops to exit
24091         and remove fake edges at the end.
24092
24093 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
24094
24095         * ipa-inline.c (edge_badness): Revert accidental checkin.
24096
24097 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
24098
24099         PR ipa/80763
24100         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24101         symbols; not inline clones.
24102
24103 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
24104
24105         PR target/83507
24106         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24107         hard registers.  Formatting fixes.
24108
24109         PR preprocessor/83722
24110         * gcc.c (try_generate_repro): Pass
24111         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24112         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24113         do_report_bug.
24114
24115 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
24116             Kito Cheng  <kito.cheng@gmail.com>
24117
24118         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24119         (riscv_leaf_function_p): Delete.
24120         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24121
24122 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24123
24124         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24125         function.
24126         (do_ifelse): New function.
24127         (do_isel): New function.
24128         (do_sub3): New function.
24129         (do_add3): New function.
24130         (do_load_mask_compare): New function.
24131         (do_overlap_load_compare): New function.
24132         (expand_compare_loop): New function.
24133         (expand_block_compare): Call expand_compare_loop() when appropriate.
24134         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24135         option description.
24136         (-mblock-compare-inline-loop-limit): New option.
24137
24138 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24139
24140         PR target/83677
24141         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24142         Reverse order of second and third operands in first alternative.
24143         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24144         of first and second elements in UNSPEC_VPERMR vector.
24145         (altivec_expand_vec_perm_le): Likewise.
24146
24147 2018-01-08  Jeff Law  <law@redhat.com>
24148
24149         PR rtl-optimizatin/81308
24150         * tree-switch-conversion.c (cfg_altered): New file scoped static.
24151         (process_switch): If group_case_labels makes a change, then set
24152         cfg_altered.
24153         (pass_convert_switch::execute): If a switch is converted, then
24154         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
24155
24156         PR rtl-optimization/81308
24157         * recog.c (split_all_insns): Conditionally cleanup the CFG after
24158         splitting insns.
24159
24160 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
24161
24162         PR target/83663 - Revert r255946
24163         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24164         generation for cases where splatting a value is not useful.
24165         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24166         across a vec_duplicate and a paradoxical subreg forming a vector
24167         mode to a vec_concat.
24168
24169 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24170
24171         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24172         -march=armv8.3-a variants.
24173         * config/arm/t-multilib: Likewise.
24174         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
24175
24176 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24177
24178         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24179         to generate rtl.
24180         (cceq_ior_compare_complement): Give it a name so I can use it, and
24181         change boolean_or_operator predicate to boolean_operator so it can
24182         be used to generate a crand.
24183         (eqne): New code iterator.
24184         (bd/bd_neg): New code_attrs.
24185         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24186         a single define_insn.
24187         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24188         decrement (bdnzt/bdnzf/bdzt/bdzf).
24189         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24190         with the new names of the branch decrement patterns, and added the
24191         names of the branch decrement conditional patterns.
24192
24193 2018-01-08  Richard Biener  <rguenther@suse.de>
24194
24195         PR tree-optimization/83563
24196         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24197         cache.
24198
24199 2018-01-08  Richard Biener  <rguenther@suse.de>
24200
24201         PR middle-end/83713
24202         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24203
24204 2018-01-08  Richard Biener  <rguenther@suse.de>
24205
24206         PR tree-optimization/83685
24207         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24208         references to abnormals.
24209
24210 2018-01-08  Richard Biener  <rguenther@suse.de>
24211
24212         PR lto/83719
24213         * dwarf2out.c (output_indirect_strings): Handle empty
24214         skeleton_debug_str_hash.
24215         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24216
24217 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
24218
24219         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24220         (emit_store_direct): Likewise.
24221         (arc_trampoline_adjust_address): Likewise.
24222         (arc_asm_trampoline_template): New function.
24223         (arc_initialize_trampoline): Use asm_trampoline_template.
24224         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24225         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24226         * config/arc/arc.md (flush_icache): Delete pattern.
24227
24228 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
24229
24230         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24231         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24232         munaligned-access.
24233
24234 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24235
24236         PR target/83681
24237         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
24238         by not USED_FOR_TARGET.
24239         (make_pass_resolve_sw_modes): Likewise.
24240
24241 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24242
24243         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
24244         USED_FOR_TARGET.
24245
24246 2018-01-08  Richard Biener  <rguenther@suse.de>
24247
24248         PR middle-end/83580
24249         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
24250
24251 2018-01-08  Richard Biener  <rguenther@suse.de>
24252
24253         PR middle-end/83517
24254         * match.pd ((t * 2) / 2) -> t): Add missing :c.
24255
24256 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
24257
24258         PR middle-end/81897
24259         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
24260         basic blocks with a small number of successors.
24261         (convert_control_dep_chain_into_preds): Improve handling of
24262         forwarder blocks.
24263         (dump_predicates): Split apart into...
24264         (dump_pred_chain): ...here...
24265         (dump_pred_info): ...and here.
24266         (can_one_predicate_be_invalidated_p): Add debugging printfs.
24267         (can_chain_union_be_invalidated_p): Improve check for invalidation
24268         of paths.
24269         (uninit_uses_cannot_happen): Avoid unnecessary if
24270         convert_control_dep_chain_into_preds yielded nothing.
24271
24272 2018-01-06  Martin Sebor  <msebor@redhat.com>
24273
24274         PR tree-optimization/83640
24275         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
24276         subtracting negative offset from size.
24277         (builtin_access::overlap): Adjust offset bounds of the access to fall
24278         within the size of the object if possible.
24279
24280 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
24281
24282         PR rtl-optimization/83699
24283         * expmed.c (extract_bit_field_1): Restrict the vector usage of
24284         extract_bit_field_as_subreg to cases in which the extracted
24285         value is also a vector.
24286
24287         * lra-constraints.c (process_alt_operands): Test for the equivalence
24288         substitutions when detecting a possible reload cycle.
24289
24290 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
24291
24292         PR debug/83480
24293         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
24294         by default if flag_selective_schedling{,2}.  Formatting fixes.
24295
24296         PR rtl-optimization/83682
24297         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
24298         if it has non-VECTOR_MODE element mode.
24299         (vec_duplicate_p): Likewise.
24300
24301         PR middle-end/83694
24302         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24303         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24304
24305 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
24306
24307         PR target/83604
24308         * config/i386/i386-builtin.def
24309         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24310         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24311         Require also OPTION_MASK_ISA_AVX512F in addition to
24312         OPTION_MASK_ISA_GFNI.
24313         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24314         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24315         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24316         to OPTION_MASK_ISA_GFNI.
24317         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24318         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24319         OPTION_MASK_ISA_AVX512BW.
24320         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24321         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24322         addition to OPTION_MASK_ISA_GFNI.
24323         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24324         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24325         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24326         to OPTION_MASK_ISA_GFNI.
24327         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24328         a requirement for all ISAs rather than any of them with a few
24329         exceptions.
24330         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24331         processing.
24332         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24333         bitmasks to be enabled with 3 exceptions, instead of requiring any
24334         enabled ISA with lots of exceptions.
24335         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24336         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24337         Change avx512bw in isa attribute to avx512f.
24338         * config/i386/sgxintrin.h: Add license boilerplate.
24339         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
24340         to __AVX512F__ and __AVX512VL to __AVX512VL__.
24341         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24342         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24343         defined.
24344         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24345         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24346         temporarily sse2 rather than sse if not enabled already.
24347
24348         PR target/83604
24349         * config/i386/sse.md (VI248_VLBW): Rename to ...
24350         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
24351         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24352         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24353         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24354         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24355         mode iterator instead of VI248_VLBW.
24356
24357 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
24358
24359         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24360         (record_modified): Skip clobbers; add debug output.
24361         (param_change_prob): Use sreal frequencies.
24362
24363 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
24364
24365         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24366         punt for user-aligned variables.
24367
24368 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
24369
24370         * tree-chrec.c (chrec_contains_symbols): Return true for
24371         POLY_INT_CST.
24372
24373 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
24374
24375         PR target/82439
24376         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24377         of (x|y) == x for BICS pattern.
24378
24379 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
24380
24381         PR tree-optimization/83605
24382         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24383         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24384         can throw.
24385
24386 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24387
24388         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24389         * config/epiphany/rtems.h: New file.
24390
24391 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24392             Uros Bizjak  <ubizjak@gmail.com>
24393
24394         PR target/83554
24395         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24396         QIreg_operand instead of register_operand predicate.
24397         * config/i386/i386.c (ix86_rop_should_change_byte_p,
24398         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24399         comments instead of -fmitigate[-_]rop.
24400
24401 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24402
24403         PR bootstrap/81926
24404         * cgraphunit.c (symbol_table::compile): Switch to text_section
24405         before calling assembly_start debug hook.
24406         * run-rtl-passes.c (run_rtl_passes): Likewise.
24407         Include output.h.
24408
24409 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24410
24411         * tree-vrp.c (extract_range_from_binary_expr_1): Check
24412         range_int_cst_p rather than !symbolic_range_p before calling
24413         extract_range_from_multiplicative_op_1.
24414
24415 2018-01-04  Jeff Law  <law@redhat.com>
24416
24417         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24418         redundant test in assertion.
24419
24420 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24421
24422         * doc/rtl.texi: Document machine_mode wrapper classes.
24423
24424 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24425
24426         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24427         using tree_to_uhwi.
24428
24429 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24430
24431         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24432         the VEC_PERM_EXPR fold to fail.
24433
24434 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24435
24436         PR debug/83585
24437         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24438         to switched_sections.
24439
24440 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24441
24442         PR target/83680
24443         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24444         test for d.testing.
24445
24446 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
24447
24448         PR target/83387
24449         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24450         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24451
24452 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24453
24454         PR debug/83666
24455         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24456         is BLKmode and bitpos not zero or mode change is needed.
24457
24458 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24459
24460         PR target/83675
24461         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24462         TARGET_VIS2.
24463
24464 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
24465
24466         PR target/83628
24467         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24468         instead of MULT rtx.  Update all corresponding splitters.
24469         (*saddl_se): Ditto.
24470         (*ssub<modesuffix>): Ditto.
24471         (*ssubl_se): Ditto.
24472         (*cmp_sadd_di): Update split patterns.
24473         (*cmp_sadd_si): Ditto.
24474         (*cmp_sadd_sidi): Ditto.
24475         (*cmp_ssub_di): Ditto.
24476         (*cmp_ssub_si): Ditto.
24477         (*cmp_ssub_sidi): Ditto.
24478         * config/alpha/predicates.md (const23_operand): New predicate.
24479         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24480         Look for ASHIFT, not MULT inner operand.
24481         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24482
24483 2018-01-04  Martin Liska  <mliska@suse.cz>
24484
24485         PR gcov-profile/83669
24486         * gcov.c (output_intermediate_file): Add version to intermediate
24487         gcov file.
24488         * doc/gcov.texi: Document new field 'version' in intermediate
24489         file format. Fix location of '-k' option of gcov command.
24490
24491 2018-01-04  Martin Liska  <mliska@suse.cz>
24492
24493         PR ipa/82352
24494         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24495
24496 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24497
24498         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24499
24500 2018-01-03  Martin Sebor  <msebor@redhat.com>
24501
24502         PR tree-optimization/83655
24503         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24504         checking calls with invalid arguments.
24505
24506 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24507
24508         * tree-vect-stmts.c (vect_get_store_rhs): New function.
24509         (vectorizable_mask_load_store): Delete.
24510         (vectorizable_call): Return false for masked loads and stores.
24511         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
24512         instead of gimple_assign_rhs1.
24513         (vectorizable_load): Handle IFN_MASK_LOAD.
24514         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24515
24516 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24517
24518         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24519         split out from..,
24520         (vectorizable_mask_load_store): ...here.
24521         (vectorizable_load): ...and here.
24522
24523 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24524
24525         * tree-vect-stmts.c (vect_build_all_ones_mask)
24526         (vect_build_zero_merge_argument): New functions, split out from...
24527         (vectorizable_load): ...here.
24528
24529 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24530
24531         * tree-vect-stmts.c (vect_check_store_rhs): New function,
24532         split out from...
24533         (vectorizable_mask_load_store): ...here.
24534         (vectorizable_store): ...and here.
24535
24536 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24537
24538         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24539         split out from...
24540         (vectorizable_mask_load_store): ...here.
24541
24542 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24543
24544         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24545         (vect_model_store_cost): Take a vec_load_store_type instead of a
24546         vect_def_type.
24547         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24548         (vect_model_store_cost): Take a vec_load_store_type instead of a
24549         vect_def_type.
24550         (vectorizable_mask_load_store): Update accordingly.
24551         (vectorizable_store): Likewise.
24552         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24553
24554 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24555
24556         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24557         IFN_MASK_LOAD calls here rather than...
24558         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24559
24560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24561             Alan Hayward  <alan.hayward@arm.com>
24562             David Sherwood  <david.sherwood@arm.com>
24563
24564         * expmed.c (extract_bit_field_1): For vector extracts,
24565         fall back to extract_bit_field_as_subreg if vec_extract
24566         isn't available.
24567
24568 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24569             Alan Hayward  <alan.hayward@arm.com>
24570             David Sherwood  <david.sherwood@arm.com>
24571
24572         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24573         they are variable or constant sized.
24574         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24575         slots for constant-sized data.
24576
24577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24578             Alan Hayward  <alan.hayward@arm.com>
24579             David Sherwood  <david.sherwood@arm.com>
24580
24581         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24582         handling COND_EXPRs with boolean comparisons, try to find a better
24583         basis for the mask type than the boolean itself.
24584
24585 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24586
24587         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24588         is calculated and how it can be overridden.
24589         * genmodes.c (max_bitsize_mode_any_mode): New variable.
24590         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24591         if defined.
24592         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24593         if nonzero.
24594
24595 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24596             Alan Hayward  <alan.hayward@arm.com>
24597             David Sherwood  <david.sherwood@arm.com>
24598
24599         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24600         Remove the mode argument.
24601         (aarch64_simd_valid_immediate): Remove the mode and inverse
24602         arguments.
24603         * config/aarch64/iterators.md (bitsize): New iterator.
24604         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24605         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24606         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24607         aarch64_simd_valid_immediate.
24608         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24609         (aarch64_reg_or_bic_imm): Likewise.
24610         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24611         with an insn_type enum and msl with a modifier_type enum.
24612         Replace element_width with a scalar_mode.  Change the shift
24613         to unsigned int.  Add constructors for scalar_float_mode and
24614         scalar_int_mode elements.
24615         (aarch64_vect_float_const_representable_p): Delete.
24616         (aarch64_can_const_movi_rtx_p)
24617         (aarch64_simd_scalar_immediate_valid_for_move)
24618         (aarch64_simd_make_constant): Update call to
24619         aarch64_simd_valid_immediate.
24620         (aarch64_advsimd_valid_immediate_hs): New function.
24621         (aarch64_advsimd_valid_immediate): Likewise.
24622         (aarch64_simd_valid_immediate): Remove mode and inverse
24623         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
24624         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24625         and aarch64_float_const_representable_p on the result.
24626         (aarch64_output_simd_mov_immediate): Remove mode argument.
24627         Update call to aarch64_simd_valid_immediate and use of
24628         simd_immediate_info.
24629         (aarch64_output_scalar_simd_mov_immediate): Update call
24630         accordingly.
24631
24632 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24633             Alan Hayward  <alan.hayward@arm.com>
24634             David Sherwood  <david.sherwood@arm.com>
24635
24636         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24637         (mode_nunits): Likewise CONST_MODE_NUNITS.
24638         * machmode.def (ADJUST_NUNITS): Document.
24639         * genmodes.c (mode_data::need_nunits_adj): New field.
24640         (blank_mode): Update accordingly.
24641         (adj_nunits): New variable.
24642         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24643         parameter.
24644         (emit_mode_size_inline): Set need_bytesize_adj for all modes
24645         listed in adj_nunits.
24646         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24647         listed in adj_nunits.  Don't emit case statements for such modes.
24648         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24649         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
24650         nothing if adj_nunits is nonnull.
24651         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24652         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24653         (emit_mode_fbit): Update use of print_maybe_const_decl.
24654         (emit_move_size): Likewise.  Treat the array as non-const
24655         if adj_nunits.
24656         (emit_mode_adjustments): Handle adj_nunits.
24657
24658 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24659
24660         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24661         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24662         (VECTOR_MODES): Use it.
24663         (make_vector_modes): Take the prefix as an argument.
24664
24665 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24666             Alan Hayward  <alan.hayward@arm.com>
24667             David Sherwood  <david.sherwood@arm.com>
24668
24669         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24670         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24671         for MODE_VECTOR_BOOL.
24672         * machmode.def (VECTOR_BOOL_MODE): Document.
24673         * genmodes.c (VECTOR_BOOL_MODE): New macro.
24674         (make_vector_bool_mode): New function.
24675         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24676         MODE_VECTOR_BOOL.
24677         * lto-streamer-in.c (lto_input_mode_table): Likewise.
24678         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24679         Likewise.
24680         * stor-layout.c (int_mode_for_mode): Likewise.
24681         * tree.c (build_vector_type_for_mode): Likewise.
24682         * varasm.c (output_constant_pool_2): Likewise.
24683         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24684         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
24685         for MODE_VECTOR_BOOL.
24686         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24687         of mode class checks.
24688         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24689         instead of a list of mode class checks.
24690         (expand_vector_scalar_condition): Likewise.
24691         (type_for_widest_vector_mode): Handle BImode as an inner mode.
24692
24693 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24694             Alan Hayward  <alan.hayward@arm.com>
24695             David Sherwood  <david.sherwood@arm.com>
24696
24697         * machmode.h (mode_size): Change from unsigned short to
24698         poly_uint16_pod.
24699         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24700         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24701         or if measurement_type is not polynomial.
24702         (fixed_size_mode::includes_p): Check for constant-sized modes.
24703         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24704         return a poly_uint16 rather than an unsigned short.
24705         (emit_mode_size): Change the type of mode_size from unsigned short
24706         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
24707         (emit_mode_adjustments): Cope with polynomial vector sizes.
24708         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24709         for GET_MODE_SIZE.
24710         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24711         for GET_MODE_SIZE.
24712         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24713         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24714         * caller-save.c (setup_save_areas): Likewise.
24715         (replace_reg_with_saved_mem): Likewise.
24716         * calls.c (emit_library_call_value_1): Likewise.
24717         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24718         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24719         (gen_lowpart_for_combine): Likewise.
24720         * convert.c (convert_to_integer_1): Likewise.
24721         * cse.c (equiv_constant, cse_insn): Likewise.
24722         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24723         (cselib_subst_to_values): Likewise.
24724         * dce.c (word_dce_process_block): Likewise.
24725         * df-problems.c (df_word_lr_mark_ref): Likewise.
24726         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24727         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24728         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24729         (rtl_for_decl_location): Likewise.
24730         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24731         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24732         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24733         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24734         (expand_expr_real_1): Likewise.
24735         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24736         (pad_below): Likewise.
24737         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24738         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24739         * ira.c (get_subreg_tracking_sizes): Likewise.
24740         * ira-build.c (ira_create_allocno_objects): Likewise.
24741         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24742         (ira_sort_regnos_for_alter_reg): Likewise.
24743         * ira-costs.c (record_operand_costs): Likewise.
24744         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24745         (resolve_simple_move): Likewise.
24746         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24747         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24748         (lra_constraints): Likewise.
24749         (CONST_POOL_OK_P): Reject variable-sized modes.
24750         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24751         (add_pseudo_to_slot, lra_spill): Likewise.
24752         * omp-low.c (omp_clause_aligned_alignment): Likewise.
24753         * optabs-query.c (get_best_extraction_insn): Likewise.
24754         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24755         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24756         (expand_mult_highpart, valid_multiword_target_p): Likewise.
24757         * recog.c (offsettable_address_addr_space_p): Likewise.
24758         * regcprop.c (maybe_mode_change): Likewise.
24759         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24760         * regrename.c (build_def_use): Likewise.
24761         * regstat.c (dump_reg_info): Likewise.
24762         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24763         (find_reloads, find_reloads_subreg_address): Likewise.
24764         * reload1.c (eliminate_regs_1): Likewise.
24765         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24766         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24767         (simplify_binary_operation_1, simplify_subreg): Likewise.
24768         * targhooks.c (default_function_arg_padding): Likewise.
24769         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24770         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24771         (verify_gimple_assign_ternary): Likewise.
24772         * tree-inline.c (estimate_move_cost): Likewise.
24773         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24774         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24775         (get_address_cost_ainc): Likewise.
24776         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24777         (vect_supportable_dr_alignment): Likewise.
24778         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24779         (vectorizable_reduction): Likewise.
24780         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24781         (vectorizable_operation, vectorizable_load): Likewise.
24782         * tree.c (build_same_sized_truth_vector_type): Likewise.
24783         * valtrack.c (cleanup_auto_inc_dec): Likewise.
24784         * var-tracking.c (emit_note_insn_var_location): Likewise.
24785         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24786         (ADDR_VEC_ALIGN): Likewise.
24787
24788 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24789             Alan Hayward  <alan.hayward@arm.com>
24790             David Sherwood  <david.sherwood@arm.com>
24791
24792         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24793         unsigned short.
24794         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24795         or if measurement_type is polynomial.
24796         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24797         * combine.c (make_extraction): Likewise.
24798         * dse.c (find_shift_sequence): Likewise.
24799         * dwarf2out.c (mem_loc_descriptor): Likewise.
24800         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24801         (extract_bit_field, extract_low_bits): Likewise.
24802         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24803         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24804         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24805         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24806         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24807         * reload.c (find_reloads): Likewise.
24808         * reload1.c (alter_reg): Likewise.
24809         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24810         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24811         * tree-if-conv.c (predicate_mem_writes): Likewise.
24812         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24813         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24814         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24815         * valtrack.c (dead_debug_insert_temp): Likewise.
24816         * varasm.c (mergeable_constant_section): Likewise.
24817         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24818
24819 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24820             Alan Hayward  <alan.hayward@arm.com>
24821             David Sherwood  <david.sherwood@arm.com>
24822
24823         * expr.c (expand_assignment): Cope with polynomial mode sizes
24824         when assigning to a CONCAT.
24825
24826 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24827             Alan Hayward  <alan.hayward@arm.com>
24828             David Sherwood  <david.sherwood@arm.com>
24829
24830         * machmode.h (mode_precision): Change from unsigned short to
24831         poly_uint16_pod.
24832         (mode_to_precision): Return a poly_uint16 rather than an unsigned
24833         short.
24834         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24835         or if measurement_type is not polynomial.
24836         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
24837         in which the mode is already known to be a scalar_int_mode.
24838         * genmodes.c (emit_mode_precision): Change the type of mode_precision
24839         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
24840         initializer.
24841         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24842         for GET_MODE_PRECISION.
24843         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24844         for GET_MODE_PRECISION.
24845         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24846         as polynomial.
24847         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24848         (expand_field_assignment, make_extraction): Likewise.
24849         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24850         (get_last_value): Likewise.
24851         * convert.c (convert_to_integer_1): Likewise.
24852         * cse.c (cse_insn): Likewise.
24853         * expr.c (expand_expr_real_1): Likewise.
24854         * lra-constraints.c (simplify_operand_subreg): Likewise.
24855         * optabs-query.c (can_atomic_load_p): Likewise.
24856         * optabs.c (expand_atomic_load): Likewise.
24857         (expand_atomic_store): Likewise.
24858         * ree.c (combine_reaching_defs): Likewise.
24859         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24860         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24861         * tree.h (type_has_mode_precision_p): Likewise.
24862         * ubsan.c (instrument_si_overflow): Likewise.
24863
24864 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24865             Alan Hayward  <alan.hayward@arm.com>
24866             David Sherwood  <david.sherwood@arm.com>
24867
24868         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24869         polynomial numbers of units.
24870         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24871         (valid_vector_subparts_p): New function.
24872         (build_vector_type): Remove temporary shim and take the number
24873         of units as a poly_uint64 rather than an int.
24874         (build_opaque_vector_type): Take the number of units as a
24875         poly_uint64 rather than an int.
24876         * tree.c (build_vector_from_ctor): Handle polynomial
24877         TYPE_VECTOR_SUBPARTS.
24878         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24879         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24880         (build_vector_from_val): If the number of units is variable,
24881         use build_vec_duplicate_cst for constant operands and
24882         VEC_DUPLICATE_EXPR otherwise.
24883         (make_vector_type): Remove temporary is_constant ().
24884         (build_vector_type, build_opaque_vector_type): Take the number of
24885         units as a poly_uint64 rather than an int.
24886         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24887         VECTOR_CST_NELTS.
24888         * cfgexpand.c (expand_debug_expr): Likewise.
24889         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24890         (store_constructor, expand_expr_real_1): Likewise.
24891         (const_scalar_mask_from_tree): Likewise.
24892         * fold-const-call.c (fold_const_reduction): Likewise.
24893         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24894         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24895         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24896         (fold_relational_const): Likewise.
24897         (native_interpret_vector): Likewise.  Change the size from an
24898         int to an unsigned int.
24899         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24900         TYPE_VECTOR_SUBPARTS.
24901         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24902         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24903         duplicating a non-constant operand into a variable-length vector.
24904         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24905         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24906         * ipa-icf.c (sem_variable::equals): Likewise.
24907         * match.pd: Likewise.
24908         * omp-simd-clone.c (simd_clone_subparts): Likewise.
24909         * print-tree.c (print_node): Likewise.
24910         * stor-layout.c (layout_type): Likewise.
24911         * targhooks.c (default_builtin_vectorization_cost): Likewise.
24912         * tree-cfg.c (verify_gimple_comparison): Likewise.
24913         (verify_gimple_assign_binary): Likewise.
24914         (verify_gimple_assign_ternary): Likewise.
24915         (verify_gimple_assign_single): Likewise.
24916         * tree-pretty-print.c (dump_generic_node): Likewise.
24917         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24918         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24919         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24920         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24921         (vect_shift_permute_load_chain): Likewise.
24922         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24923         (expand_vector_condition, optimize_vector_constructor): Likewise.
24924         (lower_vec_perm, get_compute_type): Likewise.
24925         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24926         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24927         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24928         (vect_recog_mask_conversion_pattern): Likewise.
24929         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24930         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24931         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24932         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24933         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24934         (vectorizable_shift, vectorizable_operation, vectorizable_store)
24935         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24936         (supportable_widening_operation): Likewise.
24937         (supportable_narrowing_operation): Likewise.
24938         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24939         Likewise.
24940         * varasm.c (output_constant): Likewise.
24941
24942 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24943             Alan Hayward  <alan.hayward@arm.com>
24944             David Sherwood  <david.sherwood@arm.com>
24945
24946         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24947         so that both the length == 3 and length != 3 cases set up their
24948         own permute vectors.  Add comments explaining why we know the
24949         number of elements is constant.
24950         (vect_permute_load_chain): Likewise.
24951
24952 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24953             Alan Hayward  <alan.hayward@arm.com>
24954             David Sherwood  <david.sherwood@arm.com>
24955
24956         * machmode.h (mode_nunits): Change from unsigned char to
24957         poly_uint16_pod.
24958         (ONLY_FIXED_SIZE_MODES): New macro.
24959         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24960         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24961         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24962         New typedefs.
24963         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24964         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24965         or if measurement_type is not polynomial.
24966         * genmodes.c (ZERO_COEFFS): New macro.
24967         (emit_mode_nunits_inline): Make mode_nunits_inline return a
24968         poly_uint16.
24969         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24970         Use ZERO_COEFFS when emitting initializers.
24971         * data-streamer.h (bp_pack_poly_value): New function.
24972         (bp_unpack_poly_value): Likewise.
24973         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24974         for GET_MODE_NUNITS.
24975         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24976         for GET_MODE_NUNITS.
24977         * tree.c (make_vector_type): Remove temporary shim and make
24978         the real function take the number of units as a poly_uint64
24979         rather than an int.
24980         (build_vector_type_for_mode): Handle polynomial nunits.
24981         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24982         * emit-rtl.c (const_vec_series_p_1): Likewise.
24983         (gen_rtx_CONST_VECTOR): Likewise.
24984         * fold-const.c (test_vec_duplicate_folding): Likewise.
24985         * genrecog.c (validate_pattern): Likewise.
24986         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24987         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24988         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24989         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24990         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24991         * rtlanal.c (subreg_get_info): Likewise.
24992         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24993         (vect_grouped_load_supported): Likewise.
24994         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24995         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24996         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24997         (simplify_const_unary_operation, simplify_binary_operation_1)
24998         (simplify_const_binary_operation, simplify_ternary_operation)
24999         (test_vector_ops_duplicate, test_vector_ops): Likewise.
25000         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
25001         instead of CONST_VECTOR_NUNITS.
25002         * varasm.c (output_constant_pool_2): Likewise.
25003         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
25004         explicit-encoded elements in the XVEC for variable-length vectors.
25005
25006 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25007
25008         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
25009
25010 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25011             Alan Hayward  <alan.hayward@arm.com>
25012             David Sherwood  <david.sherwood@arm.com>
25013
25014         * coretypes.h (fixed_size_mode): Declare.
25015         (fixed_size_mode_pod): New typedef.
25016         * builtins.h (target_builtins::x_apply_args_mode)
25017         (target_builtins::x_apply_result_mode): Change type to
25018         fixed_size_mode_pod.
25019         * builtins.c (apply_args_size, apply_result_size, result_vector)
25020         (expand_builtin_apply_args_1, expand_builtin_apply)
25021         (expand_builtin_return): Update accordingly.
25022
25023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25024
25025         * cse.c (hash_rtx_cb): Hash only the encoded elements.
25026         * cselib.c (cselib_hash_rtx): Likewise.
25027         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25028         CONST_VECTOR encoding.
25029
25030 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25031             Jeff Law  <law@redhat.com>
25032
25033         PR target/83641
25034         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25035         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25036         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25037         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25038
25039         PR target/83641
25040         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25041         explicitly probe *sp in a noreturn function if there were any callee
25042         register saves or frame pointer is needed.
25043
25044 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25045
25046         PR debug/83621
25047         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25048         BLKmode for ternary, binary or unary expressions.
25049
25050         PR debug/83645
25051         * var-tracking.c (delete_vta_debug_insn): New inline function.
25052         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25053         insns from get_insns () to NULL instead of each bb separately.
25054         Use delete_vta_debug_insn.  No longer static.
25055         (vt_debug_insns_local, variable_tracking_main_1): Adjust
25056         delete_vta_debug_insns callers.
25057         * rtl.h (delete_vta_debug_insns): Declare.
25058         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25059         instead of variable_tracking_main.
25060
25061 2018-01-03  Martin Sebor  <msebor@redhat.com>
25062
25063         PR tree-optimization/83603
25064         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25065         arguments past the endof the argument list in functions declared
25066         without a prototype.
25067         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25068         Avoid checking when arguments are null.
25069
25070 2018-01-03  Martin Sebor  <msebor@redhat.com>
25071
25072         PR c/83559
25073         * doc/extend.texi (attribute const): Fix a typo.
25074         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25075         issuing -Wsuggest-attribute for void functions.
25076
25077 2018-01-03  Martin Sebor  <msebor@redhat.com>
25078
25079         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25080         offset_int::from instead of wide_int::to_shwi.
25081         (maybe_diag_overlap): Remove assertion.
25082         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25083         * gimple-ssa-sprintf.c (format_directive): Same.
25084         (parse_directive): Same.
25085         (sprintf_dom_walker::compute_format_length): Same.
25086         (try_substitute_return_value): Same.
25087
25088 2018-01-03  Jeff Law  <law@redhat.com>
25089
25090         PR middle-end/83654
25091         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25092         non-constant residual for zero at runtime and avoid probing in
25093         that case.  Reorganize code for trailing problem to mirror handling
25094         of the residual.
25095
25096 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
25097
25098         PR tree-optimization/83501
25099         * tree-ssa-strlen.c (get_string_cst): New.
25100         (handle_char_store): Call get_string_cst.
25101
25102 2018-01-03  Martin Liska  <mliska@suse.cz>
25103
25104         PR tree-optimization/83593
25105         * tree-ssa-strlen.c: Include tree-cfg.h.
25106         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25107         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25108         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25109         to false.
25110         (strlen_dom_walker::before_dom_children): Call
25111         gimple_purge_dead_eh_edges. Dump tranformation with details
25112         dump flags.
25113         (strlen_dom_walker::before_dom_children): Update call by adding
25114         new argument cleanup_eh.
25115         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25116
25117 2018-01-03  Martin Liska  <mliska@suse.cz>
25118
25119         PR ipa/83549
25120         * cif-code.def (VARIADIC_THUNK): New enum value.
25121         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25122         thunks.
25123
25124 2018-01-03  Jan Beulich  <jbeulich@suse.com>
25125
25126         * sse.md (mov<mode>_internal): Tighten condition for when to use
25127         vmovdqu<ssescalarsize> for TI and OI modes.
25128
25129 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25130
25131         Update copyright years.
25132
25133 2018-01-03  Martin Liska  <mliska@suse.cz>
25134
25135         PR ipa/83594
25136         * ipa-visibility.c (function_and_variable_visibility): Skip
25137         functions with noipa attribure.
25138
25139 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25140
25141         * gcc.c (process_command): Update copyright notice dates.
25142         * gcov-dump.c (print_version): Ditto.
25143         * gcov.c (print_version): Ditto.
25144         * gcov-tool.c (print_version): Ditto.
25145         * gengtype.c (create_file): Ditto.
25146         * doc/cpp.texi: Bump @copying's copyright year.
25147         * doc/cppinternals.texi: Ditto.
25148         * doc/gcc.texi: Ditto.
25149         * doc/gccint.texi: Ditto.
25150         * doc/gcov.texi: Ditto.
25151         * doc/install.texi: Ditto.
25152         * doc/invoke.texi: Ditto.
25153
25154 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25155
25156         * vector-builder.h (vector_builder::m_full_nelts): Change from
25157         unsigned int to poly_uint64.
25158         (vector_builder::full_nelts): Update prototype accordingly.
25159         (vector_builder::new_vector): Likewise.
25160         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25161         (vector_builder::operator ==): Likewise.
25162         (vector_builder::finalize): Likewise.
25163         * int-vector-builder.h (int_vector_builder::int_vector_builder):
25164         Take the number of elements as a poly_uint64 rather than an
25165         unsigned int.
25166         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25167         from unsigned int to poly_uint64.
25168         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25169         (vec_perm_indices::new_vector): Likewise.
25170         (vec_perm_indices::length): Likewise.
25171         (vec_perm_indices::nelts_per_input): Likewise.
25172         (vec_perm_indices::input_nelts): Likewise.
25173         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25174         number of elements per input as a poly_uint64 rather than an
25175         unsigned int.  Use the original encoding for variable-length
25176         vectors, rather than clamping each individual element.
25177         For the second and subsequent elements in each pattern,
25178         clamp the step and base before clamping their sum.
25179         (vec_perm_indices::series_p): Handle polynomial element counts.
25180         (vec_perm_indices::all_in_range_p): Likewise.
25181         (vec_perm_indices_to_tree): Likewise.
25182         (vec_perm_indices_to_rtx): Likewise.
25183         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25184         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25185         (tree_vector_builder::new_binary_operation): Handle polynomial
25186         element counts.  Return false if we need to know the number
25187         of elements at compile time.
25188         * fold-const.c (fold_vec_perm): Punt if the number of elements
25189         isn't known at compile time.
25190
25191 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25192
25193         * vec-perm-indices.h (vec_perm_builder): Change element type
25194         from HOST_WIDE_INT to poly_int64.
25195         (vec_perm_indices::element_type): Update accordingly.
25196         (vec_perm_indices::clamp): Handle polynomial element_types.
25197         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25198         (vec_perm_indices::all_in_range_p): Likewise.
25199         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25200         than shwi trees.
25201         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25202         polynomial vec_perm_indices element types.
25203         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25204         * fold-const.c (fold_vec_perm): Likewise.
25205         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25206         * tree-vect-generic.c (lower_vec_perm): Likewise.
25207         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25208         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25209         element type to HOST_WIDE_INT.
25210
25211 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25212             Alan Hayward  <alan.hayward@arm.com>
25213             David Sherwood  <david.sherwood@arm.com>
25214
25215         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25216         rather than an int.  Use plus_constant.
25217         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25218         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
25219
25220 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25221             Alan Hayward  <alan.hayward@arm.com>
25222             David Sherwood  <david.sherwood@arm.com>
25223
25224         * calls.c (emit_call_1, expand_call): Change struct_value_size from
25225         a HOST_WIDE_INT to a poly_int64.
25226
25227 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25228             Alan Hayward  <alan.hayward@arm.com>
25229             David Sherwood  <david.sherwood@arm.com>
25230
25231         * calls.c (load_register_parameters): Cope with polynomial
25232         mode sizes.  Require a constant size for BLKmode parameters
25233         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
25234         forces a parameter to be padded at the lsb end in order to
25235         fill a complete number of words, require the parameter size
25236         to be ordered wrt UNITS_PER_WORD.
25237
25238 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25239             Alan Hayward  <alan.hayward@arm.com>
25240             David Sherwood  <david.sherwood@arm.com>
25241
25242         * reload1.c (spill_stack_slot_width): Change element type
25243         from unsigned int to poly_uint64_pod.
25244         (alter_reg): Treat mode sizes as polynomial.
25245
25246 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25247             Alan Hayward  <alan.hayward@arm.com>
25248             David Sherwood  <david.sherwood@arm.com>
25249
25250         * reload.c (complex_word_subreg_p): New function.
25251         (reload_inner_reg_of_subreg, push_reload): Use it.
25252
25253 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25254             Alan Hayward  <alan.hayward@arm.com>
25255             David Sherwood  <david.sherwood@arm.com>
25256
25257         * lra-constraints.c (process_alt_operands): Reject matched
25258         operands whose sizes aren't ordered.
25259         (match_reload): Refer to this check here.
25260
25261 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25262             Alan Hayward  <alan.hayward@arm.com>
25263             David Sherwood  <david.sherwood@arm.com>
25264
25265         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
25266         that the mode size is in the set {1, 2, 4, 8, 16}.
25267
25268 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25269             Alan Hayward  <alan.hayward@arm.com>
25270             David Sherwood  <david.sherwood@arm.com>
25271
25272         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
25273         Use plus_constant instead of gen_rtx_PLUS.
25274
25275 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25276             Alan Hayward  <alan.hayward@arm.com>
25277             David Sherwood  <david.sherwood@arm.com>
25278
25279         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
25280         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
25281         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
25282         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
25283         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
25284         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25285         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
25286         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
25287         * config/i386/i386.c (ix86_push_rounding): ...this new function.
25288         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
25289         a poly_int64.
25290         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
25291         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
25292         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
25293         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
25294         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
25295         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
25296         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
25297         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
25298         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25299         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
25300         function.
25301         * expr.c (emit_move_resolve_push): Treat the input and result
25302         of PUSH_ROUNDING as a poly_int64.
25303         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25304         (emit_push_insn): Likewise.
25305         * lra-eliminations.c (mark_not_eliminable): Likewise.
25306         * recog.c (push_operand): Likewise.
25307         * reload1.c (elimination_effects): Likewise.
25308         * rtlanal.c (nonzero_bits1): Likewise.
25309         * calls.c (store_one_arg): Likewise.  Require the padding to be
25310         known at compile time.
25311
25312 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25313             Alan Hayward  <alan.hayward@arm.com>
25314             David Sherwood  <david.sherwood@arm.com>
25315
25316         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25317         Use plus_constant instead of gen_rtx_PLUS.
25318
25319 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25320             Alan Hayward  <alan.hayward@arm.com>
25321             David Sherwood  <david.sherwood@arm.com>
25322
25323         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25324         rather than an int.
25325
25326 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25327             Alan Hayward  <alan.hayward@arm.com>
25328             David Sherwood  <david.sherwood@arm.com>
25329
25330         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25331         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25332         via stack temporaries.  Treat the mode size as polynomial too.
25333
25334 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25335             Alan Hayward  <alan.hayward@arm.com>
25336             David Sherwood  <david.sherwood@arm.com>
25337
25338         * expr.c (expand_expr_real_2): When handling conversions involving
25339         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25340         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
25341         as a poly_uint64 too.
25342
25343 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25344             Alan Hayward  <alan.hayward@arm.com>
25345             David Sherwood  <david.sherwood@arm.com>
25346
25347         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25348
25349 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25350             Alan Hayward  <alan.hayward@arm.com>
25351             David Sherwood  <david.sherwood@arm.com>
25352
25353         * combine.c (can_change_dest_mode): Handle polynomial
25354         REGMODE_NATURAL_SIZE.
25355         * expmed.c (store_bit_field_1): Likewise.
25356         * expr.c (store_constructor): Likewise.
25357         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25358         and polynomial REGMODE_NATURAL_SIZE.
25359         (gen_lowpart_common): Likewise.
25360         * reginfo.c (record_subregs_of_mode): Likewise.
25361         * rtlanal.c (read_modify_subreg_p): Likewise.
25362
25363 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25364             Alan Hayward  <alan.hayward@arm.com>
25365             David Sherwood  <david.sherwood@arm.com>
25366
25367         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25368         numbers of elements.
25369
25370 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25371             Alan Hayward  <alan.hayward@arm.com>
25372             David Sherwood  <david.sherwood@arm.com>
25373
25374         * match.pd: Cope with polynomial numbers of vector elements.
25375
25376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25377             Alan Hayward  <alan.hayward@arm.com>
25378             David Sherwood  <david.sherwood@arm.com>
25379
25380         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25381         in a POINTER_PLUS_EXPR.
25382
25383 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25384             Alan Hayward  <alan.hayward@arm.com>
25385             David Sherwood  <david.sherwood@arm.com>
25386
25387         * omp-simd-clone.c (simd_clone_subparts): New function.
25388         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25389         (ipa_simd_modify_function_body): Likewise.
25390
25391 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25392             Alan Hayward  <alan.hayward@arm.com>
25393             David Sherwood  <david.sherwood@arm.com>
25394
25395         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25396         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25397         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25398         (expand_vector_condition, vector_element): Likewise.
25399         (subparts_gt): New function.
25400         (get_compute_type): Use subparts_gt.
25401         (count_type_subparts): Delete.
25402         (expand_vector_operations_1): Use subparts_gt instead of
25403         count_type_subparts.
25404
25405 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25406             Alan Hayward  <alan.hayward@arm.com>
25407             David Sherwood  <david.sherwood@arm.com>
25408
25409         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25410         (vect_compile_time_alias): ...this new function.  Do the calculation
25411         on poly_ints rather than trees.
25412         (vect_prune_runtime_alias_test_list): Update call accordingly.
25413
25414 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25415             Alan Hayward  <alan.hayward@arm.com>
25416             David Sherwood  <david.sherwood@arm.com>
25417
25418         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25419         numbers of units.
25420         (vect_schedule_slp_instance): Likewise.
25421
25422 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25423             Alan Hayward  <alan.hayward@arm.com>
25424             David Sherwood  <david.sherwood@arm.com>
25425
25426         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25427         constant and extern definitions for variable-length vectors.
25428         (vect_get_constant_vectors): Note that the number of units
25429         is known to be constant.
25430
25431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25432             Alan Hayward  <alan.hayward@arm.com>
25433             David Sherwood  <david.sherwood@arm.com>
25434
25435         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25436         of units as polynomial.  Choose between WIDE and NARROW based
25437         on multiple_p.
25438
25439 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25440             Alan Hayward  <alan.hayward@arm.com>
25441             David Sherwood  <david.sherwood@arm.com>
25442
25443         * tree-vect-stmts.c (simd_clone_subparts): New function.
25444         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25445
25446 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25447             Alan Hayward  <alan.hayward@arm.com>
25448             David Sherwood  <david.sherwood@arm.com>
25449
25450         * tree-vect-stmts.c (vectorizable_call): Treat the number of
25451         vectors as polynomial.  Use build_index_vector for
25452         IFN_GOMP_SIMD_LANE.
25453
25454 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25455             Alan Hayward  <alan.hayward@arm.com>
25456             David Sherwood  <david.sherwood@arm.com>
25457
25458         * tree-vect-stmts.c (get_load_store_type): Treat the number of
25459         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25460         for variable-length vectors.
25461         (vectorizable_mask_load_store): Treat the number of units as
25462         polynomial, asserting that it is constant if the condition has
25463         already been enforced.
25464         (vectorizable_store, vectorizable_load): Likewise.
25465
25466 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25467             Alan Hayward  <alan.hayward@arm.com>
25468             David Sherwood  <david.sherwood@arm.com>
25469
25470         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25471         of units as polynomial.  Punt if we can't tell at compile time
25472         which vector contains the final result.
25473
25474 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25475             Alan Hayward  <alan.hayward@arm.com>
25476             David Sherwood  <david.sherwood@arm.com>
25477
25478         * tree-vect-loop.c (vectorizable_induction): Treat the number
25479         of units as polynomial.  Punt on SLP inductions.  Use an integer
25480         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
25481         cast of such a series for variable-length floating-point
25482         reductions.
25483
25484 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25485             Alan Hayward  <alan.hayward@arm.com>
25486             David Sherwood  <david.sherwood@arm.com>
25487
25488         * tree.h (build_index_vector): Declare.
25489         * tree.c (build_index_vector): New function.
25490         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25491         of units as polynomial, forcibly converting it to a constant if
25492         vectorizable_reduction has already enforced the condition.
25493         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
25494         to create a {1,2,3,...} vector.
25495         (vectorizable_reduction): Treat the number of units as polynomial.
25496         Choose vectype_in based on the largest scalar element size rather
25497         than the smallest number of units.  Enforce the restrictions
25498         relied on above.
25499
25500 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25501             Alan Hayward  <alan.hayward@arm.com>
25502             David Sherwood  <david.sherwood@arm.com>
25503
25504         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25505         number of units as polynomial.
25506
25507 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25508             Alan Hayward  <alan.hayward@arm.com>
25509             David Sherwood  <david.sherwood@arm.com>
25510
25511         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25512         * target.def (autovectorize_vector_sizes): Return the vector sizes
25513         by pointer, using vector_sizes rather than a bitmask.
25514         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25515         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25516         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25517         Likewise.
25518         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25519         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25520         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25521         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25522         * omp-general.c (omp_max_vf): Likewise.
25523         * omp-low.c (omp_clause_aligned_alignment): Likewise.
25524         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25525         * tree-vect-loop.c (vect_analyze_loop): Likewise.
25526         * tree-vect-slp.c (vect_slp_bb): Likewise.
25527         * doc/tm.texi: Regenerate.
25528         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25529         to a poly_uint64.
25530         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25531         the vector size as a poly_uint64 rather than an unsigned int.
25532         (current_vector_size): Change from an unsigned int to a poly_uint64.
25533         (get_vectype_for_scalar_type): Update accordingly.
25534         * tree.h (build_truth_vector_type): Take the size and number of
25535         units as a poly_uint64 rather than an unsigned int.
25536         (build_vector_type): Add a temporary overload that takes
25537         the number of units as a poly_uint64 rather than an unsigned int.
25538         * tree.c (make_vector_type): Likewise.
25539         (build_truth_vector_type): Take the number of units as a poly_uint64
25540         rather than an unsigned int.
25541
25542 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25543             Alan Hayward  <alan.hayward@arm.com>
25544             David Sherwood  <david.sherwood@arm.com>
25545
25546         * target.def (get_mask_mode): Take the number of units and length
25547         as poly_uint64s rather than unsigned ints.
25548         * targhooks.h (default_get_mask_mode): Update accordingly.
25549         * targhooks.c (default_get_mask_mode): Likewise.
25550         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25551         * doc/tm.texi: Regenerate.
25552
25553 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25554             Alan Hayward  <alan.hayward@arm.com>
25555             David Sherwood  <david.sherwood@arm.com>
25556
25557         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25558         * omp-general.c (omp_max_vf): Likewise.
25559         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25560         (expand_omp_simd): Handle polynomial safelen.
25561         * omp-low.c (omplow_simd_context): Add a default constructor.
25562         (omplow_simd_context::max_vf): Change from int to poly_uint64.
25563         (lower_rec_simd_input_clauses): Update accordingly.
25564         (lower_rec_input_clauses): Likewise.
25565
25566 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25567             Alan Hayward  <alan.hayward@arm.com>
25568             David Sherwood  <david.sherwood@arm.com>
25569
25570         * tree-vectorizer.h (vect_nunits_for_cost): New function.
25571         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25572         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25573         (vect_analyze_slp_cost): Likewise.
25574         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25575         (vect_model_load_cost): Likewise.
25576
25577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25578             Alan Hayward  <alan.hayward@arm.com>
25579             David Sherwood  <david.sherwood@arm.com>
25580
25581         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25582         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25583         from an unsigned int * to a poly_uint64_pod *.
25584         (calculate_unrolling_factor): New function.
25585         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
25586
25587 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25588             Alan Hayward  <alan.hayward@arm.com>
25589             David Sherwood  <david.sherwood@arm.com>
25590
25591         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25592         from an unsigned int to a poly_uint64.
25593         (_loop_vec_info::slp_unrolling_factor): Likewise.
25594         (_loop_vec_info::vectorization_factor): Change from an int
25595         to a poly_uint64.
25596         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25597         (vect_get_num_vectors): New function.
25598         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25599         (vect_get_num_copies): Use vect_get_num_vectors.
25600         (vect_analyze_data_ref_dependences): Change max_vf from an int *
25601         to an unsigned int *.
25602         (vect_analyze_data_refs): Change min_vf from an int * to a
25603         poly_uint64 *.
25604         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25605         than an unsigned HOST_WIDE_INT.
25606         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25607         (vect_analyze_data_ref_dependence): Change max_vf from an int *
25608         to an unsigned int *.
25609         (vect_analyze_data_ref_dependences): Likewise.
25610         (vect_compute_data_ref_alignment): Handle polynomial vf.
25611         (vect_enhance_data_refs_alignment): Likewise.
25612         (vect_prune_runtime_alias_test_list): Likewise.
25613         (vect_shift_permute_load_chain): Likewise.
25614         (vect_supportable_dr_alignment): Likewise.
25615         (dependence_distance_ge_vf): Take the vectorization factor as a
25616         poly_uint64 rather than an unsigned HOST_WIDE_INT.
25617         (vect_analyze_data_refs): Change min_vf from an int * to a
25618         poly_uint64 *.
25619         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25620         vfm1 as a poly_uint64 rather than an int.  Make the same change
25621         for the returned bound_scalar.
25622         (vect_gen_vector_loop_niters): Handle polynomial vf.
25623         (vect_do_peeling): Likewise.  Update call to
25624         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25625         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25626         be constant.
25627         * tree-vect-loop.c (vect_determine_vectorization_factor)
25628         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25629         (vect_get_known_peeling_cost): Likewise.
25630         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25631         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25632         (vect_transform_loop): Likewise.  Use the lowest possible VF when
25633         updating the upper bounds of the loop.
25634         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
25635         rather than an int.
25636         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25637         polynomial unroll factors.
25638         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25639         (vect_make_slp_decision): Likewise.
25640         (vect_supported_load_permutation_p): Likewise, and polynomial
25641         vf too.
25642         (vect_analyze_slp_cost): Handle polynomial vf.
25643         (vect_slp_analyze_node_operations): Likewise.
25644         (vect_slp_analyze_bb_1): Likewise.
25645         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25646         than an unsigned HOST_WIDE_INT.
25647         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25648         (vectorizable_load): Handle polynomial vf.
25649         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25650         a poly_uint64.
25651         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25652
25653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25654             Alan Hayward  <alan.hayward@arm.com>
25655             David Sherwood  <david.sherwood@arm.com>
25656
25657         * match.pd: Handle bit operations involving three constants
25658         and try to fold one pair.
25659
25660 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25661
25662         * tree-vect-loop-manip.c: Include gimple-fold.h.
25663         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25664         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
25665         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25666         Add a path that uses a step of VF instead of 1, but disable it
25667         for now.
25668         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25669         and niters_no_overflow parameters.  Update calls to
25670         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25671         Create a new SSA name if the latter choses to use a ste other
25672         than zero, and return it via niters_vector_mult_vf_var.
25673         * tree-vect-loop.c (vect_transform_loop): Update calls to
25674         vect_do_peeling, vect_gen_vector_loop_niters and
25675         slpeel_make_loop_iterate_ntimes.
25676         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25677         (vect_gen_vector_loop_niters): Update declarations after above changes.
25678
25679 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
25680
25681         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25682         128-bit round to integer instructions.
25683         (ceil<mode>2): Likewise.
25684         (btrunc<mode>2): Likewise.
25685         (round<mode>2): Likewise.
25686
25687 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25688
25689         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25690         unaligned VSX load/store on P8/P9.
25691         (expand_block_clear): Allow the use of unaligned VSX
25692         load/store on P8/P9.
25693
25694 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25695
25696         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25697         New function.
25698         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25699         swap associated with both a load and a store.
25700
25701 2018-01-02  Andrew Waterman  <andrew@sifive.com>
25702
25703         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25704         * config/riscv/riscv.md (clear_cache): Use it.
25705
25706 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
25707
25708         * web.c: Remove out-of-date comment.
25709
25710 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25711
25712         * expr.c (fixup_args_size_notes): Check that any existing
25713         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25714         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25715         (emit_single_push_insn): ...here.
25716
25717 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25718
25719         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25720         (const_vector_encoded_nelts): New function.
25721         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25722         (const_vector_int_elt, const_vector_elt): Declare.
25723         * emit-rtl.c (const_vector_int_elt_1): New function.
25724         (const_vector_elt): Likewise.
25725         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25726         of CONST_VECTOR_ELT.
25727
25728 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25729
25730         * expr.c: Include rtx-vector-builder.h.
25731         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25732         directly on the tree encoding.
25733         (const_vector_from_tree): Likewise.
25734         * optabs.c: Include rtx-vector-builder.h.
25735         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25736         sequence of "u" values.
25737         * vec-perm-indices.c: Include rtx-vector-builder.h.
25738         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25739         directly on the vec_perm_indices encoding.
25740
25741 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25742
25743         * doc/rtl.texi (const_vector): Describe new encoding scheme.
25744         * Makefile.in (OBJS): Add rtx-vector-builder.o.
25745         * rtx-vector-builder.h: New file.
25746         * rtx-vector-builder.c: Likewise.
25747         * rtl.h (rtx_def::u2): Add a const_vector field.
25748         (CONST_VECTOR_NPATTERNS): New macro.
25749         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25750         (CONST_VECTOR_DUPLICATE_P): Likewise.
25751         (CONST_VECTOR_STEPPED_P): Likewise.
25752         (CONST_VECTOR_ENCODED_ELT): Likewise.
25753         (const_vec_duplicate_p): Check for a duplicated vector encoding.
25754         (unwrap_const_vec_duplicate): Likewise.
25755         (const_vec_series_p): Check for a non-duplicated vector encoding.
25756         Say that the function only returns true for integer vectors.
25757         * emit-rtl.c: Include rtx-vector-builder.h.
25758         (gen_const_vec_duplicate_1): Delete.
25759         (gen_const_vector): Call gen_const_vec_duplicate instead of
25760         gen_const_vec_duplicate_1.
25761         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25762         (gen_const_vec_duplicate): Use rtx_vector_builder.
25763         (gen_const_vec_series): Likewise.
25764         (gen_rtx_CONST_VECTOR): Likewise.
25765         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25766         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25767         Build a new vector rather than modifying a CONST_VECTOR in-place.
25768         (handle_special_swappables): Update call accordingly.
25769         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25770         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25771         Build a new vector rather than modifying a CONST_VECTOR in-place.
25772         (handle_special_swappables): Update call accordingly.
25773
25774 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25775
25776         * simplify-rtx.c (simplify_const_binary_operation): Use
25777         CONST_VECTOR_ELT instead of XVECEXP.
25778
25779 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25780
25781         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25782         the selector elements to be different from the data elements
25783         if the selector is a VECTOR_CST.
25784         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25785         ssizetype for the selector.
25786
25787 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25788
25789         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25790         before testing each element individually.
25791         * tree-vect-generic.c (lower_vec_perm): Likewise.
25792
25793 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25794
25795         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25796         * selftest-run-tests.c (selftest::run_tests): Call it.
25797         * vector-builder.h (vector_builder::operator ==): New function.
25798         (vector_builder::operator !=): Likewise.
25799         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25800         (vec_perm_indices::all_from_input_p): New function.
25801         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25802         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25803         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25804         instead of reading the VECTOR_CST directly.  Detect whether both
25805         vector inputs are the same before constructing the vec_perm_indices,
25806         and update the number of inputs argument accordingly.  Use the
25807         utility functions added above.  Only construct sel2 if we need to.
25808
25809 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25810
25811         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25812         the broadcast of the low byte.
25813         (expand_mult_highpart): Use an explicit encoding for the permutes.
25814         * optabs-query.c (can_mult_highpart_p): Likewise.
25815         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25816         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25817         (vectorizable_bswap): Likewise.
25818         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25819         explicit encoding for the power-of-2 permutes.
25820         (vect_permute_store_chain): Likewise.
25821         (vect_grouped_load_supported): Likewise.
25822         (vect_permute_load_chain): Likewise.
25823
25824 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25825
25826         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25827         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25828         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25829         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25830         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25831         (vect_gen_perm_mask_any): Likewise.
25832
25833 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25834
25835         * int-vector-builder.h: New file.
25836         * vec-perm-indices.h: Include int-vector-builder.h.
25837         (vec_perm_indices): Redefine as an int_vector_builder.
25838         (auto_vec_perm_indices): Delete.
25839         (vec_perm_builder): Redefine as a stand-alone class.
25840         (vec_perm_indices::vec_perm_indices): New function.
25841         (vec_perm_indices::clamp): Likewise.
25842         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25843         (vec_perm_indices::new_vector): New function.
25844         (vec_perm_indices::new_expanded_vector): Update for new
25845         vec_perm_indices class.
25846         (vec_perm_indices::rotate_inputs): New function.
25847         (vec_perm_indices::all_in_range_p): Operate directly on the
25848         encoded form, without computing elided elements.
25849         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25850         encoding.  Update for new vec_perm_indices class.
25851         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25852         the given vec_perm_builder.
25853         (expand_vec_perm_var): Update vec_perm_builder constructor.
25854         (expand_mult_highpart): Use vec_perm_builder instead of
25855         auto_vec_perm_indices.
25856         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25857         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
25858         or double series encoding as appropriate.
25859         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25860         vec_perm_indices instead of auto_vec_perm_indices.
25861         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25862         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25863         (vect_permute_store_chain): Likewise.
25864         (vect_grouped_load_supported): Likewise.
25865         (vect_permute_load_chain): Likewise.
25866         (vect_shift_permute_load_chain): Likewise.
25867         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25868         (vect_transform_slp_perm_load): Likewise.
25869         (vect_schedule_slp_instance): Likewise.
25870         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25871         (vectorizable_mask_load_store): Likewise.
25872         (vectorizable_bswap): Likewise.
25873         (vectorizable_store): Likewise.
25874         (vectorizable_load): Likewise.
25875         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25876         vec_perm_indices instead of auto_vec_perm_indices.  Use
25877         tree_to_vec_perm_builder to read the vector from a tree.
25878         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25879         vec_perm_builder instead of a vec_perm_indices.
25880         (have_whole_vector_shift): Use vec_perm_builder and
25881         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
25882         truncation to calc_vec_perm_mask_for_shift.
25883         (vect_create_epilog_for_reduction): Likewise.
25884         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25885         from auto_vec_perm_indices to vec_perm_indices.
25886         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25887         instead of changing individual elements.
25888         (aarch64_vectorize_vec_perm_const): Use new_vector to install
25889         the vector in d.perm.
25890         * config/arm/arm.c (expand_vec_perm_d::perm): Change
25891         from auto_vec_perm_indices to vec_perm_indices.
25892         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25893         instead of changing individual elements.
25894         (arm_vectorize_vec_perm_const): Use new_vector to install
25895         the vector in d.perm.
25896         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25897         Update vec_perm_builder constructor.
25898         (rs6000_expand_interleave): Likewise.
25899         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25900         (rs6000_expand_interleave): Likewise.
25901
25902 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25903
25904         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25905         to qimode could truncate the indices.
25906         * optabs.c (expand_vec_perm_var): Likewise.
25907
25908 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25909
25910         * Makefile.in (OBJS): Add vec-perm-indices.o.
25911         * vec-perm-indices.h: New file.
25912         * vec-perm-indices.c: Likewise.
25913         * target.h (vec_perm_indices): Replace with a forward class
25914         declaration.
25915         (auto_vec_perm_indices): Move to vec-perm-indices.h.
25916         * optabs.h: Include vec-perm-indices.h.
25917         (expand_vec_perm): Delete.
25918         (selector_fits_mode_p, expand_vec_perm_var): Declare.
25919         (expand_vec_perm_const): Declare.
25920         * target.def (vec_perm_const_ok): Replace with...
25921         (vec_perm_const): ...this new hook.
25922         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25923         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25924         * doc/tm.texi: Regenerate.
25925         * optabs.def (vec_perm_const): Delete.
25926         * doc/md.texi (vec_perm_const): Likewise.
25927         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25928         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25929         expand_vec_perm for constant permutation vectors.  Assert that
25930         the mode of variable permutation vectors is the integer equivalent
25931         of the mode that is being permuted.
25932         * optabs-query.h (selector_fits_mode_p): Declare.
25933         * optabs-query.c: Include vec-perm-indices.h.
25934         (selector_fits_mode_p): New function.
25935         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25936         is defined, instead of checking whether the vec_perm_const_optab
25937         exists.  Use targetm.vectorize.vec_perm_const instead of
25938         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
25939         fit in the vector mode before using a variable permute.
25940         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25941         vec_perm_indices instead of an rtx.
25942         (expand_vec_perm): Replace with...
25943         (expand_vec_perm_const): ...this new function.  Take the selector
25944         as a vec_perm_indices rather than an rtx.  Also take the mode of
25945         the selector.  Update call to shift_amt_for_vec_perm_mask.
25946         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25947         Use vec_perm_indices::new_expanded_vector to expand the original
25948         selector into bytes.  Check whether the indices fit in the vector
25949         mode before using a variable permute.
25950         (expand_vec_perm_var): Make global.
25951         (expand_mult_highpart): Use expand_vec_perm_const.
25952         * fold-const.c: Includes vec-perm-indices.h.
25953         * tree-ssa-forwprop.c: Likewise.
25954         * tree-vect-data-refs.c: Likewise.
25955         * tree-vect-generic.c: Likewise.
25956         * tree-vect-loop.c: Likewise.
25957         * tree-vect-slp.c: Likewise.
25958         * tree-vect-stmts.c: Likewise.
25959         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25960         Delete.
25961         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25962         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25963         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25964         (aarch64_vectorize_vec_perm_const): ...this new function.
25965         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25966         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25967         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25968         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25969         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25970         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25971         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25972         into...
25973         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
25974         check for NEON modes.
25975         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25976         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25977         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25978         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25979         into...
25980         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
25981         the old VEC_PERM_CONST conditions.
25982         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25983         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25984         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25985         (ia64_vectorize_vec_perm_const_ok): Merge into...
25986         (ia64_vectorize_vec_perm_const): ...this new function.
25987         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25988         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25989         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25990         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25991         * config/mips/mips.c (mips_expand_vec_perm_const)
25992         (mips_vectorize_vec_perm_const_ok): Merge into...
25993         (mips_vectorize_vec_perm_const): ...this new function.
25994         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25995         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25996         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25997         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25998         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25999         (rs6000_expand_vec_perm_const): Delete.
26000         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
26001         Delete.
26002         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26003         (altivec_expand_vec_perm_const_le): Take each operand individually.
26004         Operate on constant selectors rather than rtxes.
26005         (altivec_expand_vec_perm_const): Likewise.  Update call to
26006         altivec_expand_vec_perm_const_le.
26007         (rs6000_expand_vec_perm_const): Delete.
26008         (rs6000_vectorize_vec_perm_const_ok): Delete.
26009         (rs6000_vectorize_vec_perm_const): New function.
26010         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26011         an element count and rtx array.
26012         (rs6000_expand_extract_even): Update call accordingly.
26013         (rs6000_expand_interleave): Likewise.
26014         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
26015         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
26016         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
26017         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
26018         (rs6000_expand_vec_perm_const): Delete.
26019         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26020         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26021         (altivec_expand_vec_perm_const_le): Take each operand individually.
26022         Operate on constant selectors rather than rtxes.
26023         (altivec_expand_vec_perm_const): Likewise.  Update call to
26024         altivec_expand_vec_perm_const_le.
26025         (rs6000_expand_vec_perm_const): Delete.
26026         (rs6000_vectorize_vec_perm_const_ok): Delete.
26027         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
26028         reference to the SPE evmerge intructions.
26029         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26030         an element count and rtx array.
26031         (rs6000_expand_extract_even): Update call accordingly.
26032         (rs6000_expand_interleave): Likewise.
26033         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
26034         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26035         new function.
26036         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26037
26038 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26039
26040         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26041         vector mode and that that mode matches the mode of the data
26042         being permuted.
26043         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26044         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
26045         directly using expand_vec_perm_1 when forcing selectors into
26046         registers.
26047         (expand_vec_perm_var): New function, split out from expand_vec_perm.
26048
26049 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26050
26051         * optabs-query.h (can_vec_perm_p): Delete.
26052         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26053         * optabs-query.c (can_vec_perm_p): Split into...
26054         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26055         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26056         particular selector is valid.
26057         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26058         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26059         (vect_grouped_load_supported): Likewise.
26060         (vect_shift_permute_load_chain): Likewise.
26061         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26062         (vect_transform_slp_perm_load): Likewise.
26063         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26064         (vectorizable_bswap): Likewise.
26065         (vect_gen_perm_mask_checked): Likewise.
26066         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
26067         implementations of variable permutation vectors into account
26068         when deciding which selector to use.
26069         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26070         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26071         with a false third argument.
26072         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26073         to test whether the constant selector is valid and can_vec_perm_var_p
26074         to test whether a variable selector is valid.
26075
26076 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26077
26078         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26079         * optabs-query.c (can_vec_perm_p): Likewise.
26080         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26081         instead of vec_perm_indices.
26082         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26083         (vect_gen_perm_mask_checked): Likewise,
26084         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26085         (vect_gen_perm_mask_checked): Likewise,
26086
26087 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26088
26089         * optabs-query.h (qimode_for_vec_perm): Declare.
26090         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26091         (qimode_for_vec_perm): ...this new function.
26092         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26093
26094 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26095
26096         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26097         does not have a conditional at the top.
26098
26099 2018-01-02  Richard Biener  <rguenther@suse.de>
26100
26101         * ipa-inline.c (big_speedup_p): Fix expression.
26102
26103 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
26104
26105         PR target/81616
26106         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26107         for generic 4->6.
26108
26109 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
26110
26111         PR target/81616
26112         Generic tuning.
26113         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26114         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26115         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26116         cond_taken_branch_cost 3->4.
26117
26118 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
26119
26120         PR tree-optimization/83581
26121         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26122         TODO_cleanup_cfg if any changes have been made.
26123
26124         PR middle-end/83608
26125         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26126         convert_modes if target mode has the right side, but different mode
26127         class.
26128
26129         PR middle-end/83609
26130         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26131         last argument when extracting from CONCAT.  If either from_real or
26132         from_imag is NULL, use expansion through memory.  If result is not
26133         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26134         the parts directly to inner mode, if even that fails, use expansion
26135         through memory.
26136
26137         PR middle-end/83623
26138         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26139         check for bswap in mode rather than HImode and use that in expand_unop
26140         too.
26141 \f
26142 Copyright (C) 2018 Free Software Foundation, Inc.
26143
26144 Copying and distribution of this file, with or without modification,
26145 are permitted in any medium without royalty provided the copyright
26146 notice and this notice are preserved.