gcc/
[official-gcc.git] / gcc / ChangeLog
blob62f6383195a4b8e75aac682c0a7c89d5ea056a75
1 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
3         PR rtl-optimization/87507
4         * lower-subreg.c (operand_for_swap_move_operator): New function.
5         (simple_move): Strip simple operators.
6         (find_pseudo_copy): Likewise.
7         (resolve_operand_for_swap_move_operator): New function.
8         (resolve_simple_move): Strip simple operators and swap operands.
10 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
12         PR tree-optimization/87898
13         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
14         (ipa_simd_modify_function_body): Remove debug stmts where the first
15         argument was changed into a non-decl.
17 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
19         PR rtl-optimization/87899
20         * lra-lives.c (start_living): Update white space in comment.
21         (enum point_type): New.
22         (sparseset_contains_pseudos_p): New function.
23         (update_pseudo_point): Likewise.
24         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
25         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
26         handling.  Move early exit after adding conflicts.
27         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
28         if regno is already live.  Remove all handling of program points.
29         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
30         after adding conflicts.  Remove all handling of program points and
31         ignore_reg_for_conflicts.
32         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
33         and do not guard call to mark_pseudo_live.
34         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
35         and do not guard call to mark_pseudo_dead.
36         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
37         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
38         Use new function update_pseudo_point.  Handle register copies by
39         removing the source register from the live set.  Handle INOUT operands.
40         Update to the next program point using the unused_set, dead_set and
41         start_dying sets.
42         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
44 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
46         * builtins.c: Replace "source_location" with "location_t".
47         * diagnostic-show-locus.c: Likewise.
48         * diagnostic.c: Likewise.
49         * dumpfile.c: Likewise.
50         * gcc-rich-location.h: Likewise.
51         * genmatch.c: Likewise.
52         * gimple.h: Likewise.
53         * gimplify.c: Likewise.
54         * input.c: Likewise.
55         * input.h: Likewise.  Eliminate the typedef.
56         * omp-expand.c: Likewise.
57         * selftest.h: Likewise.
58         * substring-locations.h (get_source_location_for_substring):
59         Rename to..
60         (get_location_within_string): ...this.
61         * tree-cfg.c: Replace "source_location" with "location_t".
62         * tree-cfgcleanup.c: Likewise.
63         * tree-diagnostic.c: Likewise.
64         * tree-into-ssa.c: Likewise.
65         * tree-outof-ssa.c: Likewise.
66         * tree-parloops.c: Likewise.
67         * tree-phinodes.c: Likewise.
68         * tree-phinodes.h: Likewise.
69         * tree-ssa-loop-ivopts.c: Likewise.
70         * tree-ssa-loop-manip.c: Likewise.
71         * tree-ssa-phiopt.c: Likewise.
72         * tree-ssa-phiprop.c: Likewise.
73         * tree-ssa-threadupdate.c: Likewise.
74         * tree-ssa.c: Likewise.
75         * tree-ssa.h: Likewise.
76         * tree-vect-loop-manip.c: Likewise.
78 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
80         * doc/extend.texi: Fix typo in the weakref description.
82 2018-11-13  Richard Biener  <rguenther@suse.de>
84         PR tree-optimization/86991
85         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
86         group building until we have successfully detected the SLP
87         reduction.
88         (vect_is_simple_reduction): Remove fixup code here.
90 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
92         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
93         (dump_gimple_stmt): Use it.
94         (dump_gimple_stmt_loc): Likewise.
95         (dump_gimple_expr): Likewise.
96         (dump_gimple_expr_loc): Likewise.
97         (dump_generic_expr): Likewise.
98         (dump_generic_expr_loc): Likewise.
99         (dump_printf): Likewise.
100         (dump_printf_loc): Likewise.
101         (dump_dec): Likewise.
102         (dump_dec): Likewise.
103         (dump_hex): Likewise.
104         (dump_symtab_node): Likewise.
105         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
106         Guard dump call with dump_enabled_p.
107         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
108         * graphite-optimize-isl.c (optimize_isl): Likewise.
109         * graphite.c (graphite_transform_loops): Likewise.
110         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
111         * tree-parloops.c (parallelize_loops): Likewise.
112         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
113         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
114         (vect_prune_runtime_alias_test_list): Likewise.
115         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
116         (vect_estimate_min_profitable_iters): Likewise.
117         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
118         (vect_build_slp_tree_2): Likewise.
119         (vect_supported_load_permutation_p): Likewise.
120         (vect_slp_analyze_operations): Likewise.
121         (vect_slp_analyze_bb_1): Likewise.
122         (vect_slp_bb): Likewise.
123         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
124         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
125         (pass_slp_vectorize::execute): Likewise.
126         (increase_alignment): Likewise.
128 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
130         PR ipa/87955
131         * ipa-inline.c (report_inline_failed_reason): Guard calls to
132         cl_target_option_print_diff and cl_optimization_print_diff with
133         if (dump_file).
135 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
137         * doc/invoke.texi (-fsave-optimization-record): Note that the
138         output is compressed.
139         * optinfo-emit-json.cc: Include <zlib.h>.
140         (optrecord_json_writer::write): Compress the output.
142 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
144         * tree-vrp.c (value_range_base::dump): Dump type.
145         Do not use INF nomenclature for 1-bit types.
146         (dump_value_range): Group all variants to common dumping code.
147         (debug): New overloaded functions for value_ranges.
148         (value_range_base::dump): Remove no argument version.
149         (value_range::dump): Same.
151 2018-11-13  Richard Biener  <rguenther@suse.de>
153         PR tree-optimization/87931
154         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
155         nested cycles we support to latch computations vectorizable_reduction
156         handles.
158 2018-11-13  Martin Liska  <mliska@suse.cz>
160         PR tree-optimization/87885
161         * cfghooks.c (account_profile_record): Rename
162         to ...
163         (profile_record_check_consistency): ... this.
164         Calculate missing num_mismatched_freq_in.
165         (profile_record_account_profile): New function
166         that calculates time and size of a function.
167         * cfghooks.h (struct profile_record): Remove
168         all tuples.
169         (struct cfg_hooks): Remove after_pass flag.
170         (account_profile_record): Rename to ...
171         (profile_record_check_consistency): ... this.
172         (profile_record_account_profile): New.
173         * cfgrtl.c (rtl_account_profile_record): Remove
174         after_pass flag.
175         * passes.c (check_profile_consistency): Do only
176         checking.
177         (account_profile): Calculate size and time of
178         function only.
179         (pass_manager::dump_profile_report): Reformat
180         output.
181         (execute_one_ipa_transform_pass): Call
182         consistency check before clean upand call account_profile
183         after a clean up is done.
184         (execute_one_pass): Call check_profile_consistency and
185         account_profile instead of using after_pass flag..
186         * tree-cfg.c (gimple_account_profile_record): Likewise.
188 2018-11-13  Martin Liska  <mliska@suse.cz>
190         PR sanitizer/87930
191         * config/i386/i386.c (ix86_option_override_internal): Error
192         about usage -mabi=ms and -fsanitize={,kernel-}address.
194 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
196         * config/aarch64/aarch64.c
197         (aarch64_vectorize_preferred_vector_alignment): Change return type to
198         poly_uint64.
199         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
200         alignment being a poly int.
201         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
202         return type to poly_uint64.
203         * target.def (default_preferred_vector_alignment): Likewise.
204         * targhooks.c (default_preferred_vector_alignment): Likewise.
205         * targhooks.h (default_preferred_vector_alignment): Likewise.
206         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
207         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
208         poly int.
209         (vect_update_misalignment_for_peel): Likewise.
210         (vect_enhance_data_refs_alignment): Likewise.
211         (vect_find_same_alignment_drs): Likewise.
212         (vect_duplicate_ssa_name_ptr_info): Likewise.
213         (vect_setup_realignment): Likewise.
214         (vect_can_force_dr_alignment_p): Change alignment parameter type to
215         poly_uint64.
216         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
217         mask with a compile time variable vector alignment.
218         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
219         int.
220         (vect_do_peeling): Exit early if vector alignment is not constant.
221         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
222         a poly int.
223         (vectorizable_store): Likewise.
224         (vectorizable_load): Likweise.
225         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
226         poly_uint64.
227         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
228         poly int.
229         (vect_can_force_dr_alignment_p): Change alignment parameter type to
230         poly_uint64.
232 2018-11-13  Richard Biener  <rguenther@suse.de>
234         PR tree-optimization/87962
235         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
236         detect outer reduction for disqualifying in-loop uses.
238 2018-11-13  Richard Biener  <rguenther@suse.de>
240         PR tree-optimization/87967
241         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
242         for constants for the scalar loop.
244 2018-11-13  Alan Modra  <amodra@gmail.com>
246         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
247         integer constants when -mcmodel=medium.
249 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
251         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
252         * config/arc/elf.h (PROFILE_HOOK): Define.
253         * config/arc/linux.h (PROFILE_HOOK): Likewise.
255 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
257         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
258         jump table data in the text section.
260 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
262         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
263         to fit the eh_return pattern.
264         * config/arc/arc.md (eh_return): Define.
265         (VUNSPEC_ARC_EH_RETURN): Likewise.
266         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
267         new implementation.
268         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
271 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
273         * common/config/arc/arc-common.c (arc_option_optimization_table):
274         Millicode optimization is default on for size optimizations.
275         * config/arc/arc-protos.h (arc_check_multi): New function.
276         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
277         (ENTER_LEAVE_START_REG): Define.
278         (ENTER_LEAVE_END_REG): Likewise.
279         (arc_override_options): Disable millicode when long calls option
280         is on.
281         (arc_frame_info): Change it from int to bool.
282         (arc_compute_frame_size): Clean up.
283         (arc_save_restore): Remove.
284         (frame_save_reg): New function.
285         (frame_restore_reg): Likewise.
286         (arc_enter_leave_p): Likewise.
287         (arc_save_callee_saves): Likewise.
288         (arc_restore_callee_saves): Likewise.
289         (arc_save_callee_enter): Likewise.
290         (arc_restore_callee_leave): Likewise.
291         (arc_save_callee_milli): Likewise.
292         (arc_restore_callee_milli): Likewise.
293         (arc_expand_prologue): Reimplement to emit enter/leave
294         instructions.
295         (arc_expand_epilogue): Likewise.
296         (arc_check_multi): New function.
297         * config/arc/arc.md (push_multi_fp): New pattern.
298         (push_multi_fp_blink): Likewise.
299         (pop_multi_fp): Likewise.
300         (pop_multi_fp_blink): Likewise.
301         (pop_multi_fp_ret): Likewise.
302         (pop_multi_fp_blink_ret): Likewise.
303         * config/arc/arc.opt (mmillicode): Update option.
304         (mcode-density-frame): New option.
305         * config/arc/predicates.md (push_multi_operand): New predicate.
306         (pop_multi_operand): Likewise.
307         * doc/invoke.texi (ARC): Update ARC options information.
310 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
312         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
313         * config/arc/arc.c (operands_ok_ldd_std): New function.
314         (mem_ok_for_ldd_std): Likewise.
315         (gen_operands_ldd_std): Likewise.
316         * config/arc/arc.md: Add peephole2 rules for std/ldd.
318 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
320         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
321         test on stack_usage_file.
322         (lang_dependent_init): Do not open the .su file if generating LTO.
324 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
326         PR rtl-optimization/87918
327         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
328         simplify_gen_relational rather than simplify_gen_binary.
330 2018-11-13  Richard Biener  <rguenther@suse.de>
332         * tree-ssanames.h (set_range_info): Use value_range_base.
333         (get_range_info): Likewise.
334         * tree-ssanames.c (set_range_info): Likewise.
335         (get_range_info): Likewise.
336         * tree-vrp.c (value_range_base::union_helper): Split
337         out common parts of value_range[_base]::union_.
338         (value_range_base::union_): Update.
339         (value_range::union_): Likewise.
340         (determine_value_range_1): Use value_range_base.
341         (determine_value_range): Likewise.
342         * tree-vrp.h (value_range_base::union_helper): Move ...
343         (value_range::union_helper): ... from here.
345 2018-11-13  Alan Modra  <amodra@gmail.com>
347         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
348         offset for PRE_DEC.
349         (rs6000_secondary_reload_gpr): Don't call find_replacement.
351 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
353         PR middle-end/59634
354         * doc/invoke.texi (Optimize Options): Clarify that the
355         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
356         apply to data cache size.
358 2018-11-13  Alan Modra  <amodra@gmail.com>
360         * config/rs6000/predicates.md (logical_const_operand),
361         (logical_operand): Correct comment.
362         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
364 2018-11-13  Alan Modra  <amodra@gmail.com>
366         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
368 2018-11-13  Alan Modra  <amodra@gmail.com>
370         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
371         description.
372         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
373         covered by alternative.
374         (movcc_internal1): Ignore h for register preference.
375         (mov<mode>_hardfloat64): Likewise.
376         (mov<mode>_softfloat): Ignore c, l, h for register preference.
378 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
380         PR preprocessor/47823
381         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
382         documentation to...
383         (Pragmas): ...here.
384         * doc/extend.texi (Pragmas): Note additional pragmas documented
385         in the CPP manual.
387 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
389         PR middle-end/86677
390         PR middle-end/87528
391         * tree-scalar-evolution.c (expression_expensive_p): Make BUILTIN POPCOUNT
392         as expensive when backend does not define it.
394 2018-11-12  Fredrik Noring  <noring@nocrew.org>
396         * config.gcc: Update with-llsc defaults for MIPS r5900.
398 2018-11-12  Martin Liska  <mliska@suse.cz>
400         PR gcov-profile/87442
401         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
402         options.
403         * doc/invoke.texi: Document them.
404         * tree-profile.c (parse_profile_filter): New.
405         (parse_profile_file_filtering): Likewise.
406         (release_profile_file_filtering): Likewise.
407         (include_source_file_for_profile): Likewise.
408         (tree_profiling): Filter source files based on the
409         newly added options.
411 2018-11-12  Martin Liska  <mliska@suse.cz>
413         PR target/87903
414         * doc/extend.texi: Add missing values for __builtin_cpu_is and
415         __builtin_cpu_supports for x86 target.
417 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
419         PR middle-end/21110
420         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
421         have to be Pmode.
422         (Arithmetic): Likewise for "lo_sum".
424 2018-11-12  Renlin Li  <renlin.li@arm.com>
426         PR target/87815
427         * dse.c (get_stored_val): Add check for compile-time constantness
428         of gap.
430 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
432         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
433         (ARMv8_5a): New fgroup.
434         (armv8.5-a): New arch.
435         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
436         sb and predres.
437         * config/arm/arm-tables.opt: Regenerate.
438         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
439         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
440         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
441         Add matching rules for -march=armv8.5-a and extensions.
442         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
443         Add sb and predres to all armv8-a except armv8.5-a.
445 2018-11-12  Richard Biener  <rguenther@suse.de>
447         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
448         overload for single value.
449         (value_range[_base]::set_nonnull): New.
450         (value_range[_base]::set_null): Likewise.
451         (value_range): Document bitmap copying behavior, mark
452         copy constructor and assignment operator deleted.
453         (value_range::move): New.
454         (value_range::set_and_canonicalize): Default bitmap to zero.
455         (set_value_range_to_nonnull): Remove.
456         (set_value_range_to_null): Likewise.
457         (set_value_range): Likewise.
458         (set_value_range_to_value): Likewise.
459         (extract_range_from_unary_expr): Work on value_range_base.
460         (extract_range_from_binary_expr_1): Likewise.  Rename to...
461         (extract_range_from_binary_expr): ... this.
462         * tree-vrp.c (value_range::update): Clear equiv bitmap
463         if required.
464         (value_range::move): New, move equiv bitmap.
465         (value_range_base::set_undefined): Avoid assignment.
466         (value_range::set_undefined): Likewise.
467         (value_range_base::set_varying): Likewise.
468         (value_range::set_varying): Likewise.
469         (set_value_range): Remove.
470         (value_range_base::set): New overload for value.
471         (value_range::set): Likewise.
472         (set_value_range_to_nonnull): Remove.
473         (value_range_base::set_nonnull): New.
474         (value_range::set_nonnull): Likewise.
475         (set_value_range_to_null): Remove.
476         (value_range_base::set_null): New.
477         (value_range::set_null): Likewise.
478         (range_is_null): Work on value_range_base.
479         (range_is_nonnull): Likewise.
480         (ranges_from_anti_range): Likewise.
481         (extract_range_into_wide_ints): Likewise.
482         (extract_range_from_multiplicative_op): Likewise.
483         (extract_range_from_binary_expr): Likewise.  Update for API changes.
484         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
485         handling.
486         (value_range::intersect_helper): Avoid copy and assignment.
487         (value_range::union_helper): Likewise.
488         (determine_value_range_1): Adjust.
489         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
490         Avoid assignment by using move.
491         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
492         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
493         Likewise.
494         * tree-ssanames.c (get_range_info): Likewise.
495         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
496         * vr-values.c (vr_values::get_value_range): Adjust.
497         (vr_values::update_value_range): Likewise.
498         (symbolic_range_based_on_p): Work on value_range_base.
499         (vr_values::extract_range_from_binary_expr): Use value_range_base.
500         (vr_values::extract_range_from_unary_expr): Likewise.
501         (vr_values::extract_range_from_cond_expr): Avoid assignment.
502         (vr_values::extract_range_from_comparison): Adjust.
503         (vr_values::check_for_binary_op_overflow): Use value_range_base.
504         (vr_values::extract_range_basic): Adjust.
505         (vr_values::adjust_range_with_scev): Likewise.
506         (vr_values::vrp_visit_assignment_or_call): Likewise.
507         (vr_values::get_vr_for_comparison): Change API to avoid
508         assignment and copy construction.
509         (vr_values::compare_name_with_value): Adjust accordingly.
510         (vr_values::compare_names): Likewise.
511         (vr_values::extract_range_from_phi_node): Avoid assignment and
512         bogus in-place modify of equiv bitmap.
513         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
514         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
515         for extract_range_from_unary_expr API change.
516         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
518 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
520         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
522 2018-11-12  Richard Biener  <rguenther@suse.de>
524         * tree-vrp.h (value_range_base::symbolic_p,
525         value_range_base::constant_p, value_range_base::zero_p,
526         value_range_base::singleton_p): Move from value_range.
527         (value_range::dump): Add.
528         * gimple-ssa-evrp-analyze.c
529         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
530         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
531         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
532         Use set_varying.
533         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
534         (value_range::constant_p): Likewise.
535         (value_range::singleton_p): Likewise.
536         (value_range_base::dump): Add.
537         (set_value_range_to_undefined): Remove.
538         (set_value_range_to_varying): Likewise.
539         (range_int_cst_p): Take value_range_base argument.
540         (range_int_cst_singleton_p): Likewise.
541         (value_range_constant_singleton): Likewise.
542         (vrp_set_zero_nonzero_bits): Likewise.
543         (extract_range_from_multiplicative_op): Use set_varying.
544         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
545         (extract_range_from_unary_expr): Likewise.
546         (dump_value_range_base): Change to overload of dump_value_range.
547         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
548         (vrp_prop::visit_stmt): Likewise.
549         (value_range::intersect_helper): Likewise.
550         (value_range::union_helper): Likewise.
551         (determine_value_range_1): Likewise.
553 2018-11-12  Richard Biener  <rguenther@suse.de>
555         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
556         (set_value_range_to_null): Likewise.
557         * vr-values.c (vr_values::extract_range_from_comparison):
558         Clear equiv for constant singleton ranges.
560 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
562         * config/i386/sse.md: Combine VFIXUPIMM* patterns
563         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
564         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
565         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
566         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
567         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
568         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
570 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
572         PR c/69502
573         * doc/extend.texi (Common Type Attributes): For the align type
574         attribute, copy language about decreasing alignment from the
575         corresponding variable attribute.
577 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
579         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
580         -frounding-math.
582 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
584         PR c++/43105
585         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
586         -frtti and -fno-rtti code.
588 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
590         PR c/26366
591         * doc/extend.texi (Other Builtins): Document probability associated
592         with __builtin_expect.
594 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
596         PR target/87928
597         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
598         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
599         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
600         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
602 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
604         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
606 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
608         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
610 2018-11-11  Richard Biener  <rguenther@suse.de>
612         * tree-vrp.h (class value_range_base): New base class for
613         value_range containing all but the m_equiv member.
614         (dump_value_range_base): Add.
615         (range_includes_zero_p): Work on value_range_base.
616         * tree-vrp.c (value_range_base::set): Split out base handling
617         from...
618         (value_range::set): this.
619         (value_range::set_equiv): New.
620         (value_range_base::value_range_base): New constructors.
621         (value_range_base::check): Split out base handling from...
622         (value_range::check): this.
623         (value_range::equal_p): Refactor in terms of
624         ignore_equivs_equal_p which is now member of the base.
625         (value_range_base::set_undefined): New.
626         (value_range_base::set_varying): Likewise.
627         (value_range_base::dump):Split out base handling from...
628         (value_range::dump): this.
629         (value_range_base::set_and_canonicalize): Split out base handling
630         from...
631         (value_range::set_and_canonicalize): this.
632         (value_range_base::union_): New.
633         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
634         for m_vr.
635         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
636         instead of value_range everywhere.
637         (ipcp_vr_lattice::print): Use dump_value_range_base.
638         (ipcp_vr_lattice::meet_with): Adjust.
639         (ipcp_vr_lattice::meet_with_1): Likewise.
640         (ipa_vr_operation_and_type_effects): Likewise.
641         (propagate_vr_across_jump_function): Likewise.
642         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
643         (ipa_get_value_range): Likewise.
644         (ipa_set_jfunc_vr): Likewise.
645         (ipa_compute_jump_functions_for_edge): Likewise.
647 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
649         PR middle-end/65703
650         * doc/invoke.texi (Optimize Options): Add @opindex entries
651         for the positive forms of -fno-xxx and -mno-xxx options
652         that were lacking them.
654 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
656         * combine.c (make_more_copies): Only make an intermediate copy if the
657         dest of a move is a pseudo.
659 2018-11-09  Maya Rashish  <coypu@sdf.org>
661         PR target/87221
662         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
663         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
665 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
667         PR driver/41179
668         PR middle-end/65703
669         * doc/invoke.texi (Optimize Options): Clarify default behavior
670         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
672 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
674         PR tree-optimization/87940
675         * expr.c (string_constant): Don't strip NOPS in subexpressions.
676         Fold PLUS_EXPR correctly.
678 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
680         PR target/87762
681         * config/s390/s390.c (s390_safe_relative_long_p): New function.
682         (annotate_constant_pool_refs): Skip insns which support
683         relative addressing.
684         (annotate_constant_pool_refs_1): New helper function.
685         (find_constant_pool_ref): Skip insns which support relative
686         addression.
687         (find_constant_pool_ref_1): New helper function.
688         (replace_constant_pool_ref): Skip insns which support
689         relative addressing.
690         (replace_constant_pool_ref_1): New helper function.
691         (s390_mainpool_start): Adapt to the new signature.
692         (s390_mainpool_finish): Likewise.
693         (s390_chunkify_start): Likewise.
694         (s390_chunkify_finish): Likewise.
695         (pass_s390_early_mach::execute): Likewise.
696         (s390_prologue_plus_offset): Likewise.
697         (s390_emit_prologue): Likewise.
698         (s390_emit_epilogue): Likewise.
700 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
702         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
703         but unsupported lastprivate with conditional modifier.
705 2018-11-09  Jeff Law  <law@redhat.com>
707         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
708         unused argument better.  Add gcc_unreachable to silence warning.
710 2018-11-09  Martin Sebor  <msebor@redhat.com>
712         PR middle-end/81824
713         * attribs.c (has_attribute): New helper function.
714         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
715         * attribs.h (decls_mismatched_attributes): Declare.
716         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
717         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
718         * common.opt (-Wattribute-alias): Take an argument.
719         (-Wno-attribute-alias): New option.
720         * doc/extend.texi (Common Function Attributes): Document copy.
721         (Common Variable Attributes): Same.
722         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
723         (-Wattribute-alias): Document new option argument.
725 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
727         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
728         matching criteria.  Remove unused array initializer.
730 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
731             Jinsong Ji  <jji@us.ibm.com>
733         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
734         constraints by introducing a new temporary.
735         (_mm_cvtss_si64): Likewise.
737 2018-11-09  Martin Liska  <mliska@suse.cz>
739         * common.opt: Add -fipa-stack-alignment flag.
740         * doc/invoke.texi: Document it.
741         * final.c (rest_of_clean_state): Guard stack
742         shrinking with flag.
744 2018-11-09  Martin Liska  <mliska@suse.cz>
746         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
747         to ...
748         (ipa_discover_variable_flags): ... this.
749         * common.opt: Come up with new flag -fipa-reference-addressable.
750         * doc/invoke.texi: Document it.
751         * ipa-reference.c (propagate): Call the renamed fn.
752         * ipa-visibility.c (whole_program_function_and_variable_visibility):
753         Likewise.
754         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
755         ...
756         (ipa_discover_variable_flags): ... this.  Discover
757         non-addressable variables only with the newly added flag.
758         * opts.c: Enable the newly added flag with -O1 and higher
759         optimization level.
761 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
763         * json.cc (selftest::test_writing_literals): Fix comment.
765 2018-11-09  Martin Liska  <mliska@suse.cz>
767         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
768         string to a stack buffer.
769         (aarch64_parse_cpu): Likewise.
770         (aarch64_parse_tune): Likewise.
772 2018-11-09  Richard Biener  <rguenther@suse.de>
774         PR tree-optimization/87953
775         * tree-vect-loop.c (vectorizable_reduction): For analysis
776         always pass ops[0] to vectorizable_condition.
778 2018-11-09  Stafford Horne  <shorne@gmail.com>
779             Richard Henderson  <rth@twiddle.net>
780             Joel Sherrill  <joel@rtems.org>
782         * common/config/or1k/or1k-common.c: New file.
783         * config/or1k/*: New.
784         * config.gcc (or1k*-*-*): New.
785         * configure.ac (or1k*-*-*): New test for openrisc tls.
786         * configure: Regenerated.
787         * doc/install.texi: Document OpenRISC triplets.
788         * doc/invoke.texi: Document OpenRISC arguments.
789         * doc/md.texi: Document OpenRISC.
791 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
793         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
794         (arm7tdmi-s): Delete CPU.
795         (arm710t): Add aliases for arm720t and arm740t.
796         (arm720t, arm740t): Delete CPUs.
797         (arm920t): Add aliases for arm920, arm922t and arm940t.
798         (arm920, arm922t, arm940t): Delete CPUs.
799         (arm10tdmi): Add alias for arm1020t.
800         (arm1020t): Delete CPU.
801         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
802         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
803         (arm10e): Add aliases for arm1020e and arm1022e.
804         (arm1020e, arm1022e): Delete CPU.
805         * config/arm/arm.md (generic_sched): Remove entries that are now
806         handled by aliases.
807         (generic_vfp): Likewise.
808         * config/arm/arm1020e.md: Simplify tuning selection based on alias
809         changes.
810         * config/arm/arm-tune.md: Regenerated.
811         * config/arm/arm-tables.opt: Regenerated.
813 2018-11-09  Richard Biener  <rguenther@suse.de>
815         PR tree-optimization/87621
816         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
817         op with only phi inputs.
818         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
819         (ch_base::copy_headers): Run CSE on copied loop headers.
820         (pass_ch_vect::process_loop_p): Simplify.
822 2018-11-09  Alexandre Oliva <oliva@adacore.com>
824         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
825         for non-w64 x86_64 biarch.
827 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
829         PR rtl-optimization/86438
830         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
831         of in_b for the compare if in_b is SET_DEST.
833         PR target/87793
834         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
835         non-toplevel UNSPEC.
837 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
839         * tree-vrp.c (value_range::check): Do not access internals
840         directly.
841         (value_range::singleton_p): Same.
842         (value_range::type): Same.
843         (vrp_finalize): Use value_range API.
845 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
847         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
849 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
851         * vr-values.c (vr_values::get_value_range): Use value_range API
852         instead of piecing together ranges.
853         (vr_values::update_value_range): Same.
855 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
857         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
858         of performing ad-hoc calculations.
859         * tree-ssanames.c (set_range_info): New overloaded function
860         accepting value_range &.
861         (get_range_info): Same.
862         * tree-ssanames.h (set_range_info_raw): Remove.
863         (set_range_info): New prototype.
864         (get_range_info): Same.
865         * tree-vrp.h (value_range::null_p): Rename to zero_p.
866         * tree-vrp.c (value_range::null_p): Same.
868 2018-11-09  Jan Hubicka  <jh@suse.cz>
870         * tree.c (fld_type_variant_equal_p): Test user align flag.
871         (flt_type_variant): Copy user align flag.
872         (fld_incomplete_type_of): Clear it.
874 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
876         * config/arm/neon.md (div<mode>3): New pattern.
878 2018-11-08  Andi Kleen  <ak@linux.intel.com>
880         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
881         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
882         (ix86_handle_option): Handle OPT_mptwrite.
883         * config/i386/cpuid.h (bit_PTWRITE): Add.
884         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
885         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
886         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
887         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
888         (ix86_option_override_internal): Handle PTA_PTWRITE.
889         (ix86_valid_target_attribute_inner_p): Define ptwrite.
890         (def_builtin2): Force UINT64 to be 64bit only.
891         * config/i386/i386.h (TARGET_PTWRITE): Add.
892         (TARGET_PTWRITE_P): Add.
893         (PTA_PTWRITE): Add.
894         * config/i386/i386.md: Define ptwrite.
895         * config/i386/i386.opt: Add -mptwrite.
896         * config/i386/immintrin.h (_ptwrite64): Add.
897         (_ptwrite32): Add
898         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
899         * doc/invoke.texi: Document -mptwrite.
901 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
903         PR rtl-optimization/87600
904         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
905         * lra-constraints.c (process_alt_operands): Skip illegal hard
906         register usage.  Prefer reloading non hard register operands.
908 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
910         PR other/36572
911         * doc/invoke.texi (Optimize Options): Clarify default behavior
912         for -fno-sched-interblock and -fno-sched-spec.
914 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
916         * collect2.c (linker_select):  Add USE_LLD_LD.
917         (ld_suffixes): Add ld.lld.
918         (main): Handle -fuse-ld=lld.
919         * common.opt (-fuse-ld=lld): New option.
920         * doc/invoke.texi (-fuse-ld=lld): Document.
921         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
923 2018-11-08  Paul Koning  <ni1d@arrl.net>
925         * config/pdp11/constraints.md: Add "Z" series constraints for use
926         with pre-dec and post-inc addressing.
927         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
928         (pdp11_expand_operands): Add int argument (word count).
929         (pdp11_sp_frame_offset): Delete.
930         (pdp11_cmp_length): New function.
931         (pushpop_regeq): New function.
932         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
933         Add hook.
934         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
935         frame layout.
936         (pdp11_initial_elimination_offset): Ditto.
937         (pdp11_expand_operands): Add word count argument.  Bugfixes.
938         (output_move_multiple): Change how pointer adjustment is done.
939         (pdp11_gen_int_label): Correct format.
940         (output_ascii): Ditto.
941         (pdp11_asm_output_var): Add code for DEC assembler case.
942         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
943         value.
944         (legitimate_const_double_p): Ditto.
945         (pdp11_register_move_cost): Adjust for new register classes.
946         (pdp11_regno_reg_class): Ditto.
947         (expand_block_move): Delete.
948         (pushpop_regeq): New function.
949         (pdp11_legitimate_address_p): Bugfix in check for constant
950         offset.
951         (pdp11_sp_frame_offset): Delete.
952         (pdp11_reg_save_size): New helper function for new frame layout.
953         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
954         (pdp11_expand_shift): Bugfix in check for constant shift count.
955         (pdp11_shift_length): Ditto.
956         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
957         (pdp11_cmp_length): New function.
958         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
959         some compile options.
960         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
961         (CALL_USED_REGISTERS): Ditto.
962         (ELIMINABLE_REGS): Ditto.
963         (REGISTER_NAMES): Ditto.
964         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
965         constraints.
966         (REG_CLASS_NAMES): Ditto.
967         (REG_CLASS_CONTENTS): Ditto.  Also remove
968         HARD_FRAME_POINTER_REGNUM.
969         (CPU_REG_CLASS): New macro.
970         (CLASS_MAX_NREGS): Adjust for new register classes.
971         (FUNCTION_PROFILER): Make no-op.
972         (may_call_alloca): Remove unused declaration.
973         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
974         (ASM_OUTPUT_SKIP): Fix format.
975         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
976         (HARD_FRAME_POINTER_REGNUM): Remove.
977         (return): Delete.
978         (*rts): Rename.  Remove epilogue related checks.
979         (cmpsi, cmpdi): New insn.
980         (cbranch<mode>4): Change to apply to SI and DI modes as well.
981         (mov<mode>): Change constraints to enforce that push/pop
982         destination cannot use the same register as source.
983         (*mov<mode><cc_cc>): Ditto.
984         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
985         at assembly output rather than as RTL expander.
986         (zero_extendqihi2): Bugfix in check for same registers.
987         (adddi3_nocc): Bugfix in check for constant operand.
988         (addsi3_nocc): Ditto.
989         (subdi3_nocc): Ditto.
990         (subsi3_nocc): Ditto.
991         (negdi2_nocc): Copy input to pdp11_expand_operands.
992         (negsi2_nocc): Ditto.
993         (bswap2_nocc): Ditto.
994         * config/pdp11/pdp11.opt (mlra): Fix documentation.
995         * config/pdp11/t-pdp11: Use -Os.
997 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
999         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1000         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1001         (gen_comm_data): Print out CPU alias tables.
1002         (check_cpu): Match aliases when checking the CPU name.
1003         * config/arm/arm-protos.h (cpu_alias): New structure.
1004         (cpu_option): Add entry for aliases.
1005         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1006         strongarm1100 and strongarm1110.
1007         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1008         (config/arm/arm-generic.md): Remove redundant references to
1009         strongarm110, strongarm1100 and strongarm1110.
1010         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1011         Scan aliases for additional hints.
1012         (arm_parse_cpu_option_name): Also match a cpu name against the list
1013         of aliases.
1014         * config/arm/arm-tables.opt: Regenerated.
1015         * config/arm/arm-tune.md: Regenerated.
1017 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1019         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1020         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1021         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1022         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1023         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1024         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1025         * gengtype.c (open_base_files): Add omp-general.h.
1026         * gimple.c (gimple_build_omp_critical):
1027         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1028         gimple_omp_taskgroup_set_clauses.
1029         (gimple_build_omp_atomic_load): Add mo argument, call
1030         gimple_omp_atomic_set_memory_order.
1031         (gimple_build_omp_atomic_store): Likewise.
1032         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1033         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1034         instead of GSS_OMP.
1035         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1036         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1037         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1038         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1039         different value for GF_OMP_ATOMIC_NEED_VALUE.
1040         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1041         comments.
1042         (struct gimple_statement_omp_single_layout): And remove here.
1043         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1044         than gimple_statement_omp_single_layout.
1045         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1046         GIMPLE_OMP_TEAMS.
1047         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1048         (gimple_omp_subcode): Formatting fix.
1049         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1050         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1051         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1052         gimple_omp_teams_host, gimple_omp_teams_set_host,
1053         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1054         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1055         gimple_omp_taskgroup_set_clauses): New inline functions.
1056         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1057         (gimple_build_omp_atomic_store): Likewise.
1058         (gimple_omp_atomic_seq_cst_p): Remove.
1059         (gimple_omp_atomic_memory_order): New function.
1060         (gimple_omp_atomic_set_seq_cst): Remove.
1061         (gimple_omp_atomic_set_memory_order): New function.
1062         (gimple_build_omp_taskgroup): Add clauses argument.
1063         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1064         (dump_gimple_omp_task): Print taskwait with depend clauses.
1065         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1066         dump_omp_atomic_memory_order.
1067         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1068         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1069         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1070         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1071         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1072         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1073         ORT_UNTIED_TASKLOOP enumerators.
1074         (enum gimplify_defaultmap_kind): New.
1075         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1076         target_map_pointers_as_0len_arrays members, add defaultmap.
1077         (new_omp_context): Initialize defaultmap member.
1078         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1079         (maybe_fold_stmt): Don't fold even in host teams regions.
1080         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1081         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1082         ctx->omp_firstprivatize_variable.
1083         (omp_add_variable): Don't add private/firstprivate for VLAs in
1084         ORT_TASKGROUP.
1085         (omp_default_clause): Print "taskloop" rather than "task" if
1086         ORT_*TASKLOOP.
1087         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1088         Handle new defaultmap clause kinds.
1089         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1090         iterator to be lastprivate or private.  Fix up diagnostics if linear
1091         is used on collapse>1 simd iterator.
1092         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1093         (gimplify_omp_depend): New function.
1094         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1095         combined parallel master taskloop{, simd} if taskloop has
1096         firstprivate, lastprivate or reduction clause.  Handle
1097         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1098         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1099         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1100         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1101         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1102         lastprivate.
1103         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1104         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
1105         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1106         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1107         (gimplify_omp_task): Handle taskwait with depend clauses.
1108         (gimplify_omp_for): Add shared clause on parallel for combined
1109         parallel master taskloop{, simd} if taskloop has firstprivate,
1110         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1111         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1112         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1113         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1114         __for_end and __for_range temporaries on OMP_PARALLEL for
1115         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1116         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1117         sandwiched in between two taskloops.
1118         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1119         instead of ctx->omp_firstprivatize_variable.
1120         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1121         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1122         host teams, use gimplify_and_return_first etc. for body like
1123         for target or target data constructs, and at the end call
1124         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1125         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1126         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1127         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1128         gimple_omp_atomic_set_seq_cst calls.
1129         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1130         case, handle taskgroup clauses.
1131         * lto-streamer-out.c (hash_tree): Handle
1132         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1133         * Makefile.in (GTFILES): Add omp-general.h.
1134         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1135         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1136         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1137         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1138         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1139         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1140         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1141         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1142         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1143         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1144         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1145         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1146         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1147         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1148         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1149         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1150         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1151         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1152         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1153         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1154         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1155         non-worksharing loops.
1156         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1157         (determine_parallel_type): Don't combine parallel with worksharing
1158         which has _reductemp_ clause.
1159         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1160         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1161         if there is nonmonotonic modifier or if there is no modifier and no
1162         ordered clause.  For dynamic and guided schedule without monotonic
1163         and nonmonotonic modifier, default to nonmonotonic.
1164         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1165         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1166         task reductions.
1167         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1168         there are any reduction clauses.
1169         (expand_taskwait_call): New function.
1170         (expand_teams_call): New function.
1171         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1172         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1173         depend clauses.
1174         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1175         of worksharing loops with task reductions.
1176         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1177         expansion of worksharing loops with task reductions.
1178         (expand_omp_sections): Handle expansion of sections with task
1179         reductions.
1180         (expand_omp_synch): For host teams call expand_omp_taskreg.
1181         (omp_memory_order_to_memmodel): New function.
1182         (expand_omp_atomic_load, expand_omp_atomic_store,
1183         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1184         instead of gimple_omp_atomic_seq_cst_p.
1185         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1186         depend clauses as a standalone directive.
1187         * omp-general.c (enum omp_requires): New variable.
1188         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1189         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1190         GT_EXPR loops depending on incr sign.  Formatting fixes.
1191         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1192         (enum omp_requires): New enum.
1193         (omp_requires_mask): Declare.
1194         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1195         Fix comment typos.
1196         * omp-low.c (struct omp_context): Add task_reductions and
1197         task_reduction_map fields.
1198         (is_host_teams_ctx): New function.
1199         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1200         (use_pointer_for_field): Use is_global_var instead of
1201         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1202         in outer contexts.
1203         (build_outer_var_ref): Ignore taskgroup outer contexts.
1204         (delete_omp_context): Release task_reductions and task_reduction_map.
1205         (scan_sharing_clauses): Don't add any fields for reduction clause on
1206         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1207         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1208         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1209         Handle OMP_CLAUSE_NONTEMPORAL.
1210         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1211         needed.
1212         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1213         clauses with task modifier.
1214         (scan_omp_task): Handle taskwait with depend clauses.
1215         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1216         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1217         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1218         (scan_omp_for): Fix comment formatting.
1219         (scan_omp_teams): Handle host teams constructs.
1220         (check_omp_nesting_restrictions): Allow teams with no outer
1221         OpenMP context.  Adjust diagnostics for teams strictly nested into
1222         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1223         inside of simd regions.
1224         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1225         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1226         taskreg_nesting_level while scanning host teams construct.
1227         (task_reduction_read): New function.
1228         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1229         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1230         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1231         second argument create_tmp_var if it is NULL.  Don't ignore shared
1232         clauses in is_host_teams_ctx contexts.  Handle
1233         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1234         clauses.
1235         (lower_reduction_clauses): Ignore reduction clauses with task
1236         modifier.  Remove second argument create_tmp_var if it is NULL.
1237         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1238         (lower_send_clauses): Ignore reduction clauses with task modifier.
1239         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1240         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1241         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1242         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1243         contexts.
1244         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1245         lower_omp_task_reductions): New functions.
1246         (lower_omp_sections): Handle reduction clauses with taskgroup
1247         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1248         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1249         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1250         modifiers.
1251         (lower_omp_taskgroup): Handle taskgroup reductions.
1252         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1253         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1254         (lower_depend_clauses): If there are any
1255         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1256         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1257         seen, assume lowering is done already and return early.  Set kind
1258         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1259         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1260         parallel construct.  Handle reduction clause on taskloop construct.
1261         Handle taskwait with depend clauses.
1262         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1263         for host teams constructs.
1264         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1265         nontemporal and _reductemp_ clause entries.
1266         (omp_clause_code_name): Likewise.
1267         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1268         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1269         * tree-core.h (enum omp_clause_code): Add
1270         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1271         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1272         (struct tree_base): Add omp_atomic_memory_order field into union.
1273         Remove OMP_ATOMIC_SEQ_CST comment.
1274         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1275         and OMP_CLAUSE_DEPEND_DEPOBJ.
1276         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1277         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1278         OpenMP constructs with body and clauses operands.
1279         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1280         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1281         (OMP_TASKGROUP_CLAUSES): Define.
1282         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1283         OMP_CLAUSE__LOOPTEMP_.
1284         (OMP_ATOMIC_SEQ_CST): Remove.
1285         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1286         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1287         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1288         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1289         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1290         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1291         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1292         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1293         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1294         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1295         Define.
1296         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1297         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1298         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1299         (convert_local_omp_clauses): Likewise.  Remove useless test.
1300         * tree-parloops.c (create_call_for_reduction_1): Pass
1301         OMP_MEMORY_ORDER_RELAXED as new argument to
1302         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1303         * tree-pretty-print.c (dump_omp_iterators): New function.
1304         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1305         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1306         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1307         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1308         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1309         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1310         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1311         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1312         (dump_omp_atomic_memory_order): New function.
1313         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1314         taskwait with depend clauses.
1315         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1316         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1317         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1318         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1319         write_ts_omp_clause_tree_pointers): Likewise.
1321 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1323         PR ipa/86395
1324         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1325         "-missed", "-note", and "-all" sub-options.
1326         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1327         API.
1328         (can_early_inline_edge_p): Likewise.
1329         (want_early_inline_function_p): Likewise.
1330         (want_inline_self_recursive_call_p): Likewise.
1331         (recursive_inlining): Likewise.
1332         (inline_small_functions): Likewise.
1333         (flatten_function): Likewise.
1334         (ipa_inline): Likewise.
1335         (inline_always_inline_functions): Likewise.
1336         (early_inline_small_functions): Likewise.
1337         (early_inliner): Likewise.
1338         * tree-inline.c (expand_call_inline): Likewise.
1340 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1342         * pretty-print.c (pp_format): Handle %f.
1343         (selftest::test_pp_format): Add test of %f.
1344         * pretty-print.h (pp_double): New macro.
1346 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1348         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1349         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1350         (dump_pretty_printer::decode_format): Implement "%C" for
1351         cgraph_node *.
1352         (selftest::test_capture_of_dump_calls): Rename "where" to
1353         "stmt_loc".  Convert test_decl to a function decl and set its
1354         location.  Add a symbol_table_test RAII instance and a
1355         cgraph_node, using it to test "%C" and dump_symtab_node.
1357 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1359         PR middle-end/87916
1360         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1362 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1364         * cgraph.c: Include "selftest.h".
1365         (saved_symtab): New variable.
1366         (selftest::symbol_table_test::symbol_table_test): New ctor.
1367         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1368         (selftest::test_symbol_table_test): New test.
1369         (selftest::cgraph_c_tests): New.
1370         * cgraph.h (saved_symtab): New decl.
1371         (selftest::symbol_table_test): New class.
1372         * selftest-run-tests.c (selftest::run_tests): Call
1373         selftest::cgraph_c_tests.
1374         * selftest.h (selftest::cgraph_c_tests): New decl.
1376 2018-11-08  Richard Biener  <rguenther@suse.de>
1378         * tree-data-ref.h (lambda_int): New typedef.
1379         (lambda_vector_gcd): Adjust.
1380         (lambda_vector_new): Likewise.
1381         (lambda_matrix_new): Likewise.
1382         * tree-data-ref.c  (print_lambda_vector): Adjust.
1384 2018-11-08  Richard Biener  <rguenther@suse.de>
1386         PR tree-optimization/87929
1387         * tree-complex.c (expand_complex_comparison): Clean EH.
1389 2018-11-08  Martin Liska  <mliska@suse.cz>
1391         * doc/extend.texi: Reword.
1392         * predict.c (expr_expected_value_1): Likewise.
1394 2018-11-08  Richard Biener  <rguenther@suse.de>
1396         PR tree-optimization/87913
1397         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1398         of extreme values to ordered comparisons.
1400 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1402         PR middle-end/42726
1403         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1405 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1407         * doc/invoke.texi: Remove leading dash from @opindex entries
1408         throughout the file.
1410 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1412         PR driver/80828
1413         * doc/invoke.texi (Option Summary): Add -e and --entry.
1414         (Link Options): Likewise.
1416 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1418         PR 87926
1419         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1420         --disable-checking bootstrap.
1422 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1424         * configure: Regenerated.
1426 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1428         PR c/87691
1429         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1430         to the mode of the widest field iff the widest field has mode class
1431         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1432         reference.
1434 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1436         * common.opt: Add -Wattribute-warning.
1437         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1438         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1439         call to allow user configure behavior of "warning" attribute.
1441 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1443         * target.def: Put @: after every vs., e.g., and i.e. where it is
1444         followed by whitespace.
1445         * doc/extend.texi: Ditto.
1446         * doc/fragments.texi: Ditto.
1447         * doc/gimple.texi: Ditto.
1448         * doc/implement-c.texi: Ditto.
1449         * doc/install.texi: Ditto.
1450         * doc/invoke.texi: Ditto.
1451         * doc/md.texi: Ditto.
1452         * doc/plugins.texi: Ditto.
1453         * doc/rtl.texi: Ditto.
1454         * doc/sourcebuild.texi: Ditto.
1455         * doc/tm.texi.in: Ditto.
1456         * doc/ux.texi: Ditto.
1457         * doc/tm.texi: Regenerate.
1459 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1461         * config/arm/arm-cpus.in (ares): New entry.
1462         * config/arm/arm-tables.opt: Regenerate.
1463         * config/arm/arm-tune.md: Likewise.
1464         * doc/invoke.texi (ARM Options): Document ares.
1466 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1468         * config/aarch64/aarch64-cores.def (ares): Define.
1469         * config/aarch64/aarch64-tune.md: Regenerate.
1470         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1472 2018-11-07  Jan Hubicka  <jh@suse.cz>
1474         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1475         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1476         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1477         TYPE_DECL.
1479 2018-11-07  Richard Biener  <rguenther@suse.de>
1481         PR tree-optimization/87914
1482         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1483         of nested cycles.
1484         (vectorizable_reduction): Handle shifts and rotates by dispatching
1485         to vectorizable_shift.
1486         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1487         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1488         def cases.
1489         (vectorizable_shift): Export and handle being called as
1490         vect_nested_cycle.
1491         (vect_analyze_stmt): Call vectorizable_shift after
1492         vectorizable_reduction.
1493         * tree-vectorizer.h (vectorizable_shift): Declare.
1495 2018-11-07  Jan Hubicka  <jh@suse.cz>
1497         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1498         than const decls in TREE_VALUE of enum.
1499         (dump_type_inheritance_graph): Improve duplicate dumping.
1500         (free_enum_values): New.
1501         (build_type_inheritance_graph): Use it.
1502         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
1503         which are not main variants or not ODR types.
1504         (verify_type_variant): Expect variants to have no TYPE_VALUES.
1506 2018-11-07  Richard Biener  <rguenther@suse.de>
1508         * ipa-inline.c (want_inline_small_function_p): Compute
1509         big_speedup_p lazily and last.
1511 2018-11-07  Jan Hubicka  <jh@suse.cz>
1513         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
1514         building incomplete variant of complete type.
1515         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
1516         variant of complete type.
1518 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1520         * config/mips/mips.c: Fix typo in documentation of
1521         mips_loongson_ext2_prefetch_cookie.
1522         (mips_option_override): fix brain twister logical.
1523         * config/mips/mips.h: Fix typo in documentation of
1524         ISA_HAS_CTZ_CTO and define pattern.
1525         * config/mips/mips.md (prefetch): Hoist EXT2 above
1526         the 2EF/EXT block.
1527         (prefetch_indexed): Hoist EXT2 above the EXT block.
1529 2018-11-07  Jan Hubicka  <jh@suse.cz>
1531         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
1532         return and parameter types of function and method types.
1533         (free_lang_data_in_cgraph): Update.
1535 2018-11-07  Martin Liska  <mliska@suse.cz>
1537         PR rtl-optimization/87868
1538         * postreload-gcse.c (eliminate_partially_redundant_load): Set
1539         threshold to max_count if we would overflow.
1540         * profile-count.h: Make max_count a public constant.
1542 2018-11-07  Martin Liska  <mliska@suse.cz>
1544         * mem-stats.h: Fix GNU coding style.
1546 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1548         * config/mips/gs264e.md: New.
1549         * config/mips/mips-cpus.def: Define gs264e.
1550         * config/mips/mips-tables.opt: Regenerate.
1551         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1552         gs264e.
1553         (mips_issue_rate): Add support for gs264e.
1554         (mips_multipass_dfa_lookahead): Likewise.
1555         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
1556         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
1557         (MIPS_ASE_MSA_SPEC): New.
1558         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
1559         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
1560         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
1561         * config/mips/mips.md: Include gs264e.md.
1562         (processor): Add gs264e.
1563         * config/mips/mips.opt (MSA): Use Mask instead of Var.
1564         * doc/invoke.texi: Add gs264e to supported architectures.
1566 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1568         * config/mips/gs464e.md: New.
1569         * config/mips/mips-cpus.def: Define gs464e.
1570         * config/mips/mips-tables.opt: Regenerate.
1571         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1572         gs464e.
1573         (mips_issue_rate): Add support for gs464e.
1574         (mips_multipass_dfa_lookahead): Likewise.
1575         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
1576         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
1577         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
1578         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
1579         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
1580         * config/mips/mips.md: Include gs464e.md.
1581         (processor): Add gs464e.
1582         * doc/invoke.texi: Add gs464e to supported architectures.
1584 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1586         * config/mips/loongson3a.md: Rename to ...
1587         * config/mips/gs464.md: ... here.
1588         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
1589         as an alias of gs464 processor.
1590         * config/mips/mips-tables.opt: Regenerate.
1591         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
1592         instead of PROCESSOR_LOONGSON_3A.
1593         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
1594         TUNE_LOONGSON_3A.
1595         (mips_option_override): Enable MMI and EXT for gs464.
1596         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
1597         Rename TUNE_LOONGSON_3A to TUNE_GS464.
1598         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
1599         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
1600         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
1601         TARGET_LOONGSON_3A.
1602         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
1603         (processor): Add gs464;
1604         * doc/invoke.texi: Add gs464 to supported architectures.
1606 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1608         * config/mips/mips-protos.h
1609         (mips_loongson_ext2_prefetch_cookie): New prototype.
1610         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
1611         (mips_option_override): Enable TARGET_LOONGSON_EXT when
1612         TARGET_LOONGSON_EXT2 is true.
1613         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
1614         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
1615         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
1616         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
1617         TARGET_LOONGSON_EXT2.
1618         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
1619         (define_insn "ctz<mode>2"): New insn pattern.
1620         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
1621         (define_insn "prefetch_indexed_<mode>"): Include
1622         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
1623         * config/mips/mips.opt (-mloongson-ext2): Add option.
1624         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
1626 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1628         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
1629         __mips_loongson_ext.
1630         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
1631         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
1632         -mloongson-ext.
1633         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
1634         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
1635         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
1636         instead of TARGET_LOONGSON_3A.
1637         * config/mips/mips.opt (-mloongson-ext): Add option.
1638         * gcc/doc/invoke.texi (-mloongson-ext): Document.
1640 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1642         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
1643         * config/mips/loongson.md: Move to ...
1644         * config/mips/loongson-mmi.md: here; Adjustment.
1645         * config/mips/loongson.h: Move to ...
1646         State as deprecated. Include loongson-mmiintrin.h for back
1647         compatibility and warning.
1648         * config/mips/loongson-mmiintrin.h: ... here.
1649         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
1650         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
1651         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1652         (mips_option_override): Make sure MMI use hard float;
1653         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
1654         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
1655         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
1656         TARGET_LOONGSON_VECTORS.
1657         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
1658         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
1659         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
1660         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
1661         -mloongson-mmi.
1662         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
1663         TARGET_LOONGSON_VECTORS.
1664         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
1665         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1666         (Loongson MMI patterns): Include loongson-mmi.md instead of
1667         loongson.md.
1668         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
1669         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
1671 2018-11-07  Richard Biener  <rguenther@suse.de>
1673         PR lto/87906
1674         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
1675         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
1677 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
1679         PR rtl-optimization/87874
1680         * lra.c (lra_substitute_pseudo): Do not create a subreg for
1681         const wide ints.
1683 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
1685         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
1686         if not in indexed or indirect form.
1687         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
1688         (bswap<mode>2_store): Ditto.
1690 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
1692         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
1693         the UNSPEC.
1695 2018-11-06  Richard Biener  <rguenther@suse.de>
1697         PR tree-optimization/86850
1698         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
1699         instead of src.m_vec.
1701 2018-11-06  Jan Hubicka  <jh@suse.cz>
1703         * tree.c (fld_simplified_type_name): Break out form ...
1704         (free_lang_data_in_type): ... here.
1705         (fld_type_variant_equal_p): Use it.
1707 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1709         * config/default-d.c: Include memmodel.h.
1711         * config/sol2-d.c: New file.
1712         * config/t-sol2 (sol2-d.o): New rule.
1713         * config.gcc <*-*-solaris2*>: Set d_target_objs,
1714         target_has_targetdm.
1716 2018-11-06  Jan Hubicka  <jh@suse.cz>
1718         * tree.c (fld_type_variant): Also copy alignment; be sure that
1719         new variant is equal.
1721 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1723         PR target/87762
1724         * config/s390/s390.md: Add relative_long attribute.
1726 2018-11-06  Jan Hubicka  <jh@suse.cz>
1728         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
1729         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
1730         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
1731         stream TYPE_NEEDS_CONSTRUCTING.
1732         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
1733         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
1735 2018-11-06  Richard Biener  <rguenther@suse.de>
1737         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
1738         dump-scope ...
1739         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
1741 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1743         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
1744         REG_CLASS_CONTENTS[GEN_REGS].
1745         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
1747 2018-11-06  Jan Hubicka  <jh@suse.cz>
1749          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
1751 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1753         PR sanitizer/80953
1754         * config/sol2.h (ASAN_CC1_SPEC): Define.
1755         (LD_WHOLE_ARCHIVE_OPTION): Define.
1756         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
1757         (ASAN_REJECT_SPEC): Provide default.
1758         (LIBASAN_EARLY_SPEC): Define.
1759         (LIBTSAN_EARLY_SPEC): Define.
1760         (LIBLSAN_EARLY_SPEC): Define.
1761         * config/i386/sol2.h (CC1_SPEC): Redefine.
1762         (ASAN_REJECT_SPEC): Define.
1764         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
1765         (TARGET_ASAN_SHADOW_OFFSET): Define.
1766         (sparc_asan_shadow_offset): New function.
1767         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
1768         (ASAN_REJECT_SPEC): Define.
1770 2018-11-06  Jan Hubicka  <jh@suse.cz>
1772         * tree.c (fld_type_variant): Copy canonical type.
1773         (fld_incomplete_type_of): Check that canonical types looks sane;
1774         copy canonical type.
1775         (verify_type): Accept when incomplete type has complete canonical type.
1777 2018-11-06  Jan Hubicka  <jh@suse.cz>
1779         * tree.c (free_lang_data): Reset overwite_assembler_name,
1780         print_xnode, print_decl, print_type and print_identifier of
1781         langhooks.
1783 2018-11-06  Richard Biener  <rguenther@suse.de>
1785         PR tree-optimization/87889
1786         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1787         Do nothing if old and new arg are the same
1789 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
1791         PR target/87723
1792         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
1793         attributes for operands 3 and 4.
1795 2018-11-06  Richard Biener  <rguenther@suse.de>
1797         PR middle-end/18041
1798         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
1799         matching bitfield insertion.
1801 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
1803         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
1804         (reg_next_debug_use): New.
1805         (attempt_change): Propagate adjusted expression into affected
1806         debug insns.
1807         (merge_in_block): Track uses in debug insns.
1808         (pass_inc_dec::execute): Allocate and release
1809         reg_next_debug_use.
1811 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
1813         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
1814         (_mm512_fixupimm_round_pd): Update parameters and builtin.
1815         (_mm512_maskz_fixupimm_round_pd): Ditto.
1816         (_mm512_fixupimm_round_ps): Ditto.
1817         (_mm512_maskz_fixupimm_round_ps): Ditto.
1818         (_mm_fixupimm_round_sd): Ditto.
1819         (_mm_maskz_fixupimm_round_sd): Ditto.
1820         (_mm_fixupimm_round_ss): Ditto.
1821         (_mm_maskz_fixupimm_round_ss): Ditto.
1822         (_mm512_fixupimm_pd): Ditto.
1823         (_mm512_maskz_fixupimm_pd): Ditto.
1824         (_mm512_fixupimm_ps): Ditto.
1825         (_mm512_maskz_fixupimm_ps): Ditto.
1826         (_mm_fixupimm_sd): Ditto.
1827         (_mm_maskz_fixupimm_sd): Ditto.
1828         (_mm_fixupimm_ss): Ditto.
1829         (_mm_maskz_fixupimm_ss): Ditto.
1830         (_mm512_mask_fixupimm_round_pd): Update builtin.
1831         (_mm512_mask_fixupimm_round_ps): Ditto.
1832         (_mm_mask_fixupimm_round_sd): Ditto.
1833         (_mm_mask_fixupimm_round_ss): Ditto.
1834         (_mm512_mask_fixupimm_pd): Ditto.
1835         (_mm512_mask_fixupimm_ps): Ditto.
1836         (_mm_mask_fixupimm_sd): Ditto.
1837         (_mm_mask_fixupimm_ss): Ditto.
1838         * config/i386/avx512vlintrin.h:
1839         (_mm256_fixupimm_pd): Update parameters and builtin.
1840         (_mm256_maskz_fixupimm_pd): Ditto.
1841         (_mm256_fixupimm_ps): Ditto.
1842         (_mm256_maskz_fixupimm_ps): Ditto.
1843         (_mm_fixupimm_pd): Ditto.
1844         (_mm_maskz_fixupimm_pd): Ditto.
1845         (_mm_fixupimm_ps): Ditto.
1846         (_mm_maskz_fixupimm_ps): Ditto.
1847         (_mm256_mask_fixupimm_pd): Update builtin.
1848         (_mm256_mask_fixupimm_ps): Ditto.
1849         (_mm_mask_fixupimm_pd): Ditto.
1850         (_mm_mask_fixupimm_ps): Ditto.
1851         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
1852         * config/i386/i386-builtin.def: Update builtin definitions.
1853         * config/i386/i386.c: Handle new builtin types and remove useless ones.
1854         * config/i386/sse.md: Update VFIXUPIMM* patterns.
1855         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1856         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1857         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
1858         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1859         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1860         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
1861         * config/i386/subst.md:
1862         (round_saeonly_sd_mask_operand4): Add new subst_attr.
1863         (round_saeonly_sd_mask_op4): Ditto.
1864         (round_saeonly_expand_operand5): Ditto.
1865         (round_saeonly_expand): Update.
1867 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
1869         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1871 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
1873         PR rtl-optimization/87871
1874         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
1876 2018-11-05  Paul Koning  <ni1d@arrl.net>
1878         * doc/sourcebuild.texi (target attributes): Document new "inf"
1879         effective target keyword.
1881 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1883         * config/s390/s390.c (s390_register_move_cost): Increase costs for
1884         moves involving the CC reg.
1886 2018-11-05  Richard Biener  <rguenther@suse.de>
1888         PR tree-optimization/87873
1889         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
1890         argument.
1891         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
1892         * tree-vect-loop.c (vect_transform_loop): When splitting the
1893         loop exit also create forwarder PHIs for constants.
1894         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1895         Handle constant to_arg, add extra checking we match up the correct
1896         PHIs.
1898 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1900         * config/s390/s390.md: QImode and HImode for load on condition.
1902 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1904         * config/s390/predicates.md: Fix typo.
1905         * config/s390/s390.md: Allow immediates for load on condition.
1907 2018-11-05  Martin Liska  <mliska@suse.cz>
1909         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
1910         * bitmap.h (struct bitmap_usage): Likewise.
1911         * ggc-common.c (SCALE): Remove.
1912         (LABEL): Likewise.
1913         (struct ggc_usage): Use SIZE_AMOUNT. And update
1914         compare method.
1915         * ggc-page.c (SCALE): Remove.
1916         (STAT_LABEL): Remove.
1917         (ggc_print_statistics): Use SIZE_AMOUNT.
1918         * gimple.h (SCALE): Remove.
1919         (LABEL): Likewise.
1920         * input.c (ONE_K): Remove.
1921         (ONE_M): Likewise.
1922         (SCALE): Likewise.
1923         (STAT_LABEL): Likewise.
1924         (FORMAT_AMOUNT): Likewise.
1925         (dump_line_table_statistics): Use SIZE_AMOUNT.
1926         * mem-stats.h (struct mem_usage): Likewise.
1927         * rtl.c (dump_rtx_statistics): Likewise.
1928         (rtx_alloc_counts): Change type to size_t.
1929         (rtx_alloc_sizes): Likewise.
1930         (rtx_count_cmp): New.
1931         (dump_rtx_statistics): Sort first based on counts.
1932         * tree.c (tree_nodes_cmp): New.
1933         (tree_codes_cmp): New.
1934         (dump_tree_statistics): Sort first based on counts.
1935         * system.h (ONE_K): New.
1936         (ONE_M): Likewise.
1937         (SIZE_SCALE): Likewise.
1938         (SIZE_LABEL): Likewise.
1939         (SIZE_AMOUNT): Likewise.
1940         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
1941         * tree-dfa.c (dump_dfa_stats): Likewise.
1942         * tree-phinodes.c (phinodes_print_statistics): Likewise.
1943         * tree-ssanames.c (ssanames_print_statistics): Likewise.
1944         * tree.c (dump_tree_statistics): Likewise.
1945         * vec.c (struct vec_usage): Likewise.
1946         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
1947         trigger a -Werror=format-overflow with
1948         --enable-gather-detailed-stats.
1950 2018-11-05  Martin Liska  <mliska@suse.cz>
1952         * mem-stats.h (mem_alloc_description::release_instance_overhead):
1953         Return T *.
1954         * vec.c (struct vec_usage): Register m_element_size.
1955         (vec_prefix::register_overhead): New arguments: elements and
1956         element_size.
1957         (vec_prefix::release_overhead): Subtract elements.
1958         * vec.h (struct vec_prefix): Change signature.
1959         (va_heap::reserve): Pass proper arguments.
1960         (va_heap::release): Likewise.
1962 2018-11-05  Martin Liska  <mliska@suse.cz>
1964         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
1965         style.
1966         * vec.c: Likewise.
1968 2018-11-05  Richard Biener  <rguenther@suse.de>
1970         * tree-scalar-evolution.h (final_value_replacement_loop): Update
1971         prototype.
1972         * tree-scalar-evolution.c (final_value_replacement_loop): Return
1973         whether anything was done.
1974         (scev_const_prop): Remove constant propagation part, fold
1975         remains into ...
1976         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
1977         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
1978         conditionally.
1980 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
1982         PR tree-optimization/87859
1983         * gimple-ssa-store-merging.c (struct merged_store_group): Add
1984         only_constants and first_nonmergeable_order members.
1985         (merged_store_group::merged_store_group): Initialize them.
1986         (merged_store_group::do_merge): Clear only_constants member if
1987         adding something other than INTEGER_CST store.
1988         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
1989         stores with order >= first_nonmergeable_order.  Use
1990         merged_store->only_constants instead of always recomputing it.
1991         Set merged_store->first_nonmergeable_order if we've skipped any
1992         stores.  Attempt to merge overlapping INTEGER_CST stores that
1993         we would otherwise skip.
1995         PR sanitizer/87837
1996         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
1998 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2000         PR target/87853
2001         * config/i386/emmintrin.h (__v16qs): New to cope with option
2002         -funsigned-char.
2003         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2004         (_mm_cmplt_epi8): Likewise.
2005         (_mm_cmpgt_epi8): Likewise.
2007 2018-11-05  Richard Biener  <rguenther@suse.de>
2009         PR rtl-optimization/87852
2010         * fwprop.c (use_killed_between): Only consider single-defs of the
2011         use whose definition statement dominates the use.
2013 2018-11-05  Martin Liska  <mliska@suse.cz>
2015         PR web/87829
2016         * doc/invoke.texi: Remove options that are
2017         not disabled with -Os.
2019 2018-11-05  Martin Liska  <mliska@suse.cz>
2021         PR c/87811
2022         * doc/extend.texi: Update constrain about the last argument
2023         of __builtin_expect_with_probability.
2025 2018-11-05  Martin Liska  <mliska@suse.cz>
2027         PR c/87811
2028         * predict.c (expr_expected_value_1): Verify
2029         that last argument is a real constants and emit
2030         error.
2032 2018-11-05  Martin Liska  <mliska@suse.cz>
2034         PR gcov-profile/77698
2035         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2036         only in LTO mode.
2038 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2040         PR tree-optimization/86572
2041         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2043 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2045         PR tree-optimization/87672
2046         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2047         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2049 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2051         PR middle-end/58372
2052         * cfgexpand.c (pass_expand::execute): Move the call to
2053         finish_eh_generation in front of the call to expand_stack_alignment.
2055 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2057         * common/config/i386/i386-common.c (processor_alias_table): Add
2058         znver2 entry.
2059         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2060         (case ${target}): Add znver2.
2061         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2062         -march=native recognize znver2 processors.
2063         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2064         * config/i386/i386.c (m_znver2): New definition.
2065         (m_ZNVER): New definition.
2066         (m_AMD_MULTIPLE): Includes m_znver2.
2067         (processor_cost_table): Add znver2 entry.
2068         (processor_target_table): Add znver2 entry.
2069         (get_builtin_code_for_version): Set priority for
2070         PROCESSOR_ZNVER2.
2071         (processor_model): Add M_AMDFAM17H_ZNVER2.
2072         (arch_names_table): Ditto.
2073         (ix86_reassociation_width): Include znver2.
2074         * config/i386/i386.h (TARGET_znver2): New definition.
2075         (struct ix86_size_cost): Add TARGET_ZNVER2.
2076         (enum processor_type): Add PROCESSOR_ZNVER2.
2077         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2078         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2079         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2080         (ix86_adjust_cost): Add znver2.
2081         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2082         * gcc/doc/extend.texi: Add details about znver2.
2083         * gcc/doc/invoke.texi: Add details about znver2.
2085 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2087         PR target/87079
2089         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2090         pattern.
2092 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2094         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2095         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2097 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2099         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2100         shrink-wrapping when -mtrack-speculation.
2102 2018-11-02  Richard Biener  <rguenther@suse.de>
2104         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2105         (pop_cost_one_pair): Do not free pair.
2106         (pop_best_coalesce): Likewise.
2107         (create_coalesce_list): Initialize obstack.
2108         (delete_coalesce_list): Free obstack.
2109         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2110         (add_cost_one_coalesce): Likewise.
2111         (struct live_track): Remove bitmap pointer indirections.
2112         (new_live_track): Adjust.
2113         (delete_live_track): Likewise.
2114         (live_track_remove_partition): Likewise.
2115         (live_track_add_partition): Likewise.
2116         (live_track_live_p): Likewise.
2117         (live_track_process_def): Likewise.
2118         (live_track_clear_base_vars): Likewise.
2120 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2122         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2123         * configure: Regenerate.
2124         * config.in: Regenerate.
2125         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2126         HAVE_AS_REGISTER_PSEUDO_OP guard.
2127         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2128         !HAVE_AS_REGISTER_PSEUDO_OP support.
2129         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2130         guard.
2132 2018-11-02  Richard Biener  <rguenther@suse.de>
2134         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2135         with is_type_die.
2137 2018-11-02  Richard Biener  <rguenther@suse.de>
2139         PR tree-optimization/87776
2140         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2141         executable when iterating but running into rpo-vn-max-loop-depth
2142         and not eliding the iteration.
2144 2018-11-30  Jan Hubicka  <jh@suse.cz>
2146         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2147         TYPE_DECL.
2149 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2151         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2152         prototype.
2153         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2154         Combine with rs6000_address_for_fpconvert.
2155         (rs6000_address_for_fpconvert): Combine with
2156         rs6000_force_indexed_or_indirect_mem.
2157         (rs6000_expand_vector_init): Change function call from
2158         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2159         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2160         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2161         (floatsi<mode>2_lfiwax_mem): Ditto.
2162         (floatunssi<mode>2_lfiwzx): Ditto.
2163         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2164         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2165         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2166         (fix_trunc<mode>si2_stfiwx): Ditto.
2167         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2168         (float_<mode>si2_hw): Ditto.
2169         (floatuns_<mode>si2_hw): Ditto.
2170         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2171         (vsx_splat_<mode>): Ditto.
2173 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2175         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2176         at end of assembler input text.
2177         * configure: Regenerate.
2179 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2181         PR tree-optimization/87826
2182         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2183         negative or larger or equal to type's precision.
2185 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2187         * opts.c (default_options_table): Do not enable
2188         OPT_fdelayed_branch at -Og.
2189         * doc/invoke.texi (-fdelayed-branch): Document it.
2191 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2193         * optabs-libfuncs.c (build_libfunc_function_visibility):
2194         New, split out from...
2195         (build_libfunc_function): ... here.
2196         (init_one_libfunc_visibility): New, split out from ...
2197         (init_one_libfunc): ... here.
2199         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2200         scratch register need not be early-clobber.  Document the reason
2201         why we cannot use ST<OP>.
2203 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2205         PR bootstrap/82856
2206         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2207         line for second argument of AC_DEFINE_UNQUOTED.
2208         * doc/install.texi (Tools/packages necessary for modifying GCC):
2209         Update to autoconf 2.69 and automake 1.15.1.
2210         * aclocal.m4, config.in, configure: Regenerate.
2212 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2214         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2215         initialization.
2217 2018-10-31  Martin Liska  <mliska@suse.cz>
2219         PR driver/83193
2220         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2221         Add new argument invalid_extension.
2222         (aarch64_get_all_extension_candidates): New function.
2223         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2224         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2225         new argument.
2226         (aarch64_get_all_extension_candidates): New function.
2227         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2228         argument invalid_extension.
2229         (aarch64_parse_cpu): Likewise.
2230         (aarch64_print_hint_for_extensions): New function.
2231         (aarch64_validate_mcpu): Provide hint about invalid extension.
2232         (aarch64_validate_march): Likewise.
2233         (aarch64_handle_attr_arch): Pass new argument.
2234         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2235         (aarch64_handle_attr_isa_flags): Likewise.
2237 2018-10-31  Richard Biener  <rguenther@suse.de>
2239         PR middle-end/70359
2240         PR middle-end/86270
2241         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2242         copy generation to useful cases.  Place the copy before
2243         the definition of the backedge value when possible.
2245 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2247         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2248         * config/arc/arc.c (arc_active_insn): New function.
2249         (check_store_cacheline_hazard): Likewise.
2250         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2251         (arc_override_options): Disable delay slot scheduler for older
2252         A7.
2253         (arc_store_addr_hazard_p): New implementation, old one renamed to
2254         ...
2255         (arc_store_addr_hazard_internal_p): Renamed.
2256         (arc_reorg): Don't combine into brcc instructions which are part
2257         of hardware hazard solution.
2258         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2259         (tune_arc700): Likewise.
2260         * config/arc/arc.opt (arc7xx): New tune value.
2261         * config/arc/arc700.md: Improve A7 scheduler.
2263 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2265         * config/arc/arc.c (arc_override_options): Remove
2266         TARGET_COMPACT_CASESI.
2267         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2268         (CASE_VECTOR_MODE): Likewise.
2269         (CASE_VECTOR_PC_RELATIVE): Likewise.
2270         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2271         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2272         (ADDR_VEC_ALIGN): Update.
2273         (ASM_OUTPUT_CASE_LABEL): Undefine.
2274         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2275         (TARGET_BI_BIH): Define.
2276         (DEFAULT_BRANCH_INDEX): Likewise.
2277         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2278         instructions, remove compact_casesi use case.
2279         (casesi_compact_jump): Remove.
2280         (casesi_dispatch): New pattern.
2281         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2282         compact_casesi option.
2283         * doc/invoke.texi: Document mbranch-index option.
2285 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2287         * config/arc/arc.c (arc_get_tp): Remove function.
2288         (arc_emit_call_tls_get_addr): Likewise.
2289         (arc_call_tls_get_addr): New function.
2290         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2291         * config/arc/arc.md (tls_load_tp_soft): Remove.
2292         (tls_gd_get_addr): Likewise.
2294 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2296         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2297         (mulsi3_600_lib): Remove pattern.
2298         (umulsi3_highpart_600_lib_le): Likewise.
2299         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2300         (umulsidi3): Remove call to umulsidi3_600_lib.
2301         (umulsidi3_600_lib): Remove pattern.
2302         (peephole2): Remove peephole using the above deprecated patterns.
2304 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2306         PR target/87374
2307         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2308         use of -mslow-flash-data and -mword-relocations.
2309         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2310         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2311         flag_pic.
2312         * doc/invoke.texi (-mword-relocations): Mention conflict with
2313         -mslow-flash-data.
2314         (-mslow-flash-data): Reciprocally.
2316 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2318         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2319         16-byte modes held in GP registers to use an even regno.
2321         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2322         (aarch64_atomic_ldop_supported_p): Remove.
2323         (aarch64_gen_atomic_ldop): Remove.
2324         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2325         Fully expand LSE operations here.
2326         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2327         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2328         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2329         and use ATOMIC_LDOP instead; use register_operand for the input;
2330         drop the split and emit insns directly.
2331         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2332         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2333         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2335         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2336         (aarch64_gen_atomic_ldop): Don't call it.
2337         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2338         Use aarch64_reg_or_zero.
2339         (aarch64_atomic_exchange<ALLI>): Likewise.
2340         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2341         operand 0; use aarch64_reg_or_zero for input; merge ...
2342         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2344         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2345         (aarch64_split_compare_and_swap): Use it.
2346         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2347         test oldval against the proper predicate.
2348         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2349         Use nonmemory_operand for expected.
2350         (cas_short_expected_pred): New.
2351         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2352         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2353         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2354         (aarch64_plushi_operand): New.
2356         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2357         Force oldval into the rval register for TARGET_LSE; emit the compare
2358         during initial expansion so that it may be deleted if unused.
2359         (aarch64_gen_atomic_cas): Remove.
2360         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2361         Change =&r to +r for operand 0; use match_dup for operand 2;
2362         remove is_weak and mod_f operands as unused.  Drop the split
2363         and merge with...
2364         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2365         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2366         (@aarch64_atomic_cas<GPI>): Similarly.
2368 2018-10-31  Richard Biener  <rguenther@suse.de>
2370         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2371         using ABSU_EXPR.
2373 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2375         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2376         * config/aarch64/aarch64.md: Include saphira.md
2377         * config/aarch64/saphira.md: New file for pipeline description.
2379 2018-10-30  Martin Sebor  <msebor@redhat.com>
2381         PR middle-end/87041
2382         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2383         inlining context.
2384         (sprintf_dom_walker::compute_format_length):
2385         Avoid setting POSUNDER4K here.
2386         (get_destination_size): Handle null argument values.
2387         (get_user_idx_format): New function.
2388         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2389         functions, including user-defined with attribute format printf.
2390         Use %G to include inlining context.
2391         Set POSUNDER4K here.
2393 2018-10-30  Jan Hubicka  <jh@suse.cz>
2395         * params.def (lto-partitions): Bump from 32 to 128.
2397 2018-10-30  Jan Hubicka  <jh@suse.cz>
2399         * tree.c
2400         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2401         head in file.
2402         (free_lang_data_in_type): Forward declare.
2403         (fld_type_variant_equal_p): New function.
2404         (fld_type_variant): New function
2405         (fld_incomplete_types): New hash.
2406         (fld_incomplete_type_of): New function
2407         (fld_simplfied-type): New function.
2408         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2409         (free_lang_data): Allocate and free fld_incomplete_type; update call
2410         of free_lang_data_in_decl.
2412 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2414         * gcov.c (output_lines): Remove duplicate line.
2416 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2418         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2419         if not in indexed or indirect form.
2420         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2421         (bswapdi2_store): Ditto.
2422         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2423         helper function.
2424         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2425         Prototype for helper function.
2427 2018-10-30  Martin Sebor  <msebor@redhat.com>
2429         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2430         (target, pragma GCC optimize, pragma GCC target): Ditto.
2432 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2434         * doc/extend.texi: Fix prototype and description of
2435         __builtin_expect_with_probability.
2437 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2439         * cgraph.h (clone_function_name_1): Replaced by new
2440           clone_function_name_numbered that takes name as string; for
2441           privatize_symbol_name_1 use only.
2442           (clone_function_name): Renamed to
2443           clone_function_name_numbered to be explicit about numbering.
2444           (clone_function_name): New two-argument function that does
2445           not number its output.
2446           (clone_function_name): New three-argument function that
2447           takes a number to append to its output.
2448         * cgraphclones.c (duplicate_thunk_for_node):
2449           (clone_function_name_1): Renamed.
2450           (clone_function_name_numbered): Two new functions.
2451           (clone_function_name): Improved documentation.
2452           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2453         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2454         * final.c (final_scan_insn_1): Use the new clone_function_name
2455           without numbering.
2456         * multiple_target.c (create_dispatcher_calls): Ditto.
2457           (create_target_clone): Ditto.
2458         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2459         * omp-low.c (create_omp_child_function_name): Ditto.
2460         * omp-simd-clone.c (simd_clone_create): Ditto.
2461         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2462           new clone_function_name without numbering.
2464 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2466         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2467         Assert that the allocation size is not zero.
2469 2018-10-30  Richard Biener  <rguenther@suse.de>
2471         PR tree-optimization/87800
2472         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2473         non-induction or reduction PHIs.
2475 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2477         * config/aarch64/falkor-tag-collision-avoidance.c
2478         (execute_tag_collision_avoidance): Call df_note_add_problem.
2480 2018-10-30  Martin Liska  <mliska@suse.cz>
2482         * doc/extend.texi: Fix typo in documentation
2483         of __builtin_expect_with_probability.
2485 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2487         PR c++/87721
2488         * input.c (get_substring_ranges_for_loc): Detect if
2489         linemap_resolve_location gives us a NULL map, and reject
2490         this case.
2492 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2494         * config.gcc (xstormy16-*-elf): Set tm_d_file.
2496 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2497             Martin Sebor  <msebor@redhat.com>
2498             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2500         * doc/ux.texi (Quoting): New subsection, adapted from material at
2501         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
2502         MartinSebor and ManuelLopezIbanez.
2503         (Fix-it hints): Note that fix-it hints shouldn't be marked for
2504         translation.
2506 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
2508         PR middle-end/87469
2509         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
2510         max value.
2512 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2514         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
2516 2018-10-29  Paul Koning  <ni1d@arrl.net>
2518         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
2520 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2522         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
2523         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
2524         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
2525         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
2526         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
2527         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
2528         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
2529         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
2530         Change 'vector' to '__vector'.
2531         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
2532         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
2533         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
2534         _mm_avg_pu16): Likewise.  And, whitespace corrections.
2536 2018-10-29  Richard Biener  <rguenther@suse.de>
2538         PR tree-optimization/87785
2539         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
2540         internal defs.
2542 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2544         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
2546 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2548         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
2549         ports configuration macro, defaults to "gnu".
2550         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
2551         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
2553 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2555         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
2557 2018-10-29  Richard Biener  <rguenther@suse.de>
2559         PR tree-optimization/87790
2560         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
2561         (vect_make_slp_decision): Adjust.
2562         (vect_slp_analyze_bb_1): Likewise.
2563         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
2564         edges.
2566 2018-10-29  Richard Biener  <rguenther@suse.de>
2568         PR tree-optimization/87785
2569         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
2570         and processing.
2571         (vect_build_slp_tree): Likewise.
2572         (vect_gather_slp_loads): New function.
2573         (vect_analyze_slp_instance): Gather loads separately from the
2574         SLP tree build.
2576 2018-10-29  Martin Liska  <mliska@suse.cz>
2578         * Makefile.in: Make dependency to json.o.
2579         * doc/gcov.texi: Document new JSON format, remove
2580         old intermediate format documentation.
2581         * gcov.c (struct function_info): Come up with m_name and
2582         m_demangled_name.
2583         (function_info::function_info): Initialize it.
2584         (function_info::~function_info): Release it.
2585         (main): Rename flag_intermediate_format to flag_json_format.
2586         (print_usage): Describe --json-format.
2587         (process_args): Set flag_json_format.
2588         (output_intermediate_line): Remove.
2589         (output_intermediate_json_line): Likewise.
2590         (get_gcov_intermediate_filename): Return new extension
2591         ".gcov.json.gz".
2592         (output_intermediate_file): Implement JSON emission.
2593         (output_json_intermediate_file): Implement JSON emission.
2594         (generate_results): Use ::get_name for function name.
2595         Handle JSON output file.
2596         (read_graph_file): Use ::get_name instead of cplus_demangle.
2597         (read_count_file): Likewise.
2598         (solve_flow_graph): Likewise.
2599         (add_line_counts): Likewise.
2600         (accumulate_line_counts): Use new flag_json_format.
2601         (output_function_details): Use ::get_name instead of cplus_demangle.
2602         (output_lines): Likewise.
2603         * json.cc (test_writing_literals): Add new tests.
2604         * json.h (class literal): Add new boolean constructor.
2606 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
2608         PR rtl-optimization/87701
2609         PR rtl-optimization/87780
2610         * combine.c (make_more_copies): Rewrite.
2612 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
2614         * doc/generic.texi (ABSU_EXPR): Document.
2615         * match.pd (absu(x)*absu(x) -> x*x): Handle.
2616         (absu(absu(X)) -> absu(X)): Likewise.
2617         (absu(-X) -> absu(X)): Likewise.
2618         (absu(X)  where X is nonnegative -> X): Likewise.
2620 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
2622         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
2623         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
2624         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
2625         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
2626         (s-tm-texi): Also check timestamp on d-target.def.
2627         (generated_files): Add TM_D_H and d-target-hooks-def.h.
2628         (build/genhooks.o): Also depend on D_TARGET_DEF.
2629         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
2630         variables.
2631         * config/aarch64/aarch64-d.c: New file.
2632         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
2633         Define.
2634         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
2635         prototype.
2636         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
2637         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
2638         * config/arm/arm-d.c: New file.
2639         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
2640         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
2641         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2642         * config/arm/t-arm (arm-d.o): New rule.
2643         * config/default-d.c: New file.
2644         * config/glibc-d.c: New file.
2645         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2646         * config/i386/i386-d.c: New file.
2647         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
2648         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
2649         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2650         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
2651         * config/i386/t-i386 (i386-d.o): New rule.
2652         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2653         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2654         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
2655         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2656         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2657         * config/mips/mips-d.c: New file.
2658         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
2659         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
2660         * config/mips/t-mips (mips-d.o): New rule.
2661         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2662         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2663         * config/powerpcspe/powerpcspe-d.c: New file.
2664         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
2665         New prototype.
2666         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2667         Support GNU D by using 0 as the language type.
2668         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
2669         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
2670         * config/riscv/riscv-d.c: New file.
2671         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
2672         prototype.
2673         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
2674         * config/riscv/t-riscv (riscv-d.o): New rule.
2675         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2676         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2677         * config/rs6000/rs6000-d.c: New file.
2678         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
2679         prototype.
2680         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2681         Support GNU D by using 0 as the language type.
2682         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
2683         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
2684         * config/s390/s390-d.c: New file.
2685         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
2686         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
2687         * config/s390/t-s390 (s390-d.o): New rule.
2688         * config/sparc/sparc-d.c: New file.
2689         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
2690         prototype.
2691         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
2692         * config/sparc/t-sparc (sparc-d.o): New rule.
2693         * config/t-glibc (glibc-d.o): New rule.
2694         * configure: Regenerated.
2695         * configure.ac (tm_d_file): New variable.
2696         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
2697         * doc/contrib.texi (Contributors): Add self for the D frontend.
2698         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
2699         * doc/install.texi (Configuration): Mention libphobos as an option for
2700         --enable-shared.  Mention d as an option for --enable-languages.
2701         (Testing): Mention check-d as a target.
2702         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
2703         name suffixes.  Mention d as a -x option.
2704         * doc/sourcebuild.texi (Top Level): Mention libphobos.
2705         * doc/standards.texi (Standards): Add section on D language.
2706         * doc/tm.texi: Regenerated.
2707         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
2708         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
2709         * dwarf2out.c (is_dlang): New function.
2710         (gen_compile_unit_die): Use DW_LANG_D for D.
2711         (declare_in_namespace): Return module die for D, instead of adding
2712         extra declarations into the namespace.
2713         (gen_namespace_die): Generate DW_TAG_module for D.
2714         (gen_decl_die): Handle CONST_DECLSs for D.
2715         (dwarf2out_decl): Likewise.
2716         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
2717         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
2718         kinds of aggregates.
2719         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
2720         * genhooks.c: Include d/d-target.def.
2722 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
2724         PR target/85669
2725         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2726         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2727         (STACK_DYNAMIC_OFFSET): Likewise.
2729 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
2731         PR target/80024
2732         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
2733         error message.
2735 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
2737         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
2738         return type and other typos.
2740 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
2742         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2743         a shorter sequence with fewer branches.
2744         (emit_final_str_compare_gpr): Ditto.
2746 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2748         * config/rs6000/tmmintrin.h: New file.
2749         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
2751 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2753         * config/rs6000/mmintrin.h: Enable 32bit compilation.
2754         * config/rs6000/xmmintrin.h: Likewise.
2756 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2758         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2760 2018-10-26  Richard Biener  <rguenther@suse.de>
2762         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
2763         and wrapper.
2764         (vect_mark_slp_stmts_relevant): Likewise.
2765         (vect_detect_hybrid_slp_stmts): Likewise.
2766         (vect_bb_slp_scalar_cost): Likewise.
2767         (vect_remove_slp_scalar_calls): Likewise.
2769 2018-10-26  Jan Hubicka  <jh@suse.cz>
2771         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
2772         (warn_types_mismatch): Fix walk of DECL_NAME.
2773         (odr_types_equivalent_p): Fix overactive assert.
2775 2018-10-26  Richard Biener  <rguenther@suse.de>
2777         PR tree-optimization/87105
2778         * tree-vectorizer.h (_slp_tree::refcnt): New member.
2779         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
2780         refcnt.
2781         (vect_create_new_slp_node): Initialize refcnt to one.
2782         (bst_traits): Move.
2783         (scalar_stmts_set_t, bst_fail): Remove.
2784         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
2785         (vect_build_slp_tree): Add bst_map argument and lookup
2786         already created SLP nodes.
2787         (vect_print_slp_tree): Handle a SLP graph, print SLP node
2788         addresses.
2789         (vect_slp_rearrange_stmts): Handle a SLP graph.
2790         (vect_analyze_slp_instance): Adjust and free SLP nodes from
2791         the CSE map.  Fix indenting.
2792         (vect_schedule_slp_instance): Add short-cut.
2794 2018-10-26  Martin Liska  <mliska@suse.cz>
2796         PR testsuite/86158
2797         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
2798         addr_expr and not with pointers.
2800 2018-10-26  Jan Hubicka  <jh@suse.cz>
2802         * tree.c (free_lang_data_in_type): Only check main variants.
2803         * ipa-devirt.c (warn_odr): Make static.
2804         (types_same_for_odr): Drop strict variant.
2805         (types_odr_comparable): Likewise.
2806         (odr_or_derived_type_p): Look for main variants.
2807         (odr_name_hasher::equal): Cleanup comment.
2808         (odr_subtypes_equivalent): Add warn and warned arguments; check main
2809         variants.
2810         (type_variants_equivalent_p): break out from ...
2811         (odr_types_equivalent): ... here; go for main variants where needed.
2812         (warn_odr): ... here; turn static.
2813         (warn_types_mismatch): Compare mangled names of main variants.
2814         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
2815         (type_with_linkage_p): Sanity check that we look at main variant.
2816         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
2817         * tree.h (types_same_for_odr): Drop strict argument.
2819 2018-10-26  Richard Biener  <rguenther@suse.de>
2821         PR tree-optimization/87746
2822         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
2823         Simplify and fix WRT strided store groups with size not
2824         equal to step in element count.
2825         (vect_analyze_group_access_1): Dump the whole group.
2827 2018-10-25  Carl Love  <cel@us.ibm.com>
2829         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
2830         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
2831         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
2832         precicion entry for each overloaded builtin.
2833         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
2834         VSCEDPUO): Rename overloaded name.
2835         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
2836         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
2837         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
2838         define_expand for xscmpexqp instruction.
2839         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
2841 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2842             Jinsong Ji <jji@us.ibm.com>
2844         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
2845         function with vec_sl.
2846         (_mm_slli_epi32): Likewise.
2847         (_mm_slli_epi64): Likewise.
2848         (_mm_srai_epi16): Replace deprecated function with vec_sra.
2849         (_mm_srai_epi32): Likewise.
2850         (_mm_srli_epi16): Replace deprecated function with vec_sr.
2851         (_mm_srli_epi32): Likewise.
2852         (_mm_srli_epi64): Likewise.
2853         (_mm_sll_epi16): Replace deprecated function with vec_sl.
2854         (_mm_sll_epi32): Likewise.
2855         (_mm_sll_epi64): Likewise.
2856         (_mm_sra_epi16): Replace deprecated function with vec_sra.
2857         (_mm_sra_epi32): Likewise.
2858         (_mm_srl_epi16): Replace deprecated function with vec_sr.
2859         (_mm_srl_epi32): Likewise.
2860         (_mm_srl_epi64): Likewise.
2862 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2863             Jinsong Ji <jji@us.ibm.com>
2865         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
2866         comparison operators with vec_cmp* for compatibility due to
2867         unfortunate history; clean up formatting and use types more
2868         appropriately.
2869         (_mm_sll_epi32): Likewise.
2870         (_mm_sll_epi64): Likewise.
2871         (_mm_srl_epi16): Likewise.
2872         (_mm_srl_epi32): Likewise.
2873         (_mm_srl_epi64): Likewise.
2875 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2876             Jinsong Ji <jji@us.ibm.com>
2878         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
2879         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
2880         __vector __bool int.  Use vec_cmpgt in preference to deprecated
2881         function vec_vcmpgtfp.
2882         (_mm_max_ps): Likewise.
2884 2018-10-25  Jeff Law  <law@redhat.com>
2886         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
2887         if returning false.
2889 2018-10-25  Martin Sebor  <msebor@redhat.com>
2891         * doc/extend.texi (aligned): Expand attribute description.
2892         (Alignment): Rename section.  Discuss function arguments.
2894 2018-10-25  Jan Hubicka  <jh@suse.cz>
2896         * ipa-devirt.c (main_odr_variant): Remove.
2897         (hash_odr_name, types_same_for_odr, types_odr_comparable,
2898         odr_name_hasher::equal, odr_subtypes_equivalent_p):
2899         Drop use of main_odr_variant.
2900         (add_type_duplicate): Silence confused warnings on integer types.
2901         (get_odr_type): Always look for main variant.
2902         (register_odr_type): Simplify.
2904 2018-10-25  Richard Biener  <rguenther@suse.de>
2906         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
2907         Initialize ng to silence error with release checking bootstrap.
2909 2018-10-25  Richard Biener  <rguenther@suse.de>
2911         * tree-if-conv.c: Include tree-ssa-sccvn.h.
2912         (tree_if_conversion): Run CSE on the if-converted loop body.
2914 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2916         * config/s390/constraints.md (ZL): New constraint.
2917         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
2918         operands.
2919         * config/s390/s390.md (movdi_larl): Remove.
2920         (movdi_64): Add the LARL alternative.
2922 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2924         PR bootstrap/87747
2925         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
2926         (RTX_CODE_HWINT_P): New macro.
2927         (rtx_code_size): Use RTX_CODE_HWINT_P ().
2929 2018-10-25  Jan Hubicka  <jh@suse.cz>
2931         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
2932         is anonymous.
2934 2018-10-25  Richard Biener  <rguenther@suse.de>
2936         PR tree-optimization/87665
2937         PR tree-optimization/87745
2938         * tree-vectorizer.h (get_earlier_stmt): Remove.
2939         (get_later_stmt): Pick up UID from the original non-pattern stmt.
2941 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
2943         * options.texi (Deprecated): Move list to Var section.
2945 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
2946             Jinsong Ji <jji@us.ibm.com>
2948         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
2949         __vector long to __vector long long.
2950         (_mm_cvtpd_ps): Likewise.
2951         (_mm_cvttpd_epi32): Likewise.
2952         (_mm_cvtpi32_pd): Likewise.
2953         (_mm_unpackhi_epi64): Likewise.
2954         (_mm_unpacklo_epi64): Likewise.
2956 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
2958         PR rtl-optimization/87720
2959         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
2961 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
2963         * gimple-ssa-isolate-paths.c
2964         (find_implicit_erroneous_behavior): Do not change code if the
2965         pass is running for warnings only.
2966         (find_explicit_erroneous_behavior): Likewise.
2968 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
2970         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
2971         Define as rs6000_mangle_decl_assembler_name.
2972         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
2973         long double to IEEE long double, switch the names of the long
2974         double built-in functions to be <func>f128 instead of <func>l.
2976 2018-10-24  Martin Sebor  <msebor@redhat.com>
2978         * doc/extend.texi (nonnull): List no-argument form.  Reference
2979         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
2981 2018-10-24  Richard Biener  <rguenther@suse.de>
2983         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
2985 2018-10-24  Martin Liska  <mliska@suse.cz>
2987         PR tree-optimization/84436
2988         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
2989         Remove.
2990         (switch_conversion::contains_linear_function_p): New.
2991         (switch_conversion::build_one_array): Support linear
2992         transformation on input.
2993         * tree-switch-conversion.h (struct switch_conversion): Add
2994         contains_linear_function_p declaration.
2996 2018-10-24  Richard Biener  <rguenther@suse.de>
2998         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
2999         if its argument is CONSTANT_CLASS_P.
3001 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3003         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3004         it is wrong for forward declarations.
3006 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3008         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3009         SYMBOL_REF_P () check.
3011 2018-10-24  Richard Biener  <rguenther@suse.de>
3013         PR tree-optimization/87105
3014         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3015         dump classification.
3016         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3017         stores by splitting the affected group after the fact.
3018         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3019         fail the SLP build because of size constraints.
3021 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3023         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3024         * configure: Regenerate.
3025         * config.in: Regenerate.
3026         * varasm.c (mergeable_string_section): Use readonly_data_section
3027         if linker doesn't support SHF_MERGE with alignment > 8.
3028         (mergeable_constant_section): Likewise.
3030 2018-10-24  Richard Biener  <rguenther@suse.de>
3032         PR tree-optimization/84013
3033         * tree-ssa-structalias.c (struct msdi_data): New struct for
3034         marshalling data to walk_stmt_load_store_ops.
3035         (maybe_set_dependence_info): Refactor as callback for
3036         walk_stmt_load_store_ops.
3037         (compute_dependence_clique): Set restrict info on all stmt kinds.
3039 2018-10-24  Martin Liska  <mliska@suse.cz>
3041         * cgraph.c (cgraph_node::dump):
3042         Remove reduntant dumps and make tp_first_run dump more compact.
3044 2018-10-24  Richard Biener  <rguenther@suse.de>
3046         PR tree-optimization/87665
3047         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3048         to reflect reality.
3050 2018-10-12  Jeff Law  <law@redhat.com>
3052         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3053         for H8/S.
3055 2018-10-23  Richard Biener  <rguenther@suse.de>
3057         * tree-vrp.c (add_assert_info): Guard dump_printf with
3058         dump_enabled_p.
3059         * gimple-ssa-evrp-analyze.c
3060         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3061         Use value_range::ignore_equivs_equal_p.
3063 2018-10-23  Richard Biener  <rguenther@suse.de>
3065         PR tree-optimization/87105
3066         PR tree-optimization/87608
3067         * passes.def (pass_all_early_optimizations): Add early phi-opt
3068         after dce.
3069         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3070         addition to debug stmts.
3071         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3072         and abs replacement early.
3073         * tree-cfg.c (gimple_empty_block_p): Likewise.
3075 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3077         PR target/86383
3078         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3079         specified to configure.
3080         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3082 2018-10-23  Richard Biener  <rguenther@suse.de>
3084         PR tree-optimization/87700
3085         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3087 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3089         PR target/87674
3090         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3091         second argument from __mmask16 to __mmask8.
3092         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3093         _mm_mask_packs_epi32): Likewise.
3094         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3095         Likewise.
3096         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3098 2018-10-23  Richard Biener  <rguenther@suse.de>
3100         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3102 2018-10-23  Richard Biener  <rguenther@suse.de>
3104         PR tree-optimization/86144
3105         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3106         over simd attribute.
3108 2018-10-23  Richard Biener  <rguenther@suse.de>
3110         PR tree-optimization/87693
3111         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3112         the case we do not find the taken edge.
3114 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3115             Jinsong Ji  <jji@us.ibm.com>
3117         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3118         (_mm_store_pd): Use unaligned vector type for pointer cast.
3119         (_mm_maskmoveu_si128): Likewise.
3120         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3121         (_mm_store_ps): Use unaligned vector type for pointer cast.
3123 2018-10-22  Paul Koning  <ni1d@arrl.net>
3125         * symtab.c (symtab_node::increase_alignment): Correct max
3126         alignment check.
3128 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3130         PR tree-optimization/87633
3131         * match.pd: Do not generate unordered integer comparisons.
3133 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3135         PR rtl-optimization/87600
3136         * combine.c: Add include of expr.h.
3137         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3138         register to a pseudo.
3139         (make_more_copies): New function, add a copy to a new pseudo after
3140         the moves from hard registers into pseudos.
3141         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3142         later.  Call make_more_copies.
3144 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3146         * lra-constraints.c (process_alt_operands): New local array,
3147         matching_early_clobber.  Check matching_early_clobber before
3148         decrementing reject, and set matching_early_clobber after.
3150 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3152         PR target/87598
3153         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3154         call output_addr_const and hope for the best.
3156 2018-10-22  Richard Biener  <rguenther@suse.de>
3158         * gimple-ssa-evrp-analyze.c
3159         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3160         smarter about what ranges to use.
3161         * tree-vrp.c (add_assert_info): Dump here.
3162         (register_edge_assert_for_2): Instead of here at multiple but
3163         not all places.
3165         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3166         * gcc.dg/predict-6.c: Adjust.
3167         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3168         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3169         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3171 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3172         Richard Biener  <rguenther@suse.de>
3174         * bitmap.h: Update data structure documentation, including a
3175         description of bitmap views as either linked-lists or splay trees.
3176         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3177         (struct bitmap_head_def): Likewise.
3178         (bitmap_list_view, bitmap_tree_view): New prototypes.
3179         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3180         tree_form fields.
3181         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3182         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3183         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3184         released bitmap element here.
3185         (bitmap_element_free): Remove.
3186         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3187         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3188         this function similar ones such that linked-list bitmap implementation
3189         functions are grouped.
3190         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3191         and moved for grouping.
3192         (bitmap_list_insert_element_after): Renamed from
3193         bitmap_elt_insert_after, and moved for grouping.
3194         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3195         (bitmap_tree_link_left, bitmap_tree_link_right,
3196         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3197         bitmap_tree_link_element, bitmap_tree_unlink_element,
3198         bitmap_tree_find_element): New functions for splay-tree bitmap
3199         implementation.
3200         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3201         Renamed and moved, see above entries.
3202         (bitmap_tree_listify_from): New function to convert part of a splay
3203         tree bitmap to a linked-list bitmap.
3204         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3205         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3206         (bitmap_find_bit): Remove.
3207         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3208         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3209         Handle splay tree bitmaps.
3210         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3211         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3212         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3213         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3214         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3215         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3216         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3217         corresponding changes to use linked-list specific bitmap_element
3218         manipulation functions as applicable for efficiency.
3219         (bitmap_tree_to_vec): New function.
3220         (debug_bitmap_elt_file): New function split out from ...
3221         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3222         (bitmap_print): Likewise.
3224         PR tree-optimization/63155
3225         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3226         SSA edge worklists.
3227         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3228         in tree-view.
3230 2018-10-22  Martin Liska  <mliska@suse.cz>
3232         PR tree-optimization/87686
3233         Revert
3234         2018-08-29  Martin Liska  <mliska@suse.cz>
3236         * tree-switch-conversion.c (switch_conversion::expand):
3237         Strenghten assumption about gswitch statements.
3239 2018-10-22  Martin Liska  <mliska@suse.cz>
3241         * ipa-icf.c (sem_item::compare_attributes): Remove.
3242         (sem_item::compare_referenced_symbol_properties): Use
3243         attribute_list_equal instead.
3244         (sem_function::equals_wpa): Likewise.
3245         * ipa-icf.h: Remove compare_attributes.
3247 2018-10-22  Richard Biener  <rguenther@suse.de>
3249         PR middle-end/87682
3250         * mem-stats.h (mem_usage::operator==): Fix pasto.
3252 2018-10-22  Richard Biener  <rguenther@suse.de>
3254         PR tree-optimization/87640
3255         * tree-vrp.c (set_value_range_with_overflow): Decompose
3256         incomplete result.
3257         (extract_range_from_binary_expr_1): Adjust.
3259 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3261         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3262         (stmt_can_throw_external): Likewise.
3263         (stmt_can_throw_internal): Likewise.
3264         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3265         (lower_eh_constructs_2): Likewise.
3266         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3267         (stmt_can_throw_external): Likewise.
3268         (stmt_can_throw_internal): Likewise.
3269         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3270         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3271         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3272         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3273         (pass_lower_eh_dispatch::execute): Pass cfun to
3274         stmt_can_throw_external.
3275         (cleanup_empty_eh): Likewise.
3276         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3277         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3278         stmt_can_throw_external instead of pushing it to cfun.
3279         (symbol_table::create_edge): Likewise.
3280         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3281         stmt_can_throw_internal.
3282         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3283         to stmt_could_throw_p.
3284         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3285         stmt_can_throw_internal.
3286         (pass_store_merging::execute): Likewise.
3287         * gimple-ssa-strength-reduction.c
3288         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3289         stmt_could_throw_p.
3290         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3291         stmt_can_throw_internal.
3292         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3293         to stmt_can_throw_external.
3294         (check_stmt): Pass cfun to stmt_could_throw_p.
3295         (check_stmt): Pass cfun to stmt_can_throw_external.
3296         (pass_nothrow::execute): Likewise.
3297         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3298         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3299         stmt_can_throw_internal.
3300         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3301         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3302         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3303         * tree-complex.c (expand_complex_libcall): Pass cfun to
3304         stmt_could_throw_p and to stmt_can_throw_internal.
3305         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3306         * tree-inline.c (copy_edges_for_bb): Likewise.
3307         (maybe_move_debug_stmts_to_successors): Likewise.
3308         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3309         stmt_could_throw_p.
3310         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3311         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3312         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3313         stmt_can_throw_internal.
3314         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3315         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3316         stmt_could_throw_p.
3317         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3318         stmt_can_throw_internal.
3319         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3320         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3321         stmt_could_throw_p.
3322         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3323         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3324         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3325         (convert_mult_to_fma_1): Likewise.
3326         (convert_to_divmod): Likewise.
3327         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3328         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3329         * tree-ssa-propagate.c
3330         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3331         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3332         (maybe_optimize_range_tests): Likewise.
3333         (linearize_expr_tree): Likewise.
3334         (reassociate_bb): Likewise.
3335         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3336         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3337         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3338         (handle_char_store): Likewise.
3339         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3340         stmt_can_throw_internal.
3341         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3342         stmt_could_throw_p.
3343         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3344         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3345         (vectorizable_simd_clone_call): Likewise.
3346         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3347         (gimple_stringop_fixed_value): Likewise.
3349 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3351         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3352         literal pool references.
3353         (s390_check_qrst_address): Adapt to the new behavior of
3354         s390_loadrelative_operand_p ().
3356 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3358         PR target/72782
3359         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3361 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3363         PR target/72782
3364         * config/i386/sse.md (*<code><mode>3_bcst): New.
3366 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3368         PR target/72782
3369         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3370         V4DI, V16SI and V8DI.
3371         (*sub<mode>3<mask_name>_bcst): New.
3372         (*add<mode>3<mask_name>_bcst): Likewise.
3374 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3375             Jinsong Ji  <jji@us.ibm.com>
3377         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3378         __m64 with __vector unsigned long long for compatibility.
3379         (_mm_movemask_epi8): Likewise.
3380         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3381         (_mm_cvttps_pi32): Likewise.
3382         (_mm_cvtpi32_ps): Likewise.
3383         (_mm_cvtps_pi16): Likewise.
3384         (_mm_loadh_pi): Likewise.
3385         (_mm_storeh_pi): Likewise.
3386         (_mm_movehl_ps): Likewise.
3387         (_mm_movelh_ps): Likewise.
3388         (_mm_loadl_pi): Likewise.
3389         (_mm_storel_pi): Likewise.
3390         (_mm_movemask_ps): Likewise.
3391         (_mm_shuffle_pi16): Likewise.
3393 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3395         PR target/72782
3396         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3397         __builtin_ia32_vfnmsubpd512_mask.
3398         (_mm512_mask_fnmsub_round_pd): Likewise.
3399         (_mm512_fnmsub_pd): Likewise.
3400         (_mm512_mask_fnmsub_pd): Likewise.
3401         (_mm512_maskz_fnmsub_round_pd): Use
3402         __builtin_ia32_vfnmsubpd512_maskz.
3403         (_mm512_maskz_fnmsub_pd): Likewise.
3404         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3405         (_mm512_mask_fnmsub_round_ps): Likewise.
3406         (_mm512_fnmsub_ps): Likewise.
3407         (_mm512_mask_fnmsub_ps): Likewise.
3408         (_mm512_maskz_fnmsub_round_ps): Use
3409         __builtin_ia32_vfnmsubps512_maskz.
3410         (_mm512_maskz_fnmsub_ps): Likewise.
3411         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3412         __builtin_ia32_vfnmsubpd256_mask.
3413         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3414         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3415         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3416         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3417         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3418         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3419         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3420         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3421         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3422         __builtin_ia32_vfnmsubpd.
3423         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3424         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3425         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3426         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3427         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3428         * config/i386/i386-builtin.def: Add
3429         __builtin_ia32_vfnmsubpd256_mask,
3430         __builtin_ia32_vfnmsubpd256_maskz,
3431         __builtin_ia32_vfnmsubpd128_mask,
3432         __builtin_ia32_vfnmsubpd128_maskz,
3433         __builtin_ia32_vfnmsubps256_mask,
3434         __builtin_ia32_vfnmsubps256_maskz,
3435         __builtin_ia32_vfnmsubps128_mask,
3436         __builtin_ia32_vfnmsubps128_maskz,
3437         __builtin_ia32_vfnmsubpd512_mask,
3438         __builtin_ia32_vfnmsubpd512_maskz,
3439         __builtin_ia32_vfnmsubps512_mask,
3440         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3441         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3442         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3443         __builtin_ia32_vfnmsubpd256.
3444         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3445         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3446         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3447         Likewise.
3448         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3449         Likewise.
3450         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3451         Likewise.
3452         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3454 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3456         PR target/72782
3457         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3458         __builtin_ia32_vfnmaddpd512_mask.
3459         (_mm512_mask_fnmadd_round_pd): Likewise.
3460         (_mm512_fnmadd_pd): Likewise.
3461         (_mm512_mask_fnmadd_pd): Likewise.
3462         (_mm512_maskz_fnmadd_round_pd): Use
3463         __builtin_ia32_vfnmaddpd512_maskz.
3464         (_mm512_maskz_fnmadd_pd): Likewise.
3465         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3466         (_mm512_mask_fnmadd_round_ps): Likewise.
3467         (_mm512_fnmadd_ps): Likewise.
3468         (_mm512_mask_fnmadd_ps): Likewise.
3469         (_mm512_maskz_fnmadd_round_ps): Use
3470         __builtin_ia32_vfnmaddps512_maskz.
3471         (_mm512_maskz_fnmadd_ps): Likewise.
3472         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3473         __builtin_ia32_vfnmaddpd256_mask.
3474         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3475         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3476         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3477         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3478         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3479         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3480         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3481         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3482         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3483         __builtin_ia32_vfnmaddpd.
3484         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3485         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3486         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3487         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3488         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3489         * config/i386/i386-builtin.def: Add
3490         __builtin_ia32_vfnmaddpd256_mask,
3491         __builtin_ia32_vfnmaddpd256_maskz,
3492         __builtin_ia32_vfnmaddpd128_mask,
3493         __builtin_ia32_vfnmaddpd128_maskz,
3494         __builtin_ia32_vfnmaddps256_mask,
3495         __builtin_ia32_vfnmaddps256_maskz,
3496         __builtin_ia32_vfnmaddps128_mask,
3497         __builtin_ia32_vfnmaddps128_maskz,
3498         __builtin_ia32_vfnmaddpd512_mask,
3499         __builtin_ia32_vfnmaddpd512_maskz,
3500         __builtin_ia32_vfnmaddps512_mask,
3501         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
3502         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
3503         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
3504         __builtin_ia32_vfnmaddpd256.
3505         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
3506         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
3507         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
3508         Likewise.
3509         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
3510         Likewise.
3511         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
3512         Likewise.
3513         (fmai_vmfnmadd_<mode><round_name>): Likewise.
3515 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3517         PR target/72782
3518         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
3519         __builtin_ia32_vfmsubpd512_mask.
3520         (_mm512_mask_fmsub_round_pd): Likewise.
3521         (_mm512_fmsub_pd): Likewise.
3522         (_mm512_mask_fmsub_pd): Likewise.
3523         (_mm512_maskz_fmsub_round_pd): Use
3524         __builtin_ia32_vfmsubpd512_maskz.
3525         (_mm512_maskz_fmsub_pd): Likewise.
3526         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
3527         (_mm512_mask_fmsub_round_ps): Likewise.
3528         (_mm512_fmsub_ps): Likewise.
3529         (_mm512_mask_fmsub_ps): Likewise.
3530         (_mm512_maskz_fmsub_round_ps): Use
3531         __builtin_ia32_vfmsubps512_maskz.
3532         (_mm512_maskz_fmsub_ps): Likewise.
3533         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
3534         __builtin_ia32_vfmsubpd256_mask.
3535         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
3536         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3537         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
3538         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3539         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3540         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
3541         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
3542         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
3543         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
3544         __builtin_ia32_vfmsubpd.
3545         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
3546         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
3547         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
3548         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
3549         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
3550         * config/i386/i386-builtin.def: Add
3551         __builtin_ia32_vfmsubpd256_mask,
3552         __builtin_ia32_vfmsubpd256_maskz,
3553         __builtin_ia32_vfmsubpd128_mask,
3554         __builtin_ia32_vfmsubpd128_maskz,
3555         __builtin_ia32_vfmsubps256_mask,
3556         __builtin_ia32_vfmsubps256_maskz,
3557         __builtin_ia32_vfmsubps128_mask,
3558         __builtin_ia32_vfmsubps128_maskz,
3559         __builtin_ia32_vfmsubpd512_mask,
3560         __builtin_ia32_vfmsubpd512_maskz,
3561         __builtin_ia32_vfmsubps512_mask,
3562         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
3563         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
3564         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
3565         __builtin_ia32_vfmsubpd256.
3566         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
3567         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
3568         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
3569         Likewise.
3570         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
3571         Likewise.
3572         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
3573         Likewise.
3574         (fmai_vmfmsub_<mode><round_name>): Likewise.
3576 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3578         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
3579         Remove plus.  Renamed to ...
3580         (*sub<mode>3<mask_name>_bcst): This.
3581         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
3582         (*add<mode>3<mask_name>_bcst): This.
3584 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3586         PR target/72782
3587         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
3589 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3591         PR target/87662
3592         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
3593         (_mm_or_epi32): Likewise.
3594         (_mm256_xor_epi32): Likewise.
3595         (_mm_xor_epi32): Likewise.
3596         (_mm256_or_epi64): Likewise.
3597         (_mm_or_epi64): Likewise.
3598         (_mm256_xor_epi64): Likewise.
3599         (_mm_xor_epi64): Likewise.
3601 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
3603         PR target/72782
3604         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
3606 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
3608         PR middle-end/87647
3609         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
3611 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
3613         * doc/ux.texi: Move @section directly after @node.
3615 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
3617         PR middle-end/85488
3618         PR middle-end/87649
3619         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
3620         depend closely nested inside of loop with ordered clause with
3621         a parameter.
3623 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
3625         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
3626         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
3627         * doc/ux.texi: New file.
3629 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
3631         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
3632         be the first CR field allocated.
3634 2018-10-19  Richard Biener  <rguenther@suse.de>
3636         PR target/87657
3637         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
3638         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
3640 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3642         PR target/72782
3643         * config/i386/sse.md
3644         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
3645         (*add<mode>3<mask_name>_bcst_2): Likewise.
3647 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3649         * config/i386/sse.md
3650         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3651         Replace nonimmediate_operand with register_operand.
3652         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3653         Likewise.
3654         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3655         Likewise.
3657 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3659         PR rtl-optimization/87596
3660         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
3661         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
3662         for instructions in FROM..TO range.
3664 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
3666         * cfgexpand.c (expand_one_var): Use specific wording in error message
3667         for non-local frame variables.
3668         * stor-layout.c (layout_decl): Do not issue a warning for them.
3670 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
3672         * haifa-sched.c (priority): Add force_recompute parameter.
3673         (apply_replacement): Call priority () with force_recompute = true.
3674         (restore_pattern): Likewise.
3676 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3678         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
3679         HOST_BITS_PER_WIDE_INT.
3680         (test_vector_ops_duplicate): Likewise.
3682 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3684         PR target/72782
3685         * config/i386/sse.md (VF_AVX512): New.
3686         (avx512bcst): Likewise.
3687         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3688         Likewise.
3689         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3690         Likewise.
3691         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3692         Likewise.
3694 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
3696         * doc/invoke.texi (-dumpversion): Improve grammar.
3697         (-dumpfullversion): Make more consistent with -dumpversion.
3699 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
3701         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
3702         Set JUMP_LABEL to the jump insn.
3703         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
3704         Predict emitted jump and add label to jump insn.
3706 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
3708         PR tree-optimization/87562
3709         * input.c (get_substring_ranges_for_loc): Use
3710         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
3711         getting the linemap for the endpoint.  Verify that it's either
3712         in the same linemap as the start point's spelling location, or
3713         at least in the same file.
3715 2018-10-18  Richard Biener  <rguenther@suse.de>
3717         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
3718         feed width-specific load/store costs through ix86_vec_cost.
3719         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
3720         (k8_cost): Likewise.
3721         (bdver_cost): Likewise.
3722         (znver1_cost): Likewise.
3723         (btver1_cost): Likewise.
3724         (btver2_cost): Likewise.
3726 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3728         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
3729         to simplify subreg of vec_merge.
3731 2018-10-18  Richard Biener  <rguenther@suse.de>
3733         * config/i386/i386.c: Fix costing of vector FMA.
3735 2018-10-18  Richard Biener  <rguenther@suse.de>
3737         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
3738         and argument.
3739         (ix86_builtin_vectorization_cost): For vec_construct properly
3740         cost insertion into SSE regs.
3741         (...): Adjust calls to ix86_vec_cost.
3743 2018-10-18  Richard Biener  <rguenther@suse.de>
3745         PR middle-end/87087
3746         Revert
3747         2018-02-07  Richard Biener  <rguenther@suse.de>
3749         PR tree-optimization/84204
3750         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3751         this place.
3753 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3755         PR target/87537
3756         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
3757         of vec_duplicate.
3758         (test_vector_ops_duplicate): Add test for a scalar subreg of a
3759         VEC_MERGE of a VEC_DUPLICATE.
3761 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3763         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
3764         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
3765         * doc/standards.texi (C Language): Document C2X.
3766         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
3767         (rl78_option_override): Handle "GNU C2X" language name.
3769 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3771         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
3772         Document C17 as published in 2018.
3774 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
3776         PR middle-end/87623
3777         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
3778         bail out if both sides do not have the same storage order.
3780 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3782         * bitmap.c (bitmap_head::dump): New.
3783         * bitmap.h (bitmap_head): Add dump().
3784         * gimple-ssa-evrp-analyze.c
3785         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
3786         (evrp_range_analyzer::set_ssa_range_info): Same.
3787         (evrp_range_analyzer::record_ranges_from_phis): Same.
3788         (evrp_range_analyzer::record_ranges_from_stmt): Same.
3789         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
3790         * gimple-ssa-sprintf.c (get_int_range): Same.
3791         (format_integer): Same.
3792         (sprintf_dom_walker::handle_gimple_call): Same.
3793         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
3794         (ipcp_vr_lattice::top_p): Same.
3795         (ipcp_vr_lattice::bottom_p): Same.
3796         (ipcp_vr_lattice::set_to_bottom): Same.
3797         (ipa_vr_operation_and_type_effects): Same.
3798         (propagate_vr_across_jump_function): Same.
3799         (ipcp_store_vr_results): Same.
3800         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
3801         (ipa_print_node_jump_functions_for_edge): Same.
3802         (ipa_get_value_range): Same.
3803         (ipa_compute_jump_functions_for_edge): Same.
3804         (ipa_write_jump_function): Same.
3805         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
3806         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3807         Same.
3808         * vr-values.c (set_value_range_to_nonnegative): Same.
3809         (set_value_range_to_truthvalue): Same.
3810         (vr_values::get_value_range): Same.
3811         (vr_values::set_defs_to_varying): Same.
3812         (vr_values::update_value_range): Same.
3813         (symbolic_range_based_on_p): Same.
3814         (vr_values::op_with_boolean_value_range_p): Same.
3815         (vr_values::extract_range_for_var_from_comparison_expr): Same.
3816         (vr_values::extract_range_from_ssa_name): Same.
3817         (vr_values::extract_range_from_binary_expr): Same.
3818         (vr_values::extract_range_from_unary_expr): Same.
3819         (vr_values::extract_range_from_cond_expr): Same.
3820         (vr_values::extract_range_from_comparison): Same.
3821         (vr_values::check_for_binary_op_overflow): Same.
3822         (vr_values::extract_range_basic): Same.
3823         (vr_values::extract_range_from_assignment): Same.
3824         (compare_ranges): Same.
3825         (compare_range_with_value): Same.
3826         (vr_values::adjust_range_with_scev): Same.
3827         (vrp_valueize): Same.
3828         (vrp_valueize_1): Same.
3829         (vr_values::get_vr_for_comparison): Same.
3830         (vr_values::compare_name_with_value): Same.
3831         (vr_values::compare_names): Same.
3832         (vr_values::vrp_evaluate_conditional): Same.
3833         (find_case_label_ranges): Same.
3834         (vr_values::vrp_visit_switch_stmt): Same.
3835         (vr_values::extract_range_from_phi_node): Same.
3836         (vr_values::simplify_div_or_mod_using_ranges): Same.
3837         (vr_values::simplify_bit_ops_using_ranges): Same.
3838         (test_for_singularity): Same.
3839         (range_fits_type_p): Same.
3840         (vr_values::simplify_cond_using_ranges_1): Same.
3841         (vr_values::simplify_switch_using_ranges): Same.
3842         (vr_values::simplify_float_conversion_using_ranges): Same.
3843         (vr_values::two_valued_val_range_p): Same.
3844         (vr_values::add_equivalence): Move to value_range::equiv_add.
3845         * vr-values.h (vr_values::add_equivalence): Remove.
3846         (VR_INITIALIZER): Remove.
3847         * tree-vrp.c (value_range::set): New.
3848         (value_range::equiv_add): New.
3849         (value_range::value_range): New.
3850         (value_range::deep_copy): New.
3851         (value_range::check): New.
3852         (value_range::equal_p): New.
3853         (value_range::ignore_equivs_equal_p): New.
3854         (value_range::operator==): New.
3855         (value_range::operator!=): New.
3856         (value_range::symbolic_p): New.
3857         (value_range::numeric_p): New.
3858         (value_range::set_undefined): New.
3859         (value_range::set_varying): New.
3860         (value_range::may_contain_p): New.
3861         (value_range::equiv_clear): New.
3862         (value_range::singleton_p): New.
3863         (value_range::intersect): New.
3864         (value_range::dump): New.
3865         (value_range::set_and_canonicalize): New.
3866         (set_value_range): Adjust for value_range API.
3867         (set_value_range_to_undefined): Same.
3868         (set_value_range_to_varying): Same.
3869         (set_and_canonicalize_value_range): Same.
3870         (set_value_range_to_nonnull): Same.
3871         (set_value_range_to_null): Same.
3872         (range_is_null): Same.
3873         (range_is_nonnull): Same.
3874         (range_int_cst_p): Same.
3875         (range_int_cst_singleton_p): Same.
3876         (symbolic_range_p): Same.
3877         (range_includes_zero_p): Same.
3878         (value_range_constant_singleton): Same.
3879         (vrp_set_zero_nonzero_bits): Same.
3880         (ranges_from_anti_range): Same.
3881         (extract_range_into_wide_ints): Same.
3882         (extract_range_from_multiplicative_op): Same.
3883         (set_value_range_with_overflow): Same.
3884         (extract_range_from_binary_expr_1): Same.
3885         (extract_range_from_unary_expr): Same.
3886         (dump_value_range): Same.
3887         (debug_value_range): Same.
3888         (vrp_prop::check_array_ref): Same.
3889         (vrp_prop::check_mem_ref): Same.
3890         (vrp_prop::vrp_initialize): Same.
3891         (vrp_prop::visit_stmt): Same.
3892         (intersect_ranges): Same.
3893         (vrp_prop::visit_phi): Same.
3894         (vrp_prop::vrp_finalize): Same.
3895         (determine_value_range_1): Same.
3896         (determine_value_range): Same.
3897         (vrp_intersect_ranges_1): Rename to...
3898         (vrp_intersect_1): this.
3899         (vrp_intersect_ranges): Rename to...
3900         (value_range::intersect_helper): ...this.
3901         (vrp_meet_1): Rename to...
3902         (value_range::union_helper): ...this.
3903         (vrp_meet): Rename to...
3904         (value_range::union_): ...this.
3905         (copy_value_range): Remove.
3906         * tree-vrp.h (struct value_range): Rewrite into a proper class.
3907         (value_range::vrtype): New.
3908         (value_range::type): New.
3909         (value_range::equiv): New.
3910         (value_range::min): New.
3911         (value_range::max): New.
3912         (value_range::varying_p): New.
3913         (value_range::undefined_p): New.
3914         (value_range::null_p): New.
3915         (value_range::equiv_add): New.
3916         (copy_value_range): Remove.
3918 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
3920         * Makefile.in (SELFTEST_TARGETS): New.
3921         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
3922         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
3923         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
3924         c/Make-lang.in.
3925         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
3926         (selftest-c++-gdb, selftest-c++-valgrind): Move to
3927         cp/Make-lang.in.
3928         * configure: Regenerate.
3929         * configure.ac (selftest_languages): New.
3931 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3933         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
3934         overflow wraps argument.
3935         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
3936         wide_int_range_multiplicative_op.
3937         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
3938         overflow wraps argument.
3939         (wide_int_range_multiplicative_op): Same.
3940         (wide_int_range_lshift): Same.
3941         (wide_int_range_div): Same.
3942         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
3943         (wide_int_range_lshift): Same.
3944         (wide_int_range_div): Same.
3946 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3948         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
3949         use sign as argument.
3950         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
3951         wide_int_range_shift_undefined_p.
3953 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3955         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
3956         Rename to...
3957         (@despeculate_copy<ALLI_TI:mode>): ... This.
3958         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
3959         switch statement.
3961 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3963         * config.gcc: Obsolete *-*-solaris2.10*.
3964         * doc/install.texi (Specific, *-*-solaris2*): Document it.
3966 2018-10-12  Jeff Law  <law@redhat.com>
3968         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
3969         reg + sym +- const_int addressing modes.
3971 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3973         * common.opt (fdiagnostics-minimum-margin-width=): New option.
3974         * diagnostic-show-locus.c (layout::layout): Apply the minimum
3975         margin width.
3976         (layout::start_annotation_line): Only print up to 3 of the
3977         margin character, to avoid touching the left-hand side.
3978         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
3979         minimum margin width, as set by test_diagnostic_context's ctor.
3980         (selftest::test_fixit_insert_containing_newline): Likewise.
3981         (selftest::test_fixit_insert_containing_newline_2): Likewise.
3982         (selftest::test_line_numbers_multiline_range): Clear
3983         dc.min_margin_width.
3984         * diagnostic.c (diagnostic_initialize): Initialize
3985         min_margin_width.
3986         * diagnostic.h (struct diagnostic_context): Add field
3987         "min_margin_width".
3988         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
3989         * opts.c (common_handle_option): Handle
3990         OPT_fdiagnostics_minimum_margin_width_.
3991         * selftest-diagnostic.c
3992         (selftest::test_diagnostic_context::test_diagnostic_context):
3993         Initialize min_margin_width to 6.
3994         * toplev.c (general_init): Initialize global_dc->min_margin_width.
3996 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3998         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3999         Fix usage of "error_at_rich_loc" in the comment.
4001 2018-10-15  Renlin Li  <renlin.li@arm.com>
4003         PR target/87563
4004         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4005         if-conversioned loop when it contains ifn with types not
4006         supported by backend.
4007         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4008         (direct_internal_fn_supported_p): New helper function.
4009         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4011 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4013         PR target/87572
4014         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4015         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4016         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4018 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4020         PR tree-optimization/87022
4021         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4022         bits in dist vector rather than the first one.
4024 2018-10-15  Richard Biener  <rguenther@suse.de>
4026         PR middle-end/87610
4027         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4028         (visit_loadstore): When a used restrict tag escaped verify that
4029         the points-to solution of "other" pointers do not include
4030         escaped.
4031         (compute_dependence_clique): If a used restrict tag escaped
4032         communicated that down to visit_loadstore.
4034 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4036         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4037         into reg if it isn't a general operand.
4039 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4041         PR target/87599
4042         * config/i386/sse.md (*vec_dupv2di): Add register source to
4043         movddup.
4045 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4047         PR target/87572
4048         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4049         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4050         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4051         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4053 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4055         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4056         (notice_args_size): Set it in the current trace if no insn that can
4057         throw internally has been seen yet.
4058         (connect_traces): When connecting args_size between traces, allow the
4059         incoming values not to match if there is an insn setting it before the
4060         first insn that can throw internally; in that case, force the creation
4061         of a CFI note on this latter insn.
4063 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4065         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4066         the base class, because using the injected-class-name was not clearly
4067         specified until DR 176.
4069 2018-10-12  Paul Koning  <ni1d@arrl.net>
4071         * config/pdp11/pdp11.md (doloop_end): New expander.
4072         (doloop_end_insn): renamed from "doloop_end".
4073         (addqi3): New pattern.
4074         (subqi3): New pattern.
4075         * config/pdp11/predicates.md (incdec_operand): New predicate.
4077 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4079         PR middle-end/81376
4080         * real.c (format_helper::can_represent_integral_type_p): New function
4081         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4082         * match.pd: New pattern.
4084 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4086         * configure.ac: Introduce --enable-large-address-aware
4087         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4088         * doc/install.texi: Document it.
4089         * configure, config.in: Rebuilt.
4090         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4091         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4092         (LINK_SPEC): Insert it.
4093         * config/i386/mingw-w64.h: Likewise.
4095         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4097 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4099         PR rtl-optimization/87600
4100         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4102 2018-10-12  Paul Koning  <ni1d@arrl.net>
4104         * doc/md.texi (doloop_end): Document that the pattern code may
4105         need to check operand mode.
4107 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4109         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4110         to zero-extend between int and floating-point registers.
4111         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4112         ldp into floating-point registers.  Add type and arch attributes.
4113         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4114         Use f_loads for type attribute.
4116 2018-10-11  Martin Sebor  <msebor@redhat.com>
4118         * doc/extend.texi (attribute packed): Correct typos.
4120 2018-10-11  Martin Sebor  <msebor@redhat.com>
4122         * doc/extend.texi (attribute flatten): Mention interaction with
4123         noinline.
4125 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4127         PR target/87156
4128         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4129         Set new_decl virtual flag to zero.
4131 2018-10-11  Martin Sebor  <msebor@redhat.com>
4133         PR middle-end/87593
4134         * doc/extend.texi (attribute format_arg): Discuss using multiple
4135         attributes on a single function.
4137 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4139         PR tree-optimization/86829
4140         * match.pd (sin (atan (x))): New simplification rules.
4141         (cos (atan (x))): Likewise.
4142         * real.c (build_sinatan_real): New function.
4143         * real.h (build_sinatan_real): Prototype.
4145 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4147         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4148         function.
4149         (fold_mergeeo_helper): New helper function.
4150         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4151         intrinsics.  Correct some whitespace indentation issues.
4153 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4155         PR target/87511
4156         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4157         Use HOST_WIDE_INT_1U for shift.
4159 2018-10-11  Doug Rupp  <rupp@adacore.com>
4160             Olivier Hainque  <hainque@adacore.com>
4162         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4163         Pass --relax to the linker for RTPs.
4164         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4166 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4167             Jan Hubicka  <jh@suse.cz>
4168             Martin Jambor  <mjambor@suse.cz>
4170         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4171         the same elements are repeated rather than printing all of them.
4172         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4173         "repeated" elements.
4174         * read-rtl-function.c (test_loading_repeat): New function.
4175         (read_rtl_function_c_tests): Call test_loading_repeat.
4176         * rtl-tests.c (test_dumping_repeat): New function.
4177         (rtl_tests_c_tests): Call test_dumping_repeat.
4179 2018-10-11  Richard Biener  <rguenther@suse.de>
4181         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4182         bdver?_cost): Unify to ...
4183         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4184         * config/i386/i386.c (processor_cost_table): Adjust.
4186 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4188         PR middle-end/87574
4189         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4190         the thunk when expanding to GIMPLE.
4192 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4194         * varasm.c (mergeable_string_section): Don't try to move zero-length
4195         strings to the merge section.
4197 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4199         PR target/87573
4200         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4202 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4204         PR target/87550
4205         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4206         to special_args set.
4208 2018-10-10  Richard Biener  <rguenther@suse.de>
4210         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4211         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4212         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4213         and recursing and pattern terminating the recursion on SSE
4214         vector width using ix86_expand_reduc.
4215         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4216         width and recursing and SSE2 vector variant doing the final
4217         reduction with ix86_expand_reduc.
4218         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4219         with terminating the recursion at AVX level, splitting that
4220         to SSE there.
4222 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4224         * genmatch.c (error_cb): Rename to...
4225         (diagnostic_cb): ...this, converting int params to enums.
4226         (fatal_at): Update for renaming.
4227         (warning_at): Likewise.
4228         (main): Likewise.
4229         * input.c (selftest::ebcdic_execution_charset::apply):
4230         Update for renaming of...
4231         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4232         to...
4233         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4234         converting level and reason to enums.
4235         (class selftest::lexer_error_sink): Rename to...
4236         (class selftest::lexer_test_options): ...this, renaming field
4237         "m_errors" to "m_diagnostics".
4238         (selftest::lexer_test_options::apply): Update for renaming of...
4239         (selftest::lexer_test_options::on_error): ...this, renaming to...
4240         (selftest::lexer_test_options::on_diagnostic): ...this
4241         converting level and reason to enums.
4242         (selftest::test_lexer_string_locations_raw_string_unterminated):
4243         Update for renamings.
4244         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4245         "reason".
4247 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4249         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4250         * config/rs6000/pmmintrin.h: New file.
4252 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4254         PR tree-optimization/86659
4255         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4257 2018-10-09  Richard Biener  <rguenther@suse.de>
4259         PR tree-optimization/63155
4260         * tree-ssa-structalias.c: Include tree-ssa.h.
4261         (get_constraint_for_ssa_var): For undefs return nothing_id.
4262         (find_func_aliases): Cleanup PHI handling.
4264 2018-10-09  Richard Biener  <rguenther@suse.de>
4266         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4267         replacements.
4269 2018-10-09  Martin Liska  <mliska@suse.cz>
4271         * asan.c (asan_emit_stack_protection): If a stack variable
4272         is located in a same file as current function, then emit
4273         line info into variable definition string.
4275 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4277         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4278         information.
4280 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4282         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4283         on the thunk.
4285 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4287         PR tree-optimization/86659
4288         * gimple-match.h (struct gimple_match_op): Add reverse field.
4289         (gimple_match_op::set_op): New overloaded method.
4290         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4291         the REF_REVERSE_STORAGE_ORDER flag on the value.
4292         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4293         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4294         
4295 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4297         PR middle-end/63155
4298         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4299         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4301 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4303         PR target/87517
4304         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4305         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4307 2018-10-08  Richard Biener  <rguenther@suse.de>
4309         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4310         cost the same as AVX128 ones.
4312 2018-10-08  Paul Koning  <ni1d@arrl.net>
4314         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4315         (expand_block_move): New function.
4316         * config/pdp11/pdp11.c (output_block_move): Remove.
4317         (expand_block_move): New function.
4318         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4319         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4320         (*movmemhi1): Remove.
4322 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4324         * config/s390/2827.md: Increase latencies for some FP instructions.
4326 2018-10-08  Richard Biener  <rguenther@suse.de>
4328         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4329         Open a dump scope.
4330         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4331         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4332         (add_stmt_cost): Dump return value of the hook.
4334 2018-10-08  Richard Biener  <rguenther@suse.de>
4336         PR tree-optimization/63155
4337         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4338         (ssa_propagation_engine::ssa_propagate): Remove redundant
4339         bitmap bit clearing.
4341 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4343         PR rtl-optimization/86939
4344         PR rtl-optimization/87479
4345         * ira.h (non_conflicting_reg_copy_p): New prototype.
4346         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4347         (make_hard_regno_dead): Don't add conflicts for register
4348         ignore_reg_for_conflicts.
4349         (make_object_dead): Likewise.
4350         (non_conflicting_reg_copy_p): New function.
4351         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4352         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4353         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4354         (make_hard_regno_dead): Don't add conflicts for register
4355         ignore_reg_for_conflicts.  Remove special conflict handling of
4356         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4357         check_pic_pseudo_p and update callers.
4358         (mark_pseudo_dead): Don't add conflicts for register
4359         ignore_reg_for_conflicts.
4360         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4362 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4363             Jim Wilson  <jimw@sifive.com>
4365         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4366         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4367         new pattern using HONOR_SNANS that emits one extra instruction.
4369 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4371         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4372         patterns): Merge SI and DI patterns to a GPR pattern.
4373         (unnamed define_insn and define_split for record form of that): Merge
4374         to a single define_insn_and_split pattern.
4376 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4378         PR c++/56856
4379         * input.c (expand_location_to_spelling_point): Add param "aspect"
4380         and use rather than hardcoding LOCATION_ASPECT_CARET.
4381         (get_substring_ranges_for_loc): Handle the case of a single token
4382         within a macro expansion.
4383         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4384         defaulting to LOCATION_ASPECT_CARET.
4386 2018-10-05  Paul Koning  <ni1d@arrl.net>
4388         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4389         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4390         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4391         (pdp11_guard_type): New function.
4393 2018-10-05  Paul Koning  <ni1d@arrl.net>
4395         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4396         * config/pdp11/pdp11.opt (mfloat32): Remove.
4397         (mfloat64): Remove.
4398         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4399         (pdp11 -mfloat64): Remove.
4401 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4403         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4404         (*cmp<mode>_cc_i387): Ditto.
4405         (*cmpu<mode>_cc_i387): Ditto.
4406         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4407         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4408         "scratch" argument.
4409         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4410         Emit x86_sahf_1 pattern.
4411         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4412         (ix86_expand_carry_flag_compare): Ditto.
4414 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4416         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4417         to reg_or_0_operand.  Add "C" constraint.
4418         (*cmpxf_cc_i387): Ditto.
4419         (*cmp<mode>_i387): Change operand 2 predicate
4420         to nonimm_or_0_operand.  Add "C" constraint.
4421         (*cmp<mode>_cc_i387): Ditto.
4422         (*cmp<mode>_0_i387): Remove insn pattern.
4423         (*cmp<mode>_0_cc_i387): Ditto.
4425 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4427         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4428         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4429         from vector_move_operand.  Update all uses.
4431 2018-10-05  Martin Sebor  <msebor@redhat.com>
4433         PR tree-optimization/87490
4434         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4435         consistently.
4437 2018-10-05  Richard Biener  <rguenther@suse.de>
4439         PR tree-optimization/63155
4440         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4441         vertical space in dumpfiles.
4442         * tree-ssa-propagate.h
4443         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4444         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4445         (ssa_edge_worklist_back): Likewise.
4446         (curr_order): Likewise.
4447         (cfg_blocks_get): Remove abstraction.
4448         (cfg_blocks_add): Likewise.
4449         (cfg_blocks_empty_p): Likewise.
4450         (add_ssa_edge): Add to current or next worklist based on
4451         RPO index.
4452         (add_control_edge): Likewise.
4453         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4454         into ...
4455         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4456         iteration from CFG and SSA edge worklist so we process
4457         everything in RPO order, prioritizing forward progress
4458         over iteration.
4459         (ssa_prop_init): Allocate new worklists, do not dump
4460         immediate uses.
4461         (ssa_prop_fini): Free new worklists.
4463 2018-10-05  Richard Biener  <rguenther@suse.de>
4465         * tree-core.h (tree_block::abstract_flag): Remove.
4466         (tree_block::block_num): Make full 32bits.
4467         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4468         * tree.h (BLOCK_ABSTRACT): Remove.
4469         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4470         resulting from BLOCK_ABSTRACT being always false.
4471         (gen_inlined_subroutine_die): Likewise.
4472         (gen_block_die): Likewise.
4473         * tree.c (block_ultimate_origin): Likewise.
4474         * tree-pretty-print.c (dump_block_node): Remove code dealing
4475         with BLOCK_ABSTRACT.
4476         * tree-ssa-live.c (dump_scope_block): Likewise.
4477         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4478         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4480 2018-10-05   Richard Biener  <rguenther@suse.de>
4482         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4483         is asked for initialize mode to the component mode of the
4484         vector type.
4486 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4488         PR target/87522
4489         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4490         assembler for -mavx.
4491         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4493 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4495         PR target/87509
4496         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4497         RS6000_BTM_DFP.
4498         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4499         to be DImode.  When using mffscrn, force the operand to a register.
4501 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
4503         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
4504         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
4505         X87MODEF mode iterator.
4506         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
4507         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
4508         X87MODEF mode iterator.
4510 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
4512         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
4513         -Wno-prio-ctor-dtor.
4515 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4517         * Makefile.in (OBJS): Add opt-problem.o.
4518         * dump-context.h: Include "selftest.h.
4519         (selftest::temp_dump_context): New forward decl.
4520         (class dump_context): Make friend of class
4521         selftest::temp_dump_context.
4522         (dump_context::dump_loc_immediate): New decl.
4523         (class dump_pretty_printer): Move here from dumpfile.c.
4524         (class temp_dump_context): Move to namespace selftest.
4525         (temp_dump_context::temp_dump_context): Add param
4526         "forcibly_enable_dumping".
4527         (selftest::verify_dumped_text):
4528         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
4529         (selftest::verify_item):
4530         (ASSERT_IS_TEXT): Move here from dumpfile.c.
4531         (ASSERT_IS_TREE): Likewise.
4532         (ASSERT_IS_GIMPLE): Likewise.
4533         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
4534         to...
4535         (dump_context::dump_loc_immediate): ...this new function.
4536         (class dump_pretty_printer): Move to dump-context.h.
4537         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
4538         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
4539         (temp_dump_context::temp_dump_context): Move to "selftest"
4540         namespace.  Add param "forcibly_enable_dumping", and use it to
4541         conditionalize the use of m_pp;
4542         (selftest::verify_dumped_text): Make non-static.
4543         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
4544         (selftest::verify_item): Make non-static.
4545         (ASSERT_IS_TEXT): Move to dump-context.h.
4546         (ASSERT_IS_TREE): Likewise.
4547         (ASSERT_IS_GIMPLE): Likewise.
4548         (selftest::test_capture_of_dump_calls): Pass "true" for new
4549         param of temp_dump_context.
4550         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
4551         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
4552         TDF_COMPARE_DEBUG.
4553         * opt-problem.cc: New file.
4554         * opt-problem.h: New file.
4555         * optinfo-emit-json.cc
4556         (selftest::test_building_json_from_dump_calls): Pass "true" for
4557         new param of temp_dump_context.
4558         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
4559         (optinfo::emit_for_opt_problem): New function.
4560         (optinfo::emit): Clarity which emit_item is used.
4561         * optinfo.h (optinfo::get_dump_location): New accessor.
4562         (optinfo::emit_for_opt_problem): New decl.
4563         (optinfo::emit): Make const.
4564         * selftest-run-tests.c (selftest::run_tests): Call
4565         selftest::opt_problem_cc_tests.
4566         * selftest.h (selftest::opt_problem_cc_tests): New decl.
4567         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
4568         bool to opt_result, converting fprintf messages to
4569         opt_result::failure_at calls.  Add "stmt" param for use by the
4570         failure_at calls.
4571         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
4572         (runtime_alias_check_p): Convert return type from bool to
4573         opt_result, converting dump_printf calls to
4574         opt_result::failure_at, using the statement DDR_A for their
4575         location.
4576         (find_data_references_in_stmt): Convert return type from bool to
4577         opt_result, converting "return false" to opt_result::failure_at
4578         with a new message.
4579         * tree-data-ref.h: Include "opt-problem.h".
4580         (dr_analyze_innermost): Convert return type from bool to opt_result,
4581         and add a const gimple * param.
4582         (find_data_references_in_stmt): Convert return type from bool to
4583         opt_result.
4584         (runtime_alias_check_p): Likewise.
4585         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
4586         dr_analyze_innermost.
4587         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
4588         Convert return type from bool to opt_result, adding a message for
4589         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
4590         (vect_analyze_data_ref_dependence): Convert return type from bool
4591         to opt_result.  Change sense of return type from "false"
4592         effectively meaning "no problems" to "false" meaning a problem,
4593         so that "return false" becomes "return opt_result::success".
4594         Convert "return true" calls to opt_result::failure_at, using
4595         the location of statement A rather than vect_location.
4596         (vect_analyze_data_ref_dependences): Convert return type from bool
4597         to opt_result.
4598         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
4599         calls to opt_result::failure_at, using the stmt location rather
4600         than vect_location.
4601         (vect_verify_datarefs_alignment): Convert return type from bool
4602         to opt_result.
4603         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
4604         into multiple more-tightly-scoped copies.
4605         (vect_analyze_data_refs_alignment): Convert return type from bool
4606         to opt_result.
4607         (vect_analyze_data_ref_accesses): Likewise, converting a
4608         "return false" to a "return opt_result::failure_at", adding a
4609         new message.
4610         (vect_prune_runtime_alias_test_list): Convert return type from
4611         bool to opt_result, converting dump_printf_loc to
4612         opt_result::failure_at.  Add a %G to show the pertinent statement,
4613         and use the stmt's location rather than vect_location.
4614         (vect_find_stmt_data_reference): Convert return type from
4615         bool to opt_result, converting dump_printf_loc to
4616         opt_result::failure_at, using stmt's location.
4617         (vect_analyze_data_refs):  Convert return type from bool to
4618         opt_result.  Convert "return false" to "return
4619         opt_result::failure_at", adding messages as needed.
4620         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
4621         type from bool to opt_result.
4622         (vect_determine_vf_for_stmt): Likewise.
4623         (vect_determine_vectorization_factor): Likewise, converting
4624         dump_printf_loc to opt_result::failure_at, using location of phi
4625         rather than vect_location.
4626         (vect_analyze_loop_form_1): Convert return type from bool to
4627         opt_result, converting dump_printf_loc calls, retaining the use of
4628         vect_location.
4629         (vect_analyze_loop_form): Convert return type from loop_vec_info
4630         to opt_loop_vec_info.
4631         (vect_analyze_loop_operations): Convert return type from bool to
4632         opt_result, converting dump_printf_loc calls, using the location
4633         of phi/stmt rather than vect_location where available.  Convert
4634         various "return false" to "return opt_result::failure_at" with
4635         "unsupported phi" messages.
4636         (vect_get_datarefs_in_loop): Convert return type from bool to
4637         opt_result.  Add a message for the
4638         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
4639         (vect_analyze_loop_2): Convert return type from bool to
4640         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
4641         each "goto again;", adding new messages where needed.
4642         Add "unsupported grouped {store|load}" messages.
4643         (vect_analyze_loop): Convert return type from loop_vec_info to
4644         opt_loop_vec_info.
4645         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
4646         bool to opt_result.
4647         * tree-vect-stmts.c (process_use): Likewise, converting
4648         dump_printf_loc call and using stmt location, rather than
4649         vect_location.
4650         (vect_mark_stmts_to_be_vectorized): Likeise.
4651         (vect_analyze_stmt): Likewise, adding a %G.
4652         (vect_get_vector_types_for_stmt): Convert return type from bool to
4653         opt_result, converting dump_printf_loc calls and using stmt
4654         location, rather than vect_location.
4655         (vect_get_mask_type_for_stmt): Convert return type from tree to
4656         opt_tree, converting dump_printf_loc calls and using stmt location.
4657         * tree-vectorizer.c: Include "opt-problem.h.
4658         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
4659         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
4660         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
4661         enabled, use it to report at the top level "couldn't vectorize
4662         loop" followed by the problem.
4663         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
4664         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
4665         to opt_result.
4666         (vect_analyze_stmt): Likewise.
4667         (vect_get_vector_types_for_stmt): Likewise.
4668         (tree vect_get_mask_type_for_stmt): Likewise.
4669         (vect_analyze_data_ref_dependences): Likewise.
4670         (vect_enhance_data_refs_alignment): Likewise.
4671         (vect_analyze_data_refs_alignment): Likewise.
4672         (vect_verify_datarefs_alignment): Likewise.
4673         (vect_analyze_data_ref_accesses): Likewise.
4674         (vect_prune_runtime_alias_test_list): Likewise.
4675         (vect_find_stmt_data_reference): Likewise.
4676         (vect_analyze_data_refs): Likewise.
4677         (vect_analyze_loop): Convert return type from loop_vec_info to
4678         opt_loop_vec_info.
4679         (vect_analyze_loop_form): Likewise.
4680         (vect_analyze_slp): Convert return type from bool to opt_result.
4682 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4684         * doc/invoke.texi (-fopt-info): Document new "internals"
4685         sub-option.
4686         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
4687         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
4688         MSG_ALL_KINDS.
4689         (optinfo_verbosity_options): Add "internals".
4690         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
4691         (dump_context::apply_dump_filter_p): New member function.
4692         (dump_context::dump_loc): Use apply_dump_filter_p rather than
4693         explicitly masking the dump_kind.
4694         (dump_context::begin_scope): Increment the scope depth first.  Use
4695         apply_dump_filter_p rather than explicitly masking the dump_kind.
4696         (dump_context::emit_item): Use apply_dump_filter_p rather than
4697         explicitly masking the dump_kind.
4698         (dump_dec): Likewise.
4699         (dump_hex): Likewise.
4700         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
4701         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
4702         (opt_info_switch_p): Update handling of default
4703         MSG_OPTIMIZED_LOCATIONS to cope with default of
4704         MSG_PRIORITY_USER_FACING.
4705         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
4706         masking the dump_kind.
4707         (selftest::test_capture_of_dump_calls): Update test_dump_context
4708         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
4709         than MSG_ALL.  Generalize scope test to be run at all four
4710         combinations of with/without MSG_PRIORITY_USER_FACING and
4711         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
4712         for each of the two values.
4713         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
4714         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
4715         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
4716         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
4717         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
4718         with MSG_PRIORITY_*.
4719         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
4720         dump messages as MSG_PRIORITY_USER_FACING.
4721         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
4722         about the interaction with MSG_PRIORITY_*.
4724 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4726         * varasm.c (output_constant): Add new parameter merge_strings.
4727         Make strings properly zero terminated in merge string sections.
4728         (mergeable_string_section): Don't fail if the last char is non-zero.
4729         (assemble_variable_contents): Handle merge string sections.
4730         (assemble_variable): Likewise.
4731         (assemble_constant_contents): Likewise.
4732         (output_constant_def_contents): Likewise.
4733         (output_constructor_array_range,
4734         output_constructor_regular_field): Adjust call to output_constant.
4735         (output_object_block): Adjust call to assemble_constant_contents
4736         and assemble_variable_contents.
4738 2018-10-04  Martin Liska  <mliska@suse.cz>
4740         PR c/87483
4741         * cgraphunit.c (process_function_and_variable_attributes):
4742         Warn about a function with alias attribute and a body.
4744 2018-10-04  Martin Liska  <mliska@suse.cz>
4746         PR ipa/82625
4747         * multiple_target.c (redirect_to_specific_clone): New function.
4748         (ipa_target_clone): Use it.
4749         * tree-inline.c: Fix comment.
4751 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4753         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
4754         fields.
4755         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
4756         (gcc::dump_manager::register_pass): New member function, adapted
4757         from loop body in gcc::pass_manager::register_pass, adding a
4758         call to update_dfi_for_opt_info.
4759         (gcc::dump_manager::opt_info_enable_passes): Store the
4760         -fopt-info options into the new fields.  Move the loop
4761         bodies into...
4762         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
4763         function.
4764         * dumpfile.h (struct opt_pass): New forward decl.
4765         (gcc::dump_manager::register_pass): New decl.
4766         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
4767         (class gcc::dump_manager): Add fields "m_optgroup_flags",
4768         "m_optinfo_flags", and "m_optinfo_filename".
4769         * passes.c (gcc::pass_manager::register_pass): Move all of the
4770         dump-handling code to gcc::dump_manager::register_pass.
4772 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
4774         PR rtl-optimization/87466
4775         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
4776         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
4777         * doc/tm.texi: Regenerate.
4778         * ira-lives.c (process_bb_node_lives): Use the new target hook.
4779         * lra-lives.c (process_bb_lives): Likewise.
4780         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
4781         Define.
4783 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
4785         * params.c (add_params): Fix initialization.
4787 2018-10-04  Martin Liska  <mliska@suse.cz>
4789         PR gcov-profile/84107
4790         * tree-profile.c (init_ic_make_global_vars):
4791         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
4792         Come up with new ic_tuple* variables.  Emit
4793         __gcov_indirect_call{,_topn} variables.
4794         (gimple_gen_ic_profiler): Access the variable
4795         and emit gimple.
4796         (gimple_gen_ic_func_profiler): Access
4797         __gcov_indirect_call.callee field.
4798         (gimple_init_gcov_profiler): Use ptr_type_node.
4799         * value-prof.c (gimple_ic): Use ptr_type_node.
4801 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4803         PR tree-optimization/85787
4804         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
4805         into this function and add support for detecting multiple phis.
4806         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
4808 2018-10-04  Martin Liska  <mliska@suse.cz>
4810         PR ipa/87491
4811         * ipa-inline.c (inline_to_all_callers_1):
4812         Call ultimate_alias_target for node being inlined.
4814 2018-10-03  Jeff Law  <law@redhat.com>
4816         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
4817         target's wchar_t.
4818         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
4819         * tree.h (get_typenode_from_name): Prototype.
4821 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4823         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
4824         Change operand 2 predicate to nonimmediate_operand.
4825         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4827 2018-10-03  Martin Sebor  <msebor@redhat.com>
4828             Jeff Law  <law@redhat.com>
4830         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
4831         initialize it.
4832         (get_string_length): Detect unterminated arrays.
4833         (format_string): Same.
4834         (format_directive): Warn about unterminated arrays.
4835         (handle_gimple_call): Mark statements with no_warning as needed.
4837 2018-10-03  Jim Wilson  <jimw@sifive.com>
4839         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
4840         also define __riscv_abi_rve.  Delete trailing white space.
4842 2018-10-03  Paul Koning  <ni1d@arrl.net>
4844         Enable LRA register allocator for PDP11.
4845         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
4846         (R): Likewise.
4847         (D): Likewise.
4848         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
4849         * config/pdp11/pdp11.opt (-mlra): New option.
4850         * doc/invoke.texi (PDP-11 Options): Document -mlra.
4852 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4854         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
4855         (*<absneg:code>extend<mode>xf2): Ditto.
4857 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4859         PR tree-optimization/87415
4860         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
4861         precision fields.
4863 2018-10-02  Jeff Law  <law@redhat.com>
4865         * gimple-fold.c (get_range_strlen): Only set *nonstr when
4866         an unterminated string is discovered.  Bubble up range
4867         even for unterminated strings.
4868         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
4869         indicates the string was not terminated via NONSTR.
4871 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4873         * tree-vrp.c (extract_range_from_unary_expr): Special case all
4874         pointer conversions.
4875         Do not do anything special for anti-ranges.
4877 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
4879         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
4880         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
4882 2018-10-03  Martin Liska  <mliska@suse.cz>
4884         PR gcov-profile/86109
4885         * coverage.c (coverage_begin_function): Do not
4886         mark lambdas as artificial.
4887         * tree-core.h (struct GTY): Remove tm_clone_flag
4888         and introduce new lambda_function.
4889         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
4891 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4893         PR target/87474
4894         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
4895         P8_VECTOR and VSX are enabled.
4897 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4899         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
4900         0x3907 as CPU model number.
4902 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4904         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
4905         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
4906         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
4907         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
4908         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
4909         * config/s390/s390.md: Likewise. Rename also the cpu attribute
4910         value from arch12 to z14.
4912 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4914         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
4915         (isinfxf2): Ditto.
4916         (isinf<mode>2): Ditto.
4918 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4920         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
4921         before emitting fxam.  Perform calculations in XFmode.
4923 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4925         * match.pd (((X /[ex] A) +- B) * A): New transformation.
4927 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4929         PR middle-end/87319
4930         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
4931         * tree.c (signed_or_unsigned_type_for): Handle complex.
4933 2018-10-02  Jeff Law  <law@redhat.com>
4935         * gimple-fold.c (get_range_strlen): Remove dead code.
4937 2018-10-02  Martin Sebor  <msebor@redhat.com>
4938             Jeff Law  <law@redhat.com>
4940         * builtins.c (unterminated_array): Add new arguments.
4941         If argument is not terminated, bubble up size and exact
4942         state to callers.
4943         (expand_builtin_strnlen): Detect, avoid expanding
4944         and diagnose unterminated arrays.
4945         (c_strlen): Fill in offset of start of unterminated strings.
4946         * builtins.h (unterminated_array): Update prototype.
4948 2018-10-02  Richard Biener  <rguenther@suse.de>
4950         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
4951         of haddv4df, first reduce to SSE width and exploit the fact
4952         that we only need element zero with the reduction result.
4953         (reduc_plus_scal_v2df): Likewise.
4955 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4957         * dojump.h (do_jump): Delete.
4958         (do_jump_1): Likewise.
4959         (split_comparison): Move around.
4960         * dojump.c (do_jump): Make static.
4961         (do_jump_1): Likewise.
4962         (jumpifnot): Move around.
4963         (jumpifnot_1): Likewise.
4964         (jumpif): Likewise.
4965         (jumpif_1): Likewise.
4966         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
4968 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4970         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
4971         insns in the delay slot and add_insn_after for the jump insn.
4973 2018-10-02  Richard Biener  <rguenther@suse.de>
4975         * tree-inline.c (expand_call_inline): Use the location of
4976         the callee declaration for the inline-entry marker.
4977         * final.c (notice_source_line): Remove special-casing of
4978         NOTE_INSN_INLINE_ENTRY.
4980 2018-10-01  Carl Love  <cel@us.ibm.com>
4982         PR 69431
4983         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
4984         (__builtin_mtfsb0): New.
4985         (__builtin_mtfsb1): New.
4986         ( __builtin_set_fpscr_rn): New.
4987         (__builtin_set_fpscr_drn): New.
4988         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
4989         (rs6000_expand_set_fpscr_rn_builtin): Add.
4990         (rs6000_expand_set_fpscr_drn_builtin): Add.
4991         (rs6000_expand_builtin): Add case statement entries for
4992         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
4993         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
4994         RS6000_BUILTIN_MFFSL.
4995         (rs6000_init_builtins): Add ftype initialization and def_builtin
4996         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
4997         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
4998         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
4999         rs6000_mffscdrn): Add define_insn.
5000         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5001         * doc/extend.texi: Add documentation for the builtins.
5003 2018-10-01  Richard Biener  <rguenther@suse.de>
5005         PR tree-optimization/87465
5006         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5007         causing branch miscounts.
5009 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5011         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5012         aarch64_option_default_param):  New.
5013         (params.h): Include.
5014         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5015         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5016         stack-clash protection validation code.
5018 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5020         * params.c (validate_param): New.
5021         (add_params): Use it.
5022         (set_param_value): Refactor param validation into validate_param.
5023         (diagnostic.h): Include.
5024         * diagnostic.h (diagnostic_ready_p): New.
5026 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5028         * params.c (set_param_value):
5029         Add index of parameter being validated.
5030         * common/common-target.def (option_validate_param): New.
5031         * common/common-targhooks.h (default_option_validate_param): New.
5032         * common/common-targhooks.c (default_option_validate_param): New.
5033         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5034         * doc/tm.texi: Regenerate.
5036 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5038         PR target/86486
5039         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5040         Add validation for stack-clash parameters and set defaults.
5042 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5044         PR target/86486
5045         * configure.ac: Add stack-clash-protection-guard-size.
5046         * doc/install.texi: Document it.
5047         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5048         * params.def: Update comment for guard-size.
5049         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5050         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5051         * configure: Regenerate.
5053 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5055         PR target/86486
5056         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5057         STACK_DYNAMIC_OFFSET): New.
5058         * config/aarch64/aarch64.c (aarch64_layout_frame):
5059         Update outgoing args size.
5060         (aarch64_stack_clash_protection_alloca_probe_range,
5061         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5063 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5065         PR target/86486
5066         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5067         probe ranges.
5068         * target.def (stack_clash_protection_alloca_probe_range): New.
5069         (stack_clash_protection_final_dynamic_probe): Remove.
5070         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5071         (default_stack_clash_protection_final_dynamic_probe): Remove.
5072         * targhooks.c: Likewise.
5073         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5074         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5075         * doc/tm.texi: Regenerate.
5077 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5079         PR target/86486
5080         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5081         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5082         aarch64_clamp_to_uimm12_shift): New.
5083         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5084         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5086 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5088         PR target/86486
5089         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5091 2018-10-01  Jeff Law  <law@redhat.com>
5092             Richard Sandiford <richard.sandiford@linaro.org>
5093             Tamar Christina  <tamar.christina@arm.com>
5095         PR target/86486
5096         * config/aarch64/aarch64.md
5097         (probe_stack_range): Add k (SP) constraint.
5098         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5099         STACK_CLASH_MAX_UNROLL_PAGES): New.
5100         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5101         stack probes for stack clash.
5102         (aarch64_allocate_and_probe_stack_space): New.
5103         (aarch64_expand_prologue): Use it.
5104         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5105         (aarch64_sub_sp): Add emit_move_imm optional param.
5107 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5109         PR tree-optimization/87261
5110         * match.pd: Remove trailing whitespace.
5111         Add (x & y) | ~(x | y) -> ~(x ^ y),
5112         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5114 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5116         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5117         constraints.
5118         (ashlsi3_insn): Update instruction constraints.
5119         (ashrsi3_insn): Likewise.
5120         (rotrsi3): Likewise.
5121         (add_shift): Likewise.
5122         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5123         avoids placing in the limm field small constants which, otherwise,
5124         could end into a small instruction.
5126 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5128         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5129         destination register is not odd-even.
5130         (umaddsidi4_split): Likewise.
5132 2018-10-01  Richard Biener  <rguenther@suse.de>
5134         * tree-inline.c (expand_call_inline): Store origin of fn
5135         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5136         * tree.c (block_ultimate_origin): Simplify and do some
5137         checking.
5139 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5141         * config/i386/mmx.md (EMMS): New int iterator.
5142         (emms): New int attribute.
5143         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5144         EMMS int iterator.  Explicitly declare clobbers.
5145         (mmx_emms): Remove expander.
5146         (mmx_femms): Ditto.
5147         * config/i386/predicates.md (emms_operation): Remove predicate.
5148         (vzeroall_pattern): New predicate.
5149         (vzeroupper_pattern): Rename from vzeroupper_operation.
5150         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5151         vzeroupper_pattern and vzeroall_pattern predicates.
5153 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5155         PR rtl-optimization/86939
5156         * ira-lives.c (make_hard_regno_born): Rename from this...
5157         (make_hard_regno_live): ... to this.  Remove update to conflict
5158         information.  Update function comment.
5159         (make_hard_regno_dead): Add conflict information update.  Update
5160         function comment.
5161         (make_object_born): Rename from this...
5162         (make_object_live): ... to this.  Remove update to conflict information.
5163         Update function comment.
5164         (make_object_dead):  Add conflict information update.  Update function
5165         comment.
5166         (mark_pseudo_regno_live): Call make_object_live.
5167         (mark_pseudo_regno_subword_live): Likewise.
5168         (mark_hard_reg_dead): Update function comment.
5169         (mark_hard_reg_live): Call make_hard_regno_live.
5170         (process_bb_node_lives): Likewise.
5171         * lra-lives.c (make_hard_regno_born): Rename from this...
5172         (make_hard_regno_live): ... to this.  Remove update to conflict
5173         information.  Remove now uneeded check_pic_pseudo_p argument.
5174         Update function comment.
5175         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5176         to conflict information.  Update function comment.
5177         (mark_pseudo_live): Remove update to conflict information.  Update
5178         function comment.
5179         (mark_pseudo_dead): Add conflict information update.
5180         (mark_regno_live): Call make_hard_regno_live.
5181         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5182         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5184 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5186         PR target/87370
5187         * config/i386/i386.c (construct_container): Use TImode for
5188         BLKmode values in 2 integer registers.
5190 2018-09-29  Jeff Law  <law@redhat.com>
5192         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5193         c_strlen rather than just a tree *.
5194         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5195         Update recursive calls appropriately.  If caller did not provide a
5196         suitable data pointer, create a local one.  When a non-terminated
5197         string is discovered, bubble up information about the string via the
5198         c_strlen_data object.
5199         * builtins.h (c_strlen): Update prototype.
5200         (c_strlen_data): New structure.
5201         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5202         For a type 2 call, if c_strlen indicates a non-terminated string
5203         use the length of the non-terminated string.
5204         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5206 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5208         PR target/87467
5209         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5210         __m512d type for __A argument rather than __m512.
5212 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5214         * match.pd (simple_comparison): Don't optimize if either operand is
5215         a function pointer when target needs function pointer canonicalization.
5217 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5219         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5220         power5 .. power9 to remove indirection.
5221         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5222         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5223         ASM_CPU_476_SPEC): Delete.
5224         (ASM_CPU_SPEC): Adjust.
5225         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5226         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5228 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5230         * config.in: Delete HAVE_AS_DCI.
5231         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5232         * config/rs6000/rs6000.h: Ditto.
5233         * configure.ac: Delete HAVE_AS_DCI.
5234         * configure: Regenerate.
5236 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5238         * config.in (HAVE_AS_LWSYNC): Delete.
5239         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5240         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5241         do it as a .long .
5242         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5243         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5244         as a .long .
5245         * configure.ac: Delete HAVE_AS_LWSYNC.
5246         * configure: Regenerate.
5248 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5249             Pierre-Marie de Rodat  <derodat@adacore.com>
5251         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5252         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5253         (cgraph_node::create_thunk): Add indirect_offset parameter.
5254         (thunk_adjust): Likewise.
5255         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5256         and initialize the corresponding field with it.
5257         (cgraph_node::dump): Dump indirect_offset field.
5258         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5259         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5260         (thunk_adjust): Add indirect_offset parameter and deal with it.
5261         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5262         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5263         or if the thunk is external or local.  Fix formatting.  Do not chain
5264         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5265         if any, in the GIMPLE representation.
5266         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5267         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5268         (input_node): Read indirect_offset field.
5269         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5270         call to thunk_adjust.
5271         * tree-nested.c (struct nesting_info): Add thunk_p field.
5272         (create_nesting_tree): Set it.
5273         (convert_all_function_calls): Copy static chain from targets to thunks.
5274         (finalize_nesting_tree_1): Return early for thunks.
5275         (unnest_nesting_tree_1): Do not finalize thunks.
5276         (gimplify_all_functions): Do not gimplify thunks.
5278 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5280         * opt-suggestions.c (option_proposer::build_option_suggestions):
5281         Release "option_values".
5283 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5285         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5286         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5287         * dumpfile.c (kind_as_string): New function.
5288         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5289         kind_as_string to vary the prefix based on dump_kind.
5290         (selftest::test_capture_of_dump_calls): Update for above.
5292 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5294         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5295         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5297 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5299         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5300         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5301         INVALID_REGNUM instead of FPSR_REG.
5302         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5303         * config/i386/i386.md: Update comment of FP compares.
5304         (fldenv): Do not clobber FPSR_REG.
5306 2018-09-28  Richard Biener  <rguenther@suse.de>
5308         * tree.h (BLOCK_ORIGIN): New.
5309         * omp-expand.c (grid_expand_target_grid_body): Assign
5310         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5311         * tree-inline.c (remap_block): Likewise.
5312         * auto-profile.c (get_function_decl_from_block): Simplify
5313         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5314         * langhooks.c (lhd_print_error_function): Likewise.
5315         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5316         Likewise.
5317         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5318         * tree.c (block_nonartificial_location): Likewise.
5319         (block_ultimate_origin): Likewise.
5320         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5321         no longer needed LTO case.
5323 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5324             Jan Hubicka  <jh@suse.cz>
5325             Martin Jambor  <mjambor@suse.cz>
5327         * simplify-rtx.c (simplify_merge_mask): New function.
5328         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5329         same masks are used in op1 or op2.
5330         (test_vec_merge): New function.
5331         (test_vector_ops): Call test_vec_merge.
5333 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5335         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5336         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5337         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5338         (cypress_costs): Set it.
5339         (supersparc_costs): Likewise.
5340         (hypersparc_costs): Likewise.
5341         (leon_cost): Likewise.
5342         (leon3_costs): Likewise.
5343         (sparclet_costs): Likewise.
5344         (ultrasparc_costs): Likewise.
5345         (ultrasparc_costs): Likewise.
5346         (niagara_costs): Likewise.
5347         (niagara2_costs): Likewise.
5348         (niagara3_costs): Likewise.
5349         (niagara4_costs): Likewise.
5350         (niagara7_costs): Likewise.
5351         (m8_costs): Likewise.
5352         (TARGET_CAN_FOLLOW_JUMP): Define.
5353         (pass_work_around_errata::gate): Minor tweak.
5354         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5355         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5356         Automaitcally clear MASK_FSMULD mask for V7 processors.
5357         (sparc_can_follow_jump): New static function.
5358         (output_ubranch): Deal with CROSSING_JUMP_P.
5359         (sparc_use_sched_lookahead): Rewrite using switch statement.
5360         (sparc_issue_rate): Reorder.
5361         (sparc_branch_cost): New function.
5363 2018-09-27  Martin Sebor  <msebor@redhat.com>
5365         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5366         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5367         (int_fits_type_p): Same.
5369 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5371         * config/i386/i386.md (FPCR_REG): Remove.
5372         (UNSPEC_FLDCW): Remove.
5373         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5374         (x86_fldcw_1): Remove insn pattern.
5375         (fnstenv): Do not clobber FPCR_REG.
5376         (fldenv): Ditto.
5377         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5378         (CALL_USED_REGISTERS): Ditto.
5379         (REG_ALLOC_ORDER): Ditto.
5380         (REG_CLASS_CONTENTS): Ditto.
5381         (HI_REGISTER_NAMES): Ditto.
5382         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5383         of numerical constants.
5384         * config/i386/i386.c (regclass_map): Remove fpsr register.
5385         (dbx_register_map): Ditto.
5386         (dbx64_register_map): Ditto.
5387         (svr4_dbx_register_map): Ditto.
5388         (print_reg): Do not handle FPCR_REG.
5390 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5392         PR target/87149
5393         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5394         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5395         Delete, always treat as true.
5396         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5397         Ditto.  Simplify remaining code.
5398         * config/powerpcspe/powerpcspe.h: Ditto.
5399         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5400         Simplify remaining code.
5401         (rs6000_expand_builtin): Ditto.
5402         * config/rs6000/rs6000.h: Ditto.
5403         * configure.ac: Ditto.
5404         * configure: Regenerate.
5406 2018-09-27  Martin Liska  <mliska@suse.cz>
5408         * coverage.c (get_coverage_counts): Revert the formatting
5409         of missing profile opt info.
5411 2018-09-27  Richard Biener  <rguenther@suse.de>
5413         PR debug/37801
5414         PR debug/87440
5415         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5416         block as we do not output that.
5417         (gen_inlined_subroutine_die): Elide the originally outermost
5418         block, matching what we do for concrete instances.
5419         (decls_for_scope): Add parameter specifying whether to recurse
5420         to subblocks.
5422 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5423             Tom de Vries  <tom@codesourcery.com>
5425         PR 82089
5427         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5428         STORE_FLAG_VALUE == 1.
5430 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5432         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5433         constant definitions.
5434         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5435         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5436         ("speculation_barrier"): New expander definition.
5438 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5440         PR gcov-profile/86957
5441         * common.opt: New warning option -Wmissing-profile.
5442         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5443         * doc/invoke.texi: Document -Wmissing-profile.
5445 2018-09-26  Jim Wilson  <jimw@sifive.com>
5447         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5448         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5450 2018-09-26  Martin Sebor  <msebor@redhat.com>
5452         * tree.c (zerop): Change return type to bool.
5453         (integer_zerop, integer_onep, integer_each_onep): Same.
5454         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5455         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5456         (real_onep, real_minus_onep, chain_index): Same.
5457         (print_type_hash_statistics, type_list_equal): Same.
5458         * tree.h (zerop): Same.
5459         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5460         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5461         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5462         (real_onep, real_minus_onep, chain_index): Same.
5463         (print_type_hash_statistics, type_list_equal): Same.
5465 2018-09-26  Jim Wilson  <jimw@sifive.com>
5467         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5469 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5471         PR target/87414
5472         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5473         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5474         call.
5476 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5478         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5480 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5482         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5483         and FP_SECOND_SSE_REGS.
5484         (REG_CLASS_NAMES): Ditto.
5485         (REG_CLASS_CONTENTS): Ditto.
5486         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5487         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5488         (ix86_preferred_output_reload_class): Ditto.
5489         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5490         clobber constraint to "=&f".
5491         (fix_truncdi_i387): Ditto.
5492         (lrintxfdi2): Ditto.
5493         (fistdi2_<rounding>): Ditto.
5494         (fpremxf4_i387): Change "=u" constraint to "=f".
5495         (fprem1xf4_i387): Ditto.
5496         (sincosxf3): Ditto.
5497         (fptanxf4_i387): Ditto.
5498         (fxtractxf3_i387): Ditto.
5499         (fscalexf4_i387): Ditto.
5500         (atan2xf3): Change "u" constraint to "f".
5501         (fyl2xxf3_i387): Ditto.
5502         (fyl2xp1xf3_i387): Ditto.
5504 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5506         PR target/87439
5507         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5508         for removed I387_MASK_PM entity.
5511 2018-09-26  Jeff Law  <law@redhat.com>
5512         Revert
5513         2018-09-26  Alexey Neyman  <stilor@att.net>
5515         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5516         headers are no longer pulled in by <isl/val.h>.
5518 2018-09-26  Richard Biener  <rguenther@suse.de>
5520         PR debug/87443
5521         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
5522         or concrete instance DIE to the tree.  Create abstract origin
5523         attributes also for concrete instances.
5525 2018-09-26  Alexey Neyman  <stilor@att.net>
5527         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5528         headers are no longer pulled in by <isl/val.h>.
5530 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
5532         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
5533         Use new helper functions.
5534         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
5535         Use new helper functions.
5536         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
5537         aarch_mm_needs_release): New declarations.
5538         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
5539         aarch_mm_needs_release): New.
5541 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
5543         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
5544         (arm32_output_mi_thunk): Deal with long calls.
5546 2018-09-26  Richard Biener  <rguenther@suse.de>
5548         PR debug/87428
5549         PR debug/87362
5550         * tree-inline.c (expand_call_inline): When the location
5551         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
5552         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
5553         the inserted BLOCK to make inlined_function_outer_scope_p
5554         recognize it.
5555         * dwarf2out.c (add_call_src_coords_attributes): Do not add
5556         coords for reserved locations.
5558 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5560         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
5561         (*call_indirect_nonlocal_sysv<mode>): Ditto.
5562         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
5563         (*sibcall_nonlocal_sysv<mode>): Ditto.
5564         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
5565         (<bd>_<mode>): Ditto.
5566         (<bd>tf_<mode>): Ditto.
5568 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5570         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
5571         control string as a list of templates instead of as C code.
5572         (*altivec_movti): Ditto.
5573         * config/rs6000/darwin.md (movdf_low_di): Ditto.
5575 2018-09-25  Jim Wilson  <jimw@sifive.com>
5577         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
5578         when target symbol is weak.
5580 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5582         PR c/87387
5583         * builtins.c (unterminated_array): Simplify.
5584         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
5585         where pointer arithmetic is safe.
5587 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5589         PR target/86987
5590         * config/rs6000/altivec.md (altivec_vspltb): Use
5591         const_0_to_15_operand instead of u5bit_cint_operand.
5592         (*altivec_vspltb_internal): Ditto.
5593         (altivec_vspltb_direct): Ditto.
5594         (altivec_vsplth): Use const_0_to_7_operand instead of
5595         u5bit_cint_operand.
5596         (*altivec_vsplth_internal): Ditto.
5597         (altivec_vsplth_direct): Ditto.
5598         (altivec_vspltw): Use const_0_to_3_operand instead of
5599         u5bit_cint_operand.
5600         (*altivec_vspltw_internal): Ditto.
5601         (altivec_vspltw_direct): Ditto.
5602         (altivec_vspltsf): Ditto.
5603         (*altivec_vspltsf_internal): Ditto.
5604         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
5605         various splats with the proper size immediate.  Reorder the various
5606         cases by ascending size of immediate, and put all such together.
5608 2018-09-25  Richard Biener  <rguenther@suse.de>
5610         PR debug/83941
5611         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
5612         GC-ification.
5613         (maybe_create_die_with_external_ref): Do not create
5614         DW_TAG_imported_unit here.
5615         (add_abstract_origin_attribute): Handle external BLOCK refs.
5616         (dwarf2out_abstract_function): Simplify LTO case.
5617         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
5618         rather than using maybe_create_die_with_external_ref.
5620 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
5622         PR target/71278
5623         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
5624         (frndintxf2_mask_pm_i387): Ditto.
5625         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
5626         Enable for !flag_trapping_math.
5627         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
5628         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
5629         Change operand 1 predicate to nonimmediate_operand.
5630         (attr "i387_cw"): Remove mask_pm.
5631         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
5632         (enum ix86_entity): Remove I387_MASK_PM.
5633         * config/i386/i386.c (ix86_i387_mode_needed): Do not
5634         handle I387_MASK_PM.
5635         (ix86_mode_needed): Ditto.
5636         (ix86_mode_after): Ditto.
5637         (ix86_mode_entry): Ditto.
5638         (ix86_mode_exit): Ditto.
5639         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
5641 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
5643         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
5644         to_update_switch_stmts to vNULL instead of calling create on them
5645         immediately.
5647 2018-09-25  Richard Biener  <rguenther@suse.de>
5649         PR tree-optimization/87402
5650         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
5651         (visit_phi): Re-instantiate handling of supposed to be VARYING
5652         but non-VARYING backedge value.
5654 2018-09-25  Richard Biener  <rguenther@suse.de>
5656         PR debug/83941
5657         * dwarf2out.c (struct sym_off_pair): New.
5658         (external_die_map): New global.
5659         (lookup_decl_die): When in LTO create DIEs lazily from the
5660         external_die_map.
5661         (lookup_block_die): New function, create DIEs lazily in LTO.
5662         (equate_block_to_die): New function.
5663         (dwarf2out_die_ref_for_decl): During WPA get the association
5664         from the external DIE map.
5665         (dwarf2out_register_external_die): Record mapping into the
5666         external DIE map.
5667         (maybe_create_die_with_external_ref): New function split out from
5668         DIE generation part of old dwarf2out_register_external_die.
5669         (add_abstract_origin_attribute): Do not return the DIE.  When
5670         in LTO reference externals directly.
5671         (dwarf2out_abstract_function): When in LTO ignore calls for
5672         decls with external DIEs (already present abstract instances).
5673         (gen_call_site_die): Adjust.
5674         (add_high_low_attributes): Likewise.
5675         (gen_lexical_block_die): Likewise.
5676         (gen_inlined_subroutine_die): Likewie.
5677         (gen_block_die): Likewise.
5678         (dwarf2out_inline_entry): Likewise.
5679         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
5680         DIEs.
5682 2018-09-25  Martin Liska  <mliska@suse.cz>
5684         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
5685         integers and not by a float value.
5687 2018-09-25  Martin Liska  <mliska@suse.cz>
5689         PR fortran/87394
5690         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
5691         instead of NULL.
5692         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
5693         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
5694         * godump.c (go_format_type): Remove extra parenthesis.
5696 2018-09-25  Martin Liska  <mliska@suse.cz>
5698         * alias.c (set_dest_equal_p): Remove unused function.
5699         * config/i386/i386.c (def_builtin_pure2): Likewise.
5700         * diagnostic-show-locus.c (class layout): Remove
5701         unused field.
5702         (layout::layout): Likewise here.
5703         * dump-context.h (class temp_dump_context): Likewise.
5704         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
5705         (add_AT_loclistsptr): Likewise.
5706         (add_AT_offset): Likewise.
5707         (get_AT_hi_pc): Likewise.
5708         (is_comdat_die): Likewise.
5709         (type_is_enum): Likewise.
5710         (ceiling): Likewise.
5711         (add_AT_vms_delta): Likewise.
5712         (is_class_die): Likewise.
5713         * edit-context.c (class line_event): Remove unused field.
5714         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
5715         unused function.
5716         * ipa-cp.c (ipa_get_vr_lat): Likewise.
5717         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
5718         (ok_for_base_p_nonstrict): Likewise.
5719         * tree-chrec.c (is_not_constant_evolution): Likewise.
5720         (chrec_fold_poly_cst): Likewise.
5721         * tree-if-conv.c (has_pred_critical_p): Likewise.
5722         * tree-ssa-coalesce.c (print_exprs): Likewise.
5723         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
5724         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
5725         * tree-vrp.c (value_ranges_intersect_p): Likewise.
5726         (value_range_nonnegative_p): Likewise.
5728 2018-09-25  Martin Liska  <mliska@suse.cz>
5730         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5731         Do not handle "GNU Pascal".
5732         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5733         Likewise.
5734         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
5735         from documentation. Likewise.
5736         * dbxout.c (dbxout_range_type): Likewise.
5737         * doc/cpp.texi: Likewise.
5738         * doc/extend.texi: Likewise.
5739         * doc/frontends.texi: Likewise.
5740         * doc/invoke.texi: Remove Pascal entry.
5741         * tree.def (CLEANUP_POINT_EXPR): Likewise.
5742         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
5744 2018-09-25  Martin Liska  <mliska@suse.cz>
5746         PR middle-end/86078
5747         * doc/invoke.texi: Document all parameters and remove default
5748         of the parameters.
5750 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5752         PR bootstrap/87417
5753         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
5754         contains HOST_WIDE_INTs when computing its size.
5756 2018-09-24  Jim Wilson  <jimw@sifive.com>
5758         PR target/87391
5759         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
5760         not TARGET_RVE.
5761         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
5763 2018-09-24  Andrew Pinski  <apinski@marvell.com>
5765         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
5766         access prev before checking it for NULLness in the
5767         AARCH64_FUSE_CMP_BRANCH case.
5769 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
5771         PR target/82699
5772         * config/i386/i386.c (rest_of_insert_endbranch): Set
5773         endbr_queued_at_entrance to true and don't insert ENDBR if
5774         x86_function_profiler will be called.
5775         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
5776         is true.
5777         * config/i386/i386.h (machine_function): Add
5778         endbr_queued_at_entrance.
5780 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5782         * genattrtab.c (mk_attr_alt): Use alternative_mask.
5783         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
5784         types.
5785         (check_attr_test): Use alternative_mask.
5786         (get_attr_value): Likewise.
5787         (compute_alternative_mask): Use alternative_mask and XWINT.
5788         (make_alternative_compare): Use alternative_mask.
5789         (attr_alt_subset_p): Use XWINT.
5790         (attr_alt_subset_of_compl_p): Likewise.
5791         (attr_alt_intersection): Use alternative_mask and XWINT.
5792         (attr_alt_union): Likewise.
5793         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
5794         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
5795         (simplify_test_exp): Use alternative_mask and XWINT.
5796         (write_test_expr): Use alternative_mask and XWINT, adjust bit
5797         number calculation to support 64 bits.  Generate code that
5798         checks 64-bit masks.
5799         (main): Use alternative_mask.
5800         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
5802 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5804         PR target/80080
5805         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5806         RETURN+USE when returning via %r14.
5808 2018-09-24  Martin Liska  <mliska@suse.cz>
5810         * gcov.c (output_lines): Print colorization legend
5811         for both flag_use_colors and flag_use_hotness_colors.
5812         Reword the help.
5814 2018-09-24  Martin Liska  <mliska@suse.cz>
5816         * coverage.c (get_coverage_counts): Use warning_at
5817         with current_function_decl location. Use %qD in warning
5818         message.
5820 2018-09-24  Martin Liska  <mliska@suse.cz>
5822         * memory-block.h (memory_block_pool::release): Annotate with
5823         valgrind that the memory is not accessible.
5825 2018-09-24  Martin Liska  <mliska@suse.cz>
5827         PR sanitizer/85774
5828         * asan.c: Make asan_handled_variables extern.
5829         * asan.h: Likewise.
5830         * cfgexpand.c (expand_stack_vars): Make sure
5831         a representative is unpoison if another
5832         variable in the partition is handled by
5833         use-after-scope sanitization.
5835 2018-09-24  Richard Biener  <rguenther@suse.de>
5837         PR tree-optimization/63155
5838         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
5839         the worklist when the edge of the respective argument isn't
5840         executable.
5842 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5844         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
5845         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
5846         (MASK_CLASS_P): Update for rename.
5847         (MAYBE_MASK_CLASS_P): Ditto.
5848         (REG_CLASS_NAMES): Update.
5849         (REG_CLASS_CONTENT): Update.
5850         * config/i386/i386.c (regclass_map): Update for MASK_REG
5851         and ALL_MASK_REGS rename.
5852         * config/i386/constraints.md (Yk): Update for rename.
5853         (k): Ditto.
5855 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5857         * config/i386/i386.h (enum reg_class): Remove
5858         EVEX_SSE_REGS and MOD4_SSE_REGS.
5859         (REG_CLASS_NAMES): Update.
5860         (REG_CLASS_CONTENT): Update.
5861         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
5862         registers as ALL_SSE_REGS.
5863         (ix86_additional_allocno_class_p): Remove.
5864         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
5865         (ix86_register_priority): Lower priority of EVEX SSE registers.
5866         Use IN_RANGE macro where appropriate.
5867         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
5868         AVX-5124VNNIW checks.
5869         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
5870         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
5871         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
5872         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
5873         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
5874         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
5875         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
5876         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
5877         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
5878         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
5879         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
5880         * config/i386/constraints.md (Yh): Remove.
5882 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5884         * config/i386/i386.c (regclass_map): Declare integer REX registers
5885         as GENERAL_REGS.
5887 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
5889         * doc/service.texi (Service): Switch the fsf.org link to https.
5891 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5893         PR target/86798
5894         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5895         Define to speculation_safe_value_not_needed.
5897 2018-09-21  Florian Weimer  <fweimer@redhat.com>
5899         PR middle-end/81035
5900         * doc/extend.texi (Common Function Attributes): Mention that
5901         noreturn suppresses tail call optimization.
5903 2018-09-21  Jeff Law  <law@redhat.com>
5905         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
5906         vr_values::cleanup_edges_and_switches.
5907         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
5908         vr_values class.
5909         (identify_jump_threads): Remove EDGE_IGNORE handling.
5910         (execute_vrp): Move handling of to_remove_edges and
5911         to_update_switch_stmts into vr_values class member functions.
5912         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
5913         (to_update_switch_stmts): Likewise.
5914         * vr-values.c: Include cfghooks.h.
5915         (vr_values::vr_values): Initialize to_remove_edges and
5916         to_update_switch_stmts.
5917         (vr_values::~vr_values): Verify to_remove_edges and
5918         to_update_switch_stmts are empty.
5919         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
5920         (vr_values::cleanup_edges_and_switches): New member function.
5921         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
5922         function.  Add new data members.
5924 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
5926         PR tree-optimization/87309
5927         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
5928         calls with pflags and alt_flags.
5929         (selftest::test_capture_of_dump_calls): Add test of interaction of
5930         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
5932 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5934         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
5936 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5938         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
5940 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5942         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
5943         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
5944         (VXWORKS_LIBS_RTP): Minor reordering.
5946 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5948         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
5949         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
5951 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5953         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
5954         (PTRDIFF_TYPE): Likewise.
5956 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5958         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
5959         triplet, similar to support for VxWorks7.
5960         * config/vxworks-dummy.h: Provide a default definition
5961         of TARGET_VXWORKS64 to 0.
5963 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5965         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
5966         * config/vxworks-dummy.h: here.
5968 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5970         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
5972 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
5973             Bo Zhou  <zbo.zhou@hisilicon.com>
5975         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
5976         * config/aarch64/aarch64-tune.md: Regenerated.
5977         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
5978         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
5979         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
5981 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
5982             Julian Brown  <julian@codesourcery.com>
5984         * builtins.c (get_builtin_sync_mem): Handle address spaces.
5986 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
5988         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
5989         if the call takes a static chain.
5991 2018-09-21  Martin Liska  <mliska@suse.cz>
5993         * auto-profile.c (autofdo_source_profile::read): Do not
5994         set sum_all.
5995         (read_profile): Do not add working sets.
5996         (read_autofdo_file): Remove sum_all.
5997         (afdo_callsite_hot_enough_for_early_inline): Remove const
5998         qualifier.
5999         * coverage.c (struct counts_entry): Remove gcov_summary.
6000         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6001         do not support GCOV_TAG_PROGRAM_SUMMARY.
6002         (get_coverage_counts): Remove summary and expected
6003         arguments.
6004         * coverage.h (get_coverage_counts): Likewise.
6005         * doc/gcov-dump.texi: Remove -w option.
6006         * gcov-dump.c (dump_working_sets): Remove.
6007         (main): Do not support '-w' option.
6008         (print_usage): Likewise.
6009         (tag_summary): Likewise.
6010         * gcov-io.c (gcov_write_summary): Do not dump
6011         histogram.
6012         (gcov_read_summary): Likewise.
6013         (gcov_histo_index): Remove.
6014         (gcov_histogram_merge): Likewise.
6015         (compute_working_sets): Likewise.
6016         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6017         it not obsolete.
6018         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6019         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6020         (GCOV_HISTOGRAM_SIZE): Remove.
6021         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6022         (struct gcov_summary): Simplify rapidly just
6023         to runs and sum_max fields.
6024         (gcov_histo_index): Remove.
6025         (NUM_GCOV_WORKING_SETS): Likewise.
6026         (compute_working_sets): Likewise.
6027         * gcov-tool.c (print_overlap_usage_message): Remove
6028         trailing empty line.
6029         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6030         (output_lines): Remove program related line.
6031         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6032         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6033         histogram.
6034         (input_profile_summary): Do not read it.
6035         (merge_profile_summaries): And do not merge it.
6036         (input_symtab): Do not call removed function.
6037         * modulo-sched.c (sms_schedule): Do not print sum_max.
6038         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6039         removed when histogram method was invented.
6040         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6041         mode.
6042         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6043         GCOV coding style.
6044         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6045         and dump selected value.
6046         * profile.c (add_working_set): Remove.
6047         (get_working_sets): Likewise.
6048         (find_working_set): Likewise.
6049         (get_exec_counts): Do not work with working sets.
6050         (read_profile_edge_counts): Do not inform as sum_max is removed.
6051         (compute_branch_probabilities): Likewise.
6052         (compute_value_histograms): Remove argument for call of
6053         get_coverage_counts.
6054         * profile.h: Do not make gcov_summary const.
6056 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6058         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6060 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6062         PR tree-optimization/86990
6063         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6064         Check that the entire merged store group is made of constants only for
6065         overlapping stores.
6067 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6069         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6070         (VTABLE_VERIFICATION_SPEC): Likewise.
6071         (SANITIZER_EARLY_SPEC): Likewise.
6072         (SANITIZER_SPEC): Likewise.
6073         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6074         * doc/invoke.texi (Link Options): Document -r.
6076 2018-09-20  Richard Biener <rguenther@suse.de>
6078         PR middle-end/87054
6079         * gimplify.c (gimplify_expr): Retain alignment of
6080         addressable lvalue in dereference.
6082 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6084         PR bootstrap/87013
6085         * configure.ac: Check for .loc is_stmt support.
6086         * configure, config.in: Rebuilt.
6087         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6088         if not supported.
6090 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6092         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6093         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6094         -misel=no.
6096 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6098         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6099         VECTOR_OTHER.
6100         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6101         case VECTOR_OTHER.
6103 2018-09-20  Marek Polacek  <polacek@redhat.com>
6105         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6107 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6109         PR tree-optimization/87288
6110         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6111         into account when determining PEELING_FOR_NITERS.
6113 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6115         PR tree-optimization/86877
6116         * tree-vect-loop.c (vect_analyze_loop_2): Call
6117         vect_verify_datarefs_alignment.
6119 2018-09-19  Marek Polacek  <polacek@redhat.com>
6121         * doc/invoke.texi: Document -Wclass-conversion.
6123 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6125         * config/pa/pa.c (pa_adjust_priority): Delete.
6126         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6128         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6129         (atomic_storehi): Likewise.
6130         (atomic_storesi): Likewise.
6131         (atomic_loaddi): Restore compare and swap exchange loop code.
6133 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6135         PR rtl-optimization/86902
6136         * combine.c (try_combine): When changing the CC mode used, don't change
6137         an unrelated mode in other_insn to that new CC mode.
6139 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6141         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6142         with %T in place of calls to dump_generic_expr.
6143         (prune_runtime_alias_test_list): Likewise.
6144         (create_runtime_alias_checks): Likewise.
6145         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6146         (vect_analyze_data_ref_dependence): Likewise.
6147         (vect_slp_analyze_data_ref_dependence): Likewise.
6148         (vect_record_base_alignment): Likewise.  Use %G in place of call
6149         to dump_gimple_stmt.
6150         (vect_compute_data_ref_alignment): Likewise.
6151         (verify_data_ref_alignment): Likewise.
6152         (vect_find_same_alignment_drs): Likewise.
6153         (vect_analyze_group_access_1): Likewise.
6154         (vect_analyze_data_ref_accesses): Likewise.
6155         (dependence_distance_ge_vf): Likewise.
6156         (dump_lower_bound): Likewise.
6157         (vect_prune_runtime_alias_test_list): Likewise.
6158         (vect_find_stmt_data_reference): Likewise.
6159         (vect_analyze_data_refs): Likewise.
6160         (vect_create_addr_base_for_vector_ref): Likewise.
6161         (vect_create_data_ref_ptr): Likewise.
6162         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6163         (vect_can_advance_ivs_p): Likewise.
6164         (vect_update_ivs_after_vectorizer): Likewise.
6165         (vect_gen_prolog_loop_niters): Likewise.
6166         (vect_prepare_for_masked_peels): Likewise.
6167         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6168         (vect_determine_vectorization_factor): Likewise.
6169         (vect_is_simple_iv_evolution): Likewise.
6170         (vect_analyze_scalar_cycles_1): Likewise.
6171         (vect_analyze_loop_operations): Likewise.
6172         (report_vect_op): Likewise.
6173         (vect_is_slp_reduction): Likewise.
6174         (check_reduction_path): Likewise.
6175         (vect_is_simple_reduction): Likewise.
6176         (vect_create_epilog_for_reduction): Likewise.
6177         (vect_finalize_reduction:): Likewise.
6178         (vectorizable_induction): Likewise.
6179         (vect_transform_loop_stmt): Likewise.
6180         (vect_transform_loop): Likewise.
6181         (optimize_mask_stores): Likewise.
6182         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6183         (vect_split_statement): Likewise.
6184         (vect_recog_over_widening_pattern): Likewise.
6185         (vect_recog_average_pattern): Likewise.
6186         (vect_determine_min_output_precision_1): Likewise.
6187         (vect_determine_precisions_from_range): Likewise.
6188         (vect_determine_precisions_from_users): Likewise.
6189         (vect_mark_pattern_stmts): Likewise.
6190         (vect_pattern_recog_1): Likewise.
6191         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6192         (vect_record_max_nunits): Likewise.
6193         (vect_build_slp_tree_1): Likewise.
6194         (vect_build_slp_tree_2): Likewise.
6195         (vect_print_slp_tree): Likewise.
6196         (vect_analyze_slp_instance): Likewise.
6197         (vect_detect_hybrid_slp_stmts): Likewise.
6198         (vect_detect_hybrid_slp_1): Likewise.
6199         (vect_slp_analyze_operations): Likewise.
6200         (vect_slp_analyze_bb_1): Likewise.
6201         (vect_transform_slp_perm_load): Likewise.
6202         (vect_schedule_slp_instance): Likewise.
6203         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6204         (vect_mark_stmts_to_be_vectorized): Likewise.
6205         (vect_init_vector_1): Likewise.
6206         (vect_get_vec_def_for_operand): Likewise.
6207         (vect_finish_stmt_generation_1): Likewise.
6208         (vect_check_load_store_mask): Likewise.
6209         (vectorizable_call): Likewise.
6210         (vectorizable_conversion): Likewise.
6211         (vectorizable_operation): Likewise.
6212         (vectorizable_load): Likewise.
6213         (vect_analyze_stmt): Likewise.
6214         (vect_is_simple_use): Likewise.
6215         (vect_get_vector_types_for_stmt): Likewise.
6216         (vect_get_mask_type_for_stmt): Likewise.
6217         * tree-vectorizer.c (increase_alignment): Likewise.
6219 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6221         * doc/rtl.texi: Adjust vec_select description.
6222         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6223         non-constant selectors.
6225 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6227         * config/aarch64/aarch64-protos.h
6228         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6229         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6230         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6231         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6232         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6233         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6234         (AARCH64_ISA_RCPC8_4): New flag.
6235         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6236         to aarch64_offset_9bit_signed_unscaled_p.
6237         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6238         and use stlur.
6239         * config/aarch64/constraints.md (Ust): New constraint.
6240         * config/aarch64/predicates.md.
6241         (aarch64_9bit_offset_memory_operand): New predicate.
6242         (aarch64_rcpc_memory_operand): New predicate.
6244 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6246         PR rtl-optimization/87361
6247         * rtlanal.c (nonzero_bits1): Revert accidental change.
6249 2018-09-19  Richard Biener  <rguenther@suse.de>
6251         PR tree-optimization/87349
6252         PR tree-optimization/87342
6253         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6255 2018-09-18  Marek Polacek  <polacek@redhat.com>
6257         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6258         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6260 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6262         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6264 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6266         PR rtl-optimization/86882
6267         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6269 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6271         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6272         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6274 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6276         PR other/87353
6277         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6279 2018-09-18  Richard Biener  <rguenther@suse.de>
6281         PR middle-end/63155
6282         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6283         (compute_samebase_partition_bases): Likewise.
6284         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6285         (gimple_can_coalesce_p): Simplify.
6287 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6289         Handle a library implementation of ffs calling __builtin_ffs.
6290         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6291         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6292         to __ffsdi2.
6294 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6296         * diagnostic-show-locus.c (class layout_range): Add field
6297         "m_original_idx".
6298         (layout_range::layout_range): Add "original_idx" param and use it
6299         to initialize new field.
6300         (make_range): Use 0 for original_idx.
6301         (layout::layout): Pass in index to calls to
6302         maybe_add_location_range.
6303         (layout::maybe_add_location_range): Add param "original_idx" and
6304         pass it on to layout_range.
6305         (layout::print_any_labels): Pass on range->m_original_idx to
6306         get_text call.
6307         (gcc_rich_location::add_location_if_nearby): Use 0 for
6308         original_idx.
6309         * gcc-rich-location.h (text_range_label::get_text): Update for new
6310         param.
6311         (range_label_for_type_mismatch::get_text): Likewise.
6313 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6315         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6317 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6319         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6320         format_string_diagnostic_t.
6321         (fmtwarn_n): Likewise.
6322         * substring-locations.c
6323         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6324         (format_warning_n_va): Convert to...
6325         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6326         (format_warning_va): Convert to...
6327         (format_string_diagnostic_t::emit_warning_va): ...this.
6328         (format_warning_at_substring): Convert to...
6329         (format_string_diagnostic_t::emit_warning): ...this.
6330         (format_warning_at_substring_n): Convert to...
6331         (format_string_diagnostic_t::emit_warning_n): ...this.
6332         * substring-locations.h (class format_string_diagnostic_t): New
6333         class.
6334         (format_warning_va): Convert to
6335         format_string_diagnostic_t::emit_warning_va.
6336         (format_warning_n_va): Convert to
6337         format_string_diagnostic_t::emit_warning_n_va.
6338         (format_warning_at_substring): Convert to
6339         format_string_diagnostic_t::emit_warning.
6340         (format_warning_at_substring_n): Convert to
6341         format_string_diagnostic_t::emit_warning_n.
6343 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6344             Bernd Schmidt <bernds_cb1@t-online.de>
6346         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6347         SImode args.
6349 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6351         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6352         operand 0 predicate to nonimmediate operand.
6353         (rint<mode>2_frndint): Remove insn pattern.
6354         (rint<mode>2): Change operand 1 predicate to general_operand.
6355         Extend operand 1 to XFmode and generate rintxf2 insn.
6356         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6357         Do not use X87MODEF mode macro.
6358         (frndintxf2_<rounding>_i387): Rename from
6359         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6360         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6361         to XFmode and generate significandxf3 insn.
6363 2018-09-17  Richard Biener  <rguenther@suse.de>
6365         PR tree-optimization/87328
6366         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6367         visiting unexecutable backedges when not iterating.
6368         (do_rpo_vn): Mark all edges not executable even when not
6369         iterating.
6371 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6373         PR c/63886
6374         * doc/invoke.texi (Warning Options): Likewise.
6376 2018-09-17  Richard Biener  <rguenther@suse.de>
6378         PR tree-optimization/87301
6379         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6380         clean EH info from leftover copy assignments.
6382 2018-09-17  Martin Liska  <mliska@suse.cz>
6384         PR gcov-profile/85871
6385         * gcov.c (output_intermediate_file): Fix out of bounds
6386         access.
6388 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6390         * config/arc/arc.c: Object attributes for core4 not reflected
6391         correctly.
6392         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6393         core3).
6395 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6397         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6399 2018-09-17  Martin Liska  <mliska@suse.cz>
6401         * doc/gcov.texi: Document new option --use-hotness-colors.
6402         * gcov.c (struct source_info): Declare new field.
6403         (source_info::source_info): Set default for maximum_count.
6404         (print_usage): Add new -q option.
6405         (process_args): Process it.
6406         (accumulate_line_info): Save src->maximum_count.
6407         (output_line_beginning): Make color line number if
6408         flag_use_hotness_colors is set.
6409         (output_line_details): Pass default argument value.
6410         (output_lines): Pass src->maximum_count.
6412 2018-09-17  Martin Liska  <mliska@suse.cz>
6414         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6415         Use processor_names table.
6416         * config/i386/i386.c (ix86_default_align): Use
6417         processor_cost_table for alignment values.
6418         (ix86_option_override_internal): Use processor_names.
6419         (ix86_function_specific_print): Likewise.
6420         * config/i386/i386.h (struct processor_costs):
6421         Add alignment values.
6422         (struct ptt): Remove and replace with const char *.
6423         * config/i386/x86-tune-costs.h (struct processor_costs):
6424         Declare default alignments for all costs.
6426 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6428         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6429         symbolics or VR_VARYING ranges for ABS_EXPR.
6430         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6431         when range will wrap.
6433 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6435         PR middle-end/86864
6436         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6437         before and after a JUMP_TABLE_DATA.
6439 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6441         PR middle-end/87188
6442         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6443         when one operand is a function pointer.  Use POINTER_TYPE_P and
6444         FUNC_OR_METHOD_TYPE_P.
6445         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6446         * fold-const.c (build_range_check): Likewise.
6447         * match.pd (simple_comparison): Likewise.
6449 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6451         PR c/82967
6452         * spellcheck.c (get_edit_distance_cutoff): New function.
6453         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6454         (selftest::test_get_edit_distance_one_way): ...this.
6455         (selftest::test_get_edit_distance_unit): Rename to...
6456         (selftest::test_get_edit_distance_both_ways): ...this.
6457         (selftest::test_edit_distances): Move tests to this new function,
6458         and test some more pairs of strings.  Update for above renaming.
6459         (selftest::get_old_cutoff): New function.
6460         (selftest::test_get_edit_distance_cutoff): New function.
6461         (selftest::assert_suggested_for): New function.
6462         (ASSERT_SUGGESTED_FOR): New macro.
6463         (selftest::assert_not_suggested_for): New function.
6464         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6465         (selftest::test_suggestions): New function.
6466         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6467         tests to selftest::test_edit_distances and call it.  Add calls to
6468         selftest::test_get_edit_distance_cutoff and
6469         selftest::test_suggestions.
6470         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6471         (best_match::consider): Replace hard-coded cutoff calculation with
6472         a call to...
6473         (best_match::get_cutoff): New declaration.
6474         (best_match::get_best_meaningful_candidate): Likewise.
6476 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6478         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6480 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6482         revert:
6483         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6485         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6486         terminated string literal.
6488 2018-09-14  Martin Sebor  <msebor@redhat.com>
6490         * builtins.c (unterminated_array): Handle ARRAY_REF.
6491         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6492         * builtins.h (unterminated_array): Declare extern.
6493         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6494         arrays.
6495         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6496         calls.
6498 2018-09-14  Martin Sebor  <msebor@redhat.com>
6499             Jeff Law  <law@redhat.com>
6501         * builtins.c (unterminated_array): New.
6502         (expand_builtin_strcpy): Adjust.
6503         (expand_builtin_strcpy_args): Detect unterminated arrays.
6504         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
6505         unterminated arrays.
6506         * gimple-fold.h (get_maxval_strlen): Add argument.
6507         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
6509         * gimple-fold.c (get_range_strlen): Add argument.
6510         (get_maxval_strlen): Adjust.
6511         * gimple-fold.h (get_range_strlen): Add argument.
6513 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
6515         * config/i386/movdirintrin.h: Fix copyright year.
6517 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
6519         * reg-stack.c: Include regs.h.
6520         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
6521         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6522         FIRST_STACK_REG, not DFmode.
6523         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
6524         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
6525         (change stack): Default register mode to the reg_raw_mode of
6526         FIRST_STACK_REG, not DFmode.
6527         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
6528         (*swapxf): Rename from swapxf.
6530 2018-09-14  Carl Love  <cel@us.ibm.com>
6532         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
6533         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
6535 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
6537         PR target/87224
6538         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
6539         alternatives.
6541 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
6543         PR target/85628
6544         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
6546 2018-09-14  Jason Merrill  <jason@redhat.com>
6548         Fix --enable-gather-detailed-mem-stats.
6549         * hash-table.c (hash_table_usage): Change from variable to function.
6550         * hash-table.h: Adjust.
6551         * Makefile.in: Add missing dependencies on hash-table.h.
6553 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6555         PR tree-optimization/87259
6556         PR lto/87283
6557         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
6558         execute_cse_reciprocals_1 has tried transforming.
6560 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
6562         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
6563         VR_VARYING for PLUS/MINUS_EXPR.
6565 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6567         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
6568         formatting.
6570 2018-09-14  Richard Biener  <rguenther@suse.de>
6572         PR middle-end/63155
6573         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
6574         bits for the merged partition.
6576 2018-09-13  Martin Sebor  <msebor@redhat.com>
6577             Bernd Edlinger  <bernd.edlinger@hotmail.de>
6579         * builtins.h (c_srlen): Add argument.
6580         * builtins.c (warn_string_no_nul): New function.
6581         (c_strlen): Add argument and use it.  Update recursive calls.
6582         Pass DECL argument to string_constant to get info on non
6583         terminated strings.  Update *NONSTR as needed.
6584         (fold_builtin_strlen): Add argument to calls to c_strlen.
6585         Warn for unterminated arrays.
6586         (warn_string_no_null): Add prototype.
6587         * expr.c (string_constant): Update arguments.  Update recursive
6588         calls appropriately.  Detect missing NUL terminator and outermost
6589         declaration its missing in.
6590         Improve checks for arrays with nonzero lower bound or elements
6591         that are not a single byte.  Simplify offset computation.
6592         Simplify checks for non-NUL terminated strings.
6593         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
6594         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
6596 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6598         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
6599         correctly.
6600         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
6601         argument.  Fix range checks.
6602         * fold-const.h (c_getstr): Adjust protoype.
6603         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
6604         string is constant but contains no NUL byte.
6606         * expr.c (string_constant): Adjust function comment.
6607         Remove bogus check for zero termination.
6609         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
6611         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
6612         (get_constant_size): Don't make STRING_CSTs larger than they are.
6613         (check_string_literal): New check function for STRING_CSTs.
6614         (output_constant): Use it.
6616 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
6618         PR target/86812
6619         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
6621 2018-09-13  Richard Biener  <rguenther@suse.de>
6623         PR tree-optimization/87263
6624         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
6625         (struct unwind_state): Add max_rpo field.
6626         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
6627         Compute max_rpo, the max RPO number a block can be backwards reached
6628         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
6629         separating it from the iterating mode.
6631 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6633         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
6634         (rfs_decision): New scheduling decision.
6636 2018-09-13  Richard Biener  <rguenther@suse.de>
6638         PR bootstrap/87134
6639         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
6640         (vn_nary_op_insert_pieces_predicated): Do not write useless
6641         valid_dominated_by_p entry outside of the allocated storage.
6643 2018-09-13  Omar Sandoval  <osandov@osandov.com>
6644             Tom de Vries  <tdevries@suse.de>
6646         PR debug/86985
6647         * dwarf2out.c (is_c): New function.
6648         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
6650 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
6652         PR target/85628
6653         * config/aarch64/aarch64.md (*aarch64_bfxil):
6654         Define.
6655         * config/aarch64/constraints.md (Ulc): Define.
6656         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
6657         Define.
6658         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
6659         New function.
6661 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6663         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
6664         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
6665         aarch64_layout_frame call.
6666         (aarch64_expand_epilogue): Likewise.
6667         (aarch64_initial_elimination_offset): Likewise.
6668         (aarch64_get_separate_components): Likewise.
6669         (aarch64_use_return_insn_p): Likewise.
6670         (aarch64_layout_frame): Remove unneeded check.
6672 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6674         * configure.ac: Only append
6675         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
6676         gcc_config_arguments if it was never reconfigured or last reconfigure
6677         was with different arguments.
6678         * configure: Regenerated.
6680 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6681             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6683         PR middle-end/87290
6684         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
6685         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
6687 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6689         PR tree-optimization/87287
6690         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
6691         X % C == 0 to X % (unsigned) C == 0 optimization to ...
6692         * match.pd (X % C == 0): ... here.  New optimization.
6694 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6696         PR middle-end/82853
6697         * expr.h (maybe_optimize_mod_cmp): Declare.
6698         * expr.c (mod_inv): New function.
6699         (maybe_optimize_mod_cmp): New function.
6700         (do_store_flag): Use it.
6701         * cfgexpand.c (expand_gimple_cond): Likewise.
6703 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
6704             Julian Brown  <julian@codesourcery.com>
6706         PR middle-end/86336
6707         * gimplify.c (gimplify_scan_omp_clauses): Set
6708         target_firstprivatize_array_bases in OpenACC parallel and kernels
6709         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
6710         OpenACC data regions.
6712 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
6714         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
6715         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
6717 2018-09-12  Richard Biener  <rguenther@suse.de>
6719         PR tree-optimization/87280
6720         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
6721         edge but unreachable target.
6722         (do_rpo_vn): For conservatively handling a PHI only mark
6723         the backedge executable but not the block reachable.
6725 2018-09-12  Richard Biener  <rguenther@suse.de>
6727         PR tree-optimization/87266
6728         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
6729         visited blocks.
6731 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
6733         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
6734         constants.
6735         ("trunc<BFP:mode><DFP_ALL:mode>2")
6736         ("trunc<DFP_ALL:mode><BFP:mode>2")
6737         ("extend<BFP:mode><DFP_ALL:mode>2")
6738         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
6739         according to the target operand type.
6741 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6742             Andreas Krebbel  <krebbel@linux.ibm.com>
6744         PR tree-optimization/86844
6745         * gimple-ssa-store-merging.c
6746         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
6747         there are any overlapping stores in between them, make sure they are
6748         also coalesced or we give up completely.
6750 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6752         PR middle-end/87248
6753         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
6754         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
6756 2018-09-12  Tom de Vries  <tdevries@suse.de>
6758         * common.opt (gdescribe-dies): Add option.
6759         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
6760         attribute for artifical and nameless decls.
6761         (dwarf2out_register_external_die): Add description attribute to
6762         external reference die.
6763         (add_desc_attribute): New functions.
6764         (gen_subprogram_die): Add description attribute to
6765         DW_TAG_call_site_parameter.
6766         * tree-pretty-print.c (print_generic_expr_to_str): New function.
6767         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
6768         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
6769         -gno-describe-dies.
6770         (@item -gdescribe-dies): Add.
6772 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6774         * tree-vrp.c (vrp_shift_undefined_p): Remove.
6775         (extract_range_from_binary_expr_1: Call
6776         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
6777         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
6778         depend on sign.
6780 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6782         * gimple-ssa-warn-alloca.c
6783         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
6784         field for ALLOCA_BOUND_*_LARGE.
6786 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6788         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
6790 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6792         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
6793         for clobbers.  Remove obsolete comment.
6795 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6797         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
6798         mpxchk, mpxld and mpxst types.
6799         (define_attr length_immediate): Remove all processing of mpx types.
6800         (define_attr prefix_0f): Ditto.
6801         (define_attr memory): Ditto.
6803 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6805         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
6806         (log<mode>2): Change operand 1 predicate to general_operand.
6807         Extend operand 1 to XFmode and generate logxf3 insn.
6808         (log10<mode>2): Change operand 1 predicate to general_operand.
6809         Extend operand 1 to XFmode and generate log10xf3 insn.
6810         (log2<mode>2): Change operand 1 predicate to general_operand.
6811         Extend operand 1 to XFmode and generate log2xf3 insn.
6812         (fyl2xp1_extend<mode>xf3_i387): Remove.
6813         (log1p<mode>2): Change operand 1 predicate to general_operand.
6814         Extend operand 1 to XFmode and generate log1pxf3 insn.
6815         (fxtract_extend<mode>xf3_i387): Remove.
6816         (logb<mode>2): Change operand 1 predicate to general_operand.
6817         Extend operand 1 to XFmode and generate logbxf3 insn.
6818         (ilogb<mode>2): Change operand 1 predicate to general_operand.
6819         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
6820         (significand<mode>2): Change operand 1 predicate to general_operand.
6821         Extend operand 1 to XFmode and generate significandxf3 insn.
6823 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6825         * gcc.c (perror_with_name, pfatal_with_name): Delete.
6826         (load_specs): Use fatal_error.
6827         (DELETE_IF_ORDINARY, process_command): Use error.
6828         (execute, run_attempt): Use fatal_error.
6830 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
6832         * diagnostic-core.h (sorry_at): New prototype.
6833         * diagnostic.c (sorry_at): New function.
6835 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
6837         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
6838         by zero as VR_UNDEFINED.
6840 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6842         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
6843         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
6844         (<sincos>mode2): New expander.
6845         (sincos_extend<mode>xf3_i387): Remove insn pattern.
6846         (sincos -> sin, cos splitters): Remove splitter patterns.
6847         (sincos<mode>3): Change operand 2 predicate to general_operand.
6848         Extend operand 2 to XFmode and generate sincosxf3 insn.
6849         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
6850         Change operand 3 predicate to const1_operand.
6851         (fptan_extend<mode>xf4_i387): Remove insn pattern.
6852         (tanxf2): Update operands in the call to fptanxf4_i387.
6853         (tan<mode>2): Change operand 1 predicate to general_operand.
6854         Extend operand 1 to XFmode and generate tanxf3 insn.
6855         (atan2xf3): Rename from *fpatanxf3_i387.
6856         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
6857         (atan2xf3): Remove expander.
6858         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
6859         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
6860         (atan<mode>2): Change operand 1 predicate to general_operand.
6861         Extend operand 1 to XFmode and generate atanxf3 insn.
6863 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6865         * config/i386/i386.md (x87/SSE constant load splitter): Use
6866         memory_operand instead of nonimmediate_operand for input operand
6867         predicate.
6869 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
6871         * config/i386/i386.md (float partial SSE register stall splitter): Move
6872         splitter near its instruction pattern.
6873         (float_extend partial SSE register stall splitter): Ditto.
6874         (float_truncate partial SSE register stall splitter): Ditto.
6876 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
6878         PR target/86794
6879         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6880         to speculation_safe_value_not_needed.
6882         PR target/85666
6883         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
6884         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
6885         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
6886         leaf_function_p, instead use has_hard_reg_initial_val.
6888 2018-09-09  Nathan Sidwell  <nathan@acm.org>
6890         * gcc.h (pfatal_with_name): Don't declare here.
6891         * gcc.c (pfatal_with_name): Make static.
6893 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
6895         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
6896         earlyclobber.
6898 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
6900         PR rtl-optimization/85458
6901         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
6902         priority hook to reduce the priority of EXPR.
6904 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
6906         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
6907         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
6908         with X87_ENABLE_FLOAT.  Remove preparation code.
6909         (*float<SWI48:mode><MODEF:mode>2): Rename from
6910         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
6911         math using "enabled" attribute.
6912         (*floatdi<MODEF:mode>2_i387): Rename from
6913         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
6914         enable for 32bit targets only.
6915         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
6916         splitter.
6917         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
6918         as operand 1 predicate.  Rewrite as define_insn_and_split.
6919         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
6921 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6923         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
6924         to fallthru to FLOAT case.
6926 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6928         PR target/86731
6929         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
6930         around folding of vec_sl to handle out of range shift values.
6932 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6934         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
6935         Update callers to gen_fix_trunc<mode>_i387_fisttp
6936         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
6937         nonimmediate_operand.
6938         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
6939         and corresponding splitters.
6940         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
6941         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
6942         (fix_truncdi_i387_with_temp): Remove insn pattern
6943         and corresponding splitters.
6944         (fix_trunc<mode>_i387): Change operand 0 predicate to
6945         nonimmediate_operand.
6946         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
6947         and corresponding splitters.
6948         (*fistdi2_1): Remove.
6949         (fistdi2): Ditto.
6950         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
6951         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
6952         (*fist<mode>2_1): Remove.
6953         (fist<mode>2): Ditto.
6954         (fist<mode>2_with_temp): Remove insn pattern and corresponding
6955         splitters.
6956         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
6957         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
6958         (fistdi2_<rounding>): Change operand 0 predicate to
6959         nonimmediate_operand.
6960         (fistdi2_<rounding>_with_temp): Remove insn pattern
6961         and corresponding splitters.
6962         (fist<mode>2_<rounding>): Change operand 0 predicate to
6963         nonimmediate_operand.
6964         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
6965         and corresponding splitters.
6967         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
6969 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6971         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
6972         the init value.
6974 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6976         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
6977         early gimple folding of vec_splat().
6978         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
6979         * gimple-fold.h: Add an extern define for tree_vec_extract().
6981 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6983         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
6984         wrappers around TREE_TYPE comparisons.
6986 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6988         PR target/80080
6989         * config/s390/predicates.md: Add nonsym_memory_operand.
6990         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
6991         contains a SYMBOL_REF, load it into an intermediate pseudo.
6992         (s390_emit_compare_and_swap): Legitimize operand.
6993         * config/s390/s390.md: Use the new nonsym_memory_operand
6994         with UNSPECV_CAS patterns.
6996 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6998         PR target/80080
6999         * config/s390/s390-passes.def: New file.
7000         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7001         declaration.
7002         (make_pass_s390_early_mach): Add declaration.
7003         * config/s390/s390.c (make_pass_s390_early_mach):
7004         (s390_option_override): Remove dynamic registration.
7005         * config/s390/t-s390: Add s390-passes.def.
7007 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7009         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7010         Remove UNSPEC_LTREL_BASE check.
7011         (annotate_constant_pool_refs): Likewise.
7012         (find_constant_pool_ref): Likewise.
7013         (find_ltrel_base): Removed.
7014         (replace_ltrel_base): Removed.
7015         (s390_mainpool_finish): Remove replace_ltrel_base call.
7016         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7017         (s390_chunkify_finish): Remove replace_ltrel_base call.
7018         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7020 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7022         PR target/86779
7023         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7024         to speculation_safe_value_not_needed.
7026 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7027             Bernd Schmidt  <bernds_cb1@t-online.de>
7029         * config/nvptx/nvptx-opts.h: New file.
7030         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7031         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7032         (ASM_SPEC): Define.
7033         (TARGET_SM35): New macro.
7034         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7035         correct predicate.
7036         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7037         values.
7038         (misa=): New option.
7039         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7041 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7043         * config/i386/i386.md (truncdfsf2): Remove expander.
7044         (truncdfsf2_with_temp): Ditto.
7045         (truncxf<mode>2): Ditto.
7046         (*truncdfsf_fast_mixed): Remove insn pattern.
7047         (*truncdfsf_fast_i387): Ditto.
7048         (*truncdfsf_mixed): Ditto.
7049         (*truncdfsf_i387): Ditto.
7050         (*truncdfsf2_i387_1): Ditto.
7051         (*truncxfsf2_mixed): Ditto.
7052         (*truncxfdf2_mixed): Ditto.
7053         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7054         to call gen_truncxf<mode>2 instead.
7055         (*truncxf<mode>2_i387): Remove.
7056         (reg->reg splitters): Remove splitter pattern.
7057         (reg->mem splitters): Ditto.
7059         (truncdfsf2): New insn pattern.
7060         (truncxf<mode>2): Ditto.
7062 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7064         * tree-ssa-math-opts.c (is_mult_by): New function.
7065         (is_square_of): Use the above.
7066         (optimize_recip_sqrt): New function.
7067         (pass_cse_reciprocals::execute): Use the above.
7069 2018-09-05  Richard Biener  <rguenther@suse.de>
7071         PR bootstrap/87134
7072         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7073         to zero-init the emplaced vec.
7075 2018-09-05  Martin Liska  <mliska@suse.cz>
7077         PR tree-optimization/87205
7078         * tree-switch-conversion.c (pass_lower_switch::execute):
7079         Group cases for switch statements.
7081 2018-09-05  Richard Biener  <rguenther@suse.de>
7083         PR tree-optimization/87217
7084         * tree-ssa-sccvn.c (vuse_valueize): New.
7085         (vn_reference_lookup_pieces): Use it.
7086         (vn_reference_lookup): Likewise.
7088 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7090         PR c++/87137
7091         * stor-layout.c (place_field): Scan forwards to check last
7092         bitfield when ms_bitfield_placement is in effect.
7094 2018-09-05  Richard Biener  <rguenther@suse.de>
7096         PR bootstrap/87225
7097         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7098         return.
7100 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7101             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7103         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7104         * config.gcc (extra_objs): Build it.
7105         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7106         Likewise.
7107         * config/aarch64/aarch64-passes.def
7108         (pass_tag_collision_avoidance): New pass.
7109         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7110         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7111         (aarch64_classify_address): Remove static qualifier.
7112         (aarch64_address_info, aarch64_address_type): Move to...
7113         * config/aarch64/aarch64-protos.h: ... here.
7114         (make_pass_tag_collision_avoidance): New function.
7115         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7116         New tuning flag.
7118 2018-09-05  Martin Liska  <mliska@suse.cz>
7120         * doc/gcov.texi: Update documentation of humar
7121         readable mode.
7122         * gcov.c (format_count): Print one decimal place, it provides
7123         more fine number of situations like '1G' vs. '1.4G'.
7125 2018-09-05  Martin Liska  <mliska@suse.cz>
7127         PR target/87164
7128         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7129         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7130         to generate a MASK value.
7132 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7134         PR debug/86593
7135         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7136         if frame pointer isn't used.
7137         (compute_frame_pointer_to_fb_displacement): Likewise.
7139 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7141         PR target/87198
7142         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7143         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7144         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7145         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7146         and OPTION_MASK_ISA_XSAVEC_UNSET.
7148 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7150         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7151         XOR operations in NAND case.
7153 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7155         * wide-int-range.cc (wide_int_range_convert): New.
7156         * wide-int-range.h (wide_int_range_convert): New.
7157         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7158         code into wide_int_range_convert.
7159         (extract_range_into_wide_ints): Do not munge anti range constants
7160         into the entire domain.  Just return the range back.
7162 2018-09-04  Martin Liska  <mliska@suse.cz>
7164         * genmatch.c (output_line_directive): Add new argument
7165         fnargs.
7166         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7168 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7170         * doc/invoke.texi (Option Summary): Add whitespace.
7172         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7174 2018-09-04  Richard Biener  <rguenther@suse.de>
7176         PR tree-optimization/87211
7177         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7178         backedge value we're supposed to treat as VARYING also number
7179         the PHI to VARYING in case it got a different value-number already.
7181 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7183         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7184         (extract_range_from_binary_expr_1): Do not call
7185         vrp_can_optimize_bit_op.
7186         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7187         static.
7188         (wide_int_range_get_mask_and_bounds): New.
7189         (wide_int_range_optimize_bit_op): New.
7190         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7191         (wide_int_range_bit_and): Same.
7192         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7193         (wide_int_range_optimize_bit_op): New.
7194         (wide_int_range_get_mask_and_bounds): New.
7196 2018-09-04  Richard Biener  <rguenther@suse.de>
7198         PR tree-optimization/87176
7199         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7200         variable.  When value-numbering a virtual PHI node make sure
7201         to not value-number to the backedge value.
7203 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7205         * doc/extend.texi (Long Long, Hex Floats): Document support for
7206         long long and hex floats in more recent versions of ISO C++.
7208 2018-09-03  Richard Biener  <rguenther@suse.de>
7210         PR tree-optimization/87177
7211         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7212         cleanup.
7214 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7216         * bb-reorder.c (edge_order): Convert to C-qsort-style
7217         tri-state comparator.
7218         (reorder_basic_blocks_simple): Change std::stable_sort to
7219         gcc_stablesort.
7221 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7223         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7224         tri-state comparator.
7225         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7227 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7229         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7230         (mergesort): ... here as maximum count for using netsort.
7231         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7232         (gcc_stablesort): New.
7233         * system.h (gcc_stablesort): Declare.
7235 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7237         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7238         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7239         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7241 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7243         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7244         lxsdx and stxsdx alternatives.
7245         (*mov<mode>_hardfloat64): Ditto.
7246         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7248 2018-09-03  Richard Biener  <rguenther@suse.de>
7250         PR tree-optimization/87200
7251         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7252         simplify result.
7254 2018-09-03  Martin Liska  <mliska@suse.cz>
7256         PR tree-optimization/87201
7257         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7258         Fix parenthesis in an expression.
7260 2018-09-03  Richard Biener  <rguenther@suse.de>
7262         PR tree-optimization/87197
7263         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7264         visited.  CSE the VN_INFO hashtable lookup.
7266         PR tree-optimization/87169
7267         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7268         iterating make sure there's no extra backedges from irreducible
7269         regions feeding the header.  Mark the destination block
7270         executable.
7272 2018-09-03  Martin Liska  <mliska@suse.cz>
7274         PR driver/83193
7275         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7276         * common/common-targhooks.c (default_get_valid_option_values):
7277         New function.
7278         * common/common-targhooks.h (default_get_valid_option_values):
7279         Likewise.
7280         * common/config/i386/i386-common.c: Move processor_target_table
7281         from i386.c.
7282         (ix86_get_valid_option_values): New function.
7283         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7284         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7285         (PTA_*): Move all defined masks into i386-common.c.
7286         (ix86_function_specific_restore): Use new processor_cost_table.
7287         * config/i386/i386.h (struct ptt): Moved from i386.c.
7288         (struct pta): Likewise.
7289         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7290         * doc/tm.texi.in: Likewise.
7291         * opt-suggestions.c (option_proposer::suggest_option):
7292         Pass prefix to build_option_suggestions.
7293         (option_proposer::get_completions): Likewise.
7294         (option_proposer::build_option_suggestions): Use the new target
7295         hook.
7296         * opts.c (struct option_help_tuple): New struct.
7297         (print_filtered_help): Use the new target hook.
7299 2018-09-03  Martin Liska  <mliska@suse.cz>
7301         PR middle-end/59521
7302         * predict.c (set_even_probabilities): Add likely_edges
7303         argument and handle cases where we have precisely one
7304         likely edge.
7305         (combine_predictions_for_bb): Catch also likely_edges.
7306         (tree_predict_by_opcode): Handle gswitch statements.
7307         * tree-cfg.h (find_case_label_for_value): New declaration.
7308         (find_taken_edge_switch_expr): Likewise.
7309         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7310         Find pivot in decision tree based on probabily, not by number of
7311         nodes.
7313 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7315         * doc/standards.texi (Standards): Update Objective-C reference.
7317 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7319         * doc/install.texi (Prerequisites): Update link for MPC.
7321 2018-09-01  Michael Matz  <matz@suse.de>
7323         PR tree-optimization/87074
7324         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7325         PHIs for outer-loop uses.
7327 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7329         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7330         * doc/invoke.texi (C Dialect Options): Ditto.
7332 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7334         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7336 2018-08-31  Richard Biener  <rguenther@suse.de>
7338         PR tree-optimization/87168
7339         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7340         (rpo_elim::eliminate_avail): When OP was not visited it must
7341         be available.
7343 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7345         * tree-vrp.c (copy_value_range): Convert param "from" from
7346         "value_range *" to "const value_range *".
7347         (range_is_null): Likewise for param "vr".
7348         (range_int_cst_p): Likewise.
7349         (range_int_cst_singleton_p): Likewise.
7350         (symbolic_range_p): Likewise.
7351         (value_ranges_intersect_p): Likewise for both params.
7352         (value_range_nonnegative_p): Likewise for param "vr".
7353         (value_range_constant_singleton): Likewise.
7354         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7355         (extract_range_into_wide_ints): Likewise for param "vr".
7356         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7357         and "vr1".
7358         (vrp_can_optimize_bit_op): Likewise.
7359         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7360         "vr1_".
7361         (extract_range_from_unary_expr): Likewise.
7362         (debug_value_range): Likewise for param "vr".
7363         (value_range::dump): Add "const" qualifier.
7364         (vrp_prop::check_array_ref): Convert local "vr" from
7365         "value_range *" to "const value_range *".
7366         (vrp_prop::check_mem_ref): Likewise.
7367         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7368         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7369         (vrp_intersect_ranges): Likewise.
7370         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7371         (vrp_prop::vrp_finalize): Likewise.
7372         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7373         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7374         (extract_range_from_unary_expr): Likewise.
7375         (value_range_constant_singleton): Likewise.
7376         (symbolic_range_p): Likewise.
7377         (copy_value_range): Likewise.
7378         (extract_range_from_binary_expr_1): Likewise.
7379         (range_int_cst_p): Likewise.
7380         (vrp_set_zero_nonzero_bits): Likewise.
7381         (range_int_cst_singleton_p): Likewise.
7383 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7385         * config/aarch64/arm_neon.h (vabsd_s64): New.
7386         (vnegd_s64): Likewise.
7388 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7390         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7392 2018-08-31  Martin Liska  <mliska@suse.cz>
7394         * ipa-icf.c (sem_item::add_type): Use
7395         sem_item::m_type_hash_cache.
7396         * ipa-icf.h: Move the cache from sem_item_optimizer
7397         to sem_item.
7399 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7401         * doc/extend.texi (Backwards Compatibility): Remove implicit
7402         extern C leeway of () being (...).
7404 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7406         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7408 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7410         PR target/86684
7411         PR target/87149
7412         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7414 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7416         PR middle-end/87138
7417         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7418         gen_int_mode.  Formatting fixes.
7420 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7422         * target.def (custom_function_descriptors): Improve documentation.
7423         * doc/tm.texi.in (Trampolines): Expand discussion of function
7424         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7425         beginning of the section.
7426         * doc/tm.texi: Regenerated.
7428 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7430         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7431         base class in the initializer list.  This is a workaround for
7432         building with older GCC.
7433         (class auto_bb_flag): Likewise.
7435 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7437         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7438         (altivec_vcmpequ<VI_char>_p): Remove star.
7439         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7440         vector load modes.
7441         (expand_strncmp_vec_sequence): New function.
7442         (emit_final_str_compare_vec): New function.
7443         (expand_strn_compare): Add support for vector strncmp.
7444         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7445         length specification to bytes.
7446         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7447         (vcmpnezb_p): New pattern.
7448         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7449         for option -mstring-compare-inline-limit.
7451 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7453         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7454         (PTA_SKYLAKE): Add PTA_AES.
7455         (PTA_GOLDMONT): Likewise.
7457 2018-08-29  Jan Hubicka  <jh@suse.cz>
7459         PR lto/86517
7460         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7461         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7463 2018-08-29  Jan Hubicka  <jh@suse.cz>
7465         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7466         TYPE_STUB_DECL.
7467         (hash_tree): Do not visit TYPE_STUB_DECL.
7468         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7469         stream TYPE_STUB_DECL.
7470         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7471         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7472         after free_lang_data.
7473         (type_in_anonymous_namespace_p): Likewise.
7475 2018-08-29  Jan Hubicka  <jh@suse.cz>
7477         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7478         comment that it has to be even number.
7479         (class sreal): Change m_sig type to int32_t.
7480         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7481         int64_t for temporary calculations.
7482         (sreal_verify_basics): Drop one bit from minimum and maximum.
7484 2018-08-30  Richard Biener  <rguenther@suse.de>
7486         PR tree-optimization/87147
7487         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7488         (visit_phi): When the degenerate result is from the backedge and
7489         we didn't visit its definition yet drop to VARYING.
7490         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7492 2018-08-29  Jan Hubicka  <jh@suse.cz>
7494         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7495         DECL_VINDEX.
7496         (hash_tree): Likewise.
7498 2018-08-29  Jan Hubicka  <jh@suse.cz>
7500         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
7501         and TYPE_NEXT_REF_TO.
7503 2018-08-29  Jan Hubicka  <jh@suse.cz>
7505         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7506         comment that it has to be even number.
7507         (class sreal): Change m_sig type to int32_t.
7508         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7509         int64_t for temporary calculations.
7510         (sreal_verify_basics): Drop one bit from minimum and maximum.
7512 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
7514         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
7516 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
7518         PR middle-end/86995
7519         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
7520         if to_add is negative.
7522 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7524         PR middle-end/87053
7525         * builtins.c (c_strlen): Improve range checks.
7527 2018-08-29  Martin Sebor  <msebor@redhat.com>
7528             Jeff Law  <law@redhat.com>
7530         PR tree-optimization/86714
7531         PR tree-optimization/86711
7532         * builtins.c (c_strlen): Add arguments to call to string_constant.
7533         * expr.c (string_constant): Add argument.  Detect missing nul
7534         terminator and outermost declaration it's missing in.
7535         * expr.h (string_constant): Add argument.
7536         * fold-const.c (read_from_constant_string): Add arguments to call to
7537         string_constant.
7538         (c_getstr): Likewise.
7539         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7540         to string_constant.
7541         * tree-ssa-strlen.c (get_stridx): Likewise.
7543 2018-08-29  Jan Hubicka  <jh@suse.cz>
7545         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
7546         Do not stream DECL_VINDEX.
7547         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
7548         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
7549         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
7551 2018-08-29  Richard Biener  <rguenther@suse.de>
7553         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
7554         virtual operands that are not default defs to honor region
7555         boundaries.
7556         (rpo_vn_valueize): Remove ineffective code here.
7558 2018-08-29  Richard Biener  <rguenther@suse.de>
7560         PR tree-optimization/87132
7561         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
7562         when skipping defs reachable over backedges.
7564 2018-08-29  Richard Biener  <rguenther@suse.de>
7566         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
7567         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
7568         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
7569         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
7570         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
7571         (vn_reference_lookup_3): Remove use of const_parms.
7572         (free_rpo_vn): Do not free const_parms.
7573         (do_rpo_vn): Do not call init_const_parms.
7574         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
7575         SSA_NAME_POINTS_TO_READONLY_MEMORY.
7576         (call_may_clobber_ref_p_1): Likewise.
7578 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
7580         PR other/86726
7581         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
7582         (-O): Ditto.
7583         (-ftree-scev-cprop): Document.
7585 2018-08-29  Jan Hubicka  <jh@suse.cz>
7587         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
7588         parameters.
7589         (sreal constructor): Update.
7590         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
7591         sreal:operator/): Update.
7593 2018-08-29  Martin Liska  <mliska@suse.cz>
7595         * tree-switch-conversion.c (switch_conversion::expand):
7596         Strenghten assumption about gswitch statements.
7598 2018-08-29  Richard Biener  <rguenther@suse.de>
7600         PR tree-optimization/87117
7601         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
7602         re-value-number released SSA VDEFs.
7604 2018-08-29  Richard Biener  <rguenther@suse.de>
7606         PR tree-optimization/87126
7607         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
7609 2018-08-28  Jim Wilson  <jimw@sifive.com>
7611         * config/riscv/pic.md: Rewrite.
7612         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
7613         invalid address.
7614         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
7615         (SOFTF, default_load, softload, softstore): New.
7617 2018-08-28  Jeff Law  <law@redhat.com>
7619         * fold-const.c (fold_binary_loc): Remove recently added assert.
7621 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
7623         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
7624         to OP parmeter of generated function.
7626 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
7628         PR tree-optimization/87009
7629         * match.pd: Add boolean optimizations.
7631 2018-08-28  Martin Sebor  <msebor@redhat.com>
7633         PR middle-end/86631
7634         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
7635         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
7636         (pass_walloca::gate): Use it.
7637         (alloca_call_type): Same.
7638         (pass_walloca::execute): Same.
7639         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
7641 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
7643         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
7644         GCC_VERSION for usage of "__gcc_dump_printf__" format from
7645         >= 3005 to >= 9000.
7647 2018-08-28  Richard Biener  <rguenther@suse.de>
7649         PR tree-optimization/87124
7650         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
7651         constants before looking up avail.
7653 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
7655         PR middle-end/87099
7656         * calls.c (maybe_warn_nonstring_arg): Punt early if
7657         warn_stringop_overflow is zero.  Don't call get_range_strlen
7658         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
7659         Swap comparison operands to have constants on rhs.  Only use
7660         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
7661         increment lenrng[0].
7663 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
7665         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
7666         use of tree_to_shwi.  Remove duplicated test for the size being
7667         a whole number of bytes.
7669 2018-08-28  Richard Biener  <rguenther@suse.de>
7671         PR tree-optimization/87117
7672         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
7673         Handle removed stmt without LHS (GIMPLE_NOP).
7675 2018-08-28  Richard Biener  <rguenther@suse.de>
7677         PR tree-optimization/87117
7678         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
7679         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
7681 2018-08-28  Richard Biener  <rguenther@suse.de>
7683         PR tree-optimization/87117
7684         * tree-ssa-pre.c (compute_avail): Do not make expressions
7685         with predicated values available.
7686         (get_expr_value_id): Assert we do not run into predicated value
7687         expressions.
7689 2018-08-28  Richard Biener  <rguenther@suse.de>
7691         PR tree-optimization/87117
7692         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
7693         get virtual operands.
7694         (get_expr_operands): Handle STRING_CST like other decls.
7696 2018-08-28  Martin Liska  <mliska@suse.cz>
7698         * tree.h: Update documentation of fndecl_built_in_p
7699         functions.
7702 2018-08-27  Jeff Law  <law@redhat.com>
7703         PR tree-optimization/87110
7704         * tree-ssa-dse.c (compute_trims): Handle non-constant
7705         TYPE_SIZE_UNIT.
7707 2018-08-27  Martin Sebor  <msebor@redhat.com>
7709         PR tree-optimization/86914
7710         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
7712 2018-08-27  Martin Sebor  <msebor@redhat.com>
7714         PR tree-optimization/87112
7715         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
7716         the type of the bound argument.
7718 2018-08-27  Jeff Law  <law@redhat.com>
7720         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
7721         type does not have a TYPE_SIZE_UNIT.
7723 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
7725         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
7726         with include of backend.h.
7728 2018-08-27  Richard Biener  <rguenther@suse.de>
7730         PR tree-optimization/86927
7731         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
7732         use const cond reduction code.
7734 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
7736         PR tree-optimization/85758
7737         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
7739 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7741         PR c++/87091
7742         * diagnostic-show-locus.c (class layout_range): Update for
7743         conversion of show_caret_p to a tri-state.
7744         (layout_range::layout_range): Likewise.
7745         (make_range): Likewise.
7746         (layout::maybe_add_location_range): Likewise.
7747         (layout::should_print_annotation_line_p): Don't show annotation
7748         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
7749         (layout::get_state_at_point): Update for conversion of
7750         show_caret_p to a tri-state.  Bail out early for
7751         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
7752         underlining or source colorization.
7753         (gcc_rich_location::add_location_if_nearby): Update for conversion
7754         of show_caret_p to a tri-state.
7755         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
7756         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
7757         Likewise.
7758         (selftest::test_one_liner_labels): Likewise.
7759         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
7760         conversion of show_caret_p to a tri-state.
7761         * pretty-print.c (text_info::set_location): Likewise.
7762         * pretty-print.h (text_info::set_location): Likewise.
7763         * substring-locations.c (format_warning_n_va): Likewise.
7764         * tree-diagnostic.c (default_tree_printer): Likewise.
7765         * tree-pretty-print.c (newline_and_indent): Likewise.
7767 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7769         PR c++/87091
7770         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
7771         line above for line-insertion fix-it hints.
7772         (selftest::test_fixit_insert_containing_newline): Update the
7773         expected results, and add a test with line-numbering enabled.
7775 2018-08-27  Martin Liska  <mliska@suse.cz>
7777         PR sanitizer/86962
7778         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
7779         params with DECL_HAS_VALUE_EXPR_P.
7781 2018-08-27  Martin Liska  <mliska@suse.cz>
7783         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
7784         selected expansion strategy.
7786 2018-08-27  Martin Liska  <mliska@suse.cz>
7788         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
7789         * builtins.c (is_builtin_fn): Likewise.
7790         * attribs.c (diag_attr_exclusions): Use new function
7791         fndecl_built_in_p and remove check for FUNCTION_DECL if
7792         possible.
7793         (builtin_mathfn_code): Likewise.
7794         (fold_builtin_expect): Likewise.
7795         (fold_call_expr): Likewise.
7796         (fold_builtin_call_array): Likewise.
7797         (fold_call_stmt): Likewise.
7798         (set_builtin_user_assembler_name): Likewise.
7799         (is_simple_builtin): Likewise.
7800         * calls.c (gimple_alloca_call_p): Likewise.
7801         (maybe_warn_nonstring_arg): Likewise.
7802         * cfgexpand.c (expand_call_stmt): Likewise.
7803         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
7804         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
7805         (cgraph_node::verify_node): Likewise.
7806         * cgraphclones.c (build_function_decl_skip_args): Likewise.
7807         (cgraph_node::create_clone): Likewise.
7808         * config/arm/arm.c (arm_insert_attributes): Likewise.
7809         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
7810         * dse.c (scan_insn): Likewise.
7811         * expr.c (expand_expr_real_1): Likewise.
7812         * fold-const.c (operand_equal_p): Likewise.
7813         (fold_binary_loc): Likewise.
7814         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
7815         * gimple-low.c (lower_stmt): Likewise.
7816         * gimple-pretty-print.c (dump_gimple_call): Likewise.
7817         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7818         Likewise.
7819         * gimple.c (gimple_build_call_from_tree): Likewise.
7820         (gimple_call_builtin_p): Likewise.
7821         (gimple_call_combined_fn): Likewise.
7822         * gimplify.c (gimplify_call_expr): Likewise.
7823         (gimple_boolify): Likewise.
7824         (gimplify_modify_expr): Likewise.
7825         (gimplify_addr_expr): Likewise.
7826         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
7827         * ipa-cp.c (determine_versionability): Likewise.
7828         * ipa-fnsummary.c (compute_fn_summary): Likewise.
7829         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
7830         * ipa-split.c (visit_bb): Likewise.
7831         (split_function): Likewise.
7832         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7833         * lto-cgraph.c (input_node): Likewise.
7834         * lto-streamer-out.c (write_symbol): Likewise.
7835         * omp-low.c (setjmp_or_longjmp_p): Likewise.
7836         (lower_omp_1): Likewise.
7837         * predict.c (strip_predict_hints): Likewise.
7838         * print-tree.c (print_node): Likewise.
7839         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
7840         * trans-mem.c (is_tm_irrevocable): Likewise.
7841         (is_tm_load): Likewise.
7842         (is_tm_simple_load): Likewise.
7843         (is_tm_store): Likewise.
7844         (is_tm_simple_store): Likewise.
7845         (is_tm_abort): Likewise.
7846         (tm_region_init_1): Likewise.
7847         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
7848         * tree-cfg.c (verify_gimple_call): Likewise.
7849         (move_stmt_r): Likewise.
7850         (stmt_can_terminate_bb_p): Likewise.
7851         * tree-eh.c (lower_eh_constructs_2): Likewise.
7852         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
7853         * tree-inline.c (remap_gimple_stmt): Likewise.
7854         (copy_bb): Likewise.
7855         (estimate_num_insns): Likewise.
7856         (fold_marked_statements): Likewise.
7857         * tree-sra.c (scan_function): Likewise.
7858         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
7859         (optimize_stack_restore): Likewise.
7860         (pass_fold_builtins::execute): Likewise.
7861         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
7862         (mark_all_reaching_defs_necessary_1): Likewise.
7863         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
7864         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7865         (pass_forwprop::execute): Likewise.
7866         * tree-ssa-loop-im.c (stmt_cost): Likewise.
7867         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7868         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
7869         * tree-ssa-strlen.c (get_string_length): Likewise.
7870         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
7871         (find_func_aliases_for_call): Likewise.
7872         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
7873         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
7874         * tree-tailcall.c (find_tail_calls): Likewise.
7875         * tree.c (need_assembler_name_p): Likewise.
7876         (free_lang_data_in_decl): Likewise.
7877         (get_call_combined_fn): Likewise.
7878         * ubsan.c (is_ubsan_builtin_p): Likewise.
7879         * varasm.c (incorporeal_function_p): Likewise.
7880         * tree.h (DECL_BUILT_IN): Remove and replace with
7881         fndecl_built_in_p.
7882         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
7883         (fndecl_built_in_p): New.
7885 2018-08-27  Martin Liska  <mliska@suse.cz>
7887         PR tree-optimization/86847
7888         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
7889         Dump also subtree probability.
7890         (switch_decision_tree::do_jump_if_equal): New function.
7891         (switch_decision_tree::emit_case_nodes): Handle special
7892         situations in balanced tree that can be emitted much simpler.
7893         Fix calculation of probabilities that happen in tree expansion.
7894         * tree-switch-conversion.h (struct cluster): Add
7895         is_single_value_p.
7896         (struct simple_cluster): Likewise.
7897         (struct case_tree_node): Add new function has_child.
7898         (do_jump_if_equal): New.
7900 2018-08-27  Martin Liska  <mliska@suse.cz>
7902         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
7903         Add new argument to bit_test_cluster constructor.
7904         (bit_test_cluster::emit): Set bits really number of values
7905         handlel by a test.
7906         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
7907         probability argument.
7908         * tree-switch-conversion.h (struct bit_test_cluster):
7909         Add m_handles_entire_switch.
7911 2018-08-27  Martin Liska  <mliska@suse.cz>
7913         PR tree-optimization/86702
7914         * tree-switch-conversion.c (jump_table_cluster::emit):
7915         Make probabilities even for values in jump table
7916         according to number of cases handled.
7917         (switch_decision_tree::compute_cases_per_edge): Pass
7918         argument to reset_out_edges_aux function.
7919         (switch_decision_tree::analyze_switch_statement): Likewise.
7920         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
7921         Make it static.
7923 2018-08-27  Martin Liska  <mliska@suse.cz>
7925         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
7926         cfun argument explicitly.
7927         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
7928         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
7929         function gimple_switch_default_bb.
7930         (convert_switch_statements):
7931         (expand_builtins):
7932         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
7933         * stmt.c (label_to_block_fn): Use label_to_block and pass
7934         cfun argument explicitly and use gimple_switch_label_bb.
7935         (expand_case): Likewise.
7936         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
7937         cfun argument explicitly. Likewise.
7938         (make_edges_bb): Likewise.
7939         (make_cond_expr_edges): Likewise.
7940         (get_cases_for_edge): Likewise.
7941         (make_gimple_switch_edges): Likewise.
7942         (label_to_block_fn): Likewise.
7943         (label_to_block): Likewise.
7944         (make_goto_expr_edges): Likewise.
7945         (make_gimple_asm_edges): Likewise.
7946         (main_block_label): Likewise.
7947         (group_case_labels_stmt): Likewise.
7948         (find_taken_edge_computed_goto): Likewise.
7949         (find_taken_edge_switch_expr): Likewise.
7950         (gimple_verify_flow_info): Likewise.
7951         (gimple_redirect_edge_and_branch): Likewise.
7952         (gimple_switch_label_bb): New function.
7953         (gimple_switch_default_bb): Likewise.
7954         (gimple_switch_edge): Likewise.
7955         (gimple_switch_default_edge): Likewise.
7956         * tree-cfg.h (label_to_block_fn): Remove and replace ...
7957         (label_to_block): ... with this.
7958         (gimple_switch_label_bb): New.
7959         (gimple_switch_default_bb): Likewise.
7960         (gimple_switch_edge): Likewise.
7961         (gimple_switch_default_edge): Likewise.
7962         * tree-cfgcleanup.c (convert_single_case_switch): Use
7963         new gimple functions and pass new argument to label_to_block.
7964         (cleanup_control_flow_bb):
7965         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
7966         cfun argument explicitly.
7967         (make_eh_edges): Likewise.
7968         (redirect_eh_dispatch_edge): Likewise.
7969         (lower_resx): Likewise.
7970         (lower_eh_dispatch): Likewise.
7971         (maybe_remove_unreachable_handlers): Likewise.
7972         (unsplit_eh): Likewise.
7973         (cleanup_empty_eh): Likewise.
7974         (verify_eh_edges): Likewise.
7975         (verify_eh_dispatch_edge): Likewise.
7976         * tree-ssa-dom.c (record_edge_info): Likewise.
7977         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
7978         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
7979         (thread_through_normal_block): Likewise.
7980         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
7981         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
7982         * tree-switch-conversion.c (switch_conversion::collect): Use new
7983         gimple functions.
7984         (switch_conversion::check_final_bb): Likewise.
7985         (switch_conversion::gather_default_values): Pass new argument
7986         to label_to_block.
7987         (switch_conversion::build_constructors): Likewise.
7988         (switch_decision_tree::compute_cases_per_edge): Use new
7989         gimple_switch_edge function.
7990         (switch_decision_tree::analyze_switch_statement): Pass new argument
7991         to label_to_block.
7992         (switch_decision_tree::try_switch_expansion): Use
7993         gimple_switch_default_edge.
7994         * tree-vrp.c (find_switch_asserts): Pass new argument
7995         to label_to_block.
7996         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
7997         (vr_values::simplify_switch_using_ranges): Likewise.
7999 2018-08-27  Richard Biener  <rguenther@suse.de>
8001         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8002         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8004         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8005         (struct vn_nary_op_s): Add unwind_to member.  Add
8006         predicated_values flag and put result into a union together
8007         with a linked list of vn_pval.
8008         (struct vn_ssa_aux): Add name member to make maintaining
8009         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8010         needed info, dfsnum, low, visited, on_sccstack, use_processed
8011         and range_info_anti_range_p members.
8012         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8013         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8014         New functions.
8015         (vn_valueize): New global.
8016         (vn_context_bb): Likewise.
8017         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8018         VN_INFO_PTR_INFO): Remove.
8019         * tree-ssa-sccvn.c: ... (rewrite)
8020         (pass_fre::execute): For -O2+ initialize loops and run
8021         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8022         run RPO VN in non-optimistic mode.
8023         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8024         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8025         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8026         (rpo-vn-max-loop-depth): Document.
8027         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8028         when valuezing the VUSE signals we walked out of the region.
8029         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8030         (phi_translate): Set VN context block to use for availability
8031         lookup.
8032         (compute_avail): Likewise.
8033         (pre_valueize): New function.
8034         (pass_pre::execute): Adjust to the RPO VN API.
8036         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8037         (propagate_constants_for_unrolling): Remove.
8038         (tree_unroll_loops_completely): Perform value-numbering
8039         on the unrolled bodies loop parent.
8041 2018-08-27  Richard Biener  <rguenther@suse.de>
8043         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8044         for partial antic compute.
8046 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8048         PR rtl-optimization/87065
8049         * combine.c (simplify_if_then_else): Formatting fix.
8050         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8051         check.
8052         (known_cond): Don't return const_true_rtx for vector modes.  Use
8053         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8055 2018-08-27  Martin Liska  <mliska@suse.cz>
8057         PR gcov-profile/87069
8058         * gcov.c (process_file): Record files already processed
8059         and warn about a file being processed multiple times.
8061 2018-08-27  Martin Liska  <mliska@suse.cz>
8063         PR driver/83193
8064         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8065         Set default values for x_aarch64_*_string strings.
8066         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8067         prefix.  For -mabi do not print '=ABI' in help and use
8068         <option_value> format for -msve-vector-bits and -moverride
8069         options.
8071 2018-08-26  Jeff Law  <law@redhat.com>
8073         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8074         "cfg.h"
8076 2018-08-26  Marek Polacek  <polacek@redhat.com>
8078         PR c++/87029, Implement -Wredundant-move.
8079         * doc/invoke.texi: Document -Wredundant-move.
8081 2018-08-25  Martin Sebor  <msebor@redhat.com>
8083         PR tree-optimization/87059
8084         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8085         to the same type as the other.
8086         * fold-const.c (fold_binary_loc): Assert expectation.
8088 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8090         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8091         extraneous parentheses, dead code section and formatting.
8093 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8095         PR c++/87091
8096         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8097         wide enough for jumps in the line-numbering to be visible.
8098         (layout::print_gap_in_line_numbering): New member function.
8099         (layout::calculate_line_spans): When using line numbering, merge
8100         line spans that are only 1 line apart.
8101         (diagnostic_show_locus): When printing line numbers, show gaps in
8102         line numbering directly, rather than printing headers.
8103         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8104         line-numbering with multiple line spans.
8105         (selftest::test_fixit_insert_containing_newline_2): Add test of
8106         line-numbering, in which the spans are close enough to be merged.
8108 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8110         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8111         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8112         * tree-vrp.c (range_is_nonnull): Remove.
8113         (range_includes_zero_p): Accept value_range instead of min/max.
8114         (extract_range_from_binary_expr_1): Do not early bail on
8115         POINTER_PLUS_EXPR.
8116         Use range_includes_zero_p instead of range_is_nonnull.
8117         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8118         of range_is_nonnull.
8119         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8120         special case VR_ANTI_RANGE.
8121         (vrp_finalize): Same.
8122         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8123         instead of min/max.
8124         (range_is_nonnull): Remove.
8125         * vr-values.c (vrp_stmt_computes_nonzero): Use
8126         range_includes_zero_p instead of range_is_nonnull.
8127         (extract_range_basic): Pass value_range to range_includes_zero_p
8128         instead of range_is_nonnull.
8130 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8132         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8133         * emit-rtl.h (rtl_data): Remove return_bnd.
8134         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8135         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8136         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8137         (POINTER_BOUNDS_MODE): Remove definition.
8138         (make_pointer_bounds_mode): Remove.
8139         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8140         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8141         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8142         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8143         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8144         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8145         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8147         * config/i386/i386-modes.def (BND32, BND64): Remove.
8148         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8149         (dbx64_register_map): Ditto.
8150         (svr4_dbx_register_map): Ditto.
8151         (indirect_thunk_bnd_needed): Remove.
8152         (indirect_thunks_bnd_used): Ditto.
8153         (indirect_return_bnd_needed): Ditto.
8154         (indirect_return_via_cx_bnd): Ditto.
8155         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8156         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8157         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8158         (output_indirect_thunk_function): Remove handling of
8159         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8160         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8161         (ix86_save_reg): Remove handling of crtl->return_bnd.
8162         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8163         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8164         and UNSPEC_BNDLX_ADDR.
8165         (ix86_output_indirect_branch_via_reg): Remove handling of
8166         indirect_thunk_prefix_bnd.
8167         (ix86_output_indirect_branch_via_push): Ditto.
8168         (ix86_output_function_return): Ditto.
8169         (ix86_output_indirect_function_return): Ditto.
8170         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8171         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8172         (CALL_USED_REGISTERS): Ditto.
8173         (REG_ALLOC_ORDER): Update for removal of bound registers.
8174         (HI_REGISTER_NAMES): Ditto.
8175         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8176         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8177         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8178         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8179         (FIRST_PSEUDO_REG): Update.
8180         (BND): Remove mode iterator.
8181         * config/i386/predicates.md (bnd_mem_operator): Remove.
8183 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8185         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8186         vectors.
8188 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8190         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8191         the case in which the permute needs only a single element and
8192         repeats for every vector of the result.  Extend that case to
8193         handle variable-length vectors.
8194         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8196 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8198         PR debug/79342
8199         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8200         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8202 2018-08-24  Richard Biener  <rguenther@suse.de>
8204         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8205         bb_flags_allocated members.
8206         (auto_flag): New RAII class for allocating flags.
8207         (auto_edge_flag): New RAII class for allocating edge flags.
8208         (auto_bb_flag): New RAII class for allocating bb flags.
8209         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8210         flag dynamically.
8211         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8212         in favor of temporarily allocated BB flag.
8213         * hsa-brig.c: Re-order includes.
8214         * hsa-dump.c: Likewise.
8215         * hsa-regalloc.c: Likewise.
8216         * print-rtl.c: Likewise.
8217         * profile-count.c: Likewise.
8219 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8221         PR target/86989
8222         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8223         the TOC register.
8225 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8227         PR 87073/bootstrap
8228         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8229         from wide_int_range_multiplicative_op.
8231 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8233         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8234         "permutaion".
8236 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8238         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8239         to 'expanded'.
8241 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8243         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8244         full GENERIC expression used for replacement.
8246 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8248         * tree-vrp.c (abs_extent_range): Remove.
8249         (extract_range_into_wide_ints): Pass wide ints by reference.
8250         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8251         Pass wide ints by reference in all calls to
8252         extract_range_into_wide_ints.
8253         * wide-int-range.cc (wide_int_range_div): New.
8254         * wide-int-range.h (wide_int_range_div): New.
8255         (wide_int_range_includes_zero_p): New.
8256         (wide_int_range_zero_p): New.
8258 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8260         * config/aarch64/aarch64.md (arches): New enum.
8261         (arch): New enum attr.
8262         (arch_enabled): New attr.
8263         (enabled): Now uses arch_enabled only.
8264         (simd, sve, fp16): Removed attribute.
8265         (fp): Attr now defined in terms of 'arch'.
8266         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8267         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8268         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8269         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8270         attributes into 'arch'.
8271         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8272         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8273         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8274         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8275         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8276         'simd' attribute into 'arch'.
8277         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8278         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8279         Convert use of 'fp' attribute to 'arch'.
8280         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8281         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8282         into 'arch'.
8283         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8284         (different modes) Merge 'fp' and 'simd' into 'arch'.
8285         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8286         'simd' into 'arch'.
8288 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8290         PR rtl-optimization/87026
8291         * expmed.c (canonicalize_comparison): If we can no longer create
8292         pseudoregisters, don't.
8294 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8296         PR target/86951
8297         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8298         prototype.
8299         * config/arm/arm.c (speculation_barrier_libfunc): New static
8300         variable.
8301         (arm_init_libfuncs): Initialize it.
8302         (arm_emit_speculation_barrier): New function.
8303         * config/arm/arm.md (speculation_barrier): Call
8304         arm_emit_speculation_barrier for architectures that do not have
8305         DSB or ISB.
8306         (speculation_barrier_insn): Only match on Armv7 or later.
8308 2018-08-23  Richard Biener  <rguenther@suse.de>
8310         PR middle-end/87024
8311         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8312         calls.
8314 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8316         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8317         of single-vector TBLs.
8318         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8319         one input is given.
8321 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8323         PR target/85910
8324         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8325         aarch64_evpc_tbl guard.
8327 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8329         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8330         behaviour.
8332 2018-08-22  Martin Sebor  <msebor@redhat.com>
8334         PR middle-end/87052
8335         * tree-pretty-print.c (pretty_print_string): Add argument.
8336         (dump_generic_node): Call to pretty_print_string with string size.
8338 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8340         PR rtl-optimization/86771
8341         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8342         of two SETs into those two SETs, one to be placed at i2, if that SETs
8343         destination is modified between i2 and i3.
8345 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8347         PR tree-optimization/86725
8348         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8349         function.
8350         (vect_analyze_scalar_cycles_1): Check it.
8352 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8354         PR tree-optimization/86725
8355         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8356         an outer loop phi as a double reduction, make sure that the
8357         single user of the phi result is an inner loop phi.
8359 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8361         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8362         grouped stores with gaps to a strided group.
8364 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8366         * tree-vect-stmts.c (get_group_load_store_type)
8367         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8368         first statement in a group.
8370 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8372         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8373         the sequence used in gcc/gcc.c.
8375 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8377         PR other/704
8378         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8379         building it.
8381 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8383         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8384         Darwin10-specific unwinder-shim.
8385         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8386         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8387         New to cater for Darwin10 Rosetta.
8389 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8391         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8392         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8394 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8396         PR bootstrap/81033
8397         PR target/81733
8398         PR target/52795
8399         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8400         (dwarf2out_switch_text_section): Generate a local label for the
8401         second function sub-section and apply it as the second FDE start
8402         label.
8403         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8404         second sub-section start.
8406 2018-08-22  Richard Biener  <rguenther@suse.de>
8408         PR tree-optimization/86988
8409         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8411 2018-08-22  Richard Biener  <rguenther@suse.de>
8413         PR tree-optimization/86945
8414         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8416 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8418         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8419         a comment about how uses of r2 for .sdata2 come about.
8421 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8423         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8425 2018-08-21  Marek Polacek  <polacek@redhat.com>
8427         PR c++/86981, Implement -Wpessimizing-move.
8428         * doc/invoke.texi: Document -Wpessimizing-move.
8430 2018-08-21  Jan Hubicka  <jh@suse.cz>
8432         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8433         * tree.h (is_redundant_typedef): Remove.
8434         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8436 2018-08-21  Jan Hubicka  <jh@suse.cz>
8438         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8439         when possible.
8441 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8443         * expmed.c (extract_low_bits): Reject invalid subregs early.
8445 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8447         PR middle-end/86121
8448         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8449         behaviour.
8451 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8453         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8454         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8455         * config/vxworks.c: Likewise.
8457 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8459         * config/vxworks.c: Set targetm.have_ctors_dtors
8460         if HAVE_INITFINI_ARRAY_SUPPORT.
8461         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8462         if HAVE_INITFINI_ARRAY_SUPPORT.
8464 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8466         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8467         default search path for VxWorks < 7.
8469 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8471         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8472         (get_format_string): Refer to c_getstr.
8474 2018-08-21  Tom de Vries  <tdevries@suse.de>
8476         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8477         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8478         (debuginfo_early_stop): Declare.
8479         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8480         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8481         (debuginfo_early_stop): New function.
8482         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8483         and debuginfo_early_stop.
8484         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8485         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8486         (general_init): Call debuginfo_early_init.
8487         (finalize): Call debuginfo_fini.
8488         (do_compile): Call debuginfo_init.
8489         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8490         -fdump-early-debug.
8491         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8493 2018-08-21  Tom de Vries  <tdevries@suse.de>
8495         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8496         flag_dump_noaddr and flag_dump_unnumbered.
8498 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
8500         * wide-int-range.cc (wide_int_range_abs): New.
8501         (wide_int_range_order_set): Rename from wide_int_range_min_max.
8502         * wide-int-range.h (wide_int_range_abs): New.
8503         (wide_int_range_min_max): New.
8504         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
8505         case to call wide_int_range_abs.
8506         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
8507         (extract_range_from_abs_expr): Delete.
8509 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
8511         PR target/87033
8512         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
8513         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
8514         for indexed loads.
8516 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8517             Jeff Law <law@redhat.com>
8519         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
8520         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
8522 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
8524         PR other/84889
8525         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
8526         (decl_attributes): Likewise.
8527         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
8528         instance.
8529         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8530         * diagnostic-core.h (class auto_diagnostic_group): New class.
8531         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
8532         (diagnostic_report_diagnostic): Handle the first diagnostics within
8533         a group.
8534         (emit_diagnostic): Add auto_diagnostic_group instance.
8535         (inform): Likewise.
8536         (inform_n): Likewise.
8537         (warning): Likewise.
8538         (warning_at): Likewise.
8539         (warning_n): Likewise.
8540         (pedwarn): Likewise.
8541         (permerror): Likewise.
8542         (error): Likewise.
8543         (error_n): Likewise.
8544         (error_at): Likewise.
8545         (sorry): Likewise.
8546         (fatal_error): Likewise.
8547         (internal_error): Likewise.
8548         (internal_error_no_backtrace): Likewise.
8549         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
8550         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
8551         * diagnostic.h (struct diagnostic_context): Add fields
8552         "diagnostic_group_nesting_depth",
8553         "diagnostic_group_emission_count", "begin_group_cb",
8554         "end_group_cb".
8555         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
8556         Add auto_diagnostic_group instance(s).
8557         (find_explicit_erroneous_behavior): Likewise.
8558         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
8559         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
8560         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
8561         (gimplify_va_arg_expr): Likewise.
8562         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
8563         (HSA_SORRY_AT): Likewise.
8564         * ipa-devirt.c (compare_virtual_tables): Likewise.
8565         (warn_odr): Likewise.
8566         * multiple_target.c (expand_target_clones): Likewise.
8567         * opts-common.c (cmdline_handle_error): Likewise.
8568         * reginfo.c (globalize_reg): Likewise.
8569         * substring-locations.c (format_warning_n_va): Likewise.
8570         * tree-inline.c (expand_call_inline): Likewise.
8571         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
8572         * tree-ssa-loop-niter.c
8573         (do_warn_aggressive_loop_optimizations): Likewise.
8574         * tree-ssa-uninit.c (warn_uninit): Likewise.
8575         * tree.c (warn_deprecated_use): Likewise.
8577 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
8579         PR target/87014
8580         * config/i386/i386.md (eh_return): Always update EH return
8581         address in word_mode.
8583 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
8585         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
8586         TARGET_SPLIT_COMPLEX_ARG is defined.
8588 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8590         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
8592 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8594         PR target/86984
8595         * expr.c (expand_assignment): Assert that bitpos is positive.
8596         (store_field): Likewise
8597         (expand_expr_real_1): Make sure that bitpos is positive.
8598         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
8599         integer overflow.
8601 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8603         * Makefile.in (CPP_ID_DATA_H): Delete.
8604         (CPP_INTERNAL_H): Don't add it.
8605         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
8606         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
8608 2018-08-20  Richard Biener  <rguenther@suse.de>
8610         PR tree-optimization/78655
8611         * tree-vrp.c (extract_range_from_binary_expr_1): Make
8612         pointer + offset nonnull if either operand is nonnull work.
8614 2018-08-20  Tom de Vries  <tdevries@suse.de>
8616         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
8617         unless the referenced die describes the added property using
8618         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
8619         Otherwise, add a DW_AT_location to the referenced die.
8621 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
8623         PR target/86994
8624         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
8625         register_operand when calling ix86_set_reg_reg_cost.
8626         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
8627         Set *total to 0 for operands that satisfy x86_64_immediate_operand
8628         predicate and to 1 otherwise.
8630 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
8632         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
8633         emit a diagnostic that it is not supported and reset the option.
8634         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
8635         supported and consume the option.  (ASM_FINAL_SPEC): New.
8637 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8639         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
8640         a sentence.
8642 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
8644         C-SKY port: Documentation
8646         * doc/extend.texi (C-SKY Function Attributes): New section.
8647         * doc/invoke.texi (Option Summary): Add C-SKY options.
8648         (C-SKY Options): New section.
8649         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
8651 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8652             Huibin Wang  <huibin_wang@c-sky.com>
8653             Sandra Loosemore  <sandra@codesourcery.com>
8654             Chung-Lin Tang  <cltang@codesourcery.com>
8656         C-SKY port: Backend implementation
8658         * config/csky/*: New.
8659         * common/config/csky/*: New.
8661 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8662             Huibin Wang  <huibin_wang@c-sky.com>
8663             Sandra Loosemore  <sandra@codesourcery.com>
8664             Chung-Lin Tang  <cltang@codesourcery.com>
8665             Andrew Jenner  <andrew@codesourcery.com>
8667         C-SKY port: Configury
8669         * config.gcc (csky-*-*): New.
8670         * configure.ac: Add csky to targets for dwarf2 debug_line support.
8671         * configure: Regenerated.
8673 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
8675         * dump-context.h: Include "dumpfile.h".
8676         (dump_context::dump_printf_va): Convert final param from va_list
8677         to va_list *.  Convert from ATTRIBUTE_PRINTF to
8678         ATTRIBUTE_GCC_DUMP_PRINTF.
8679         (dump_context::dump_printf_loc_va): Likewise.
8680         * dumpfile.c: Include "stringpool.h".
8681         (make_item_for_dump_printf_va): Delete.
8682         (make_item_for_dump_printf): Delete.
8683         (class dump_pretty_printer): New class.
8684         (dump_pretty_printer::dump_pretty_printer): New ctor.
8685         (dump_pretty_printer::emit_items): New member function.
8686         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
8687         function.
8688         (dump_pretty_printer::emit_item): New member function.
8689         (dump_pretty_printer::stash_item): New member function.
8690         (dump_pretty_printer::format_decoder_cb): New member function.
8691         (dump_pretty_printer::decode_format): New member function.
8692         (dump_context::dump_printf_va): Reimplement in terms of
8693         dump_pretty_printer.
8694         (dump_context::dump_printf_loc_va): Convert final param from va_list
8695         to va_list *.
8696         (dump_context::begin_scope): Reimplement call to
8697         make_item_for_dump_printf.
8698         (dump_printf): Update for change to dump_printf_va.
8699         (dump_printf_loc): Likewise.
8700         (selftest::test_capture_of_dump_calls): Convert "stmt" from
8701         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
8702         with %T, %E, and %G.
8703         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
8704         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
8705         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
8706         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
8707         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
8708         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
8709         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
8710         within a dump_printf_loc call to "%wu".
8711         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
8712         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
8713         missing space after "=".
8714         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
8715         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
8716         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
8717         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
8718         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
8719         duplicate "vectorized" from message.
8721 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
8723         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
8724         polyNxK_t element's TYPE_STRING_FLAG.
8726 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8728         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
8729         (they were unnamed before).  Fix comments.
8731 2018-08-17  Nathan Sidwell  <nathan@acm.org>
8733         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
8735 2018-08-17  Richard Biener  <rguenther@suse.de>
8737         PR tree-optimization/86841
8738         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
8740 2018-08-17  Martin Liska  <mliska@suse.cz>
8742         * common.opt: Remove Warn, Init and Report for options with
8743         Ignore/Deprecated flag. Warning is done automatically for
8744         Deprecated flags.
8745         * config/i386/i386.opt: Likewise.
8746         * config/ia64/ia64.opt: Likewise.
8747         * config/rs6000/rs6000.opt: Likewise.
8748         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8749         Remove usage of flag_check_pointer_bounds.
8750         * lto-wrapper.c (merge_and_complain): Do not handle
8751         OPT_fcheck_pointer_bounds.
8752         (append_compiler_options): Likewise.
8753         * opt-functions.awk: Do not handle Deprecated.
8754         * optc-gen.awk: Check that Var, Report and Init are not
8755         used for an option with Ignore/Deprecated flag.
8756         * opts-common.c (decode_cmdline_option): Do not report
8757         CL_ERR_DEPRECATED.
8758         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
8759         options.
8760         * opts.h (struct cl_option): Remove cl_deprecated flag.
8761         (CL_ERR_DEPRECATED): Remove error enum value.
8763 2018-08-17  Richard Biener  <rguenther@suse.de>
8765         PR middle-end/86505
8766         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
8767         across a va-arg-pack using call adjust its return value accordingly.
8769 2018-08-16  Martin Sebor  <msebor@redhat.com>
8771         PR tree-optimization/86853
8772         * gimple-ssa-sprintf.c (struct format_result): Rename member.
8773         (struct fmtresult): Add member and initialize it in ctors.
8774         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
8775         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
8776         (format_directive): Set POSUNDER4K when MAYFAIL is set.
8777         (parse_directive): Handle %C same as %c.
8778         (sprintf_dom_walker::compute_format_length): Adjust.
8779         (is_call_safe): Adjust.
8781 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8783         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
8784         for determining how to count the elements.
8785         * builtins.h (c_strlen): Adjust prototype.
8786         * expr.c (string_constant): Add new parameter mem_size.
8787         Set *mem_size appropriately.
8788         * expr.h (string_constant): Adjust protoype.
8789         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
8790         * gimple-fold.h (get_range_strlen): Adjust prototype.
8791         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
8792         (format_string): Call get_string_length with eltsize.
8794 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8796         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
8797         to emit the span, rather than setting it as the prefix.
8799 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8801         * diagnostic-show-locus.c (layout::start_annotation_line): Add
8802         "margin_char" parameter, defaulting to space.  Use it in place
8803         of pp_space for the initial part of the margin.
8804         (layout::print_leading_fixits): Use '+' when filling the margin
8805         of line-insertion fix-it hints.
8807 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8809         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
8810         Delete.
8812 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8814         * config/rs6000/altivec.md: Don't set length attribute to the default
8815         value.
8816         * config/rs6000/darwin.md: Ditto.
8817         * config/rs6000/dfp.md: Ditto.
8818         * config/rs6000/htm.md: Ditto.
8819         * config/rs6000/rs6000.md: Ditto.
8820         * config/rs6000/sync.md: Ditto.
8821         * config/rs6000/vsx.md: Ditto.
8823 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8825         * config/rs6000/altivec.md: Don't set length attribute to the default
8826         value, for branch instructions.
8827         * config/rs6000/darwin.md: Ditto.
8828         * config/rs6000/rs6000.md: Ditto.
8830 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8832         * config/rs6000/rs6000.md (length): Always define as const_int 4.
8833         (unnamed conditional branch define_insn): Set length to 4 or 8
8834         depending on offset.
8835         (<bd>_<mode>): Similar, for alternative 0.
8836         (<bd>tf_<mode>): Ditto.
8838 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8840         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
8842 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
8844         * doc/rtl.texi: Replace old RTX class names with new names.
8847 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
8849         * expmed.h (canonicalize_comparison): New declaration.
8850         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
8851         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
8852         * optabs.c (prepare_cmp_insn): Likewise.
8853         * rtl.h (unsigned_condition_p): New function which checks if a
8854         comparison operator is unsigned.
8856 2018-08-16  Nathan Sidwell  <nathan@acm.org>
8858         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
8859         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
8861 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8863         PR target/84711
8864         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
8865         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
8866         (mov<mov>): ..this and enable unconditionally.
8868 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8870         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
8872 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
8874         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
8875         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
8876         "Common" with "Target".
8878 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8880         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
8881         * doc/invoke.texi (mmitigate-rop): Remove.
8882         * config/i386/i386.c: Do not include "regrename.h".
8883         (ix86_rop_should_change_byte_p, reg_encoded_number)
8884         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
8885         Remove.
8886         (ix86_reorg): Remove call to ix86_mitigate_rop.
8887         * config/i386/i386.md (attr "modrm_class"): Remove.
8888         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
8889         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
8890         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
8892 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
8894         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
8895         allow folding of mergeh() and mergel() for the float and double types.
8896         (fold_mergehl_helper): Rework to handle building a permute tree
8897         for float vectors.
8899 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8901         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
8902         for TARGET_SSE.
8904 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
8906         * common.opt (fdiagnostics-show-labels): New option.
8907         * diagnostic-show-locus.c (class layout_range): Add field
8908         "m_label".
8909         (class layout): Add field "m_show_labels_p".
8910         (layout_range::layout_range): Add param "label" and use it to
8911         initialize m_label.
8912         (make_range): Pass in NULL for new "label" param of layout_range's
8913         ctor.
8914         (layout::layout): Initialize m_show_labels_p.
8915         (layout::maybe_add_location_range): Pass in loc_range->m_label
8916         when constructing layout_range instances.
8917         (struct line_label): New struct.
8918         (layout::print_any_labels): New member function.
8919         (layout::print_line): Call it if label-printing is enabled.
8920         (selftest::test_one_liner_labels): New test.
8921         (selftest::test_diagnostic_show_locus_one_liner): Call it.
8922         * diagnostic.c (diagnostic_initialize): Initialize
8923         context->show_labels_p.
8924         * diagnostic.h (struct diagnostic_context): Add field
8925         "show_labels_p".
8926         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8927         -fno-diagnostics-show-labels.
8928         * dwarf2out.c (gen_producer_string): Add
8929         OPT_fdiagnostics_show_labels to the ignored options.
8930         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
8931         param.
8932         (gcc_rich_location::maybe_add_expr): Likewise.
8933         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
8934         label" param, defaulting to NULL.
8935         (gcc_rich_location::add_expr): Add "label" param.
8936         (gcc_rich_location::maybe_add_expr): Likewise.
8937         (class text_range_label): New class.
8938         (class range_label_for_type_mismatch): New class.
8939         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
8940         of format_warning_va.
8941         (fmtwarn_n): Likewise for new params of format_warning_n_va.
8942         * lto-wrapper.c (merge_and_complain): Add
8943         OPT_fdiagnostics_show_labels to the "pick one setting" options.
8944         (append_compiler_options): Likewise to the dropped options.
8945         (append_diag_options): Likewise to the passed-on options.
8946         * opts.c (common_handle_option): Handle the new option.
8947         * selftest-diagnostic.c
8948         (test_diagnostic_context::test_diagnostic_context): Enable
8949         show_labels_p.
8950         * substring-locations.c: Include "gcc-rich-location.h".
8951         (format_warning_n_va): Add "fmt_label" and "param_label" params
8952         and use them as appropriate.
8953         (format_warning_va): Add "fmt_label" and "param_label" params,
8954         passing them on to format_warning_n_va.
8955         (format_warning_at_substring): Likewise.
8956         (format_warning_at_substring_n): Likewise.
8957         * substring-locations.h (format_warning_va): Add "fmt_label" and
8958         "param_label" params.
8959         (format_warning_n_va): Likewise.
8960         (format_warning_at_substring): Likewise.
8961         (format_warning_at_substring_n): Likewise.
8962         * toplev.c (general_init): Initialize global_dc->show_labels_p.
8964 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
8966         PR testsuite/86519
8967         * builtins.c (expand_builtin_memcmp): Do not expand the call
8968         when overflow is detected.
8970 2018-08-15  Martin Sebor  <msebor@redhat.com>
8972         PR tree-optimization/71625
8973         * config/aarch64/aarch64-builtins.c
8974         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
8976 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
8978         * config/s390/s390.c (s390_reorg): Remove loop.
8980 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8982         * config/darwin.c
8983          (darwin_function_switched_text_sections): Delete.
8984         * gcc/config/darwin.h
8985          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
8987 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8989         PR target/81685
8990         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
8991         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
8992         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
8994 2018-08-15  Martin Liska  <mliska@suse.cz>
8996         PR tree-optimization/86925
8997         * predict.c (expr_expected_value_1): When taking
8998         later predictor, assign also probability.
8999         Use fold_build2_initializer_loc in order to fold
9000         the expression in -frounding-math.
9002 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9004         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9005         patterns.
9006         (expand_vec_perm_1): Try the new method.
9008 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9010         PR target/86547
9011         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9012         Check whether lra_live_max_point is 0 before dividing.
9014 2018-08-14  Martin Sebor  <msebor@redhat.com>
9016         PR tree-optimization/86650
9017         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9018         (vrp_prop::check_mem_ref): Same.
9020 2018-08-13  Liu Hao <lh_mouse@126.com>
9022         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9023         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9024         and clear it thereafter, as it only works for DBCS.
9026 2018-08-13  Liu Hao <lh_mouse@126.com>
9028         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9029         handle returned by _get_osf_handle().
9031 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9033         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9034         for folding vec_perm.
9036 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9038         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9039         Add support for gimple-folding of vec_pack() and vec_unpack()
9040         intrinsics.
9042 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9044         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9045         vec_xst variants to the list.
9046         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9047         vector loads and stores.
9049 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9051         * config.gcc (rs6000-ibm-aix4.x): Delete.
9052         (rs6000-ibm-aix5.1): Delete.
9053         (rs6000-ibm-aix5.2): Delete.
9054         (rs6000-ibm-aix5.3): Delete.
9055         * config/rs6000/aix43.h: Delete.
9056         * config/rs6000/aix51.h: Delete.
9057         * config/rs6000/aix52.h: Delete.
9058         * config/rs6000/t-aix43: Delete.
9060 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9062         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9063         New function.
9064         (s390_decompose_address): Factor out constant pool ref
9065         decomposition.
9067 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9069         * config/nds32/nds32-predicates.c
9070         (nds32_can_use_bclr_p): Change return type as bool.
9071         (nds32_can_use_bset_p): Ditto.
9072         (nds32_can_use_btgl_p): Ditto.
9073         (nds32_can_use_bitci_p): Ditto.
9074         * config/nds32/nds32-protos.h
9075         (nds32_can_use_bclr_p): Change declaration.
9076         (nds32_can_use_bset_p): Ditto.
9077         (nds32_can_use_btgl_p): Ditto.
9078         (nds32_can_use_bitci_p): Ditto.
9080 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9082         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9083         Support -msched-prolog-epilog option.
9084         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9086 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9088         * common/config/nds32/nds32-common.c
9089         (nds32_option_optimization_table): Enalbe -malways-align.
9091 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9093         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9094         extra_headers.
9095         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9096         OPT_misr_secure_ case.
9097         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9098         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9099         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9100         secure attribute.
9101         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9102         (nds32_isr_info): New field security_level.
9103         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9104         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9105         * config/nds32/nds32.opt (misr-secure): New option.
9106         * config/nds32/nds32_init.inc: New file.
9107         * config/nds32/nds32_isr.h: New file.
9109 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9111         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9112         Update comment for atomic instructions.
9113         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9114         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9115         Remove.
9116         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9117         load for __ATOMIC_SEQ_CST model.
9118         (atomic_loaddi_1): Remove float register target.
9119         (atomic_storedi): Handle CONST_INT values.
9120         (atomic_storedi_1): Remove float register source.  Add special case
9121         for zero value.
9122         (memory_barrier): New expander and insn.
9124 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9126         PR tree-optimization/86835
9127         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9128         new_stmt after def_gsi, make sure to insert new_square_stmt after
9129         that stmt, not 2 stmts before it.
9131 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9133         PR target/82418
9134         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9135         instead of SWI48.
9137 2018-08-10  Martin Liska  <mliska@suse.cz>
9139         PR target/83610
9140         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9141         function type.
9142         * builtins.c (expand_builtin_expect_with_probability):
9143         New function.
9144         (expand_builtin_expect_with_probability): New function.
9145         (build_builtin_expect_predicate): Add new argumnet probability
9146         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9147         (fold_builtin_expect):
9148         (fold_builtin_2):
9149         (fold_builtin_3):
9150         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9151         * builtins.h (fold_builtin_expect): Set new argument.
9152         * doc/extend.texi: Document __builtin_expect_with_probability.
9153         * doc/invoke.texi: Likewise.
9154         * gimple-fold.c (gimple_fold_call): Pass new argument.
9155         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9156         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9157         * predict.c (get_predictor_value): New function.
9158         (expr_expected_value): Add new argument probability. Assume
9159         that predictor and probability are always non-null.
9160         (expr_expected_value_1): Likewise.  For __builtin_expect and
9161         __builtin_expect_with_probability set probability.  Handle
9162         combination in binary expressions.
9163         (tree_predict_by_opcode): Simplify code by simply calling
9164         get_predictor_value.
9165         (pass_strip_predict_hints::execute): Add handling of
9166         BUILT_IN_EXPECT_WITH_PROBABILITY.
9167         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9168         new predictor.
9169         * tree.h (DECL_BUILT_IN_P): New function.
9171 2018-08-10  Martin Liska  <mliska@suse.cz>
9173         PR tree-optimization/85799
9174         * passes.def: Add argument for pass_strip_predict_hints.
9175         * predict.c (class pass_strip_predict_hints): Add new argument
9176         early_p.
9177         (strip_predictor_early): New function.
9178         (pass_strip_predict_hints::execute): Call the function to
9179         strip predictors.
9180         (strip_predict_hints): New function.
9181         * predict.def: Fix comment.
9183 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9185         * Makefile.in: Clarify which tm.texi to copy over to assert the
9186         right to grant a GFDL license for all.
9188 2018-08-09  Jeff Law  <law@redhat.com>
9190         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9191         unused variable.
9193 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9195         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9196         prototype.
9198 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9200         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9201         reductions for variable-length vectors.
9203 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9205         PR other/84889
9206         * common.opt (fdiagnostics-show-line-numbers): New option.
9207         * diagnostic-show-locus.c (class layout): Add fields
9208         "m_show_line_numbers_p" and "m_linenum_width";
9209         (num_digits): New function.
9210         (test_num_digits): New function.
9211         (layout::layout): Initialize new fields.  Update m_x_offset
9212         logic to handle any left margin.
9213         (layout::print_source_line): Print line number when requested.
9214         (layout::start_annotation_line): New member function.
9215         (layout::print_annotation_line): Call it.
9216         (layout::print_leading_fixits): Likewise.
9217         (layout::print_trailing_fixits): Likewise.  Update calls to
9218         move_to_column for new parameter.
9219         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9220         it to potentially call start_annotation_line.
9221         (layout::show_ruler): Call start_annotation_line.
9222         (selftest::test_line_numbers_multiline_range): New selftest.
9223         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9224         and selftest::test_line_numbers_multiline_range.
9225         * diagnostic.c (diagnostic_initialize): Initialize
9226         show_line_numbers_p.
9227         * diagnostic.h (struct diagnostic_context): Add field
9228         "show_line_numbers_p".
9229         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9230         -fno-diagnostics-show-line-numbers.
9231         * dwarf2out.c (gen_producer_string): Add
9232         OPT_fdiagnostics_show_line_numbers to the ignored options.
9233         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9234         one setting" options.
9235         (append_compiler_options): Likewise to the dropped options.
9236         (append_diag_options): Likewise to the passed-on options.
9237         * opts.c (common_handle_option): Handle the new option.
9238         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9240 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9242         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9243         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9244         third argument of type "const signed char" to descriptions of
9245         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9246         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9247         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9248         __builtin_bcdsub_ov functions.
9250 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9252         PR tree-optimization/86858
9253         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9254         flow_bb_inside_loop_p calls.
9256 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9258         PR tree-optimization/86871
9259         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9260         instead of gimple_assign_lhs.
9262 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9264         PR target/86887
9265         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9266         register constraint to operand 0.
9267         (add<mode>3_carryinC): Likewise.
9268         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9270 2018-08-09  Martin Liska  <mliska@suse.cz>
9272         PR c/86895
9273         * common.opt: Remove extra line.
9275 2018-08-09  Martin Liska  <mliska@suse.cz>
9277         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9278         at the end of a line, make first letter capital and end up
9279         a sentence with a dot.
9280         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9281         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9282         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9283         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9284         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9285         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9286         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9287         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9288         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9289         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9290         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9291         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9292         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9293         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9294         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9295         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9296         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9297         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9299 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9301         PR target/84332
9302         * config/s390/s390.c (s390_option_override_internal): Reduce the
9303         stack-clash-protection-probe-interval param if it would be too big
9304         for z900.
9306 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9308         PR target/46179
9309         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9310         * config/m68k/m68k.c (handle_move_double): Don't call
9311         m68k_final_prescan_insn.
9312         (m68k_adjust_decorated_operand): Renamed from
9313         m68k_final_prescan_insn, remove first and third operand and
9314         simplify.
9315         (print_operand): Call it.
9316         (print_operand_address): Call it.
9318 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9320         * diagnostic.c (diagnostic_report_current_module): Use
9321         linemap_included_from & linemap_included_from_linemap.
9323 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9325         * config/aarch64/aarch64-cores.def: Add phecda core.
9326         * config/aarch64/aarch64-tune.md: Regenerate.
9327         * doc/invoke.texi: Add phecda core.
9329 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9331         PR target/85295
9332         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9333         definitions.
9334         * config/s390/s390.md ("movti"): Add more alternatives for
9335         constant to GPR copies.
9337 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9339         * config/s390/s390.c: Fix whitespace damage throughout the file.
9340         * config/s390/s390.h: Likewise.
9341         * config/s390/tpf.h: Likewise.
9343 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9345         * config/s390/s390.c (s390_loadrelative_operand_p):
9346         Remove TARGET_CPU_ZARCH usages.
9347         (s390_rtx_costs): Likewise.
9348         (s390_legitimate_constant_p): Likewise.
9349         (s390_cannot_force_const_mem): Likewise.
9350         (legitimate_reload_constant_p): Likewise.
9351         (s390_preferred_reload_class): Likewise.
9352         (legitimize_pic_address): Likewise.
9353         (legitimize_tls_address): Likewise.
9354         (s390_split_branches): Removed.
9355         (s390_add_execute): Removed.
9356         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9357         (s390_mainpool_start): Likewise.
9358         (s390_mainpool_finish): Likewise.
9359         (s390_mainpool_cancel): Removed.
9360         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9361         (s390_chunkify_cancel): Likewise.
9362         (s390_return_addr_rtx): Likewise.
9363         (s390_register_info): Remove split_branches_pending_p uages.
9364         (s390_optimize_register_info): Likewise.
9365         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9366         split_branches_pending_p usages.
9367         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9368         (s390_load_got): Likewise.
9369         (s390_expand_split_stack_prologue): Likewise.
9370         (output_asm_nops): Likewise.
9371         (s390_function_profiler): Likewise.
9372         (s390_emit_call): Likewise.
9373         (s390_conditional_register_usage): Likewise.
9374         (s390_optimize_prologue): Likewise.
9375         (s390_reorg): Remove TARGET_CPU_ZARCH and
9376         split_branches_pending_p usages.
9377         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9378         usages.
9379         (s390_output_indirect_thunk_function): Likewise.
9380         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9381         (TARGET_CPU_ZARCH_P): Removed.
9382         (struct machine_function): Remove split_branches_pending_p.
9383         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9385 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9387         * common/config/s390/s390-common.c (processor_flags_table):
9388         Remove flags.
9389         * config.gcc: Remove with_arch/with_tune support.
9390         * config/s390/2064.md: Remove cpu attribute comparisons.
9391         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9392         Remove MTN.
9393         * config/s390/linux.h (ASM_SPEC):
9394         Remove -march support.
9395         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9396         Use a table to get an arch level.
9397         * config/s390/s390-opts.h (enum processor_type):
9398         Remove enum values.
9399         * config/s390/s390.c
9400         (processor_table): Remove entries, add arch_level values.
9401         (s390_issue_rate): Remove cases.
9402         (s390_option_override): Adjust
9403         s390_option_override_internal() call.
9404         (s390_option_override_internal): Remove deprecation warning.
9405         (s390_valid_target_attribute_tree): Adjust
9406         s390_option_override_internal() call.
9407         * config/s390/s390.h (struct s390_processor):
9408         Share with s390-c.c, add arch_level field.
9409         * config/s390/s390.md:
9410         Remove occurrences in cpu attribute.
9411         * config/s390/s390.opt: Remove -march/-mtune support.
9412         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9413         * doc/invoke.texi: Remove deprecation warning.
9415 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9417         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9418         global.
9419         (qdf24xx_tunings): Set vector cost structure to
9420         qdf24xx_vector_cost.
9422         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9423         <register_sextend>: Set to 3.
9425 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9427         PR target/86838
9428         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9429         * config/aarch64/aarch64-simd.md
9430         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9431         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9432         (aarch64_frecpx<mode>): ...this new pattern.
9433         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9434         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9436 2018-08-07  Martin Liska  <mliska@suse.cz>
9438         PR middle-end/83023
9439         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9440         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9441         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9442         * doc/extend.texi: Document that malloc attribute adds
9443         hit to compiler.
9445 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9447         PR target/86785
9448         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9449         Define to speculation_safe_value_not_needed.
9451 2018-08-06  Jeff Law  <law@redhat.com>
9453         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9454         the vr_values instance to cprop_into_stmt.
9455         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9456         (cprop_operand): Also query EVRP to determine if OP is a constant.
9458 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9460         * diagnostic.c (diagnostic_report_current_module): Reroll
9461         included-at loop.  Translate text.
9463 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9465         * function-tests.c (selftest::test_expansion_to_rtl): Call
9466         free_after_compilation.
9468 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9470         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9472 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9474         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9475         loops with memory block operations from getting unrolled.
9477 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9479         PR target/86807
9480         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9481         Define to speculation_safe_value_not_needed.
9483 2018-08-06  Jeff Law  <law@redhat.com>
9485         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9486         assert.
9488 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9490         PR target/86662
9491         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9492         with all enabled __intN types.
9494         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9496 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9498         * alias.c (record_set): Check for clobber high.
9499         * cfgexpand.c (expand_gimple_stmt): Likewise.
9500         * combine-stack-adj.c (single_set_for_csa): Likewise.
9501         * combine.c (find_single_use_1): Likewise.
9502         (set_nonzero_bits_and_sign_copies): Likewise.
9503         (get_combine_src_dest): Likewise.
9504         (is_parallel_of_n_reg_sets): Likewise.
9505         (try_combine): Likewise.
9506         (record_dead_and_set_regs_1): Likewise.
9507         (reg_dead_at_p_1): Likewise.
9508         (reg_dead_at_p): Likewise.
9509         * dce.c (deletable_insn_p): Likewise.
9510         (mark_nonreg_stores_1): Likewise.
9511         (mark_nonreg_stores_2): Likewise.
9512         * df-scan.c (df_find_hard_reg_defs): Likewise.
9513         (df_uses_record): Likewise.
9514         (df_get_call_refs): Likewise.
9515         * dwarf2out.c (mem_loc_descriptor): Likewise.
9516         * haifa-sched.c (haifa_classify_rtx): Likewise.
9517         * ira-build.c (create_insn_allocnos): Likewise.
9518         * ira-costs.c (scan_one_insn): Likewise.
9519         * ira.c (equiv_init_movable_p): Likewise.
9520         (rtx_moveable_p): Likewise.
9521         (interesting_dest_for_shprep): Likewise.
9522         * jump.c (mark_jump_label_1): Likewise.
9523         * postreload-gcse.c (record_opr_changes): Likewise.
9524         * postreload.c (reload_cse_simplify): Likewise.
9525         (struct reg_use): Add source expr.
9526         (reload_combine): Check for clobber high.
9527         (reload_combine_note_use): Likewise.
9528         (reload_cse_move2add): Likewise.
9529         (move2add_note_store): Likewise.
9530         * print-rtl.c (print_pattern): Likewise.
9531         * recog.c (decode_asm_operands): Likewise.
9532         (store_data_bypass_p): Likewise.
9533         (if_test_bypass_p): Likewise.
9534         * regcprop.c (kill_clobbered_value): Likewise.
9535         (kill_set_value): Likewise.
9536         * reginfo.c (reg_scan_mark_refs): Likewise.
9537         * reload1.c (maybe_fix_stack_asms): Likewise.
9538         (eliminate_regs_1): Likewise.
9539         (elimination_effects): Likewise.
9540         (mark_not_eliminable): Likewise.
9541         (scan_paradoxical_subregs): Likewise.
9542         (forget_old_reloads_1): Likewise.
9543         * reorg.c (find_end_label): Likewise.
9544         (try_merge_delay_insns): Likewise.
9545         (redundant_insn): Likewise.
9546         (own_thread_p): Likewise.
9547         (fill_simple_delay_slots): Likewise.
9548         (fill_slots_from_thread): Likewise.
9549         (dbr_schedule): Likewise.
9550         * resource.c (update_live_status): Likewise.
9551         (mark_referenced_resources): Likewise.
9552         (mark_set_resources): Likewise.
9553         * rtl.c (copy_rtx): Likewise.
9554         * rtlanal.c (reg_referenced_p): Likewise.
9555         (single_set_2): Likewise.
9556         (noop_move_p): Likewise.
9557         (note_stores): Likewise.
9558         * sched-deps.c (sched_analyze_reg): Likewise.
9559         (sched_analyze_insn): Likewise.
9561 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9563         * cse.c (invalidate_reg): New function extracted from...
9564         (invalidate): ...here.
9565         (canonicalize_insn): Check for clobber high.
9566         (invalidate_from_clobbers): invalidate clobber highs.
9567         (invalidate_from_sets_and_clobbers): Likewise.
9568         (count_reg_usage): Check for clobber high.
9569         (insn_live_p): Likewise.
9570         * cselib.c (cselib_expand_value_rtx_1):Likewise.
9571         (cselib_invalidate_regno): Check for clobber in setter.
9572         (cselib_invalidate_rtx): Pass through setter.
9573         (cselib_invalidate_rtx_note_stores):
9574         (cselib_process_insn): Check for clobber high.
9575         * cselib.h (cselib_invalidate_rtx): Add operand.
9577 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9579         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
9580         (mark_not_eliminable): Likewise.
9581         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
9582         * lra-lives.c (process_bb_lives): Check for clobber high.
9583         * lra.c (new_insn_reg): Remember clobber highs.
9584         (collect_non_operand_hard_regs): Check for clobber high.
9585         (lra_set_insn_recog_data): Likewise.
9586         (add_regs_to_insn_regno_info): Likewise.
9587         (lra_update_insn_regno_info): Likewise.
9589 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9591         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
9592         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
9594 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9596         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
9597         (copy_insn_1): Likewise.
9598         (gen_hard_reg_clobber_high): New gen function.
9599         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
9600         * genemit.c (gen_exp): Likewise.
9601         (gen_emit_seq): Pass through info.
9602         (gen_insn): Check for CLOBBER_HIGH.
9603         (gen_expand): Pass through info.
9604         (gen_split): Likewise.
9605         (output_add_clobbers): Likewise.
9606         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
9607         (remove_clobbers): Likewise.
9608         * rtl.h (gen_hard_reg_clobber_high): New declaration.
9610 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9612         * doc/rtl.texi (clobber_high): Add.
9613         (parallel): Add in clobber high
9614         * rtl.c (rtl_check_failed_code3): Add function.
9615         * rtl.def (CLOBBER_HIGH): Add expression.
9616         * rtl.h (RTL_CHECKC3): Add macro.
9617         (rtl_check_failed_code3): Add declaration.
9618         (XC3EXP): Add macro.
9620 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
9622         PR target/86386
9623         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
9624         cfun->machine->max_used_stack_alignment if needed.
9626 2018-08-04  Martin Sebor  <msebor@redhat.com>
9628         PR tree-optimization/86571
9629         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
9630         NaN output to 4.
9632 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
9634         PR target/86799
9635         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9636         Define.
9638 2018-08-03  Jeff Law  <law@redhat.com>
9640         PR target/86795
9641         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9642         Define to speculation_safe_value_not_needed.
9644 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
9646         * doc/gcov.texi (-x): Remove duplicate "to".
9647         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
9648         (-Wif-not-aligned): Remove duplicate "is".
9649         (-flto): Remove duplicate "the".
9650         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
9651         duplicate "v5.00.b".
9652         (MSP430 Options): Remove duplicate "and" from the description
9653         of "-mgprel-sec=regexp".
9654         (x86 Options): Remove duplicate copies of "vmldLog102" and
9655         vmlsLog104 from description of "-mveclibabi=type".
9657 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
9659         * internal-fn.h (first_commutative_argument): Declare.
9660         * internal-fn.c (first_commutative_argument): New function.
9661         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
9662         restrictions for pattern statements.  Use first_commutative_argument
9663         to look for commutative operands in calls to internal functions.
9665 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
9667         * Makefile.in (wide-int-range.o): New.
9668         * tree-vrp.c: Move all the wide_int_* functions to...
9669         * wide-int-range.cc: ...here.
9670         * tree-vrp.h: Move all the wide_int_* prototypes to...
9671         * wide-int-range.h: ...here.
9673 2018-08-03  Tom de Vries  <tdevries@suse.de>
9675         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
9676         UI_NONE.
9677         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
9678         * except.c (output_function_exception_table): Do early exit if
9679         targetm_common.except_unwind_info (&global_options) == UI_NONE.
9681 2018-08-03  Martin Liska  <mliska@suse.cz>
9683         * predict.c (dump_prediction): Change to 2 digits
9684         in fraction part.
9686 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9688         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
9689         neon_dup_q to...
9690         (falkor_am_1_gtov_gtov): ... a new insn reservation.
9692 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
9694         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
9695         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
9696         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
9697         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
9698         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
9700 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9702         * diagnostic-show-locus.c (diagnostic_show_locus): Use
9703         pp_take_prefix when saving the existing prefix.
9704         * diagnostic.c (diagnostic_append_note): Likewise.
9705         * langhooks.c (lhd_print_error_function): Likewise.
9706         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
9707         param's type.  Free the existing prefix.
9708         (pp_take_prefix): New function.
9709         (pretty_printer::pretty_printer): Drop the prefix parameter.
9710         Rename the length parameter to match the comment.
9711         (pretty_printer::~pretty_printer): Free the prefix.
9712         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
9713         parameter.
9714         (struct pretty_printer): Drop the "const" from "prefix" field's
9715         type and clarify memory management.
9716         (pp_set_prefix): Drop the "const" from the 2nd param.
9717         (pp_take_prefix): New decl.
9719 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
9721         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
9722         for word_mode_ok here instead of passing as argument.
9723         (expand_block_compare): Change select_block_compare_mode() call.
9724         (expand_strncmp_gpr_sequence): New function.
9725         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
9727 2018-08-02  Jeff Law  <law@redhat.com>
9729         PR target/86790
9730         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9731         Define to speculation_safe_value_not_needed.
9733         PR target/86784
9734         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9735         Define to speculation_safe_value_not_needed.
9737 2018-08-02  Tom de Vries  <tdevries@suse.de>
9739         PR target/86660
9740         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
9741         function.  Return UI_TARGET unconditionally.
9742         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
9743         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
9745 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9747         * genemit.c (print_overload_test): Fix typo.
9749 2018-08-02  Richard Biener  <rguenther@suse.de>
9751         PR tree-optimization/86816
9752         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
9753         which checks for value availability before querying it.
9754         (gvn_uses_equal): Use it.
9755         (same_succ_hash): Likewise.
9756         (gimple_equal_p): Likewise.
9758 2018-08-02  Nick Clifton  <nickc@redhat.com>
9760         PR target/86813
9761         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9762         Define to speculation_safe_value_not_needed.
9764         PR target/86810
9765         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9766         Define to speculation_safe_value_not_needed.
9768         PR target/86810
9769         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9770         Define to speculation_safe_value_not_needed.
9772         PR target/86803
9773         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9774         Define to speculation_safe_value_not_needed.
9776         PR target/86797
9777         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9778         Define to speculation_safe_value_not_needed.
9780         PR target/86791
9781         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9782         Define to speculation_safe_value_not_needed.
9784         PR target/86789
9785         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9786         Define to speculation_safe_value_not_needed.
9788         PR target/86787
9789         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9790         Define to speculation_safe_value_not_needed.
9792         PR target/86782
9793         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9794         speculation_safe_value_not_needed.
9796         PR target/86781
9797         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
9798         to speculation_safe_value_not_needed.
9800 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9802         * doc/md.texi: Expand the documentation of instruction names
9803         to mention port-local uses.  Document '@' in pattern names.
9804         * read-md.h (overloaded_instance, overloaded_name): New structs.
9805         (mapping): Declare.
9806         (md_reader::handle_overloaded_name): New member function.
9807         (md_reader::get_overloads): Likewise.
9808         (md_reader::m_first_overload): New member variable.
9809         (md_reader::m_next_overload_ptr): Likewise.
9810         (md_reader::m_overloads_htab): Likewise.
9811         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
9812         m_next_overload_ptr and m_overloads_htab.
9813         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
9814         (get_mode_token, get_code_token, get_int_token): New functions.
9815         (map_attr_string): Add an optional argument that passes back
9816         the associated iterator.
9817         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
9818         (md_reader::handle_overloaded_name, add_overload_instance): New
9819         functions.
9820         (apply_iterators): Handle '@' names.  Report an error if '@'
9821         is used without iterators.
9822         (initialize_iterators): Initialize the new iterator_group fields.
9823         * genopinit.c (handle_overloaded_code_for)
9824         (handle_overloaded_gen): New functions.
9825         (main): Use them to print declarations of maybe_code_for_* and
9826         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
9827         * genemit.c (print_overload_arguments, print_overload_test)
9828         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
9829         (main): Use it to print definitions of maybe_code_for_* and
9830         maybe_gen_* functions.
9831         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
9832         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
9833         instead of explicit mode checks.
9834         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
9835         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
9836         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
9837         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
9838         (aarch64_expand_compare_and_swap): Likewise
9839         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
9840         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
9841         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
9842         (aarch64_constant_pool_reload_icode): Delete.
9843         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
9844         instead of aarch64_constant_pool_reload_icode.  Use
9845         code_for_aarch64_reload_mov instead of explicit mode checks.
9846         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
9847         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
9848         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
9849         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
9850         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
9851         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
9852         (aarch64_atomic_load_op_code): Delete.
9853         (aarch64_emit_atomic_load_op): Likewise.
9854         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
9855         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
9856         instead of aarch64_emit_atomic_load_op.
9857         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
9858         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
9859         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
9860         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
9861         character before the pattern name.
9862         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
9863         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
9864         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
9865         (aarch64_frecps<mode>): Likewise.
9866         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
9867         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
9868         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
9869         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
9870         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
9872 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9874         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
9875         Allow HFmode constants if TARGET_FP_F16INST.
9877 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
9879         PR target/86014
9880         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
9881         No longer check last store for clobber of address register.
9883 2018-08-02  Martin Liska  <mliska@suse.cz>
9885         PR gcov-profile/86817
9886         * gcov.c (process_all_functions): New function.
9887         (main): Call it.
9888         (process_file): Move functions processing to
9889         process_all_functions.
9891 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9893         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
9894         "const" to the "gimple *" and "rtx_insn *" parameters.
9895         * dumpfile.h (dump_user_location_t::dump_user_location_t):
9896         Likewise.
9897         (dump_location_t::dump_location_t): Likewise.
9899 2018-08-01  Martin Sebor  <msebor@redhat.com>
9901         PR tree-optimization/86650
9902         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
9903         rather than a "gcall *".  Directly pass the data of interest
9904         to percent_K_format, rather than building a temporary CALL_EXPR
9905         to hold it.
9906         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
9907         (gimple_fold_builtin_strncat): Adjust.
9908         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
9909         gcall* argument with gimple*.
9910         * gimple-ssa-warn-restrict.c (check_call): Same.
9911         (wrestrict_dom_walker::before_dom_children): Same.
9912         (builtin_access::builtin_access): Same.
9913         (check_bounds_or_overlap): Same
9914         (maybe_diag_overlap): Same.
9915         (maybe_diag_offset_bounds): Same.
9916         * tree-diagnostic.c (default_tree_printer): Move usage of
9917         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
9918         to this callsite.
9919         * tree-pretty-print.c (percent_K_format): Add argument.
9920         * tree-pretty-print.h: Add argument.
9921         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
9922         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
9923         (maybe_diag_stxncpy_trunc): Same.
9924         (handle_builtin_stxncpy): Same.
9925         (handle_builtin_strcat): Same.
9927 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9929         * match.pd: Optimise pointer range checks.
9931 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9933         PR tree-optimization/86758
9934         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
9935         to remove pattern statements.
9937 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9939         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
9940         result of dfs_enumerate_from when constructing stmt_vec_infos,
9941         instead of additionally calling get_loop_body.
9943 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9945         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
9946         parameter.
9947         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
9948         When creating an iv, assert that the step is not known to be zero.
9949         (vect_setup_realignment): Update call accordingly.
9950         * tree-vect-stmts.c (vectorizable_store): Likewise.
9951         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
9953 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9955         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
9956         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
9957         (vectorizable_reduction): Likewise.
9958         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9959         (vect_detect_hybrid_slp_stmts): Likewise.
9960         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9962 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
9964         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
9965         (wide_int_set_zero_nonzero_bits): ...this.
9966         (zero_nonzero_bits_from_vr): Rename to...
9967         (vrp_set_zero_nonzero_bits): ...this.
9968         (extract_range_from_multiplicative_op_1): Abstract wide int
9969         code...
9970         (wide_int_range_multiplicative_op): ...here.
9971         (extract_range_from_binary_expr_1): Extract wide int binary
9972         operations into their own functions.
9973         (wide_int_range_lshift): New.
9974         (wide_int_range_can_optimize_bit_op): New.
9975         (wide_int_range_shift_undefined_p): New.
9976         (wide_int_range_bit_xor): New.
9977         (wide_int_range_bit_ior): New.
9978         (wide_int_range_bit_and): New.
9979         (wide_int_range_trunc_mod): New.
9980         (extract_range_into_wide_ints): New.
9981         (vrp_shift_undefined_p): New.
9982         (extract_range_from_multiplicative_op): New.
9983         (vrp_can_optimize_bit_op): New.
9984         * tree-vrp.h (value_range::dump): New.
9985         (wide_int_range_multiplicative_op): New.
9986         (wide_int_range_lshift):New.
9987         (wide_int_range_shift_undefined_p): New.
9988         (wide_int_range_bit_xor): New.
9989         (wide_int_range_bit_ior): New.
9990         (wide_int_range_bit_and): New.
9991         (wide_int_range_trunc_mod): New.
9992         (zero_nonzero_bits_from_bounds): Rename to...
9993         (wide_int_set_zero_nonzero_bits): ...this.
9994         (zero_nonzero_bits_from_vr): Rename to...
9995         (vrp_set_zero_nonzero_bits): ...this.
9996         (range_easy_mask_min_max): Rename to...
9997         (wide_int_range_can_optimize_bit_op): this.
9998         * vr-values.c (simplify_bit_ops_using_ranges): Rename
9999         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10001 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10003         * tree-vectorizer.h (vect_orig_stmt): New function.
10004         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10005         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10006         (vect_create_epilog_for_reduction): Likewise.
10007         (vectorizable_live_operation): Likewise.
10008         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10009         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10010         * tree-vect-stmts.c (vectorizable_call): Likewise.
10011         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10013 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10015         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10016         argument.
10017         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10018         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10019         (vect_transform_loop): Likewise.
10020         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10022 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10024         * tree-vectorizer.h (vect_schedule_slp): Return void.
10025         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10026         (vect_schedule_slp): Likewise.
10028 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10030         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10031         argument.
10032         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10033         instances before the main loop, if any exist.
10035 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10037         PR tree-optimization/86749
10038         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10039         If the lhs is used in a COND_EXPR, check that it is being used
10040         as the "then" or "else" value.
10042 2018-08-01  Tom de Vries  <tdevries@suse.de>
10044         PR target/86800
10045         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10046         speculation_safe_value_not_needed.
10048 2018-08-01  Richard Biener  <rguenther@suse.de>
10050         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10051         as base and offset.
10053 2018-08-01  Martin Liska  <mliska@suse.cz>
10055         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10056         format how successful transformation is dumped.
10057         (gimple_mod_pow2_value_transform): Likewise.
10058         (gimple_mod_subtract_transform): Likewise.
10059         (gimple_stringops_transform): Likewise.
10061 2018-08-01  Martin Liska  <mliska@suse.cz>
10063         PR value-prof/35543
10064         * value-prof.c (interesting_stringop_to_profile_p):
10065         Simplify the code and add BUILT_IN_MEMMOVE.
10066         (gimple_stringops_transform): Likewise.
10068 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10070         * config/aarch64/aarch64-simd.md
10071         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10072         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10073         use GPI iterator instead of SI mode.
10075 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10077         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10078         rs6000_speculation_barrier.
10079         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10080         new barrier pattern name.
10082 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10084         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10085         (speculation_barrier): New insn.
10087 2018-08-01  Richard Biener  <rguenther@suse.de>
10089         PR bootstrap/86724
10090         * graphite.h: Include isl/id.h and isl/space.h to allow build
10091         with ISL 0.20.
10093 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10095         PR target/86651
10096         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10097         mode for COFF targets.
10098         * defaults.h (TARGET_COFF): Define.
10099         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10100         TARGET_COFF): Define.
10101         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10102         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10103         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10105 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10107         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10108         * gimple-streamer-out.c (output_bb): Save it.
10109         * lto-streamer-in.c (input_struct_function_base): Restore
10110         instance discriminator if available.  Create map on demand.
10111         * lto-streamer-out.c (output_struct_function_base): Save it if
10112         available.
10113         * final.c (decl_to_instance_map): Document LTO strategy.
10115 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10116             Olivier Hainque  <hainque@adacore.com>
10118         * debug.h (decl_to_instance_map_t): New type.
10119         (decl_to_instance_map): Declare.
10120         (maybe_create_decl_to_instance_map): New inline function.
10121         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10122         to track basic block discriminators.
10123         (final_start_function_1): Initialize them.
10124         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10125         bb_discriminator.
10126         (decl_to_instance_map): New variable.
10127         (map_decl_to_instance, maybe_set_discriminator): New functions.
10128         (notice_source_line): Set discriminator.
10130 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10132         * targhooks.c (default_have_speculation_safe_value): Add
10133         ATTRIBUTE_UNUSED.
10135 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10137         * dump-context.h: Include "pretty-print.h".
10138         (dump_context::refresh_dumps_are_enabled): New decl.
10139         (dump_context::emit_item): New decl.
10140         (class dump_context): Add fields "m_test_pp" and
10141         "m_test_pp_flags".
10142         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10143         (temp_dump_context::get_dumped_text): New decl.
10144         (class temp_dump_context): Add field "m_pp".
10145         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10146         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10147         m_test_pp.
10148         (set_dump_file): Update for above change.
10149         (set_alt_dump_file): Likewise.
10150         (dump_loc): New overload, taking a pretty_printer *.
10151         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10152         to any test pretty-printer.
10153         (make_item_for_dump_gimple_stmt): New function, adapted from
10154         optinfo::add_gimple_stmt.
10155         (dump_context::dump_gimple_stmt): Call it, and use the result,
10156         eliminating the direct usage of dump_file and alt_dump_file in
10157         favor of indirectly using them via emit_item.
10158         (make_item_for_dump_gimple_expr): New function, adapted from
10159         optinfo::add_gimple_expr.
10160         (dump_context::dump_gimple_expr): Call it, and use the result,
10161         eliminating the direct usage of dump_file and alt_dump_file in
10162         favor of indirectly using them via emit_item.
10163         (make_item_for_dump_generic_expr): New function, adapted from
10164         optinfo::add_tree.
10165         (dump_context::dump_generic_expr): Call it, and use the result,
10166         eliminating the direct usage of dump_file and alt_dump_file in
10167         favor of indirectly using them via emit_item.
10168         (make_item_for_dump_printf_va): New function, adapted from
10169         optinfo::add_printf_va.
10170         (make_item_for_dump_printf): New function.
10171         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10172         and use the result, eliminating the direct usage of dump_file and
10173         alt_dump_file in favor of indirectly using them via emit_item.
10174         (make_item_for_dump_dec): New function.
10175         (dump_context::dump_dec): Call it, and use the result,
10176         eliminating the direct usage of dump_file and alt_dump_file in
10177         favor of indirectly using them via emit_item.
10178         (make_item_for_dump_symtab_node): New function, adapted from
10179         optinfo::add_symtab_node.
10180         (dump_context::dump_symtab_node): Call it, and use the result,
10181         eliminating the direct usage of dump_file and alt_dump_file in
10182         favor of indirectly using them via emit_item.
10183         (dump_context::begin_scope): Reimplement, avoiding direct usage
10184         of dump_file and alt_dump_file in favor of indirectly using them
10185         via emit_item.
10186         (dump_context::emit_item): New member function.
10187         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10188         Set up test pretty-printer on the underlying context.  Call
10189         refresh_dumps_are_enabled.
10190         (temp_dump_context::~temp_dump_context): Call
10191         refresh_dumps_are_enabled.
10192         (temp_dump_context::get_dumped_text): New member function.
10193         (selftest::verify_dumped_text): New function.
10194         (ASSERT_DUMPED_TEXT_EQ): New macro.
10195         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10196         and then without optinfo enabled.  Add uses of
10197         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10198         * dumpfile.h: Update comment for the dump_* API.
10199         * optinfo-emit-json.cc
10200         (selftest::test_building_json_from_dump_calls): Update for new
10201         param for temp_dump_context ctor.
10202         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10203         and "m_owned" field.
10204         (optinfo_item::~optinfo_item): Likewise.
10205         (optinfo::add_item): New member function.
10206         (optinfo::emit): Update comment.
10207         (optinfo::add_string): Delete.
10208         (optinfo::add_printf): Delete.
10209         (optinfo::add_printf_va): Delete.
10210         (optinfo::add_gimple_stmt): Delete.
10211         (optinfo::add_gimple_expr): Delete.
10212         (optinfo::add_tree): Delete.
10213         (optinfo::add_symtab_node): Delete.
10214         (optinfo::add_dec): Delete.
10215         * optinfo.h (class dump_context): New forward decl.
10216         (optinfo::add_item): New decl.
10217         (optinfo::add_string): Delete.
10218         (optinfo::add_printf): Delete.
10219         (optinfo::add_printf_va): Delete.
10220         (optinfo::add_gimple_stmt): Delete.
10221         (optinfo::add_gimple_expr): Delete.
10222         (optinfo::add_tree): Delete.
10223         (optinfo::add_symtab_node): Delete.
10224         (optinfo::add_dec): Delete.
10225         (optinfo::add_poly_int): Delete.
10226         (optinfo_item::optinfo_item): Remove "owned" param.
10227         (class optinfo_item): Remove field "m_owned".
10229 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10231         PR middle-end/86705
10232         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10233         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10234         requested variable alignment.
10235         (expand_one_ssa_partition): Likewise.
10236         (expand_one_var): Likewise.
10238 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10240         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10241         to speculation_safe_value_not_needed.
10243 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10245         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10246         * targhooks.c (speculation_safe_value_not_needed): New function.
10247         * target.def (have_speculation_safe_value): Update documentation.
10248         * doc/tm.texi: Regenerated.
10250 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10252         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10253         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10254         expand.
10255         (despeculate_copy<ALLI:mode>_insn): New insn.
10256         (despeculate_copyti_insn): New insn.
10257         (despeculate_simple<ALLI:mode>): New insn
10258         (despeculate_simpleti): New insn.
10259         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10260         function.
10261         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10262         aarch64_speculation_safe_value.
10263         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10265 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10267         * config/aarch64/aarch64-speculation.cc: New file.
10268         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10269         before pass_reorder_blocks.
10270         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10271         prototype.
10272         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10273         X14 and X15 when tracking speculation.
10274         * config/aarch64/aarch64.md (register name constants): Add
10275         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10276         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10277         (speculation_barrier): New insn attribute.
10278         (cmp<mode>): Allow SP in comparisons.
10279         (speculation_tracker): New insn.
10280         (speculation_barrier): Add speculation_barrier attribute.
10281         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10282         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10283         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10285 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10287         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10288         aarch64_track_speculation is true.
10289         (tb<optab><mode>1): Likewise.
10290         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10291         generate CB[N]Z when tracking speculation.
10292         (aarch64_split_compare_and_swap): Likewise.
10293         (aarch64_split_atomic_op): Likewise.
10295 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10297         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10299 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10301         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10302         (speculation_barrier): New insn.
10304 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10306         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10307         * config/arm/arm.md (speculation_barrier): New expand.
10308         (speculation_barrier_insn): New pattern.
10310 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10312         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10313         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10314         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10315         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10316         list.
10317         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10318         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10319         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10320         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10321         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10322         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10323         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10324         * builtins.c (expand_speculation_safe_value): New function.
10325         (expand_builtin): Call it.
10326         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10327         * doc/extend.texi: Document __builtin_speculation_safe_value.
10328         * doc/md.texi: Document "speculation_barrier" pattern.
10329         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10330         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10331         * doc/tm.texi: Regenerated.
10332         * target.def (have_speculation_safe_value, speculation_safe_value): New
10333         hooks.
10334         * targhooks.c (default_have_speculation_safe_value): New function.
10335         (default_speculation_safe_value): New function.
10336         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10337         (default_speculation_safe_value): Add prototype.
10339 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10341         * dump-context.h (dump_context::dump_loc): New decl.
10342         * dumpfile.c (dump_context::dump_loc): New member function.
10343         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10344         and dump_gimple_stmt.
10345         (dump_context::dump_gimple_expr_loc): Likewise, using
10346         dump_gimple_expr.
10347         (dump_context::dump_generic_expr_loc): Likewise, using
10348         dump_generic_expr.
10349         (dump_context::dump_printf_loc_va): Likewise, using
10350         dump_printf_va.
10351         (dump_context::begin_scope): Explicitly using the global function
10352         "dump_loc", rather than the member function.
10354 2018-07-31  Martin Sebor  <msebor@redhat.com>
10356         PR tree-optimization/86741
10357         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10359 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10361         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10362         depend on whether prefetch instructions will be emitted or not.
10363         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10364         will be emitted or not.
10365         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10366         (TARGET_SETMEM_PFD): New macros.
10368 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10370         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10371         (NULL_STMT_VEC_INFO): Delete.
10372         (stmt_vec_info::operator*): Likewise.
10373         (stmt_vec_info::operator gimple *): Likewise.
10374         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10375         of NULL_STMT_VEC_INFO.
10376         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10377         (vect_reassociating_reduction_p): Likewise.
10378         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10379         (vectorizable_store): Likewise.
10380         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10381         (vec_info::free_stmt_vec_infos): Likewise.
10383 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10385         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10386         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10388 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10390         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10391         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10392         (vec_info::free_stmt_vec_info): New private member functions.
10393         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10394         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10395         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10396         set_stmt_vec_info_vec and free_stmt_vec_infos.
10397         * tree-vect-loop.c (_loop_vec_info): Remove call to
10398         set_stmt_vec_info_vec.
10399         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10400         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10401         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10402         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10403         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10404         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10405         with the clearing in the xcalloc.
10406         (stmt_vec_info_vec): Delete.
10407         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10408         (vectorize_loops): Likewise.
10409         (vec_info::~vec_info): Remove argument from call to
10410         free_stmt_vec_infos.
10411         (vec_info::add_stmt): Remove vinfo argument from call to
10412         new_stmt_vec_info.
10414 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10416         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10417         rather than a gimple stmt.
10418         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10419         information for pattern statements when passed the original
10420         statement; instead wait to be passed the pattern statement itself.
10421         Don't call set_vinfo_for_stmt here.
10422         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10423         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10424         stmt_vec_infos here.
10425         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10426         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10427         stmt_vec_infos entry.
10429 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10431         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10432         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10433         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10434         * tree-vect-stmts.c (vectorizable_call): Likewise.
10435         (vectorizable_simd_clone_call): Likewise.
10437 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10439         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10440         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10441         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10442         * tree-vect-loop.c (vect_transform_loop): Likewise.
10443         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10444         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10446 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10448         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10449         (vect_dr_stmt): Delete.
10450         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10451         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10452         of DR_VECT_AUX.
10453         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10454         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10455         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10456         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10457         (vect_analyze_data_refs): Likewise.
10458         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10459         argument.
10460         (vect_find_same_alignment_drs): Likewise.
10461         (vect_slp_analyze_node_dependences): Update calls accordingly.
10462         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10463         instead of DR_VECT_AUX.
10464         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10465         of a vector data references.  Use vec_info::lookup_dr instead of
10466         DR_VECT_AUX.
10467         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10468         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10469         instead of DR_VECT_AUX.
10471 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10473         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10474         dr_vec_info.
10475         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10476         accordingly.
10477         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10478         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10479         (vect_gen_prolog_loop_niters): Likewise.
10481 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10483         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10484         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10485         (vect_known_alignment_in_bytes, vect_dr_behavior)
10486         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10487         instead of data_references.  Update calls to other routines for
10488         which the same change has been made.
10489         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10490         dr_vec_infos instead of stmt_vec_infos.
10491         (vect_analyze_data_ref_dependence): Update call accordingly.
10492         (vect_slp_analyze_data_ref_dependence)
10493         (vect_record_base_alignments): Use DR_VECT_AUX.
10494         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10495         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10496         (vector_alignment_reachable_p, vect_get_data_access_cost)
10497         (vect_peeling_supportable, vect_analyze_group_access_1)
10498         (vect_analyze_group_access, vect_analyze_data_ref_access)
10499         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10500         (vect_compile_time_alias, vect_small_gap_p)
10501         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
10502         (vect_supportable_dr_alignment): Take references as dr_vec_infos
10503         instead of data_references.  Update calls to other routines for
10504         which the same change has been made.
10505         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
10506         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10507         (vect_slp_analyze_and_verify_node_alignment)
10508         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10509         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10510         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
10511         above changes.
10512         (_vect_peel_info::dr): Replace with...
10513         (_vect_peel_info::dr_info): ...this new field.
10514         (vect_peeling_hash_get_most_frequent)
10515         (vect_peeling_hash_choose_best_peeling): Update accordingly.
10516         (vect_peeling_hash_get_lowest_cost):
10517         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
10518         routines for which the same change has been made.
10519         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
10520         data_reference.
10521         * tree-vect-loop-manip.c (get_misalign_in_elems)
10522         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
10523         above changes.
10524         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10525         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
10526         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
10527         (get_group_load_store_type, get_negative_load_store_type)
10528         (vect_get_data_ptr_increment, vectorizable_store)
10529         (vectorizable_load): Likewise.
10530         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
10531         Update calls to other routines for which the same change has been made.
10533 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10535         * tree-vectorizer.h (vec_info::move_dr): New member function.
10536         (dataref_aux): Rename to...
10537         (dr_vec_info): ...this and add "dr" and "stmt" fields.
10538         (_stmt_vec_info::dr_aux): Update accordingly.
10539         (_stmt_vec_info::data_ref_info): Delete.
10540         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
10541         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
10542         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
10543         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
10544         of data_ref.
10545         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
10546         (STMT_VINFO_DR_INFO): New macro.
10547         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
10548         (set_dr_misalignment): Update after rename of dataref_aux.
10549         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
10550         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
10551         initialization of STMT_VINFO_DATA_REF.
10552         * tree-vectorizer.c (vec_info::move_dr): New function.
10553         * tree-vect-patterns.c (vect_recog_bool_pattern)
10554         (vect_recog_mask_conversion_pattern)
10555         (vect_recog_gather_scatter_pattern): Use it.
10556         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
10557         the "dr" and "stmt" fields of dr_vec_info instead of
10558         STMT_VINFO_DATA_REF.
10560 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10562         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
10563         (is_pattern_stmt_p): Use it.
10564         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
10565         on pattern statements.
10567 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10569         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
10570         original stmt as a stmt_vec_info rather than a gimple stmt.
10571         (vect_pattern_recog_1): Take the statement directly as a
10572         stmt_vec_info, rather than via a gimple_stmt_iterator.
10573         Update call to vect_mark_pattern_stmts.
10574         (vect_pattern_recog): Update calls accordingly.
10576 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10578         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
10579         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
10580         a vect_def_type for the first argument.
10581         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
10582         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
10583         operand if it isn't defined by a vectorized statement.
10584         (vect_build_gather_load_calls): Remove the mask_dt argument and
10585         update calls to vect_get_vec_def_for_stmt_copy.
10586         (vectorizable_bswap): Likewise the dt argument.
10587         (vectorizable_call): Update calls to vectorizable_bswap and
10588         vect_get_vec_def_for_stmt_copy.
10589         (vectorizable_simd_clone_call, vectorizable_assignment)
10590         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
10591         (vectorizable_comparison): Update calls to
10592         vect_get_vec_def_for_stmt_copy.
10593         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
10594         vect_is_simple_use.
10595         (vect_get_loop_based_defs): Remove dt argument and update call
10596         to vect_get_vec_def_for_stmt_copy.
10597         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
10598         and vect_get_vec_def_for_stmt_copy.
10599         (vectorizable_load): Update calls to vect_build_gather_load_calls
10600         and vect_get_vec_def_for_stmt_copy.
10601         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10602         (vectorizable_reduction, vectorizable_live_operation): Update calls
10603         to vect_get_vec_def_for_stmt_copy.
10605 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10607         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
10608         and gimple stmt arguments with a stmt_vec_info.
10609         (vect_record_base_alignments): Update calls accordingly.
10610         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
10611         and gimple stmt arguments with a stmt_vec_info.
10612         (vect_build_slp_tree_1): Remove vinfo argument and update call
10613         to vect_record_max_nunits.
10614         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
10615         and vect_record_max_nunits.
10617 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10619         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
10620         file and take a stmt_vec_info instead of a gimple stmt.
10621         (supportable_widening_operation, vect_finish_replace_stmt)
10622         (vect_finish_stmt_generation, vect_get_store_rhs)
10623         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
10624         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
10625         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
10626         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
10627         (vect_create_data_ref_ptr, bump_vector_ptr)
10628         (vect_permute_store_chain, vect_setup_realignment)
10629         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
10630         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
10631         (vectorizable_reduction, vectorizable_induction)
10632         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
10633         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
10634         than gimple stmts as arguments.
10635         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
10636         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
10637         (can_group_stmts_p, vect_check_gather_scatter)
10638         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10639         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
10640         (vect_permute_load_chain, vect_shift_permute_load_chain)
10641         (vect_transform_grouped_load)
10642         (vect_record_grouped_load_vectors): Likewise.
10643         * tree-vect-loop.c (vect_fixup_reduc_chain)
10644         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
10645         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
10646         (vectorizable_reduction, vectorizable_induction)
10647         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
10648         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
10649         (vect_get_load_store_mask): Likewise.
10650         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
10651         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
10652         * tree-vect-stmts.c (vect_mark_relevant)
10653         (is_simple_and_all_uses_invariant)
10654         (exist_non_indexing_operands_for_use_p, process_use)
10655         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
10656         (vect_get_vec_def_for_operand, vect_get_vec_defs)
10657         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
10658         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
10659         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
10660         (get_negative_load_store_type, get_load_store_type)
10661         (vect_check_load_store_mask, vect_check_store_rhs)
10662         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
10663         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10664         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10665         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10666         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
10667         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
10668         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
10669         (supportable_widening_operation): Likewise.
10671 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10673         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
10674         a stmt_vec_info instead of a gcall.
10675         (vect_check_gather_scatter): Update call accordingly.
10676         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
10677         of a gphi.
10678         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
10679         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
10680         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
10681         instead of a gimple stmt.
10682         (vect_transform_loop): Update calls accordingly.
10683         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
10684         stmt_vec_infos instead of gimple stmts.
10685         (vect_analyze_slp_instance): Update use accordingly.
10686         * tree-vect-stmts.c (read_vector_array, write_vector_array)
10687         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
10688         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
10689         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
10690         (vect_gen_widened_results_half, vect_get_loop_based_defs)
10691         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
10692         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
10693         down to subroutines.
10695 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10697         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
10698         of the worklist from a vector of gimple stmts to a vector of
10699         stmt_vec_infos.
10700         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10701         (vect_mark_stmts_to_be_vectorized): Likewise
10703 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10705         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
10706         statement before passing it to vect_analyze_stmt.
10707         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
10708         the chain of phi vector definitions.  Track the exit phi via its
10709         stmt_vec_info.
10710         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
10711         STMT_VINFO_REDUC_DEF.
10712         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
10713         stmt_vec_infos to handle the statement chains.
10714         (vect_get_slp_defs): Record the first statement in the node
10715         using a stmt_vec_info.
10716         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
10717         statements here and pass their stmt_vec_info down to subroutines.
10718         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
10719         down to vect_finish_stmt_generation.
10720         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
10721         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
10722         stmt_vec_infos to subroutines.
10723         (vect_remove_stores): Use stmt_vec_infos to handle the statement
10724         chains.
10726 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10728         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
10729         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
10730         (vect_permute_store_chain, vect_setup_realignment)
10731         (vect_permute_load_chain, vect_shift_permute_load_chain)
10732         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
10733         stmts internally, and when passing values to other vectorizer routines.
10734         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10735         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
10736         (vect_analyze_loop_operations, get_initial_def_for_reduction)
10737         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
10738         (vectorizable_reduction, vectorizable_induction)
10739         (vectorizable_live_operation, vect_transform_loop_stmt)
10740         (vect_transform_loop): Likewise.
10741         * tree-vect-patterns.c (vect_reassociating_reduction_p)
10742         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
10743         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
10744         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10745         (vect_slp_analyze_node_operations_1): Likewise.
10746         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10747         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
10748         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
10749         (vect_finish_stmt_generation_1, get_group_load_store_type)
10750         (get_load_store_type, vect_build_gather_load_calls)
10751         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10752         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10753         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10754         (vectorizable_store, vectorizable_load, vectorizable_condition)
10755         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
10756         (supportable_widening_operation): Likewise.
10757         (vect_get_vector_types_for_stmt): Likewise.
10758         * tree-vectorizer.h (vect_dr_behavior): Likewise.
10760 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10762         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10763         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
10764         (vect_permute_store_chain, vect_permute_load_chain)
10765         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
10766         repeated stmt_vec_info lookups.
10767         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10768         (vect_update_ivs_after_vectorizer): Likewise.
10769         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
10770         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10771         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
10772         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10773         (vect_bb_slp_scalar_cost): Likewise.
10774         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
10776 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10778         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
10779         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
10780         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
10781         to get gassigns and gcalls, rather than operating on generc gimple
10782         stmts.
10783         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
10784         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
10785         (vectorizable_load, vect_analyze_stmt): Likewise.
10786         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
10788 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10790         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
10791         return stmt_vec_infos rather than gimple stmts.  Do not accept
10792         null arguments.
10793         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
10794         of a gimple stmt.
10795         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10796         Update use of get_later_stmt.
10797         (vect_get_constant_vectors): Update call accordingly.
10798         (vect_schedule_slp_instance): Likewise
10799         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
10800         (vect_slp_analyze_instance_dependence): Likewise.
10801         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
10803 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10805         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
10806         (stmt_info_for_cost::stmt_info): ...this new field.
10807         (add_stmt_costs): Update accordingly.
10808         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10809         (vect_get_known_peeling_cost): Likewise.
10810         (vect_estimate_min_profitable_iters): Likewise.
10811         * tree-vect-stmts.c (record_stmt_cost): Likewise.
10813 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10815         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
10816         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10817         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10818         accordingly.
10819         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
10821 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10823         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
10824         a gimple stmt to a stmt_vec_info.
10825         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10827 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10829         * tree-vectorizer.h (vec_info::grouped_stores): Change from
10830         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10831         (_loop_vec_info::reduction_chains): Likewise.
10832         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
10833         accordingly.
10834         * tree-vect-slp.c (vect_analyze_slp): Likewise.
10836 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10838         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
10839         a gimple stmt to a stmt_vec_info.
10840         (_stmt_vec_info::next_element): Likewise.
10841         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
10842         (vect_slp_analyze_and_verify_node_alignment)
10843         (vect_analyze_group_access_1, vect_analyze_group_access)
10844         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
10845         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
10846         (vect_supportable_dr_alignment): Update accordingly.
10847         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
10848         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
10849         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
10850         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10851         * tree-vect-slp.c (vect_build_slp_tree_1)
10852         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
10853         (vect_split_slp_store_group, vect_analyze_slp_instance)
10854         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
10855         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
10856         (get_group_load_store_type, get_load_store_type)
10857         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
10858         (vect_transform_stmt, vect_remove_stores): Likewise.
10860 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10862         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
10863         than a gimple stmt.
10864         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10865         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
10866         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
10867         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
10868         (vector_alignment_reachable_p, vect_get_data_access_cost)
10869         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
10870         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
10871         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10872         (vect_analyze_group_access_1, vect_analyze_group_access)
10873         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
10874         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
10875         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
10876         result of vect_dr_stmt and use the stmt_vec_info instead of
10877         the associated gimple stmt.
10878         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10879         (vect_gen_prolog_loop_niters): Likewise.
10880         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10882 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10884         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
10885         to a vec<stmt_vec_info>.
10886         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
10887         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
10888         vec<stmt_vec_info>.
10889         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
10890         to a vec<stmt_vec_info>.
10891         (bst_traits::value_type, bst_traits::value_type): Likewise.
10892         (bst_traits::hash): Update accordingly.
10893         (vect_get_and_check_slp_defs): Change the stmts parameter from
10894         a vec<gimple *> to a vec<stmt_vec_info>.
10895         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
10896         (vect_build_slp_tree): Likewise.
10897         (vect_build_slp_tree_2): Likewise.  Update uses of
10898         SLP_TREE_SCALAR_STMTS.
10899         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
10900         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
10901         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
10902         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
10903         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
10904         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
10905         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
10906         (vect_get_constant_vectors, vect_get_slp_defs)
10907         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
10908         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
10909         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
10910         instead of gimple stmts.
10911         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
10912         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
10913         (vect_slp_analyze_instance_dependence): Update uses of
10914         SLP_TREE_SCALAR_STMTS.
10915         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10916         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
10917         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
10918         (get_initial_defs_for_reduction): Likewise.
10919         (vect_create_epilog_for_reduction): Likewise.
10920         (vectorize_fold_left_reduction): Likewise.
10921         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
10922         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
10923         (can_vectorize_live_stmts): Likewise.
10925 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10927         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
10928         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10929         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
10930         than gimple stmts.
10931         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
10932         of a gimple stmt.
10933         (gather_scalar_reductions): Update after above interface changes.
10934         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
10935         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
10936         than gimple stmts.
10937         (vect_force_simple_reduction): Likewise.
10938         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
10939         LOOP_VINFO_REDUCTIONS.
10940         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10942 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10944         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
10945         a gimple stmt to a stmt_vec_info.
10946         * tree-vect-loop.c (vect_active_double_reduction_p)
10947         (vect_force_simple_reduction, vectorizable_reduction): Update
10948         accordingly.
10950 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10952         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
10953         vec<gimple *> to a vec<stmt_vec_info>.
10954         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
10955         the reduction_phis argument from a vec<gimple *> to a
10956         vec<stmt_vec_info>.
10957         (vectorizable_reduction): Likewise the phis local variable that
10958         is passed to vect_create_epilog_for_reduction.  Update for new type
10959         of SLP_TREE_VEC_STMTS.
10960         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
10961         (vectorizable_live_operation): Likewise.
10962         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
10963         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
10965 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10967         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
10968         a gimple stmt to a stmt_vec_info.
10969         (vectorizable_condition, vectorizable_live_operation)
10970         (vectorizable_reduction, vectorizable_induction): Pass back the
10971         vectorized statement as a stmt_vec_info.
10972         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
10973         use of STMT_VINFO_VEC_STMT.
10974         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
10975         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
10976         as stmt_vec_infos rather than gimple stmts.
10977         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
10978         to a stmt_vec_info.
10979         (vectorizable_live_operation): Likewise.
10980         (vectorizable_reduction, vectorizable_induction): Likewise,
10981         updating use of STMT_VINFO_VEC_STMT.
10982         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
10983         of STMT_VINFO_VEC_STMT.
10984         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
10985         (vectorizable_simd_clone_call, vectorizable_conversion)
10986         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10987         (vectorizable_store, vectorizable_load, vectorizable_condition)
10988         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
10989         from a gimple stmt to a stmt_vec_info.
10990         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
10991         pointer to a stmt_vec_info to the vectorizable_* routines.
10993 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10995         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
10996         a gimple stmt to a stmt_vec_info.
10997         (is_pattern_stmt_p): Update accordingly.
10998         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
10999         (vect_record_grouped_load_vectors): Likewise.
11000         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11001         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11002         (vect_model_reduction_cost): Likewise.
11003         (vect_create_epilog_for_reduction): Likewise.
11004         (vectorizable_reduction, vectorizable_induction): Likewise.
11005         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11006         Return the stmt_vec_info for the pattern statement.
11007         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11008         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11009         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11010         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11011         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11012         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11013         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11014         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11016 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11018         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11019         (vect_finish_stmt_generation): Likewise.
11020         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11021         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11022         (vect_build_gather_load_calls): Use the return value of the above
11023         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11024         scopes for the input gimple stmt and wider scopes for the associated
11025         stmt_vec_info.  Use vec_info::lookup_def when setting these
11026         stmt_vec_infos from an SSA_NAME definition.
11027         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11028         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11029         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11030         (vectorizable_store, vectorizable_load, vectorizable_condition)
11031         (vectorizable_comparison): Likewise.
11032         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11033         (vectorizable_reduction): Likewise.
11035 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11037         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11038         stmt_vec_info * parameter before the optional gimple **.
11039         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11040         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11041         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11042         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11043         (vectorizable_live_operation): Likewise.
11044         * tree-vect-patterns.c (type_conversion_p): Likewise.
11045         (vect_look_through_possible_promotion): Likewise.
11046         (vect_recog_rotate_pattern): Likewise.
11047         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11049 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11051         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11052         a typedef to a wrapper class.
11053         (NULL_STMT_VEC_INFO): New macro.
11054         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11055         (stmt_vec_info::operator*): New function.
11056         (stmt_vec_info::operator gimple *): Likewise.
11057         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11058         (add_stmt_costs): Likewise.
11059         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11060         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11061         (vect_get_known_peeling_cost): Likewise.
11062         (vect_estimate_min_profitable_iters): Likewise.
11063         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11064         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11065         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11066         (vectorizable_store, free_stmt_vec_infos): Likewise.
11067         (new_stmt_vec_info): Change return type of xcalloc to
11068         _stmt_vec_info *.
11070 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11072         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11073         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11074         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11075         a single_imm_use-based sequence.
11076         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11078 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11080         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11081         * tree-vectorizer.c (vec_info::lookup_def): New function.
11082         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11083         (vect_widened_op_tree): Likewise.
11084         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11085         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11086         (vectorizable_reduction): Likewise.
11087         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11088         of a gimple *.
11089         (vect_is_slp_reduction): Update calls accordingly.  Use
11090         vec_info::lookup_def.
11091         (vect_is_simple_reduction): Likewise
11092         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11094 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11096         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11097         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11098         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11099         of vinfo_for_stmt.
11100         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11101         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11102         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11103         (vect_is_slp_reduction, vectorizable_induction)
11104         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11105         * tree-vect-patterns.c (vect_init_pattern_stmt):
11106         (vect_determine_min_output_precision_1, vect_determine_precisions)
11107         (vect_pattern_recog): Likewise.
11108         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11109         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11110         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11111         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11112         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11113         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11114         info field from a loop to a loop_vec_info.
11116 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11118         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11119         (vec_info::add_stmt): Declare.
11120         * tree-vectorizer.c (vec_info::add_stmt): New function.
11121         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11122         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11123         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11124         (vectorizable_induction): Likewise.
11125         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11126         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11127         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11128         (vectorizable_load): Likewise.
11129         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11130         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11131         (vect_recog_gather_scatter_pattern): Likewise.
11132         (append_pattern_def_seq): Likewise.  Remove a check that is
11133         performed by add_stmt itself.
11135 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11137         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11138         which make_ssa_name was called with new_stmt before new_stmt
11139         had been created.
11141 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11143         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11144         split out from...
11145         (vect_is_slp_reduction): ...here...
11146         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11147         that are already known to be false.
11149 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11151         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11152         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11153         STMT_VINFO_NUM_SLP_USES when it's true.
11154         (vect_free_slp_instance): Add a final_p parameter and pass it to
11155         vect_free_slp_tree.
11156         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11157         (vect_analyze_slp_instance): Likewise.
11158         (vect_slp_analyze_operations): Likewise.
11159         (vect_slp_analyze_bb_1): Likewise.
11160         * tree-vectorizer.c (vec_info): Likewise.
11161         * tree-vect-loop.c (vect_transform_loop): Likewise.
11163 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11165         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11166         function is not called for second and subsequent members of
11167         a reduction group.
11169 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11171         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11172         cases for nested loops from here to ...
11173         (vect_create_epilog_for_reduction): ...here.  Only call
11174         vect_is_simple_use for inner-loop reductions.
11176 2018-07-31  Martin Liska  <mliska@suse.cz>
11178         PR gcov-profile/85338
11179         PR gcov-profile/85350
11180         PR gcov-profile/85372
11181         * profile.c (struct location_triplet): New.
11182         (struct location_triplet_hash): Likewise.
11183         (output_location): Do not output a BB that
11184         is already recorded for a line.
11185         (branch_prob): Use streamed_locations.
11187 2018-07-31  Martin Liska  <mliska@suse.cz>
11189         PR gcov-profile/85370
11190         * coverage.c (coverage_begin_function): Do not mark target
11191         clones as artificial functions.
11193 2018-07-31  Martin Liska  <mliska@suse.cz>
11195         PR gcov-profile/83813
11196         PR gcov-profile/84758
11197         PR gcov-profile/85217
11198         PR gcov-profile/85332
11199         * profile.c (branch_prob): Do not record GOTO expressions
11200         for GIMPLE statements which locations are already streamed.
11202 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11204         * gcc.c (handle_spec_function): Accept a soft_matched_part
11205         argument, as do_spec_1.  Pass it down to ...
11206         (eval_spec_function): Accept a soft_matched_part argument,
11207         and pass it down to ...
11208         (do_spec_2): Accept a soft_matched_part argument, and pass
11209         it down to do_spec_1.
11210         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11211         (handle_braces): Update call to handle_spec_function.
11212         (driver::set_up_specs): Update calls to do_spec_2.
11213         (compare_debug_dump_opt_spec_function): Likewise.
11214         (compare_debug_self_opt_spec_function): Likewise.
11216 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11218         * common.opt (nolibc): New option.
11219         * doc/invoke.texi (Link Options): Document it.
11220         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11221         * config/alpha/linux.h: Likewise.
11222         * config/arc/elf.h: Likewise.
11223         * config/arm/uclinux-elf.h: Likewise.
11224         * config/arm/unknown-elf.h: Likewise.
11225         * config/avr/avrlibc.h: Likewise.
11226         * config/bfin/bfin.h: Likewise.
11227         * config/bfin/linux.h: Likewise.
11228         * config/bfin/uclinux.h: Likewise.
11229         * config/darwin.h: Likewise.
11230         * config/darwin10.h: Likewise.
11231         * config/darwin12.h: Likewise.
11232         * config/gnu-user.h: Likewise.
11233         * config/lm32/uclinux-elf.h: Likewise.
11234         * config/pa/pa-hpux11.h: Likewise.
11235         * config/pa/pa64-hpux.h: Likewise.
11236         * config/sparc/sparc.h: Likewise.
11238 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11240         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11241         undefined variables.
11243 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11245         PR target/86640
11246         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11247         instead of GEN_INT.
11249 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11251         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11252         terminated string literal.
11254 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11256         PR rtl-optimization/85160
11257         * combine.c (is_just_move): New function.
11258         (try_combine): Allow combining two instructions into two if neither of
11259         the original instructions was a move.
11261 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11263         PR target/86673
11264         * doc/extend.texi (Global Register Variables): Discourage use of type
11265         qualifiers.
11266         (Local Register Variables): Likewise.
11268 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11270         PR tree-optimization/86506
11271         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11273 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11275         PR target/86547
11276         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11277         hard_regno, make sure no insn between `from` and `to` clobbers it.
11279 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11280             Tom de Vries  <tdevries@suse.de>
11282         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11283         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11284         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11285         PTX_DEFAULT_RUNTIME_DIM.
11286         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11288 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11290         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11291         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11292         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11294 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11296         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11297         used.
11298         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11299         predicate to reflect toc fusion has been deleted.
11300         (toc_fusion_mem_raw): Delete, no longer used.
11301         (toc_fusion_mem_wrapped): Likewise.
11302         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11303         fusion mask bit.
11304         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11305         Delete, no longer used.
11306         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11307         meant to be used for toc fusion.
11308         (rs6000_debug_print_mode): Delete toc fusion debugging.
11309         (rs6000_debug_reg_global): Likewise.
11310         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11311         fusion and secondary reload support that were never used.
11312         (rs6000_option_override_internal): Delete TOC fusion, that was only
11313         partially defined, and it did not work unless you also used the
11314         -mcmodel= switch.
11315         (rs6000_legitimate_address_p): Delete TOC fusion support.
11316         (rs6000_opt_masks): Likewise.
11317         (fusion_wrap_memory_address): Delete function, no longer used.
11318         (fusion_split_address); Delete TOC fusion support.
11319         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11320         longer used with toc fusion being deleted.
11321         (TARGET_TOC_FUSION_FP): Likewise.
11322         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11323         UNSPEC.
11324         (toc fusion spliter): Delete TOC fusion support.
11325         (toc_fusionload_<mode>): Likewise.
11326         (toc_fusionload_di): Likewise.
11327         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11328         longer needs to be named.  Rename predicate to delete TOC fusion.
11329         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11330         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11331         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11332         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11333         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11335 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11337         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11338         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11339         __int128 in built-in function prototypes.
11340         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11341         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11343 2018-07-27  Martin Sebor  <msebor@redhat.com>
11345         PR tree-optimization/86696
11346         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11347         types, including enums.
11348         (handle_char_store): Be prepared for the above function to fail.
11350 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11352         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11353         when optimization level is lower than 2 or optimize for size.
11355 2018-07-26  Martin Sebor  <msebor@redhat.com>
11357         PR tree-optimization/86043
11358         PR tree-optimization/86042
11359         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11360         (get_string_cst_length): Rename...
11361         (get_min_string_length): ...to this.  Add argument.
11362         (handle_char_store): Extend to handle multi-character stores by
11363         MEM_REF.
11364         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11365         * tree.h (initializer_zerop): Add argument.
11367 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11369         PR middle-end/86660
11370         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11371         declare target to variables if they have always,{to,from,tofrom} map
11372         kinds.
11374 2018-07-26  Martin Liska  <mliska@suse.cz>
11376         PR lto/86548
11377         * lto-wrapper.c: Add linker_output as prefix
11378         for ltrans_output_file.
11380 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11382         PR rtl-optimization/85805
11383         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11384         value for hard registers if that was written in the same mode.
11386 2018-07-26  Martin Liska  <mliska@suse.cz>
11388         PR gcov-profile/86536
11389         * gcov.c (format_gcov): Use printf format %.*f directly
11390         and do not handle special values.
11392 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11394         * common/config/arc/arc-common.c (arc_option_optimization_table):
11395         Update default optimizations for size.
11397 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11399         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11400         * config/arc/constraints.md (CfZ): New constraint.
11401         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11402         (subsf3_fpu): Likewise.
11403         (cmpsf_fpu): Likewise.
11404         (cmpsf_fpu_uneq): Likewise.
11406 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11408         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11409         accesses as well.
11410         (arc_is_uncached_mem_p): uncached applies to both the variable and
11411         the pointer.
11413 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11415         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11416         register names.
11418 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11420         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11421         field "m_scopes" from vec to auto_vec.
11423 2018-07-25  Martin Liska  <mliska@suse.cz>
11425         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11426         return type.
11428 2018-07-25  Richard Biener  <rguenther@suse.de>
11430         PR debug/86654
11431         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11432         special wrt context_die late.
11433         (gen_subprogram_die): Re-use DIEs in local scope.
11435 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11437         PR tree-optimization/86644
11438         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11440 2018-07-25  Martin Liska  <mliska@suse.cz>
11442         PR middle-end/86645
11443         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11444         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11446 2018-07-25  Martin Liska  <mliska@suse.cz>
11448         PR sanitizer/79635
11449         * params.def: Explain ASan abbreviation and provide
11450         a documentation link.
11452 2018-07-24  Martin Sebor  <msebor@redhat.com>
11454         PR tree-optimization/86622
11455         PR tree-optimization/86532
11456         * builtins.h (string_length): Declare.
11457         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11458         (check_access): Be prepared for non-constant length ranges.
11459         (string_length): Make extern.
11460         * expr.c (string_constant): Only handle the minor non-constant
11461         array index.  Use string_constant to compute the length of
11462         a generic string constant.
11464 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11466         PR tree-optimization/86618
11467         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11468         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11470 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11472         PR tree-optimization/86636
11473         * json.cc (json::object::set): Fix comment.  Add assertions.
11474         (json::array::append): Move here from json.h.  Add comment and an
11475         assertion.
11476         (json::string::string): Likewise.
11477         * json.h (json::array::append): Move to json.cc.
11478         (json::string::string): Likewise.
11479         * optinfo-emit-json.cc
11480         (optrecord_json_writer::impl_location_to_json): Assert that we
11481         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11482         wrapper around it.  Expand the location once, rather than three
11483         times.
11484         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11485         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11486         wrappers.
11487         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11488         places.  Fix some overlong lines.
11490 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11492         * config/aarch64/aarch64-simd.md
11493         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11494         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11495         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11496         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11497         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11498         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11499         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11500         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
11501         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
11503 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
11505         PR middle-end/86627
11506         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
11507         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
11508         and abs_d == d, do the power of two handling if profitable.
11510 2018-07-24  Richard Biener  <rguenther@suse.de>
11512         * match.pd: Add BIT_FIELD_REF canonicalizations.
11514 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11516         PR c/86617
11517         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
11519 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11521         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
11522         terminated STRING_CST object.
11524 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11526         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
11528 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11530         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
11531         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
11532         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
11533         the elements into a register.
11534         (rs6000_split_v4si_init_di_reg): Delete.
11535         (rs6000_split_v4si_init): Delete.
11536         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
11537         (vsx_init_v4si): Rewrite as a define_expand.
11539 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11541         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
11542         zero_extend argument from memory): New.
11544 2018-07-22  Martin Sebor  <msebor@redhat.com>
11546         PR bootstrap/86621
11547         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
11548         diagnosing calls with unknown arguments unless -Walloca-larger-than
11549         is restricted to less than PTRDIFF_MAX bytes.
11551 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
11553         * doc/gcov.texi (Invoking Gcov): Editorial changes.
11555 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11557         * pretty-print.c (text_info::set_location): Remove redundant
11558         "line_table" parameter from call to rich_location::set_range.
11560 2018-07-20  Martin Sebor  <msebor@redhat.com>
11562         PR middle-end/82063
11563         * builtins.c (expand_builtin_alloca): Adjust.
11564         * calls.c (alloc_max_size): Simplify.
11565         * cgraphunit.c (cgraph_node::expand): Adjust.
11566         * common.opt (larger_than_size, warn_frame_larger_than): Remove
11567         variables.
11568         (frame_larger_than_size): Same.
11569         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
11570         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
11571         Initialize.
11572         * doc/invoke.texi (GCC Command Options): Document option arguments.
11573         Explain byte-size arguments and suffixes.
11574         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
11575         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
11576         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
11577         * doc/options.texi (UInteger): Expand.
11578         (Host_Wide_Int, ByteSize): Document new properties.
11579         * final.c (final_start_function_1): Include sizes in an error message.
11580         * function.c (frame_offset_overflow): Same.
11581         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
11582         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
11583         Diagnose unbounded alloca calls only for limits of less than
11584         PTRDIFF_MAX.
11585         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
11586         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
11587         for alloca(0).
11588         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
11589         only for limits of less than PTRDIFF_MAX.
11590         * langhooks-def.h (lhd_handle_option): Change function argument
11591         to HOST_WIDE_INT.
11592         * langhooks.c (lhd_handle_option): Same.
11593         * langhooks.h (handle_option): Same.
11594         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
11595         ByteSize flags.
11596         (var_type, var_type_struct): Same.
11597         (var_set): Handle ByteSize flag.
11598         * optc-gen.awk: Add comments to output to ease debugging.  Make
11599         use of HOST_WIDE_INT where appropriate.
11600         * opts-gen-save.awk:  Use %lx to format unsigned long.
11601         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
11602         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
11603         arguments.  Parse bytes-size suffixes.
11604         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
11605         (enum_value_to_arg): Same.
11606         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
11607         (handle_option): Adjust.
11608         (generate_option): Change function argument to HOST_WIDE_INT.
11609         (cmdline_handle_error): Adjust.
11610         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
11611         (set_option): Change function argument to HOST_WIDE_INT.
11612         (option_enabled): Handle cl_host_wide_int.
11613         (get_option_state): Handle CLVC_SIZE.
11614         (control_warning_option): Same.
11615         * opts.c (common_handle_option): Change function argument to
11616         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
11617         OPT_Wvla_larger_than_.
11618         * opts.h (enum cl_var_type): Add an enumerator.
11619         * stor-layout.c (layout_decl): Print a more meaningful warning.
11620         * toplev.c (output_stack_usage): Adjust.
11622 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
11624         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
11625         call to inline_expand_builtin_string_cmp.
11626         (expand_builtin_strcmp): Likewise.
11627         (expand_builtin_strncmp): Likewise.
11628         (inline_string_cmp): Delete the last parameter, change char_type_node
11629         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
11630         two operands.
11631         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
11632         the inlining expansion on target where the type of the call has same or
11633         narrower precision than unsigned char.
11635 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11637         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
11638         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
11639         * common.opt (fsave-optimization-record): New option.
11640         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
11641         * doc/invoke.texi (-fsave-optimization-record): New option.
11642         * dumpfile.c: Include "optinfo-emit-json.h".
11643         (struct kv_pair): Move to coretypes.h.
11644         (optgroup_options): Make non-static.
11645         (dump_context::end_scope): Call
11646         optimization_records_maybe_pop_dump_scope.
11647         * dumpfile.h (optgroup_options): New decl.
11648         * json.cc: New file.
11649         * json.h: New file.
11650         * optinfo-emit-json.cc: New file.
11651         * optinfo-emit-json.h: New file.
11652         * optinfo.cc: Include "optinfo-emit-json.h".
11653         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
11654         (optinfo_enabled_p): Check optimization_records_enabled_p.
11655         (optinfo_wants_inlining_info_p): Likewise.
11656         * optinfo.h: Update comment.
11657         * profile-count.c (profile_quality_as_string): New function.
11658         * profile-count.h (profile_quality_as_string): New decl.
11659         (profile_count::quality): New accessor.
11660         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
11661         and optinfo_emit_json_cc_tests.
11662         * selftest.h (selftest::json_cc_tests): New decl.
11663         (selftest::optinfo_emit_json_cc_tests): New decl.
11664         * toplev.c: Include "optinfo-emit-json.h".
11665         (compile_file): Call optimization_records_finish.
11666         (do_compile): Call optimization_records_start.
11667         * tree-ssa-live.c: Include optinfo.h.
11668         (remove_unused_scope_block_p): Retain inlining information if
11669         optinfo_wants_inlining_info_p returns true.
11671 2018-07-20  Richard Biener  <rguenther@suse.de>
11673         PR debug/86585
11674         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
11675         to cover -flto-partition=none.
11677 2018-07-20  Martin Liska  <mliska@suse.cz>
11679         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
11680         (get_decl_source_range): Remove unused function.
11682 2018-07-20  Richard Biener  <rguenther@suse.de>
11684         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
11685         (struct vn_phi_s): Likewise.
11686         (struct vn_reference_s): Likewise.
11687         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
11688         for searching the slot of an entry known to be in the hash itself.
11689         (vn_phi_hasher::equal): Likewise.
11690         (vn_reference_hasher::equal): Likewise.
11691         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
11692         globals.
11693         (optimistic_info, current_info): Remove, keeping only valid_info.
11694         (vn_reference_lookup_1): Remove fallback lookup.
11695         (vn_reference_lookup_2): Likewise.
11696         (vn_nary_op_lookup_1): Likewise.
11697         (vn_phi_lookup): Likewise.
11698         (vn_nary_build_or_lookup_1): Make sure to not chain the built
11699         hash element.
11700         (vn_reference_insert): Adjust, chain the inserted hash element
11701         at last_inserted_ref.
11702         (vn_reference_insert_pieces): Likewise.
11703         (visit_reference_op_call): Likewise.
11704         (vn_nary_op_insert_into): Chain the inserted hash element at
11705         last_inserted_nary.
11706         (vn_nary_op_insert_pieces): Adjust.
11707         (vn_nary_op_insert): Likewise.
11708         (vn_nary_op_insert_stmt): Likewise.
11709         (vn_phi_insert): Adjust, chain the inserted hash element at
11710         last_inserted_phi.
11711         (process_scc): Remove clearing and copying the optimistic
11712         table.  Instead remove elements inserted during an optimistic
11713         iteration from the single table we maintain.
11714         (init_scc_vn): Adjust.
11715         (free_scc_vn): Likewise.
11716         (sccvn_dom_walker::record_cond): Likewise.
11717         (sccvn_dom_walker::after_dom_children): Likewise.
11719 2018-07-19  Martin Sebor  <msebor@redhat.com>
11721         PR tree-optimization/84047
11722         PR tree-optimization/83776
11723         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
11724         (check_array_bounds): Call it.
11726 2018-07-19  Martin Sebor  <msebor@redhat.com>
11728         * align.h (align_flags): Use member initialization.
11730 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
11732         * Makefile.in (OBJS): Add optinfo.o.
11733         * coretypes.h (class symtab_node): New forward decl.
11734         (struct cgraph_node): New forward decl.
11735         (class varpool_node): New forward decl.
11736         * dump-context.h: New file.
11737         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
11738         "tree-pass.h".
11739         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
11740         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
11741         (set_alt_dump_file): Likewise.
11742         (dump_context::~dump_context): New dtor.
11743         (dump_gimple_stmt): Move implementation to...
11744         (dump_context::dump_gimple_stmt): ...this new member function.
11745         Add the stmt to any pending optinfo, creating one if need be.
11746         (dump_gimple_stmt_loc): Move implementation to...
11747         (dump_context::dump_gimple_stmt_loc): ...this new member function.
11748         Start a new optinfo and add the stmt to it.
11749         (dump_gimple_expr): Move implementation to...
11750         (dump_context::dump_gimple_expr): ...this new member function.
11751         Add the stmt to any pending optinfo, creating one if need be.
11752         (dump_gimple_expr_loc): Move implementation to...
11753         (dump_context::dump_gimple_expr_loc): ...this new member function.
11754         Start a new optinfo and add the stmt to it.
11755         (dump_generic_expr): Move implementation to...
11756         (dump_context::dump_generic_expr): ...this new member function.
11757         Add the tree to any pending optinfo, creating one if need be.
11758         (dump_generic_expr_loc): Move implementation to...
11759         (dump_context::dump_generic_expr_loc): ...this new member
11760         function.  Add the tree to any pending optinfo, creating one if
11761         need be.
11762         (dump_printf): Move implementation to...
11763         (dump_context::dump_printf_va): ...this new member function.  Add
11764         the text to any pending optinfo, creating one if need be.
11765         (dump_printf_loc): Move implementation to...
11766         (dump_context::dump_printf_loc_va): ...this new member function.
11767         Start a new optinfo and add the stmt to it.
11768         (dump_dec): Move implementation to...
11769         (dump_context::dump_dec): ...this new member function.  Add the
11770         value to any pending optinfo, creating one if need be.
11771         (dump_context::dump_symtab_node): New member function.
11772         (dump_context::get_scope_depth): New member function.
11773         (dump_context::begin_scope): New member function.
11774         (dump_context::end_scope): New member function.
11775         (dump_context::ensure_pending_optinfo): New member function.
11776         (dump_context::begin_next_optinfo): New member function.
11777         (dump_context::end_any_optinfo): New member function.
11778         (dump_context::s_current): New global.
11779         (dump_context::s_default): New global.
11780         (dump_scope_depth): Delete global.
11781         (dumpfile_ensure_any_optinfo_are_flushed): New function.
11782         (dump_symtab_node): New function.
11783         (get_dump_scope_depth): Reimplement in terms of dump_context.
11784         (dump_begin_scope): Likewise.
11785         (dump_end_scope): Likewise.
11786         (selftest::temp_dump_context::temp_dump_context): New ctor.
11787         (selftest::temp_dump_context::~temp_dump_context): New dtor.
11788         (selftest::verify_item): New function.
11789         (ASSERT_IS_TEXT): New macro.
11790         (ASSERT_IS_TREE): New macro.
11791         (ASSERT_IS_GIMPLE): New macro.
11792         (selftest::test_capture_of_dump_calls): New test.
11793         (selftest::dumpfile_c_tests): Call it.
11794         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
11795         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
11796         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
11797         descriptive comment.
11798         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
11799         (dump_node, dump_bb): Move these unrelated decls.
11800         (class dump_manager): Add leading comment.
11801         * optinfo.cc: New file.
11802         * optinfo.h: New file.
11804 2018-07-19  Michael Collison  <michael.collison@arm.com>
11805             Richard Henderson <rth@redhat.com>
11807         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
11808         (subti): Handle op1 zero.
11809         (subvti4, usub4ti4): New.
11810         (*sub<GPI>3_compare1_imm): New.
11811         (sub<GPI>3_carryinCV): New.
11812         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
11813         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
11815 2018-07-19  Michael Collison  <michael.collison@arm.com>
11816             Richard Henderson <rth@redhat.com>
11818         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
11819         (addti3): Create simpler code if low part is already known to be 0.
11820         (addvti4, uaddvti4): New.
11821         (*add<GPI>3_compareC_cconly_imm): New.
11822         (*add<GPI>3_compareC_cconly): New.
11823         (*add<GPI>3_compareC_imm): New.
11824         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
11825         handle constants within this pattern..
11826         (*add<GPI>3_compareV_cconly_imm): New.
11827         (*add<GPI>3_compareV_cconly): New.
11828         (*add<GPI>3_compareV_imm): New.
11829         (add<GPI>3_compareV): New.
11830         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
11831         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
11832         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
11833         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
11834         with 'comparison' operator.
11835         (*add<GPI>3_compareV_cconly_imm): Ditto.
11836         (*add<GPI>3_compareV_cconly): Ditto.
11837         (*add<GPI>3_compareV_imm): Ditto.
11838         (add<GPI>3_compareV): Ditto.
11839         (add<mode>3_carryinC): Ditto.
11840         (*add<mode>3_carryinC_zero): Ditto.
11841         (*add<mode>3_carryinC): Ditto.
11842         (add<mode>3_carryinV): Ditto.
11843         (*add<mode>3_carryinV_zero): Ditto.
11844         (*add<mode>3_carryinV): Ditto.
11846 2018-07-19  Michael Collison  <michael.collison@arm.com>
11847             Richard Henderson <rth@redhat.com>
11849         * config/aarch64/aarch64-modes.def (CC_V): New.
11850         * config/aarch64/aarch64-protos.h
11851         (aarch64_addti_scratch_regs): Declare
11852         (aarch64_subvti_scratch_regs): Declare.
11853         (aarch64_expand_subvti): Declare.
11854         (aarch64_gen_unlikely_cbranch): Declare
11855         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
11856         for signed overflow using CC_Vmode.
11857         (aarch64_get_condition_code_1): Handle CC_Vmode.
11858         (aarch64_gen_unlikely_cbranch): New function.
11859         (aarch64_addti_scratch_regs): New function.
11860         (aarch64_subvti_scratch_regs): New function.
11861         (aarch64_expand_subvti): New function.
11863 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11865         * config/aarch64/aarch64-option-extensions.def: New entry for profile
11866         extension.
11867         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
11868         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
11869         extension.
11871 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11873         PR target/83009
11874         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11875         address check not strict.
11877 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11879         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
11880         Umq with Umn.
11881         (store_pair_lanes<mode>): Likewise.
11882         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
11883         enum value 'ADDR_QUERY_LDP_STP_N'.
11884         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
11885         (aarch64_print_address_internal): Add declaration.
11886         (aarch64_print_ldpstp_address): Remove.
11887         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
11888         (aarch64_print_operand): Change printing of 'y'.
11889         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
11890         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
11891         'true' rather than '1'.
11892         * config/aarch64/constraints.md (Uml): Likewise.
11893         (Uml): Rename to Umn.
11894         (Umq): Remove.
11896 2018-07-19  Richard Biener  <rguenther@suse.de>
11898         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
11899         a trailing array.
11900         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
11901         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
11902         (vn_reference_hasher): Likewise.
11903         (struct vn_tables_s): Remove obstack and alloc-pool members.
11904         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
11905         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
11906         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
11907         (vn_reference_insert_pieces): Likewise.
11908         (alloc_vn_nary_op_noinit): Adjust.
11909         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
11910         (vn_phi_eq): Adjust.
11911         (shared_lookup_phiargs): Remove.
11912         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
11913         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
11914         (visit_reference_op_call): Likewise.
11915         (copy_nary, copy_phi, copy_reference): Remove.
11916         (process_scc): Rewind the obstack when iterating.  Do not
11917         copy the elements to valid_info but just move them from one
11918         hashtable to the other.
11919         (allocate_vn_table): Adjust.
11920         (free_vn_table): Likewise.
11921         (init_scc_vn): Likewise.
11922         (free_scc_vn): Likewise.
11924 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
11926         PR target/86560
11927         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
11928         indirect_return as function type attribute.
11929         (ix86_attribute_table): Change indirect_return to function
11930         type attribute.
11931         * doc/extend.texi: Update indirect_return attribute.
11933 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
11935         * wide-int.h (widest2_int): New.
11936         * gimple-fold.c (arith_overflowed_p): Use it.
11937         * tree.h (widest2_int_cst): New.
11938         * tree-vrp.c (wide_int_binop_overflow): Rename from
11939         vrp_int_const_binop.
11940         Rewrite to work on trees.
11941         (extract_range_from_multiplicative_op_1): Abstract code to...
11942         (wide_int_range_min_max): ...here.
11943         (wide_int_range_cross_product): ...and here.
11944         (extract_range_from_binary_expr_1): Abstract overflow code to...
11945         (wide_int_range_mult_wrapping): ...here.
11946         * tree-vrp.h (wide_int_range_cross_product): New.
11947         (wide_int_range_mult_wrapping): New.
11949 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
11950             Julia Koval  <julia.koval@intel.com>
11952         * config/i386/x86-tune-costs.h (skylake_memcpy,
11953         skylake_memset): Replace rep_prefix with unrolling for size 512.
11955 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
11957         PR middle-end/86544
11958         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
11959         comparision with EQ_EXPR in last stmt.
11961 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11963         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
11964         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
11965         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
11966         previously known as "PowerPC AltiVec Built-in Functions".  Move
11967         some material to new subsubsections "PowerPC AltiVec Built-in
11968         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
11969         ISA 2.07".
11970         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
11971         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
11972         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
11973         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
11975 2018-07-18  Richard Biener  <rguenther@suse.de>
11977         PR tree-optimization/86557
11978         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
11979         EXACT_DIV_EXPR.
11981 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
11983         * config/s390/s390.c (s390_function_profiler): Generate CFI.
11985 2018-07-17  Jeff Law  <law@redhat.com>
11987         * config/arm/arm.c (get_label_padding): Update for recent
11988         changes to label_to_alignment.
11990         PR tree-optimization/86010
11991         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
11993         * config/mips/mips.c (vr4130_align_insns): Update for recent
11994         changes to label_to_alignment.
11996         * config/frv/frv.c (frv_label_align): Update for recent changes
11997         to label_to_alignment.
11999         * config/nios2/nios2.c (nios2_label_align): Update for recent
12000         changes which dropped ALIGN_LABELS_LOG.
12002 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12004         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12005         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12007 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12009         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12010         deprecated align_labels_log.
12012 2018-07-17  Richard Biener  <rguenther@suse.de>
12014         PR lto/86456
12015         * dwarf2out.c (init_sections_and_labels): Always generate
12016         a debug_line_str_section for early LTO debug.
12017         (dwarf2out_finish): Reset debug_line_str_hash output early.
12018         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12019         with fat LTO part.
12020         (dwarf2out_early_finish): Output debug_line_str.
12022 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12024         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12025         index register on z196 or later.
12027 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12029         * config/s390/s390.c (s390_default_align): Set default function
12030         alignment to 16.
12031         (s390_override_options_after_change): Call s390_default align.
12032         (s390_option_override_internal): Call s390_default_align.
12033         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12035 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12037         PR middle-end/86542
12038         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12039         to _looptemp_ clauses, other than the first two.
12041 2018-07-17  Martin Liska  <mliska@suse.cz>
12043         * opts.c: Do not enable OPT_falign_* for -Os.
12045 2018-07-17  Martin Liska  <mliska@suse.cz>
12047         * align.h (MAX_CODE_ALIGN): New.
12048         (MAX_CODE_ALIGN_VALUE): New.
12049         * common/config/i386/i386-common.c (ix86_handle_option):
12050         (MAX_CODE_ALIGN): Moved to align.h.
12051         * final.c (MAX_CODE_ALIGN): Likewise.
12052         * opts.c (parse_and_check_align_values):
12053         (MAX_CODE_ALIGN): Likewise.
12054         (MAX_CODE_ALIGN_VALUE): Likewise.
12056 2018-07-17  Martin Liska  <mliska@suse.cz>
12058         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12059         in order to fulfil coding style.
12060         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12061         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12062         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12063         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12064         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12065         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12066         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12067         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12068         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12070 2018-07-17  Martin Liska  <mliska@suse.cz>
12072         * align.h: New file.
12073         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12074         directly.
12075         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12076         align_flags of label_to_alignment.
12077         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12078         align_flags class.
12079         * config/m68k/m68k.c: Do not use removed align_labels_value and
12080         align_loops_value.
12081         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12082         (LOOP_ALIGN): Likewise.
12083         (LABEL_ALIGN): Likewise.
12084         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12085         Remove not used macro.
12086         (rs6000_loop_align): Change return type to align_flags.
12087         (rs6000_loop_align_max_skip): Remove.
12088         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12089         Change return type to align_flags.
12090         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12091         Remove not used macro.
12092         (rs6000_loop_align):  Change return type to align_flags.
12093         (rs6000_loop_align_max_skip): Remove.
12094         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12095         * config/rx/rx-protos.h (rx_align_for_label): Make it
12096         static function.
12097         * config/rx/rx.c (rx_align_for_label): Change return type
12098         to align_flags.
12099         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12100         macro definitions.
12101         into align_flags class.
12102         (LABEL_ALIGN): Likewise.
12103         (LOOP_ALIGN): Likewise.
12104         * config/s390/s390.c (s390_label_align): Use align_flags
12105         class member.
12106         (s390_asm_output_function_label): Likewise.
12107         * config/sh/sh.c (sh_override_options_after_change):
12108         Use align_flags class directly without macros.
12109         (find_barrier): Likewise.
12110         (barrier_align): Likewise.
12111         (sh_loop_align): Likewise.
12112         * config/spu/spu.c (spu_option_override):
12113         Use align_flags_tuple::get_value instead of removed macros.
12114         (spu_sched_init): Likewise.
12115         * config/spu/spu.h (GTY): Likewise.
12116         * config/visium/visium.c (visium_option_override):
12117         Set "8" as default secondary alignment.
12118         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12119         in order to guarantee secondary alignment of 8.
12120         * coretypes.h: Include align.h header file.
12121         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12122         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12123         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12124         * doc/tm.texi.in: Likewise.
12125         * final.c (struct label_alignment): Remove not used structure.
12126         (LABEL_ALIGN): Change type to align_flags.
12127         (LOOP_ALIGN): Likewise.
12128         (JUMP_ALIGN): Likewise.
12129         (default_loop_align_max_skip): Remove.
12130         (default_label_align_max_skip): Likewise.
12131         (default_jump_align_max_skip): Likewise.
12132         (default_label_align_after_barrier_max_skip):
12133         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12134         (LABEL_TO_MAX_SKIP): Remove.
12135         (label_to_alignment): Return align_flags type instead of integer.
12136         (label_to_max_skip): Remove.
12137         (align_fuzz): Use align_flags type.
12138         (compute_alignments): Use align_flags type and use align_flags::max
12139         to combine multiple alignments.
12140         (grow_label_align): Grow vec instead of C array.
12141         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12142         (shorten_branches):  Use align_flags type and use align_flags::max
12143         to combine multiple alignments.
12144         (final_scan_insn_1): Remove usage of secondary alignment that comes
12145         from label alignment, but instead use proper secondary alignment
12146         which is computed in grow_label_align.
12147         * flags.h (struct align_flags_tuple): Move to align.h.
12148         (struct align_flags): Likewise.
12149         (state_align_loops): Rename to align_loops.
12150         (state_align_jumps): Rename to align_jumps.
12151         (state_align_labels): Rename to align_labels.
12152         (state_align_functions): Rename to align_functions.
12153         (align_loops_log): Remove.
12154         (align_jumps_log): Remove.
12155         (align_labels_log): Remove.
12156         (align_functions_log): Remove.
12157         (align_loops_max_skip): Remove.
12158         (align_jumps_max_skip): Remove.
12159         (align_labels_max_skip): Remove.
12160         (align_functions_max_skip): Remove.
12161         (align_loops_value): Remove.
12162         (align_jumps_value): Remove.
12163         (align_labels_value): Remove.
12164         (align_functions_value): Remove.
12165         * output.h (label_to_alignment): Change return type to align_flags.
12166         (label_to_max_skip): Remove.
12167         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12168         jump_align_max_skip macros.
12169         * targhooks.h (default_loop_align_max_skip): Remove.
12170         (default_label_align_max_skip): Likewise.
12171         (default_jump_align_max_skip): Likewise.
12172         (default_label_align_after_barrier_max_skip): Remove.
12173         * toplev.c (read_log_maxskip): Use ::normalize function.
12174         (parse_N_M): Remove not used argument and also call ::normalize.
12175         (parse_alignment_opts): Do not pass unused arguments.
12176         * varasm.c (assemble_start_function): Use directly align_functions
12177         instead of removed macros.
12178         * system.h: Do not poison removed macros.
12180 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12182         PR middle-end/86539
12183         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12184         and cond temporaries don't have reference type if iterator has
12185         pointer type.  For init use &for_pre_body instead of pre_p if
12186         for_pre_body is non-empty.
12188 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12190         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12191         double-double modes to SFmode directly directly.
12192         (trunc<mode>sf2_fprs): Delete.
12194 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12196         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12197         for conversions between IFmode and the decimal floating point modes.
12198         (init_float128_ieee): Use the correct names for conversions between
12199         KFmode and the decimal floating point modes.
12201 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12203         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12204         for the conversions between TDmode and IFmode.
12205         (init_float128_ieee): Use more correct names for the conversions
12206         between TDmode and KFmode.
12208 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12210         PR tree-optimization/86526
12211         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12212         (inline_expand_builtin_string_cmp): Likewise.
12213         (inline_string_cmp): Likewise.  Use c_readstr instead of
12214         builtin_memcpy_read_str.  Add unit_mode temporary.
12216 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12218         PR middle-end/86528
12219         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12220         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12222 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12224         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12225         Alphabetize prototypes of built-in functions, separating out
12226         built-in functions that are listed in this section but should be
12227         described elsewhere.
12229 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12231         PR target/86511
12232         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12233         conditional move when trapping comparison was split to a
12234         non-trapping one (and vice versa).
12236 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12238         * config/s390/s390.c (s390_function_profiler): Generate nops
12239         instead of profiler call sequences.
12240         * config/s390/s390.opt: Add the new option.
12242 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12244         * config/s390/s390.c (s390_function_profiler): Generate
12245         __mcount_loc section.
12246         * config/s390/s390.opt: Add the new option.
12248 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12250         * common.opt: Add the new warning.
12251         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12252         %r0,__fentry__" when -mfentry is specified.
12253         (s390_option_override_internal): Disallow -mfentry for 31-bit
12254         CPUs.
12255         * config/s390/s390.opt: Add the new option.
12257 2018-07-16  Richard Biener  <rguenther@suse.de>
12259         PR lto/86523
12260         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12261         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12263 2018-07-16  Martin Liska  <mliska@suse.cz>
12265         PR ipa/86529
12266         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12267         to ::get_create.
12269 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12271         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12273 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12275         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12276         for ARCHS4x.
12277         * config/arc/arc-cpus.def (hs4x): New cpu.
12278         (hs4xd): Likewise.
12279         * config/arc/arc-tables.opt: Regenerate.
12280         * config/arc/arc.c (arc_sched_issue_rate): New function.
12281         (TARGET_SCHED_ISSUE_RATE): Define.
12282         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12283         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12284         fpu_cvt.
12285         (attr tune): Add ARCHS4x tune values.
12286         (attr tune_dspmpy): Define.
12287         (*tst): Correct instruction type.
12288         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12289         * config/arc/arcHS4x.md: New file.
12290         * config/arc/fpu.md: Update instruction type attributes.
12291         * config/arc/t-multilib: Regenerate.
12293 2018-07-16  Tom de Vries  <tdevries@suse.de>
12295         PR debug/86455
12296         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12298 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12300         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12301         early clobber.
12303 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12305         PR tree-optimization/86514
12306         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12307         conversion to a boolean type from a type with greater precision.
12309 2018-07-16  Tom de Vries  <tdevries@suse.de>
12311         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12312         dump_flags request TDF_SLIM.
12314 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12316         * fold-const.c (int_const_binop_1): Abstract...
12317         (wide_int_binop): ...wide int code here.
12318         (poly_int_binop): ...poly int code here.
12319         Abstract the rest of int_const_binop_1 into int_const_binop.
12320         * fold-const.h (wide_int_binop): New.
12321         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12322         Remove useless PLUS/MINUS_EXPR case.
12323         (zero_nonzero_bits_from_vr): Move wide int code...
12324         (zero_nonzero_bits_from_bounds): ...here.
12325         (extract_range_from_binary_expr_1): Move mask optimization code...
12326         (range_easy_mask_min_max): ...here.
12327         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12328         (range_easy_mask_min_max): New.
12330 2018-07-15  Jeff Law  <law@redhat.com>
12332         PR target/85993
12333         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12334         block.
12336 2018-07-14  Jim Wilson  <jimw@sifive.com>
12338         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12340 2018-07-14  Paul Koning  <ni1d@arrl.net>
12342         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12344 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12346         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12347         copying section.
12349 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12350             Steve Munroe  <munroesj52@gmail.com>
12352         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12353         (_mm_andnot_si128): Likewise.
12354         (_mm_or_si128): Likewise.
12355         (_mm_xor_si128): Likewise.
12357 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12359         PR middle-end/78809
12360         * builtins.c (expand_builtin_memcmp): Inline the calls first
12361         when result_eq is false.
12362         (expand_builtin_strcmp): Inline the calls first.
12363         (expand_builtin_strncmp): Likewise.
12364         (inline_string_cmp): New routine. Expand a string compare
12365         call by using a sequence of char comparison.
12366         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12367         a call to str(n)cmp/memcmp.
12368         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12369         New option.
12370         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12372 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12374         * config/arm/driver-arm.c: Include arm-native.h.
12375         (host_detect_local_cpu): Use auto-generated data tables.
12376         (vendors, arm_cpu_table): Delete.  Move part information to ...
12377         * config/arm/arm-cpus.in: ... here.
12378         * config/arm/parsecpu.awk (gen_native): New function.
12379         (vendor, part): New CPU fields.
12380         (END): Add support for building the native CPU detection tables.
12381         * config/arm/t-arm (arm-native.h): Add build rule.
12382         (driver-arm.o): Add dependency on arm-native.h.
12384 2018-07-13  Richard Biener  <rguenther@suse.de>
12386         PR middle-end/85974
12387         * match.pd (addr1 - addr2): Allow either of the operand to
12388         have a conversion.
12390 2018-07-13  Tom de Vries  <tdevries@suse.de>
12392         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12393         in remap_ssa_name.
12395 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12397         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12398         arrays instead of numbered variables.
12400 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12402         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12403         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12405 2018-07-13  Richard Biener  <rguenther@suse.de>
12407         PR debug/86452
12408         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12409         instead of get_context_die.
12411 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12412             Richard Biener  <rguenther@suse.de>
12414         PR middle-end/86489
12415         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12416         that the loop latch destination where phi is defined.
12418 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12420         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12421         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12422         (riscv_set_current_function): Extract getting interrupt type to new
12423         function.
12424         (riscv_get_interrupt_type): New function.
12425         (riscv_merge_decl_attributes): New function, checking interrupt type is
12426         same.
12427         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12429 2018-07-12  Paul Koning  <ni1d@arrl.net>
12431         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12432         directive.
12434 2018-07-12  Paul Koning  <ni1d@arrl.net>
12436         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12437         zero reference, add doloop_end instead.
12438         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12439         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12440         detail for doloop_end.
12442 2018-07-12  Martin Sebor  <msebor@redhat.com>
12444         PR c/86453
12445         * attribs.c (decl_attributes): Reject conflicting attributes before
12446         calling attribute handlers.
12448 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12450         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12451          parameter.
12452         (gcc::dump_manager::get_dump_file_name): likewise.
12453         (dump_begin): Likewise.
12454         * dumpfile.h (dump_begin): Update prototype.
12455         (gcc::dump_manager::get_dump_file_name,
12456         gcc::dump_manager::get_dump_file_name): Update prototype.
12458 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12460         * internal-fn.h (vectorizable_internal_fn_p): New function.
12461         * tree-vect-slp.c (compatible_calls_p): Likewise.
12462         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12463         to internal functions.
12464         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12466 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12468         * fold-const.h (inverse_conditions_p): Declare.
12469         * fold-const.c (inverse_conditions_p): New function.
12470         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12471         that test the inverse condition of a conditional internal function.
12472         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12473         * internal-fn.c (internal_fn_mask_index): Handle conditional
12474         internal functions.
12475         (vectorized_internal_fn_supported_p): New function.
12476         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12477         (any_pred_load_store): Replace with...
12478         (need_to_predicate): ...this new variable.
12479         (redundant_ssa_names): New variable.
12480         (ifcvt_can_use_mask_load_store): Move initial checks to...
12481         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12482         for which a conditional internal function exists.
12483         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12484         instead of ifcvt_can_use_mask_load_store.  Update after variable
12485         name change.
12486         (predicate_load_or_store): New function, split out from
12487         predicate_mem_writes.
12488         (check_redundant_cond_expr): New function.
12489         (value_available_p): Likewise.
12490         (predicate_rhs_code): Likewise.
12491         (predicate_mem_writes): Rename to...
12492         (predicate_statements): ...this.  Use predicate_load_or_store
12493         and predicate_rhs_code.
12494         (combine_blocks, tree_if_conversion): Update after above name changes.
12495         (ifcvt_local_dce): Handle redundant_ssa_names.
12496         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12497         general conditional functions.
12498         * tree-vect-stmts.c (vectorizable_call): Likewise.
12500 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12501             Alan Hayward  <alan.hayward@arm.com>
12502             David Sherwood  <david.sherwood@arm.com>
12504         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
12505         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
12506         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
12507         plus and minus and convert them into IFN_COND_FMA-based sequences.
12508         (convert_mult_to_fma): Handle conditional plus and minus.
12510 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12512         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
12513         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
12514         (cond_fnms_optab): New optabs.
12515         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
12516         internal functions.
12517         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
12518         * internal-fn.h (get_conditional_internal_fn): Declare.
12519         (get_unconditional_internal_fn): Likewise.
12520         * internal-fn.c (cond_ternary_direct): New macro.
12521         (expand_cond_ternary_optab_fn): Likewise.
12522         (direct_cond_ternary_optab_supported_p): Likewise.
12523         (FOR_EACH_COND_FN_PAIR): Likewise.
12524         (get_conditional_internal_fn): New function.
12525         (get_unconditional_internal_fn): Likewise.
12526         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
12527         (gimple_match_op::gimple_match_op): Add a new overload for 5
12528         operands.
12529         (gimple_match_op::set_op): Likewise.
12530         (gimple_resimplify5): Declare.
12531         * genmatch.c (decision_tree::gen): Generate simplifications for
12532         5 operands.
12533         * gimple-match-head.c (gimple_simplify): Define an overload for
12534         5 operands.  Handle calls with 5 arguments in the top-level overload.
12535         (convert_conditional_op): Handle conversions from unconditional
12536         internal functions to conditional ones.
12537         (gimple_resimplify5): New function.
12538         (build_call_internal): Pass a fifth operand.
12539         (maybe_push_res_to_seq): Likewise.
12540         (try_conditional_simplification): Try converting conditional
12541         internal functions to unconditional internal functions.
12542         Handle 3-operand unconditional forms.
12543         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
12544         Define ternary equivalents of the current rules for binary conditional
12545         internal functions.
12546         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
12547         ternary operations.
12548         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
12549         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
12550         (optab): Handle them.
12551         (SVE_COND_FP_TERNARY): New int iterator.
12552         (sve_fmla_op, sve_fmad_op): New int attributes.
12553         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
12554         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
12555         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
12557 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12559         * target.def (preferred_else_value): New target hook.
12560         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
12561         * doc/tm.texi: Regenerate.
12562         * targhooks.h (default_preferred_else_value): Declare.
12563         * targhooks.c (default_preferred_else_value): New function.
12564         * internal-fn.h (conditional_internal_fn_code): Declare.
12565         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
12566         (get_conditional_internal_fn): Use it.
12567         (conditional_internal_fn_code): New function.
12568         * gimple-match.h (gimple_match_cond): New struct.
12569         (gimple_match_op): Add a cond member function.
12570         (gimple_match_op::gimple_match_op): Update all forms to take a
12571         gimple_match_cond.
12572         * genmatch.c (expr::gen_transform): Use the same condition as res_op
12573         for the suboperation, but don't specify a particular else_value.
12574         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
12575         (visit_nary_op, visit_reference_op_load): Pass
12576         gimple_match_cond::UNCOND to the gimple_match_op constructor.
12577         * gimple-match-head.c: Include tree-eh.h
12578         (convert_conditional_op): New function.
12579         (maybe_resimplify_conditional_op): Likewise.
12580         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
12581         (gimple_resimplify2): Likewise.
12582         (gimple_resimplify3): Likewise.
12583         (gimple_resimplify4): Likewise.
12584         (maybe_push_res_to_seq): Return null for conditional operations.
12585         (try_conditional_simplification): New function.
12586         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
12587         constructor.
12588         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
12589         IFN_COND_* call.
12590         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
12591         function.
12592         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
12594 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12596         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12597         DECL_FCONTEXT
12598         (hash_tree): Do not hash DECL_FCONTEXT
12599         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
12600         Do not stream DECL_FCONTEXT.
12601         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
12602         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
12604 2018-07-12  Richard Biener  <rguenther@suse.de>
12606         PR debug/86462
12607         * dwarf2out.c (gen_block_die): Only output blocks when they have
12608         at least one !DECL_IGNORED_P variable.
12610 2018-07-12  Richard Biener  <rguenther@suse.de>
12612         PR target/84829
12613         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
12614         Remove -mieee-fp handling.
12616 2018-07-12  Richard Biener  <rguenther@suse.de>
12618         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
12619         left-over from last patch.
12621 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
12623         PR tree-optimization/86492
12624         * gimple-ssa-store-merging.c
12625         (imm_store_chain_info::coalesce_immediate_stores): Call
12626         check_no_overlap even for the merge_overlapping case.  Formatting fix.
12628 2018-07-12  Richard Biener  <rguenther@suse.de>
12630         PR middle-end/86479
12631         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
12632         move possibly trapping operations into the conditional.
12634 2018-07-12  Richard Biener  <rguenther@suse.de>
12636         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
12637         (vn_lookup_simplify_result): Remove recursion limit applied
12638         here.
12639         (vn_nary_build_or_lookup_1): Adjust.
12640         (try_to_simplify): Likewise.
12641         * gimple-match-head.c (gimple_resimplify1): Instead apply one
12642         here.
12643         (gimple_resimplify2): Likewise.
12644         (gimple_resimplify3): Likewise.
12645         (gimple_resimplify4): Likewise.
12647 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
12649         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
12650         Use __mmask64 type instead of __mmask8 for __M argument.
12651         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
12652         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
12653         __U argument.
12654         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
12655         __mmask16 for __M argument.
12656         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
12657         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12658         to __mmask16 instead of __mmask8.
12659         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
12660         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
12661         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
12662         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
12663         instead of __mmask16 for __U argument.
12664         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
12665         __mmask16 instead of __mmask8 for __U argument.
12666         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
12667         __U argument.
12668         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
12669         __mmask16.
12670         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
12671         argument.
12672         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
12673         __U argument.
12674         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
12675         __mmask16.
12676         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
12677         of __mmask16.
12678         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
12679         __U argument.
12680         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
12681         __U argument.
12682         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
12683         __U argument.
12684         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
12685         __U argument.
12686         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
12687         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
12688         return type as well as __M argument type and all casts from __mmask8
12689         to __mmask32.
12690         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
12691         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
12692         return type as well as __M argument type and all casts from __mmask8
12693         to __mmask16.
12694         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
12695         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
12696         return type as well as __M argument type and all casts from __mmask8
12697         to __mmask32.
12698         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
12699         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
12700         return type as well as __M argument type and all casts from __mmask8
12701         to __mmask16.
12702         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
12703         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
12704         __mmask16.
12706 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
12708         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
12709         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
12710         for __U argument.
12712 2018-07-11  Paul Koning  <ni1d@arrl.net>
12714         * doc/md.texi (define_subst): Document how multiple occurrences of
12715         the same argument in the replacement pattern are handled.
12717 2018-07-11  Paul Koning  <ni1d@arrl.net>
12719         * doc/extend.texi (Common Variable Attributes): Move "mode" into
12720         alphabetical order.
12721         (Common Type Attributes): Add "mode" attribute.
12723 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
12725         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
12726         stream DECL_ORIGINAL_TYPE.
12727         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
12728         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
12729         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
12730         Do not walk original type.
12731         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
12732         external decls.
12733         (write_ts_decl_non_common_tree_pointers): Do not stream
12734         DECL_ORIGINAL_TYPE
12735         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
12736         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
12738 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12740         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
12741         thread twice from the same starting edge.
12743 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12745         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
12746         * gimple.c (gimple_call_nonnull_result_p): ...here...
12747         (gimple_call_nonnull_arg): ...and here.
12748         * gimple.h (gimple_call_nonnull_result_p): New.
12749         (gimple_call_nonnull_arg): New.
12751 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
12753         * config/arm/arm-cpus.in: Move information from fpu field of each
12754         cpu definition to the isa field.
12755         * config/arm/parsecpu.awk (fpu): Delete match rule.
12756         (gen_comm_data): Don't add bits from the CPU's FPU entry.
12758 2018-07-11  Richard Biener  <rguenther@suse.de>
12760         PR debug/86457
12761         * dwarf2out.c (init_sections_and_labels): Use
12762         output_asm_line_debug_info consistently.
12763         (dwarf2out_early_finish): Likewise.
12764         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
12765         type units.
12767 2018-07-11  Richard Biener  <rguenther@suse.de>
12769         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
12770         Rework father_bb setting in a way to avoid propagating constants
12771         multiple times on a loop body.
12773 2018-07-10  Mark Wielaard  <mark@klomp.org>
12775         PR debug/86459
12776         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
12778 2018-07-10  Richard Biener  <rguenther@suse.de>
12780         * hash-map.h (hash_map::iterator::operator*): Return
12781         references to key and value.
12783 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
12785         PR c++/86443
12786         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
12787         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
12788         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
12789         (gimplify_omp_for): For composite loops, move outer
12790         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
12791         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
12792         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
12793         TREE_LIST for both the original class iterator and the "last" helper
12794         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
12795         loop, remember has_decl_expr from outer composite loops for the
12796         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
12798 2018-07-09  Martin Sebor  <msebor@redhat.com>
12800         PR middle-end/77357
12801         PR middle-end/86428
12802         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
12803         accessing implicitly initialized array elements.
12804         * expr.c (string_constant): Handle string initializers of
12805         character arrays within aggregates.
12806         * gimple-fold.c (fold_array_ctor_reference): Add argument.
12807         Store element offset.  As a special case, handle zero size.
12808         (fold_nonarray_ctor_reference): Same.
12809         (fold_ctor_reference): Add argument.  Store subobject offset.
12810         * gimple-fold.h (fold_ctor_reference): Add argument.
12812 2018-07-09  Paul Koning  <ni1d@arrl.net>
12814         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
12815         (pdp11_insn_cost): New function.
12816         (pdp11_md_asm_adjust): New function.
12817         (TARGET_INVALID_WITHIN_DOLOOP): Define.
12818         (pdp11_rtx_costs): Update to match machine better.
12819         (output_addr_const_pdp11): Correct format mismatch warnings.
12820         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
12821         * config/pdp11/pdp11.md: General change to add base_cost and/or
12822         length attributes for use by new pdp11_insn_cost function.
12823         (MIN_BRANCH): Correct definition.
12824         (MIN_SOB): Ditto.
12825         (doloop_end): Use standard pattern name for looping pattern.
12826         (doloop_end_nocc): New.
12827         (movsf): Add another constraint alternative.
12828         (zero_extendqihi2): Add constraint alternatives for not in place
12829         extend.
12830         (zero_extendhisi2): Remove.
12831         (shift patterns): Add CC handling variants.
12832         (bswaphi2): New.
12833         (bswapsi2): New.
12834         (rothi3): New.
12835         (define_peephole2): New peephole to recognize mov that sets CC for
12836         subsequent test.
12838 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12840         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
12841         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
12842         wi::add.
12844 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
12846         PR c/86420
12847         * real.c (real_nextafter): Return true if result is denormal.
12849 2018-07-09  Martin Liska  <mliska@suse.cz>
12851         * common.opt: Add back wrongly removed attribute.
12853 2018-07-09  Richard Biener  <rguenther@suse.de>
12855         PR debug/86413
12856         * dwarf2out.c (gen_block_die): For an early generated DIE
12857         always output high/low PC attributes.
12859 2018-07-09  Tom de Vries  <tdevries@suse.de>
12861         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
12862         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
12863         onto VAR_DECL with abstract origin.
12865 2018-07-07  Jim Wilson  <jimw@sifive.com>
12867         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
12869 2018-07-07  Tom de Vries  <tdevries@suse.de>
12871         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
12873 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
12875         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
12876         overflow_type.
12877         (combine_bound): Use wide-int overflow calculation instead of
12878         rolling our own.
12879         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
12880         overflow_type.
12881         * fold-const.c (int_const_binop_2): Same.
12882         (extract_muldiv_1): Same.
12883         (fold_div_compare): Same.
12884         (fold_abs_const): Same.
12885         * match.pd: Same.
12886         * poly-int.h (add): Same.
12887         (sub): Same.
12888         (neg): Same.
12889         (mul): Same.
12890         * predict.c (predict_iv_comparison): Same.
12891         * profile-count.c (slow_safe_scale_64bit): Same.
12892         * simplify-rtx.c (simplify_const_binary_operation): Same.
12893         * tree-chrec.c (tree_fold_binomial): Same.
12894         * tree-data-ref.c (split_constant_offset_1): Same.
12895         * tree-if-conv.c (idx_within_array_bound): Same.
12896         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
12897         * tree-ssa-phiopt.c (minmax_replacement): Same.
12898         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
12899         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
12900         * vr-values.c (vr_values::adjust_range_with_scev): Same.
12901         * wide-int.cc (wi::add_large): Same.
12902         (wi::mul_internal): Same.
12903         (wi::sub_large): Same.
12904         (wi::divmod_internal): Same.
12905         * wide-int.h: Change overflow type to overflow_type for neg, add,
12906         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
12907         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
12908         mul_internal, divmod_internal.
12909         (overflow_type): New enum.
12910         (accumulate_overflow): New.
12912 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
12914         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
12915         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
12917 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12919         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
12920         argument is checked for zero before entering loop, avoid checking again.
12922 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12924         * gimplify.h (generic_expr_could_trap_p): Set as global function.
12925         * gimplify.h (generic_expr_could_trap_p): Likwise.
12926         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
12928 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
12930         PR tree-optimization/86401
12931         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
12932         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
12933         (fold_bit_and_mask): ... here.  New helper function for match.pd.
12934         * fold-const.h (fold_bit_and_mask): Declare.
12935         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
12937 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
12939         PR target/86324
12940         * target.def (translate_mode_attribute): New hook.
12941         * targhooks.h (default_translate_mode_attribute): Declare.
12942         * targhooks.c (default_translate_mode_attribute): New function.
12943         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
12944         * doc/tm.texi: Regenerate.
12945         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
12946         (rs6000_translate_mode_attribute): New function.
12948 2018-07-06  Paul Koning  <ni1d@arrl.net>
12950         * doc/md.texi (define_split): Document DONE and FAIL.
12951         (define_peephole2): Ditto.
12953 2018-07-05  Jeff Law  <law@redhat.com>
12955         PR tree-optimization/86010
12956         * tree-ssa-dse.c (compute_trims): More aggressively trim at
12957         both the head and tail of mem* and str* calls.
12959 2018-07-05  Jim Wilson  <jimw@sifive.com>
12961         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
12963 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
12965         * config/aarch64/aarch64-simd.md: correct flags text for
12966         MIN_EXPR replacement.
12968 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
12970         * configure: Regenerated.
12972 2018-07-05  Carl Love  <cel@us.ibm.com>
12974         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
12975         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
12976         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
12977         VSX_BUILTIN_DOUBLEL_V4SF.
12979 2018-07-05  Martin Sebor  <msebor@redhat.com>
12981         PR c++/86400
12982         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
12983         than its domain to compute its the upper bound of a char array.
12985 2018-07-05  Nathan Sidwell  <nathan@acm.org>
12987         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
12988         * doc/cpp.texi: Update comment.
12989         * doc/tm.texi: Rebuilt.
12990         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
12991         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
12992         * doc/extend.texi (Backwards Compatibility): Clarify it is system
12993         headers affected by extern "C".
12994         * system.h: Poison NO_IMPLICIT_EXTERN_C.
12995         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
12996         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
12997         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
12998         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
12999         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13000         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13001         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13002         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13003         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13004         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13005         config/sparc/sp64-elf.h, config/spu/spu.h,
13006         config/stormy16/stormy16.h, config/v850/v850.h,
13007         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13008         define NO_IMPLICIT_EXTERN_C.
13009         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13011 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13013         PR target/84711
13014         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13015         instead of GET_MODE_SIZE when comparing Units.
13017 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13019         PR target/84711
13020         * rtlanal.c (set_noop_p): Constrain on mode change,
13021         include hard-reg-set.h
13023 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13025         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13027 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13029         Revert
13030         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13032         PR sanitizer/84250
13033         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13034         libasan.
13035         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13037 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13039         PR sanitizer/84250
13040         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13041         libasan.
13042         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13044 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13046         PR middle-end/86380
13047         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13049 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13051         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13052         neg_*_op* variables.
13054 2018-07-04  Martin Liska  <mliska@suse.cz>
13056         * tree-switch-conversion.c: Define
13057         max_ratio_for_speed and max_ratio_for_size constants.
13059 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13060             Martin Liska  <mliska@suse.cz>
13062         PR middle-end/66240
13063         PR target/45996
13064         PR c/84100
13065         * common.opt: Rename align options with 'str_' prefix.
13066         * common/config/i386/i386-common.c (set_malign_value): New
13067         function.
13068         (ix86_handle_option): Use it to set -falign-* options/
13069         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13070         type from int to string.
13071         * config/aarch64/aarch64.c: Update default values from int
13072         to string.
13073         * config/alpha/alpha.c (alpha_override_options_after_change):
13074         Likewise.
13075         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13076         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13077         max skip conditionally.
13078         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13079         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13080         max skip conditionally.
13081         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13082         max skip conditionally.
13083         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13084         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13085         max skip conditionally.
13086         * config/i386/i386.c (struct ptt): Change type from int to
13087         string.
13088         (ix86_default_align): Set default values.
13089         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13090         max skip conditionally.
13091         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13092         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13093         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13094         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13095         max skip conditionally.
13096         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13097         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13098         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13099         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13100         max skip conditionally.
13101         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13102         * config/ia64/ia64.c (ia64_option_override): Set default values
13103         for alignment options.
13104         * config/m68k/m68k.c: Handle new str_align_* options.
13105         * config/mips/mips.c (mips_set_compression_mode): Change
13106         type of constants.
13107         (mips_option_override): Set default values for options.
13108         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13109         Likewise.
13110         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13111         Likewise.
13112         * config/rx/rx.c (rx_option_override): Likewise.
13113         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13114         (LABEL_ALIGN): Use align_labels_log.
13115         (LOOP_ALIGN): Use align_loops_align.
13116         * config/s390/s390.c (s390_asm_output_function_label): Use new
13117         macros.
13118         * config/sh/sh.c (sh_override_options_after_change):
13119         Change type of constants.
13120         * config/spu/spu.c (spu_sched_init): Likewise.
13121         * config/sparc/sparc.c (sparc_option_override): Set default
13122         values for options.
13123         * config/visium/visium.c (visium_option_override): Likewise.
13124         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13125         emit p2align format with last argument if it's not needed.
13126         * doc/invoke.texi: Document extended format of -falign-*.
13127         * final.c: Use align_labels alignment.
13128         * flags.h (struct target_flag_state): Change type to use
13129         align_flags.
13130         (struct align_flags_tuple): New.
13131         (struct align_flags): Likewise.
13132         (align_loops_log): Redefine macro to use new types.
13133         (align_loops_max_skip): Redefine macro to use new types.
13134         (align_jumps_log): Redefine macro to use new types.
13135         (align_jumps_max_skip): Redefine macro to use new types.
13136         (align_labels_log): Redefine macro to use new types.
13137         (align_labels_max_skip): Redefine macro to use new types.
13138         (align_functions_log): Redefine macro to use new types.
13139         (align_loops): Redefine macro to use new types.
13140         (align_jumps): Redefine macro to use new types.
13141         (align_labels): Redefine macro to use new types.
13142         (align_functions): Redefine macro to use new types.
13143         (align_functions_max_skip): Redefine macro to use new types.
13144         (align_loops_value): New macro.
13145         (align_jumps_value): New macro.
13146         (align_labels_value): New macro.
13147         (align_functions_value): New macro.
13148         * function.c (invoke_set_current_function_hook): Propagate
13149         alignment values from flags to global variables default in
13150         topleev.h.
13151         * ipa-icf.c (sem_function::equals_wpa): Use
13152         cl_optimization_option_eq instead of memcmp.
13153         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13154         of string types.
13155         (cl_optimization_stream_in): Likewise.
13156         * optc-save-gen.awk: Support strings in cl_optimization.
13157         * opth-gen.awk: Likewise.
13158         * opts.c (finish_options): Remove error checking of invalid
13159         value ranges.
13160         (MAX_CODE_ALIGN): Remove.
13161         (MAX_CODE_ALIGN_VALUE): Likewise.
13162         (parse_and_check_align_values): New function.
13163         (check_alignment_argument): Likewise.
13164         (common_handle_option): Use check_alignment_argument.
13165         * opts.h (parse_and_check_align_values): Declare.
13166         * toplev.c (init_alignments): Remove.
13167         (read_log_maxskip): New.
13168         (parse_N_M): Likewise.
13169         (parse_alignment_opts): Likewise.
13170         (backend_init_target): Remove usage of init_alignments.
13171         * toplev.h (parse_alignment_opts): Declare.
13172         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13173         argument.
13174         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13175         * tree.c (cl_option_hasher::equal): New.
13176         * varasm.c: Use new global macros.
13178 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13180         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13181         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13182         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13183         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13184         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13185         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13186         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13187         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13188         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13189         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13190         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13191         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13192         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13194 2018-07-04  Martin Liska  <mliska@suse.cz>
13195             Jonathan Wakely  <jwakely@redhat.com>
13197         * coverage.c: Use correct type.
13198         * doc/invoke.texi: Language correction.
13200 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13202         PR target/85620
13203         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13204         ENDBRANCH for non-tail call which may return via indirect branch.
13205         * doc/extend.texi: Document indirect_return attribute.
13207 2018-07-03  Martin Sebor  <msebor@redhat.com>
13209         PR tree-optimization/86274
13210         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13211         precondition.
13212         (format_floating): Correct handling of infinities and NaNs.
13214 2018-07-03  Martin Sebor  <msebor@redhat.com>
13216         * print-tree.c (print_real_cst): New function.
13217         (print_node_brief): Call it.
13218         (print_node): Ditto.
13220 2018-07-03  Jeff Law  <law@redhat.com>
13222         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13223         into a single pattern.
13225         * config/h8300/h8300.md (ors code_iterator): New.
13226         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13227         a single pattern and single splitter.
13228         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13229         (iorqi3_1, xorqi3_1): Likewise.
13230         (iorqi3, xorqi3 expanders): Similarly.
13232         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13233         (movmd_internal) into a single pattern using the P mode iterator.
13234         (movmd splitters): Similarly.
13235         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13236         (movsd splitters): Similarly.
13238         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13239         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13240         right table based on the size of the operand.
13241         * config/h8300/h8300.md (length_table): Corresponding changes. All
13242         references to "addb", "addw" and "addl" changed to "add".
13243         (btst patterns): Merge two variants into a single pattern.
13244         (tstqi, tsthi): Likewise.
13245         (addhi3_incdec, addsi3_incdec): Likewise.
13246         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13247         (mulhi3, mulsi3): Likewise.
13248         (udivhi3, udivsi3): Likewise.
13249         (divhi3, divsi3): Likewise.
13250         (andorqi3, andorhi3, andorsi3): Likewise.
13252 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13254         PR target/85694
13255         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13256         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13258 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13260         PR tree-optimization/85694
13261         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13262         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13263         UNSPEC_URHADD.
13264         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13265         (<u>avg<mode>3_ceil): New patterns.
13267 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13269         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13270         scan-tree-dump directive.
13271         * gcc.dg/vect/slp-perm-2.c: Likewise.
13272         * gcc.dg/vect/slp-perm-3.c: Likewise.
13273         * gcc.dg/vect/slp-perm-5.c: Likewise.
13274         * gcc.dg/vect/slp-perm-6.c: Likewise.
13275         * gcc.dg/vect/slp-perm-7.c: Likewise.
13276         * gcc.dg/vect/slp-perm-8.c: Likewise.
13278 2018-07-03  Marek Polacek  <polacek@redhat.com>
13280         PR middle-end/86202
13281         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13283 2018-07-03  Richard Biener  <rguenther@suse.de>
13285         PR ipa/86389
13286         * tree-ssa-structalias.c (find_func_clobbers): Properly
13287         handle indirect calls.
13289 2018-07-03  Jeff Law  <law@redhat.com>
13291         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13292         (shifts): New code iterator.
13293         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13294         expander.  Fix HImode handling on H8/SX.
13295         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13296         (subqi3, subhi3, subsi3 expanders): Likewise.
13297         (andqi3, andhi3, andsi3 expanders): Likewise.
13298         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13299         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13300         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13301         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13302         (zero_extendqihi2, zero_extendqisi2): Likewise.
13303         (extendqihi2, extendqisi2): Likewise.
13304         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13305         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13306         (rotlqi3_1, rotlhi3_1): Likewise.
13307         (logicalhi3_sn, logicalsi3_sn): Likewise.
13308         (logicalhi3, logicalsi3): Likewise.
13310 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13312         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13313         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13314         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13315         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13316         type to append_pattern_def_seq instead of creating a stmt_vec_info
13317         directly.
13318         (build_mask_conversion): Likewise.  Remove vinfo argument.
13319         (vect_add_conversion_to_patterm): Likewise, renaming to...
13320         (vect_add_conversion_to_pattern): ...this.
13321         (vect_recog_mask_conversion_pattern): Update call to
13322         build_mask_conversion.  Pass the vector type to
13323         append_pattern_def_seq here too.
13324         (vect_recog_gather_scatter_pattern): Update call to
13325         vect_add_conversion_to_pattern.
13327 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13329         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13330         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13331         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13332         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13333         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13334         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13335         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13336         append_pattern_def_seq instead of new_pattern_def_seq.
13337         (vect_recog_divmod_pattern): Do both of the above.
13338         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13339         is null.
13341 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13343         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13344         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13345         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13346         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13347         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13348         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13349         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13350         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13351         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13352         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13353         parameter with a single stmt_vec_info.
13354         (vect_recog_func_ptr): Likewise.
13355         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13356         (vect_try_gather_scatter_pattern): ...this.
13357         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13358         the stmt_vec_info of the statement to be matched.  Don't clear
13359         STMT_VINFO_RELATED_STMT.
13360         (vect_pattern_recog): Update call accordingly.
13362 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13364         PR tree-optimization/85694
13365         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13366         (uavgM3_ceil): Document new optabs.
13367         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13368         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13369         functions.
13370         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13371         (savg_ceil_optab): New optabs.
13372         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13373         (vect_vect_recog_func_ptrs): Add it.
13374         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13375         constant directly from the associated lhs.
13377 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13379         * tree-vect-patterns.c (vect_split_statement): New function.
13380         (vect_convert_input): Use it to try to split an existing cast.
13382 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13384         * poly-int.h (print_hex): New function.
13385         * dumpfile.h (dump_dec, dump_hex): Declare.
13386         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13387         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13388         min_input_precision, operation_precision and operation_sign.
13389         * tree-vect-patterns.c (vect_get_range_info): New function.
13390         (vect_same_loop_or_bb_p, vect_single_imm_use)
13391         (vect_operation_fits_smaller_type): Delete.
13392         (vect_look_through_possible_promotion): Add an optional
13393         single_use_p parameter.
13394         (vect_recog_over_widening_pattern): Rewrite to use new
13395         stmt_vec_info infomration.  Handle one operation at a time.
13396         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13397         (vect_truncatable_operation_p, vect_set_operation_type)
13398         (vect_set_min_input_precision): New functions.
13399         (vect_determine_min_output_precision_1): Likewise.
13400         (vect_determine_min_output_precision): Likewise.
13401         (vect_determine_precisions_from_range): Likewise.
13402         (vect_determine_precisions_from_users): Likewise.
13403         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13404         (vect_vect_recog_func_ptrs): Put over_widening first.
13405         Add cast_forwprop.
13406         (vect_pattern_recog): Call vect_determine_precisions.
13408 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13410         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13411         statements that have been replaced by further pattern statements.
13412         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13414 2018-07-03  Richard Biener  <rguenther@suse.de>
13416         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13417         always set *dt.  Dump vectype in vectype overload.
13418         * dumpfile.h (dump_gimple_expr): New function.
13419         (dump_gimple_expr_loc): Likewise.
13420         * dumpfile.c (dump_gimple_expr): New function.
13421         (dump_gimple_expr_loc): Likewise.
13423 2018-07-02  Jeff Law  <law@redhat.com>
13425         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13426         the H8/300, H8/300H and H8/S variants into a single pattern.
13427         (movhi_h8300, movqi_h8300hs): Similarly.
13428         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13429         (QHI mode iterator): New.
13431         * config/h8300/h8300.md: Remove trailing whitespace.
13433 2018-07-02  Jim Wilson  <jimw@sifive.com>
13435         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13436         instead of emit_insn for interrupt returns.
13437         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13438         (riscv_sret, riscv_uret): Likewise.
13440 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13442         * pretty-print.c (selftest::test_pp_format): Move save and restore
13443         of quotes to class auto_fix_quotes, and add an instance.
13444         * selftest.c: Include "intl.h".
13445         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13446         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13447         * selftest.h (selftest::auto_fix_quotes): New class.
13449 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13451         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13452         (aarch64_sve_prepare_conditional_op): Remove.
13453         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13454         Allow aarch64_simd_reg_or_zero as select operand; remove
13455         the aarch64_sve_prepare_conditional_op call.
13456         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13457         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13458         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13459         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13460         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13461         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13462         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13463         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13464         and a splitters to match all of the *_any patterns.
13465         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13467         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13468         (SVE_COND_FP_BINARY_REV): Remove.
13469         (sve_int_op_rev, sve_fp_op_rev): New.
13470         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13471         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13472         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13473         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13474         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13475         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13476         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13477         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13478         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13480         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13481         Remove match_dup 1 from the inner unspec.
13482         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13484         * config/aarch64/aarch64.md (movprfx): New attr.
13485         (length): Default movprfx to 8.
13486         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13487         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13488         (*<su>mul<SVE_I>3_highpart): Likewise.
13489         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13490         (*v<ASHIFT><SVE_I>3): Likewise.
13491         (*<su><MAXMIN><SVE_I>3): Likewise.
13492         (*<su><MAXMIN><SVE_F>3): Likewise.
13493         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13494         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13495         (*div<SVE_F>4): Likewise.
13497 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
13499         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13500         in dump string.
13502 2018-07-02  Richard Biener  <rguenther@suse.de>
13504         PR tree-optimization/86363
13505         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
13506         memset argument refers to a non-variable address.
13508 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13510         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
13511         {PLUS,MINUS}_EXPR code to...
13512         (adjust_symbolic_bound): ...here,
13513         (combine_bound): ...here,
13514         (set_value_range_with_overflow): ...and here.
13516 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13518         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
13519         code...
13520         (extract_range_from_abs_expr): ...here.
13522 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
13524         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
13525         -fno-omit-frame-pointer when not optimizing.
13527 2018-07-02  Martin Liska  <mliska@suse.cz>
13529         PR ipa/86279
13530         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
13531         (propagate_nothrow): Likewise.
13533 2018-07-02  Martin Liska  <mliska@suse.cz>
13535         PR ipa/86323
13536         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
13538 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13540         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
13541         function in r262149, changing "loc" param from source_location to
13542         const dump_location_t &.
13543         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
13544         declaration, as above.
13546 2018-07-01  Paul Koning  <ni1d@arrl.net>
13548         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13549         -munit-asm, -mgnu-asm, -mdec-asm.
13550         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
13551         (pdp11_output_labelref): New.
13552         (pdp11_output_def): New.
13553         (pdp11_output_addr_vec_elt): New.
13554         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
13555         %# and %@ format codes.
13556         (pdp11_option_override): New.
13557         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
13558         (pdp11_output_ident): New.
13559         (pdp11_asm_named_section): New.
13560         (pdp11_asm_init_sections): New.
13561         (pdp11_file_start): New.
13562         (pdp11_file_end): New.
13563         (output_ascii): Use .ascii/.asciz for -mdec-asm.
13564         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
13565         %o, like %c but octal.
13566         (pdp11_option_override): New.
13567         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
13568         -mdec-asm.
13569         (DATA_SECTION_ASM_OP): Ditto.
13570         (READONLY_DATA_SECTION_ASM_OP): New.
13571         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
13572         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
13573         (ASM_OUTPUT_LABELREF): Ditto.
13574         (ASM_OUTPUT_DEF): Ditto.
13575         (ASM_OUTPUT_EXTERNAL): New.
13576         (ASM_OUTPUT_SOURCE_FILENAME): New.
13577         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
13578         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
13579         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
13580         %# and %@ format codes.
13581         * config/pdp11/pdp11.opt (mgnu-asm): New.
13582         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
13583         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
13584         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
13586 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
13588         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
13589         dereferencing path[] beyond its length.
13590         (debug_path): New.
13591         (debug_all_paths): New.
13592         (rewire_first_differing_edge): New.
13593         (adjust_paths_after_duplication): New.
13594         (duplicate_thread_path): Call adjust_paths_after_duplication.
13595         Add new argument.
13596         (thread_through_all_blocks): Add new argument to
13597         duplicate_thread_path.
13599 2018-06-30  Jim Wilson  <jimw@sifive.com>
13601         * config/riscv/predicates.md (p2m1_shift_operand): New.
13602         (high_mask_shift_operand): New.
13603         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
13604         pattern using p2m1_shift_operand.
13605         (lshsi3_zero_extend_3+2): New combiner pattern using
13606         high_mask_shift_operand.
13608 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13610         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
13611         split out from...
13612         (vect_recog_rotate_pattern): ...here.
13613         (vect_convert_input): Try to insert casts of invariants in the
13614         preheader.
13615         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
13616         preheader to be empty.
13618 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13620         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
13621         vector type.  If given, install it in the new statement's
13622         STMT_VINFO_VECTYPE.
13623         (vect_element_precision): New function.
13624         (vect_unpromoted_value): New struct.
13625         (vect_unpromoted_value::vect_unpromoted_value): New function.
13626         (vect_unpromoted_value::set_op): Likewise.
13627         (vect_look_through_possible_promotion): Likewise.
13628         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
13629         (vect_widened_op_tree, vect_convert_input): Likewise.
13630         (vect_convert_inputs, vect_convert_output): Likewise.
13631         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
13632         to handle the optional cast of the multiplication result and
13633         vect_widened_op_tree to detect the widened multiplication itself.
13634         Do not require the input and output of promotion casts to have
13635         the same sign, but base the signedness of the operation on the
13636         input rather than the result.  If the pattern includes two
13637         promotions, check that those promotions have the same sign.
13638         Do not restrict the MULT_EXPR handling to a double-width result;
13639         handle quadruple-width results and wider.  Use vect_convert_inputs
13640         to convert the inputs to the common type.
13641         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
13642         to handle the optional cast of the ABS result.  Also allow a sign
13643         change or a sign extension between the ABS and MINUS.
13644         Use vect_widened_op_tree to detect the widened subtraction and use
13645         vect_convert_inputs to convert the inputs to the common type.
13646         (vect_handle_widen_op_by_const): Delete.
13647         (vect_recog_widen_op_pattern): New function.
13648         (vect_recog_widen_mult_pattern): Use it.
13649         (vect_recog_widen_shift_pattern): Likewise.
13650         (vect_recog_widen_sum_pattern): Use
13651         vect_look_through_possible_promotion to handle the promoted
13652         PLUS_EXPR operand.
13654 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13656         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
13657         the containing gimple_seq *.
13658         * gimple-iterator.h (gsi_for_stmt): Declare it.
13659         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
13660         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
13661         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13662         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13663         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
13664         checks.
13665         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
13666         split out from...
13667         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
13668         statement being replaced is part of an existing pattern
13669         definition sequence, inserting the new pattern statements before
13670         the original one.
13671         (vect_pattern_recog_1): Don't return a bool.  If the statement
13672         is already part of a pattern, instead apply pattern matching
13673         to the pattern definition statements.  Don't clear the
13674         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
13675         (vect_pattern_recog): Don't break after the first match;
13676         continue processing the pattern definition statements instead.
13677         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
13679 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13681         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13682         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13683         (vect_recog_widen_sum_pattern): Use it.
13685 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13687         * tree-vect-loop.c (vectorizable_reduction): Assert that the
13688         phi is not a pattern statement and has not been replaced by
13689         a pattern statement.
13690         * tree-vect-patterns.c (type_conversion_p): Don't check
13691         STMT_VINFO_IN_PATTERN_P.
13692         (vect_recog_vector_vector_shift_pattern): Likewise.
13693         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
13694         the pattern statement rather than the original statement; check
13695         directly for a WIDEN_MULT_EXPR here.
13696         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
13697         vect_is_simple_use to return the pattern statement rather
13698         than the original statement; use is_pattern_stmt_p to check
13699         for such a pattern statement.
13700         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
13701         to return the pattern statement rather than the original statement;
13702         don't do the same transformation here.
13703         (vect_is_simple_use): If the defining statement has been replaced
13704         by a pattern statement, return the pattern statement instead.
13705         Remove the corresponding (local) transformation from the vectype
13706         overload.
13708 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13710         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
13711         end and default to null.
13712         * tree-vect-loop.c (vect_create_epilog_for_reduction)
13713         (vectorizable_reduction): Update calls accordingly, dropping the
13714         gimple ** argument if the passed-back statement isn't needed.
13715         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
13716         (vect_recog_rotate_pattern): Likewise.
13717         (vect_recog_mask_conversion_pattern): Likewise.
13718         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13719         (vect_mask_constant_operand_p): Likewise.
13720         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
13721         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
13722         (get_group_load_store_type, get_load_store_type): Likewise.
13723         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
13724         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
13725         (vectorizable_conversion, vectorizable_assignment): Likewise.
13726         (vectorizable_shift, vectorizable_operation): Likewise.
13727         (vectorizable_store, vect_is_simple_cond): Likewise.
13728         (vectorizable_condition, vectorizable_comparison): Likewise.
13729         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
13730         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
13731         and move it to the end.  Cope with null def_stmt_outs.
13733 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13735         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
13737 2018-06-29  Jeff Law  <law@redhat.com>
13739         * config/v850/v850.c (v850_legitimate_address_p): Handle large
13740         displacements for TARGET_V850E2V3 and newer.
13741         (TARGET_LRA_P): Remove.  Defaults to LRA now.
13742         * config/v850/v850.md (sign23byte_load): Remove.
13743         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
13744         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
13746 2018-06-29  Martin Liska  <mliska@suse.cz>
13748         PR lto/85759
13749         * coverage.c (coverage_init): Mangle full path name.
13750         * doc/invoke.texi: Document the change.
13751         * gcov-io.c (mangle_path): New.
13752         * gcov-io.h (mangle_path): Likewise.
13753         * gcov.c (mangle_name): Use mangle_path for path mangling.
13755 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13757         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
13758         if starting source register is not even.
13760 2018-06-29  Martin Liska  <mliska@suse.cz>
13762         PR tree-optimization/86263
13763         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
13764         Make edge redirection.
13766 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
13768         * dumpfile.c (dump_loc): Add indentation based on scope depth.
13769         (dump_scope_depth): New variable.
13770         (get_dump_scope_depth): New function.
13771         (dump_begin_scope): New function.
13772         (dump_end_scope): New function.
13773         * dumpfile.h (get_dump_scope_depth): New declaration.
13774         (dump_begin_scope): New declaration.
13775         (dump_end_scope): New declaration.
13776         (class auto_dump_scope): New class.
13777         (AUTO_DUMP_SCOPE): New macro.
13778         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
13779         AUTO_DUMP_SCOPE.
13781 2018-06-29  Richard Biener  <rguenther@suse.de>
13783         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
13784         compute_all_dependences succeeds.
13785         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
13786         exceed --param loop-max-datarefs-for-datadeps.
13788 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
13790         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
13792 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
13794         PR target/86348
13795         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
13796         alternative 0 in preferred_for_speed attribute.
13798 2018-06-28  Paul Koning  <ni1d@arrl.net>
13800         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
13801         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
13802         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
13803         * config/pdp11/pdp11.md: Correct "length" attribute calculation
13804         for shift insn patterns.
13806 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
13808         * cgraph.c (cgraph_node::get_body): Replace assignments to
13809         "dump_file" with calls to set_dump_file.
13810         * dumpfile.c (alt_dump_file): Make static, and group with...
13811         (alt_flags): ...this definition.
13812         (dumps_are_enabled): New variable.
13813         (refresh_dumps_are_enabled): New function.
13814         (set_dump_file): New function.
13815         (set_alt_dump_file): New function.
13816         (gcc::dump_manager::dump_start): Replace assignments to
13817         "dump_file" and "alt_dump_file" with calls to set_dump_file and
13818         set_alt_dump_file.
13819         (gcc::dump_manager::dump_finish): Likewise.
13820         * dumpfile.h (alt_dump_file): Delete decl.
13821         (dumps_are_enabled): New variable decl.
13822         (set_dump_file): New function decl.
13823         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
13824         global.
13825         * tree-nested.c (lower_nested_functions): Replace assignments to
13826         "dump_file" with calls to set_dump_file.
13828 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13830         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
13831         goto_locus of each outgoing edge of each basic block.
13833 2018-06-28  Richard Biener  <rguenther@suse.de>
13835         * dwarf2out.c (decl_scope_table): Remove.
13836         (push_decl_scope): Likewise.
13837         (pop_decl_scope): Likewise.
13838         (gen_type_die_for_member): Do not call push/pop_decl_scope.
13839         (gen_struct_or_union_type_die): Likewise.
13840         (gen_tagged_type_die): Likewise.
13841         (dwarf2out_init): Do not initialize decl_scope_table.
13842         (dwarf2out_c_finalize): Do not free it.
13844 2018-06-28  Richard Biener  <rguenther@suse.de>
13846         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
13847         deciding whether to not re-use a DIE.
13849 2018-06-28  Richard Biener  <rguenther@suse.de>
13851         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
13852         DW_AT_abstract_origin attribute.
13854 2018-06-28  Martin Liska  <mliska@suse.cz>
13856         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13857         Use newly introduced constants.
13858         * tree-switch-conversion.h (struct jump_table_cluster):
13859         Define max_ratio_for_size and max_ratio_for_speed.
13861 2018-06-28  Martin Liska  <mliska@suse.cz>
13863         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13864         Add new checking assert to catch invalid state.
13865         (jump_table_cluster::can_be_handled): Handle single case
13866         clusters.
13867         (jump_table_cluster::is_beneficial): Bail out for such case.
13868         (bit_test_cluster::find_bit_tests):
13869         Add new checking assert to catch invalid state.
13870         (bit_test_cluster::can_be_handled): Handle single case
13871         clusters.
13872         (bit_test_cluster::is_beneficial): Bail out for such case.
13873         (switch_decision_tree::analyze_switch_statement):
13874         Fix comment.
13876 2018-06-28  Martin Liska  <mliska@suse.cz>
13878         * common.opt: Introduce -completion option.
13879         * gcc.c (driver_handle_option): Handle it.
13880         (driver::main): Print completions if completion
13881         is set.
13882         * opt-suggestions.c (option_proposer::get_completions):
13883         New function.
13884         (option_proposer::suggest_completion): Likewise.
13885         (option_proposer::find_param_completions): Likewise.
13886         (verify_autocompletions): Likewise.
13887         (test_completion_valid_options): Likewise.
13888         (test_completion_valid_params): Likewise.
13889         (in_completion_p): Likewise.
13890         (empty_completion_p): Likewise.
13891         (test_completion_partial_match): Likewise.
13892         (test_completion_garbage): Likewise.
13893         (opt_proposer_c_tests): Likewise.
13894         * opt-suggestions.h: Declare new functions.
13895         * opts.c (common_handle_option): Handle OPT__completion_.
13896         * selftest-run-tests.c (selftest::run_tests): Add
13897         opt_proposer_c_tests.
13898         * selftest.c (assert_str_startswith): New.
13899         * selftest.h (assert_str_startswith): Likewise.
13900         (opt_proposer_c_tests): New.
13901         (ASSERT_STR_STARTSWITH): Likewise.
13903 2018-06-28  Martin Liska  <mliska@suse.cz>
13905         * Makefile.in: Add opt-suggestions.o.
13906         * gcc-main.c: Include opt-suggestions.h.
13907         * gcc.c (driver::driver): Likewise.
13908         (driver::~driver): Remove m_option_suggestions.
13909         (driver::build_option_suggestions): Moved to option_proposer.
13910         (driver::suggest_option): Likewise.
13911         (driver::handle_unrecognized_options): Use option_proposer.
13912         * gcc.h (class driver): Add new memver m_option_proposer.
13913         * opt-suggestions.c: New file.
13914         * opt-suggestions.h: New file.
13916 2018-06-28  Martin Liska  <mliska@suse.cz>
13918         * vec.h (class auto_string_vec): New (moved from auto_argvec).
13919         (auto_string_vec::~auto_string_vec): Likewise.
13921 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13923         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
13924         prevent_decl_creation_for_types fields up and add reset_location field.
13925         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
13926         statement if id->reset_location is true.
13927         (copy_edges_for_bb): Do not set goto_locus on the new edges if
13928         id->reset_location is true.
13929         (copy_phis_for_bb): Force input_location on the arguments if
13930         id->reset_location is true.
13931         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
13932         is set on the function to be inlined.
13934 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
13936         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
13938 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
13940         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
13941         registers for Pmode.
13942         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
13943         hard registers for the clobbered pseudo.
13945 2018-06-27  Paul Koning  <ni1d@arrl.net>
13947         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13948         mutually exclusive options.
13949         * config/pdp11/constraints.md (h): New constraint.
13950         (O): Update definition to match shift code generation.
13951         (D): New constraint.
13952         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
13953         (CCFP): Remove.
13954         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
13955         function.
13956         (output_jump): Change arguments.
13957         (pdp11_fixed_cc_regs): New function.
13958         (pdp11_cc_mode): Ditto.
13959         (pdp11_expand_shift): Ditto.
13960         (pdp11_assemble_shift): Ditto.
13961         (pdp11_small_shift): Ditto.
13962         (pdp11_branch_cost): Remove.
13963         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
13964         from output.
13965         (pdp11_register_move_cost): Update for CC registers.
13966         (pdp11_rtx_costs): Add case for LSHIFTRT.
13967         (pdp11_output_jump): Add CCNZ mode conditional branches.
13968         (notice_update_cc_on_set): Remove.
13969         (pdp11_cc_mode): New function.
13970         (simple_memory_operand): Correct pre/post decrement case.
13971         (no_side_effect_operand): New function.
13972         (pdp11_regno_reg_class): Add CC_REGS class.
13973         (pdp11_fixed_cc_regs): New function.
13974         (pdp11_small_shift): New function.
13975         (pdp11_expand_shift): New function to expand shift insns.
13976         (pdp11_assemble_shift): New function to output shifts.
13977         (pdp11_branch_cost): Remove.
13978         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
13979         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
13980         (WCHAR_TYPE): Ditto.
13981         (PTRDIFF_TYPE): Ditto.
13982         (ADJUST_INSN_LENGTH): New macro.
13983         (FIXED_REGISTERS): Add CC registers.
13984         (CALL_USED_REGISTERS): Ditto.
13985         (reg_class): Ditto.
13986         (REG_CLASS_NAMES): Ditto.
13987         (REG_CLASS_CONTENTS): Ditto.
13988         (SELECT_CC_MODE): Use new function.
13989         (TARGET_FLAGS_REGNUM): New macro.
13990         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
13991         (cc0_reg_rtx): Remove.
13992         (CC_STATUS_MDEP): Remove.
13993         (CC_STATUS_MDEFP_INIT): Remove.
13994         (CC_IN_FPU): Remove.
13995         (NOTICE_UPDATE_CC): Remove.
13996         (REGISTER_NAMES): Add CC registers.
13997         (BRANCH_COST): Change to constant 1.
13998         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
13999         handling.
14000         * config/pdp11/pdp11.opt (mbcopy): Remove.
14001         (mbcopy-builtin): Remove.
14002         (mbranch-cheap): Remove.
14003         (mbranch-expensive): Remove.
14004         * config/pdp11/predicates.md (expand_shift_operand): Update to
14005         match shift code generation.
14006         (ccnz_operator): New predicate.
14007         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14008         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14009         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14010         exclusive options.
14011         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14012         description of O constraint.
14014 2018-06-27  Jeff Law  <law@redhat.com>
14015             Austin Law  <austinklaw@gmail.com>
14017         * config/v850/v850.md (addsi3_set_flags): New pattern.
14018         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14019         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14020         (zero_extendhisi2_v850_set_flags): Likewise.
14021         (zero_extendqisi2_v850_set_flags): Likewise.
14022         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14023         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14024         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14026         * config/v850/v850-protos.h (notice_update_cc): Remove.
14027         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14028         (v850_print_operand): Handle 'D' and "d".
14029         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14030         Add handling of arithmetic/logical operations compared against zero.
14031         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14032         Do not look at v850_compare_op, instead get mode from last argument.
14033         (v850_gen_compare): Remove
14034         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14035         after reload for prologue insns.
14036         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14037         patterns.
14038         (construct_save_jarl): Likewise.
14039         (TARGET_FLAGS_REGNUM): Define.
14040         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14041         (NOTICE_UPDATE_CC): Remove.
14042         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14043         than cc0.  Conditionalize on reload_completed.
14044         (cmpsi_insn, setfcc_insn): Likewise.
14045         (tst1 splitter): Turn into define_and_split which sets the flags
14046         after reload.
14047         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14048         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14049         (cstoresf4, cstoredf4): Clobber the flags.
14050         (cmpsi, cmpsf, cmpdf): Remove expanders.
14051         (setf_insn): Remove pattern.
14052         (addsi3): Turn into define_and_split which clobbers the flags after
14053         reload and a suitable pattern (addsi3_clobber_flags) for use after
14054         reload.
14055         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14056         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14057         (ashrsi3, ashrsi3_v850e2): Likewise.
14058         (bins): Clobber the flags.
14059         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14060         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14061         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14062         (call_value_internal_short, call_value_internal_long): Likewise.
14063         (callt_save_interrupt, callt_return_interrupt): Likewise.
14064         (save_interrupt, return_interrupt): Likewise.
14065         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14066         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14067         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14068         (All FP comparisons): Only allow after reload has completed.
14069         (trfsr): Likewise.
14070         (divh, divhu): Tweak output template.
14071         (branch_z_normal, branch_z_invert): Remove
14072         (branch_nz_normal, branch_nz_invert): Likewise.
14073         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14075         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14076         * config/v850/v850.c (notice_update_cc): Remove.
14077         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14078         (CC_NO_CARRY): Likewise.
14079         (NOTICE_UPDATE_CC): Define to nothing.
14080         * config/v850/v850.md: Remove block comment on cc0 handling
14081         Remove "cc" attribute from all patterns.  Remove cc_status handling
14082         from all patterns.  Minor formatting fixes.
14084 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14086         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14087         (cortex-a76.cortex-a55): Likewise.
14088         * config/aarch64/aarch64-tune.md: Regenerate.
14089         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14090         cortex-a76.cortex-a55.
14092 2018-06-27  Jeff Law  <law@redhat.com>
14094         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14095         (MULTILIB_DIRNAMES): Similarly.
14097 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14099         * gimple.h (gimple_return_retbnd): Delete.
14100         (gimple_return_set_retbnd): Likewise.
14101         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14102         gimple_return_set_retbnd.
14103         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14104         gimple_return_retbnd and adjust.
14105         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14106         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14107         Explicitly return NULL in a couple more cases.  Move assertion
14108         on debug statements and remove unreachable code.
14109         (reset_debug_binding): Do not test id->retbnd.
14110         (expand_call_inline): Do not set it.
14112 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14114         * configure.ac: Add --disable-gcov option.
14115         * configure: Regenerate.
14116         * Makefile.in: Honour @enable_gcov@.
14117         * doc/install.texi: Document --disable-gcov.
14119 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14121         * config/arm/arm-cpus.in (cortex-a76): New entry.
14122         (cortex-a76.cortex-a55): Likewise.
14123         * config/arm/arm-tables.opt: Regenerate.
14124         * config/arm/arm-tune.md: Likewise.
14125         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14126         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14127         cortex-a76.cortex-a55.
14129 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14131         PR target/85769
14132         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14134 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14136         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14137         comment.
14138         (EPILOGUE_USES): Likewise.
14140 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14142         * tree-inline.c (remap_location): New function extracted from...
14143         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14144         (copy_phis_for_bb): ...here.  Call remap_location.
14145         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14147 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14149         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14150         unaligned vsx for 16B memset.
14152 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14154         PR target/86285
14155         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14156         ieee128_float_type_node to long_double_type_node unless
14157         TARGET_LONG_DOUBLE_128 is set.
14159 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14161         * cfgloop.c (get_loop_location): Convert return type from
14162         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14163         by implicit construction from rtx_insn *, and using
14164         dump_user_location_t::from_function_decl for the fallback case.
14165         * cfgloop.h (get_loop_location): Convert return type from
14166         location_t to dump_user_location_t.
14167         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14168         dump_printf_loc to pass in a dump_location_t rather than a
14169         location_t, via the gimple stmt.
14170         * coverage.c (get_coverage_counts): Update calls to
14171         dump_printf_loc to pass in dump_location_t rather than a
14172         location_t.
14173         * doc/optinfo.texi (Dump types): Convert example of
14174         dump_printf_loc from taking "locus" to taking "insn".  Update
14175         description of the "_loc" calls to cover dump_location_t.
14176         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14177         "selftest.h".
14178         (dump_user_location_t::dump_user_location_t): New constructors,
14179         from gimple *stmt and rtx_insn *.
14180         (dump_user_location_t::from_function_decl): New function.
14181         (dump_loc): Make static.
14182         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14183         const dump_location_t &.
14184         (dump_generic_expr_loc): Delete.
14185         (dump_printf_loc): Convert param "loc" from location_t to
14186         const dump_location_t &.
14187         (selftest::test_impl_location): New function.
14188         (selftest::dumpfile_c_tests): New function.
14189         * dumpfile.h: Include "profile-count.h".
14190         (class dump_user_location_t): New class.
14191         (struct dump_impl_location_t): New struct.
14192         (class dump_location_t): New class.
14193         (dump_printf_loc): Convert 2nd param from source_location to
14194         const dump_location_t &.
14195         (dump_generic_expr_loc): Delete.
14196         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14197         const dump_location_t &.
14198         * gimple-fold.c (fold_gimple_assign): Update call to
14199         dump_printf_loc to pass in a dump_location_t rather than a
14200         location_t, via the gimple stmt.
14201         (gimple_fold_call): Likewise.
14202         * gimple-loop-interchange.cc
14203         (loop_cand::analyze_iloop_reduction_var): Update for change to
14204         check_reduction_path.
14205         (tree_loop_interchange::interchange): Update for change to
14206         find_loop_location.
14207         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14208         change in return-type of find_loop_location.
14209         (graphite_regenerate_ast_isl): Likewise.
14210         * graphite-optimize-isl.c (optimize_isl): Likewise.
14211         * graphite.c (graphite_transform_loops): Likewise.
14212         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14213         pass in a dump_location_t rather than a location_t, via the
14214         gimple stmt.
14215         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14216         * ipa.c (walk_polymorphic_call_targets): Likewise.
14217         * loop-unroll.c (report_unroll): Convert "locus" param from
14218         location_t to dump_location_t.
14219         (decide_unrolling): Update for change to get_loop_location's
14220         return type.
14221         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14222         location_t to dump_user_location_t.
14223         (grid_find_single_omp_among_assignments_1): Updates calls to
14224         dump_printf_loc to pass in a dump_location_t rather than a
14225         location_t, via the gimple stmt.
14226         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14227         location_t to dump_location_t.  Updates calls to dump_printf_loc
14228         to pass in a dump_location_t rather than a location_t, via the
14229         gimple stmt.
14230         (grid_inner_loop_gridifiable_p): Likewise.
14231         (grid_dist_follows_simple_pattern): Likewise.
14232         (grid_gfor_follows_tiling_pattern): Likewise.
14233         (grid_target_follows_gridifiable_pattern): Likewise.
14234         (grid_attempt_target_gridification): Convert initialization
14235         of local "grid" from memset to zero-initialization; FIXME: does
14236         this require C++11?  Update call to dump_printf_loc to pass in a
14237         optinfo_location rather than a location_t, via the gimple stmt.
14238         * profile.c (read_profile_edge_counts): Updates call to
14239         dump_printf_loc to pass in a dump_location_t rather than a
14240         location_t
14241         (compute_branch_probabilities): Likewise.
14242         * selftest-run-tests.c (selftest::run_tests): Call
14243         dumpfile_c_tests.
14244         * selftest.h (dumpfile_c_tests): New decl.
14245         * tree-loop-distribution.c (pass_loop_distribution::execute):
14246         Update for change in return type of find_loop_location.
14247         * tree-parloops.c (parallelize_loops): Likewise.
14248         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14249         "locus" from location_t to dump_user_location_t.
14250         (canonicalize_loop_induction_variables): Likewise.
14251         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14252         for change in return type of find_loop_location.
14253         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14254         to dump_printf_loc to pass in a dump_location_t rather than a
14255         location_t, via the stmt.
14256         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14257         Likewise.
14258         * tree-vect-loop-manip.c (find_loop_location): Convert return
14259         type from source_location to dump_user_location_t.
14260         (vect_do_peeling): Update for above change.
14261         (vect_loop_versioning): Update for change in type of
14262         vect_location.
14263         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14264         from location_t to dump_user_location_t.
14265         (vect_estimate_min_profitable_iters): Update for change in type
14266         of vect_location.
14267         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14268         location_t to dump_location_t.
14269         (vect_slp_bb): Update for change in type of vect_location.
14270         * tree-vectorizer.c (vect_location): Convert from source_location
14271         to dump_user_location_t.
14272         (try_vectorize_loop_1): Update for change in vect_location's type.
14273         (vectorize_loops): Likewise.
14274         (increase_alignment): Likewise.
14275         * tree-vectorizer.h (vect_location): Convert from source_location
14276         to dump_user_location_t.
14277         (find_loop_location): Convert return type from source_location to
14278         dump_user_location_t.
14279         (check_reduction_path): Convert 1st param from location_t to
14280         dump_user_location_t.
14281         * value-prof.c (check_counter): Update call to dump_printf_loc to
14282         pass in a dump_user_location_t rather than a location_t; update
14283         call to error_at for change in type of "locus".
14284         (check_ic_target): Update call to dump_printf_loc to
14285         pass in a dump_user_location_t rather than a location_t, via the
14286         call_stmt.
14288 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14290         * config/s390/s390.h (enum processor_flags): Do not use
14291         default tune parameter when -march was specified.
14293 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14295         PR target/86314
14296         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14297         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14298         operands.
14300 2018-06-26  Richard Biener  <rguenther@suse.de>
14302         PR tree-optimization/86287
14303         PR bootstrap/86316
14304         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14305         (vect_analyze_loop): Initialize n_stmts.
14307 2018-06-26  Richard Biener  <rguenther@suse.de>
14309         PR middle-end/86271
14310         * fold-const.c (fold_convertible_p): Pointer extension
14311         isn't valid.
14313 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14315         PR debug/86064
14316         * dwarf2out.c (loc_list_has_views): Adjust comments.
14317         (dw_loc_list): Split single cross-partition range with
14318         nonzero locview.
14320 2018-06-25  Jeff Law  <law@redhat.com>
14322         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14323         on -mbig-switch by default.
14325         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14326         test.
14327         (const_float_0_operand): Remove unused predicate.
14328         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14329         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14330         (recipsf2): New expander.  Original pattern now called
14331         (recipsf2_insn).
14332         (recipdf2, recipdf2_insn): Similarly.
14333         (rsqrtsf2, rsqrtsf2_insn): Similarly
14334         (rsqrtdf2, rsqrtdf2_insn): Similarly
14336 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14338         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14339         Simplify logic for FreeBSD (twice).
14341 2018-06-25  Martin Sebor  <msebor@redhat.com>
14343         PR tree-optimization/86204
14344         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14345         a strnlen result if it's less than the length of the string.
14347 2018-06-25  Martin Sebor  <msebor@redhat.com>
14349         PR tree-optimization/85700
14350         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14351         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14352         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14354 2018-06-25  Martin Sebor  <msebor@redhat.com>
14356         * doc/extend.texi (Zero-length arrays): Update and clarify.
14358 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14360         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14361         added IEEE/IBM long double multilib support on PowerPC little
14362         endian Linux systems.
14363         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14364         (MULTILIB_DEFAULTS): Likewise.
14365         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14366         Likewise.
14367         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14368         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14369         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14371 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14373         PR middle-end/86311
14374         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14375         (REORDER_45): Likewise.
14377 2018-06-25  Jeff Law  <law@redhat.com>
14379         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14380         dividend to 32 bits.  Adjust length.
14381         (udivmodhi4): Cleanup output template.  Fix length.
14383 2018-06-25  Carl Love  <cel@us.ibm.com>
14385         * config/rs6000/vsx.md: Change word selector to prefered location.
14387 2018-06-25  Richard Biener  <rguenther@suse.de>
14389         PR tree-optimization/86304
14390         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14391         epilogue-if-converted loops as well.
14393 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14395         * lto-section-out.c (lto_begin_section): Do not print section
14396         name for noaddr and unnumbered dumps.
14398 2018-06-25  Richard Biener  <rguenther@suse.de>
14400         * tree-vectorizer.h (struct vec_info_shared): New structure
14401         with parts split out from struct vec_info and loop_nest from
14402         struct _loop_vec_info.
14403         (struct vec_info): Adjust accordingly.
14404         (struct _loop_vec_info): Likewise.
14405         (LOOP_VINFO_LOOP_NEST): Adjust.
14406         (LOOP_VINFO_DATAREFS): Likewise.
14407         (LOOP_VINFO_DDRS): Likewise.
14408         (struct _bb_vec_info): Likewise.
14409         (BB_VINFO_DATAREFS): Likewise.
14410         (BB_VINFO_DDRS): Likewise.
14411         (struct _stmt_vec_info): Add dr_aux member.
14412         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14413         (DR_MISALIGNMENT_UNINITIALIZED): New.
14414         (set_dr_misalignment): Adjust.
14415         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14416         (vect_analyze_loop): Adjust prototype.
14417         (vect_analyze_loop_form): Likewise.
14418         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14419         Compute dependences lazily.
14420         (vect_record_base_alignments): Use shared datarefs/ddrs.
14421         (vect_verify_datarefs_alignment): Likewise.
14422         (vect_analyze_data_refs_alignment): Likewise.
14423         (vect_analyze_data_ref_accesses): Likewise.
14424         (vect_analyze_data_refs): Likewise.
14425         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14426         constructor parameter for shared part.
14427         (vect_analyze_loop_form): Pass in shared part and adjust.
14428         (vect_analyze_loop_2): Pass in storage for the number of
14429         stmts.  Move loop nest finding to the caller.  Compute
14430         datarefs lazily.
14431         (vect_analyze_loop): Pass in shared part.
14432         (vect_transform_loop): Verify shared datarefs are unchanged.
14433         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14434         constructor parameter for shared part.
14435         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14436         (vect_slp_bb): Verify shared datarefs are unchanged before
14437         transform.
14438         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14439         change.
14440         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14441         DR_MISALIGNMENT_UNINITIALIZED.
14442         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14443         parameter for shared part.
14444         (vec_info::~vec_info): Adjust.
14445         (vec_info_shared::vec_info_shared): New.
14446         (vec_info_shared::~vec_info_shared): Likewise.
14447         (vec_info_shared::save_datarefs): Likewise.
14448         (vec_info_shared::check_datarefs): Likewise.
14449         (try_vectorize_loop_1): Construct shared part live for analyses
14450         of a single loop for multiple vector sizes.
14451         * tree-parloops.c (gather_scalar_reductions): Adjust.
14453 2018-06-25  Richard Biener  <rguenther@suse.de>
14455         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14456         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14457         (vect_analyze_data_refs): Remove similar code from here and
14458         simplify accordingly.
14460 2018-06-25  Richard Biener  <rguenther@suse.de>
14462         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14463         for reverse storage order accesses rather than asserting
14464         they cannot happen here.
14466 2018-06-25  Tom de Vries  <tdevries@suse.de>
14468         PR debug/86257
14469         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14470         Use data16 instead of .byte for insn prefix.
14472 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14474         PR C++/86082
14475         * parser.c (make_char_string_pack): Pass this literal chars
14476         through cpp_interpret_string.
14477         (cp_parser_userdef_numeric_literal): Check the result of
14478         make_char_string_pack.
14480 2018-06-24  Maya Rashish  <coypu@sdf.org>
14482         * ginclude/stddef.h: Simplify conditions around avoiding
14483         re-definition of __size_t.
14485 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14487         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14488         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14490 2018-06-22  Maya Rashish  <coypu@sdf.org>
14492         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14493         made to OpenBSD and NetBSD through the years.
14495 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14497         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14498         behavior of vec_pack (vector double, vector double) to match
14499         behavior of vec_float2 (vector double, vector double).
14501 2018-06-22  Olivier Hainque  <hainque@adacore.com>
14503         * gimplify.c (gimplify_function_tree): Prevent creation
14504         of a trampoline for the address of the current function
14505         passed to entry/exit instrumentation hooks.
14507 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
14509         PR target/86222
14510         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
14511         correctly.
14513 2018-06-22  Martin Liska  <mliska@suse.cz>
14515         PR tree-optimization/86263
14516         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14517         Bail out if is_enabled is false.
14518         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
14519         New declaration.
14520         (jump_table_cluster::is_enabled): New function.
14522 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14524         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14525         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
14526         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
14527         (lto_input_ts_binfo_tree_pointers): Likewise.
14528         * tree-streamer-out.c (streamer_write_tree_bitfields,
14529         write_ts_binfo_tree_pointers): Likewise.
14530         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
14532 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14534         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
14536 2018-06-22  Martin Liska  <mliska@suse.cz>
14538         * symbol-summary.h (get): Make it pure and inline move
14539         functionality from ::get function.
14540         (get): Remove and inline into ::get and ::get_create.
14541         (get_create): Move code from ::get function.
14543 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14545         PR target/85994
14546         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
14547         -x assembler-with-cpp.
14549 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14551         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
14552         _FILE_OFFSET_BITS=64 for C++.
14554 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
14556         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
14557         conversion insn that shows up when pr85657-3.c is compiled using
14558         IEEE 128-bit long double.
14559         (neg<mode>2_internal): Use the correct mode to check whether the
14560         mode is IBM extended.
14561         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
14562         multiply and divide external functions from being created more
14563         than once.
14565 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14567         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
14568         functions.
14569         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
14570         the edge can be forwarded.
14571         (cfg_layout_merge_blocks): Likewise.
14573 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14575         * except.c (finish_eh_generation): Commit edge insertions only after
14576         the EH edges have been redirected from post-landing to landing pads.
14578 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14580         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
14581         create_tmp_var_for to create the FRAME decl.
14582         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
14584 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14586         * tree-inline.c (copy_edges_for_bb): Minor tweak.
14587         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
14588         debug statement when resetting its value.
14589         (expand_call_inline): Copy the locus of the call onto the assignment
14590         of the return value, if any.  Use local variable in more cases.
14592 2018-06-21  Martin Liska  <mliska@suse.cz>
14594         * ipa-pure-const.c (propagate_nothrow): Use
14595         funct_state_summaries->get.
14596         (dump_malloc_lattice): Likewise.
14597         (propagate_malloc): Likewise.
14599 2018-06-21  Richard Biener  <rguenther@suse.de>
14601         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
14602         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
14603         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
14604         comment.
14605         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
14606         BLOCK_ABSTRACT_ORIGIN unconditionally.
14608 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
14610         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
14611         deleting it.
14612         * ipa-reference.c (ipa_reference_c_finalize): Delete
14613         ipa_ref_opt_sum_summaries and set it to NULL.
14615 2018-06-21  Tom de Vries  <tdevries@suse.de>
14617         PR tree-optimization/85859
14618         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
14619         test with comment from bb_no_side_effects_p.
14621 2018-06-21  Richard Biener  <rguenther@suse.de>
14623         PR tree-optimization/86232
14624         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
14625         max for constant niter.
14627 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14629         * config/aarch64/aarch64-simd.md
14630         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
14632 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14634         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
14635         Make opernads of the unspec commutative.
14637 2018-06-21  Richard Biener  <rguenther@suse.de>
14639         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
14640         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14641         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
14642         (vect_analyze_data_ref_dependence): Re-order checks to deal with
14643         NULL DR_STEP.
14644         (vect_record_base_alignments): Do not record base alignment
14645         for gathers or scatters.
14646         (vect_compute_data_ref_alignment): Drop return value that is always
14647         true.  Bail out early for gathers or scatters.
14648         (vect_enhance_data_refs_alignment): Bail out early for gathers
14649         or scatters.
14650         (vect_find_same_alignment_drs): Likewise.
14651         (vect_analyze_data_refs_alignment): Remove dead code.
14652         (vect_slp_analyze_and_verify_node_alignment): Likewise.
14653         (vect_analyze_data_refs): For possible gathers or scatters do
14654         not create an alternate DR, just check their possible validity
14655         and mark them.  Adjust DECL_NONALIASED handling to not rely
14656         on DR_BASE_ADDRESS.
14657         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
14658         update inits of gathers or scatters.
14659         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
14660         Also copy gather/scatter flag to pattern vinfo.
14662 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14664         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14665         behavior of vec_packsu (vector unsigned long long, vector unsigned
14666         long long) to match behavior of vec_packs with same signature.
14668 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
14669             Thomas Schwinge <thomas@codesourcery.com>
14670             Cesar Philippidis  <cesar@codesourcery.com>
14672         * gimplify.c (gimplify_scan_omp_clauses): Add support for
14673         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14674         (gimplify_adjust_omp_clauses): Likewise.
14675         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
14676         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
14677         (gimplify_omp_target_update): Update handling of acc update and
14678         enter/exit data.
14679         * omp-low.c (install_var_field): Remove unused parameter
14680         base_pointers_restrict.
14681         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
14682         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
14683         FINALIZE}
14684         (omp_target_base_pointers_restrict_p): Delete.
14685         (scan_omp_target): Update call to scan_sharing_clauses.
14686         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
14687         FINALIZE}.
14688         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
14689         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14690         (convert_local_omp_clauses): Likewise.
14691         * tree-pretty-print.c (dump_omp_clause): Likewise.
14692         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
14693         FINALIZE}.
14694         (omp_clause_code_name): Likewise.
14696 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14698         PR debug/86194
14699         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
14700         be narrowed.
14702         PR tree-optimization/86231
14703         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
14704         anti-range don't overwrite *vr0min before using it to compute *vr0max.
14706 2018-06-20  Tom de Vries  <tdevries@suse.de>
14708         PR tree-optimization/86097
14709         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
14710         iv type if signedness of iv type is not the same as that of *nit.
14712 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14714         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
14715         EDGE_EH edges, verify they are all EDGE_EH.
14717 2018-06-20  Maya Rashish  <coypu@sdf.org>
14719         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
14721 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14723         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
14724         * config/aarch64/aarch64.c (xgene1_tunings): Add
14725         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
14726         (aarch64_mode_valid_for_sched_fusion_p):
14727         Allow 16-byte modes.
14728         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
14729         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
14730         128-bit modes.
14731         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
14732         New pattern.
14733         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
14734         * config/aarch64/iterators.md (VQ2): New mode iterator.
14736 2018-06-20  Martin Liska  <mliska@suse.cz>
14738         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14739         Change default ratio from 10 to 8.
14741 2018-06-20  Martin Liska  <mliska@suse.cz>
14743         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14744         New.
14745         (bit_test_cluster::find_bit_tests): Likewise.
14746         (switch_decision_tree::analyze_switch_statement): Find clusters.
14747         * tree-switch-conversion.h (struct jump_table_cluster): Document
14748         hierarchy.
14750 2018-06-20  Martin Liska  <mliska@suse.cz>
14752         * tree-switch-conversion.c (switch_conversion::collect):
14753         Record m_uniq property.
14754         (switch_conversion::expand): Bail out for special conditions.
14755         (group_cluster::~group_cluster): New.
14756         (group_cluster::group_cluster): Likewise.
14757         (group_cluster::dump): Likewise.
14758         (jump_table_cluster::emit): New.
14759         (switch_decision_tree::fix_phi_operands_for_edges): New.
14760         (struct case_node): Remove struct.
14761         (jump_table_cluster::can_be_handled): New.
14762         (case_values_threshold): Moved to header.
14763         (reset_out_edges_aux): Likewise.
14764         (jump_table_cluster::is_beneficial): New.
14765         (bit_test_cluster::can_be_handled): Likewise.
14766         (add_case_node): Remove.
14767         (bit_test_cluster::is_beneficial): New.
14768         (case_bit_test::cmp): New.
14769         (bit_test_cluster::emit): New.
14770         (expand_switch_as_decision_tree_p): Remove.
14771         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
14772         (fix_phi_operands_for_edge): Likewise.
14773         (switch_decision_tree::analyze_switch_statement): New.
14774         (compute_cases_per_edge): Move ...
14775         (switch_decision_tree::compute_cases_per_edge): ... here.
14776         (try_switch_expansion): Likewise.
14777         (switch_decision_tree::try_switch_expansion): Likewise.
14778         (record_phi_operand_mapping): Likewise.
14779         (switch_decision_tree::record_phi_operand_mapping): Likewise.
14780         (emit_case_decision_tree): Likewise.
14781         (switch_decision_tree::emit): Likewise.
14782         (balance_case_nodes): Likewise.
14783         (switch_decision_tree::balance_case_nodes): Likewise.
14784         (dump_case_nodes): Likewise.
14785         (switch_decision_tree::dump_case_nodes): Likewise.
14786         (emit_jump): Likewise.
14787         (switch_decision_tree::emit_jump): Likewise.
14788         (emit_cmp_and_jump_insns): Likewise.
14789         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
14790         (emit_case_nodes): Likewise.
14791         (switch_decision_tree::emit_case_nodes): Likewise.
14792         (conditional_probability): Remove.
14793         * tree-switch-conversion.h (enum cluster_type): New.
14794         (PRINT_CASE): New.
14795         (struct cluster): Likewise.
14796         (cluster::cluster): Likewise.
14797         (struct simple_cluster): Likewise.
14798         (simple_cluster::simple_cluster): Likewise.
14799         (struct group_cluster): Likewise.
14800         (struct jump_table_cluster): Likewise.
14801         (struct bit_test_cluster): Likewise.
14802         (struct min_cluster_item): Likewise.
14803         (struct case_tree_node): Likewise.
14804         (case_tree_node::case_tree_node): Likewise.
14805         (jump_table_cluster::case_values_threshold): Likewise.
14806         (struct case_bit_test): Likewise.
14807         (struct switch_decision_tree): Likewise.
14808         (struct switch_conversion): Likewise.
14809         (switch_decision_tree::reset_out_edges_aux): Likewise.
14811 2018-06-20  Martin Liska  <mliska@suse.cz>
14813         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
14814         (hoist_edge_and_branch_if_true): Likewise.
14815         (expand_switch_using_bit_tests_p): Likewise.
14816         (struct case_bit_test): Likewise.
14817         (case_bit_test_cmp): Likewise.
14818         (emit_case_bit_tests): Likewise.
14819         (switch_conversion::switch_conversion): New class.
14820         (struct switch_conv_info): Remove old struct.
14821         (collect_switch_conv_info): More to ...
14822         (switch_conversion::collect): ... this.
14823         (check_range): Likewise.
14824         (switch_conversion::check_range): Likewise.
14825         (check_all_empty_except_final): Likewise.
14826         (switch_conversion::check_all_empty_except_final): Likewise.
14827         (check_final_bb): Likewise.
14828         (switch_conversion::check_final_bb): Likewise.
14829         (create_temp_arrays): Likewise.
14830         (switch_conversion::create_temp_arrays): Likewise.
14831         (free_temp_arrays): Likewise.
14832         (gather_default_values): Likewise.
14833         (switch_conversion::gather_default_values): Likewise.
14834         (build_constructors): Likewise.
14835         (switch_conversion::build_constructors): Likewise.
14836         (constructor_contains_same_values_p): Likewise.
14837         (switch_conversion::contains_same_values_p): Likewise.
14838         (array_value_type): Likewise.
14839         (switch_conversion::array_value_type): Likewise.
14840         (build_one_array): Likewise.
14841         (switch_conversion::build_one_array): Likewise.
14842         (build_arrays): Likewise.
14843         (switch_conversion::build_arrays): Likewise.
14844         (gen_def_assigns): Likewise.
14845         (switch_conversion::gen_def_assigns): Likewise.
14846         (prune_bbs): Likewise.
14847         (switch_conversion::prune_bbs): Likewise.
14848         (fix_phi_nodes): Likewise.
14849         (switch_conversion::fix_phi_nodes): Likewise.
14850         (gen_inbound_check): Likewise.
14851         (switch_conversion::gen_inbound_check): Likewise.
14852         (process_switch): Use the newly created class.
14853         (switch_conversion::expand): New.
14854         (switch_conversion::~switch_conversion): New.
14855         * tree-switch-conversion.h: New file.
14857 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14859         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
14860         tree-vect-patterns.c.
14861         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
14862         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
14863         (vect_recog_sad_pattern): Likewise.
14864         (vect_recog_widen_sum_pattern): Likewise.
14865         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
14866         (vect_recog_widen_shift_pattern): Remove the type_in argument.
14867         (vect_recog_rotate_pattern): Likewise.
14868         (vect_recog_mult_pattern): Likewise.
14869         (vect_recog_vector_vector_shift_pattern): Likewise.
14870         (vect_recog_divmod_pattern): Likewise.
14871         (vect_recog_mixed_size_cond_pattern): Likewise.
14872         (vect_recog_bool_pattern): Likewise.
14873         (vect_recog_mask_conversion_pattern): Likewise.
14874         (vect_try_gather_scatter_pattern): Likewise.
14875         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
14876         (vect_recog_over_widening_pattern): Likewise.
14877         (vect_recog_gather_scatter_pattern): Likewise.
14878         (vect_recog_func_ptr): Move from tree-vectorizer.h
14879         (vect_vect_recog_func_ptrs): Move further down the file.
14880         (vect_recog_func): Likewise.  Remove the third argument.
14881         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
14882         (vect_pattern_recog_1): Expect the pattern function to do any
14883         necessary target tests.  Also expect it to provide a vector type.
14884         Remove the type_in handling.
14886 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14888         * tree-vect-patterns.c (vect_pattern_detected): New function.
14889         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
14890         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
14891         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
14892         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
14893         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14894         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14895         (vect_recog_mask_conversion_pattern)
14896         (vect_try_gather_scatter_pattern): Likewise.
14898 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14900         * tree-vect-patterns.c (vect_get_internal_def): New function.
14901         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14902         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
14903         (search_type_for_mask_1): Use it.
14905 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14907         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14908         redundant WIDEN_SUM_EXPR handling.
14909         (vect_recog_sad_pattern): Likewise.
14911 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14913         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14914         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
14915         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
14916         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
14917         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
14919 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14921         * tree-vect-stmts.c (vectorizable_call): Make sure that we
14922         use the stmt_vec_info of the original bb statement for the
14923         new zero assignment, even if the call is part of a pattern.
14925 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14927         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
14928         that the sequence is attached to the original statement rather
14929         than the pattern statement.
14930         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
14931         PATTERN_DEF_SEQ from the original statement rather than
14932         the main pattern statement.
14933         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
14934         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
14935         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
14937 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14939         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
14940         definition statements before the early exit for statements that aren't
14941         live or relevant.
14942         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
14943         split out from...
14944         (vect_transform_loop): ...here.  Process pattern definition
14945         statements without first checking whether the main pattern
14946         statement is live or relevant.
14948 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
14950         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
14951         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
14953 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
14955         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
14956         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
14957         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
14958         (expand_block_compare): Change select_block_compare_mode call.
14959         (expand_strncmp_align_check): Use new functions, fix comment.
14960         (emit_final_str_compare_gpr): New function.
14961         (expand_strn_compare): Refactor and clean up code.
14962         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
14964 2018-06-19  Tony Reix  <tony.reix@atos.com>
14965             Damien Bergamini  <damien.bergamini@atos.com>
14966             David Edelsohn  <dje.gcc@gmail.com>
14968         * collect2.c (static_obj): New variable.
14969         (static_libs): New variable.
14970         (is_in_list): Uncomment declaration.
14971         (main): Track AIX libraries linked statically.
14972         (is_in_list): Uncomment definition.
14973         (scan_prog_file): Don't add AIX shared libraries initializer
14974         to constructor list if linking statically.
14976 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
14978         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
14979         constant.
14980         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
14982 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
14984         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
14985         blocks.
14987 2018-06-19  Martin Liska  <mliska@suse.cz>
14989         * config/i386/i386.c (ix86_can_inline_p): Do not use
14990         ipa_fn_summaries::get_create.
14991         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
14992         get.
14993         (devirtualization_time_bonus): Likewise.
14994         (ipcp_propagate_stage): Likewise.
14995         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14996         (edge_set_predicate): Likewise.
14997         (evaluate_conditions_for_known_args): Likewise.
14998         (evaluate_properties_for_edge): Likewise.
14999         (ipa_call_summary::reset): Tranform to ...
15000         (ipa_call_summary::~ipa_call_summary): ... this.
15001         (ipa_fn_summary::reset): Transform to ...
15002         (ipa_fn_summary::~ipa_fn_summary): ... this.
15003         (ipa_fn_summary_t::remove): Rename to ...
15004         (ipa_fn_summary_t::remove_callees): ... this.
15005         (ipa_fn_summary_t::duplicate): Use placement new
15006         instead of memory copy.
15007         (ipa_call_summary_t::duplicate): Likewise.
15008         (ipa_call_summary_t::remove): Remove.
15009         (dump_ipa_call_summary): Change get_create to get.
15010         (ipa_dump_fn_summary): Dump only when summary exists.
15011         (analyze_function_body): Use symbol_summary::get instead
15012         of get_create.
15013         (compute_fn_summary): Likewise.
15014         (estimate_edge_devirt_benefit): Likewise.
15015         (estimate_edge_size_and_time): Likewise.
15016         (inline_update_callee_summaries): Likewise.
15017         (remap_edge_change_prob): Likewise.
15018         (remap_edge_summaries): Likewise.
15019         (ipa_merge_fn_summary_after_inlining): Likewise.
15020         (write_ipa_call_summary): Likewise.
15021         (ipa_fn_summary_write): Likewise.
15022         (ipa_free_fn_summary): Likewise.
15023         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15024         (struct ipa_call_summary): Likewise.
15025         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15026         of get_create.
15027         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15028         (estimate_size_after_inlining): Likewise.
15029         (estimate_growth): Likewise.
15030         (growth_likely_positive): Likewise.
15031         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15032         (inline_call): Likewise.
15033         * ipa-inline.c (caller_growth_limits): Likewise.
15034         (can_inline_edge_p): Likewise.
15035         (can_inline_edge_by_limits_p): Likewise.
15036         (compute_uninlined_call_time): Likewise.
15037         (compute_inlined_call_time): Likewise.
15038         (want_inline_small_function_p): Likewise.
15039         (edge_badness): Likewise.
15040         (update_caller_keys): Likewise.
15041         (update_callee_keys): Likewise.
15042         (inline_small_functions): Likewise.
15043         (inline_to_all_callers_1): Likewise.
15044         (dump_overall_stats): Likewise.
15045         (early_inline_small_functions): Likewise.
15046         (early_inliner): Likewise.
15047         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15048         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15049         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15050         * ipa-split.c (execute_split_functions): Likewise.
15051         * symbol-summary.h: Likewise.
15052         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15054 2018-06-19  Richard Biener  <rguenther@suse.de>
15056         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15057         (vectorize_loops): ... here.  Fix dbgcnt handling.
15058         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15060 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15062         PR target/86197
15063         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15064         ieee128 argument takes up only one (vector) register, not two (floating
15065         point) registers.
15067 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15069         * gimplify.c (gimplify_init_constructor): Really never clear for an
15070         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15072 2018-06-19  Richard Biener  <rguenther@suse.de>
15074         PR tree-optimization/86179
15075         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15076         after failed recognition.
15078 2018-06-18  Martin Sebor  <msebor@redhat.com>
15080         PR middle-end/85602
15081         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15082         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15083         Handle integer subtraction.
15084         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15085         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15087 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15089         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15090         param from rtx to rtx_insn *.
15091         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15092         param.
15093         (frv_ifcvt_modify_insn): Likwise.
15094         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15095         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15096         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15097         as_a <rtx_insn *> cast to local "unprotected_region" once
15098         it's been established that it's not NULL or pc_rtx.
15099         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15100         param "sethi" from rtx to rtx_insn *.
15101         (nds32_group_float_insns): Likewise for param "insn".
15102         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15103         param.
15104         (vax_output_int_subtract): Likewise.
15105         * config/vax/vax.c (vax_output_int_add): Likewise for param
15106         "insn".
15107         (vax_output_int_subtract): Likewise.
15108         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15109         (emit_pattern_after): Likewise for param "after".
15110         (emit_insn_after): Likewise.
15111         (emit_jump_insn_after): Likewise.
15112         (emit_call_insn_after): Likewise.
15113         (emit_debug_insn_after): Likewise.
15114         (emit_pattern_before): Likewise for param "before".
15115         (emit_insn_before): Likewise.
15116         (emit_jump_insn_before): Likewise.
15117         * final.c (get_insn_template): Likewise for param "insn", removing
15118         a cast.
15119         * output.h (get_insn_template): Likewise for 2nd param.
15120         * rtl.h (emit_insn_before): Likewise.
15121         (emit_jump_insn_before): Likewise.
15122         (emit_debug_insn_before_noloc): Likewise.
15123         (emit_insn_after): Likewise.
15124         (emit_jump_insn_after): Likewise.
15125         (emit_call_insn_after): Likewise.
15126         (emit_debug_insn_after): Likewise.
15127         (set_insn_deleted): Likewise for param.
15129 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15131         PR target/85358
15132         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15133         floating point modes, so that IFmode is numerically greater than
15134         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15135         to declare the ordering.  This prevents IFmode from being
15136         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15137         machine.  Include rs6000-modes.h to share the fractional values
15138         between genmodes* and the rest of the compiler.
15139         (IFmode): Likewise.
15140         (KFmode): Likewise.
15141         (TFmode): Likewise.
15142         * config/rs6000/rs6000-modes.h: New file.
15143         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15144         meaning of rs6000_long_double_size so that 126..128 selects an
15145         appropriate 128-bit floating point type.
15146         (rs6000_option_override_internal): Likewise.
15147         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15148         (TARGET_LONG_DOUBLE_128): Change the meaning of
15149         rs6000_long_double_size so that 126..128 selects an appropriate
15150         128-bit floating point type.
15151         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15152         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15153         source and destination to match the standard usage.
15154         (truncifkf2): Likewise.
15155         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15156         ISA 2.07 to use an explicit clobber, instead of passing in a
15157         temporary.
15158         (copysign<mode>3_soft): Likewise.
15160 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15162         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15163         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15164         (vect_slp_analyze_instance_dependence): Likewise.
15165         (vect_enhance_data_refs_alignment): Likewise.
15166         (vect_analyze_data_refs_alignment): Likewise.
15167         (vect_slp_analyze_and_verify_instance_alignment
15168         (vect_analyze_data_ref_accesses): Likewise.
15169         (vect_prune_runtime_alias_test_list): Likewise.
15170         (vect_analyze_data_refs): Likewise.
15171         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15172         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15173         (vect_analyze_scalar_cycles_1): Likewise.
15174         (vect_get_loop_niters): Likewise.
15175         (vect_analyze_loop_form_1): Likewise.
15176         (vect_update_vf_for_slp): Likewise.
15177         (vect_analyze_loop_operations): Likewise.
15178         (vect_analyze_loop): Likewise.
15179         (vectorizable_induction): Likewise.
15180         (vect_transform_loop): Likewise.
15181         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15182         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15183         (vect_make_slp_decision): Likewise.
15184         (vect_detect_hybrid_slp): Likewise.
15185         (vect_slp_analyze_operations): Likewise.
15186         (vect_slp_bb): Likewise.
15187         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15188         (vectorizable_bswap): Likewise.
15189         (vectorizable_call): Likewise.
15190         (vectorizable_simd_clone_call): Likewise.
15191         (vectorizable_conversion): Likewise.
15192         (vectorizable_assignment): Likewise.
15193         (vectorizable_shift): Likewise.
15194         (vectorizable_operation): Likewise.
15195         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15197 2018-06-18  Martin Sebor  <msebor@redhat.com>
15199         PR tree-optimization/81384
15200         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15201         * builtins.c (expand_builtin_strnlen): New function.
15202         (expand_builtin): Call it.
15203         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15204         * builtins.def (BUILT_IN_STRNLEN): New.
15205         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15206         Warn for bounds in excess of maximum object size.
15207         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15208         single-value ranges.  Handle strnlen.
15209         (handle_builtin_strlen): Handle strnlen.
15210         (strlen_check_and_optimize_stmt): Same.
15211         * doc/extend.texi (Other Builtins): Document strnlen.
15213 2018-06-18  Maya Rashish  <coypu@sdf.org>
15215         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15216         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15217         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15219         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15220         here to ...
15221         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15223 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15225         * tree.c (escaped_string::escape): Replace cast to char * by
15226         const_cast<char *> (unescaped).
15228 2018-06-18  Nick Clifton  <nickc@redhat.com>
15230         PR 84195
15231         * tree.c (escaped_string): New class.  Converts an unescaped
15232         string into its escaped equivalent.
15233         (warn_deprecated_use): Use the new class to convert the
15234         deprecation message, if present.
15235         (test_escaped_strings): New self test.
15236         (test_c_tests): Add test_escaped_strings.
15237         * doc/extend.texi (deprecated): Add a note that the
15238         deprecation message is affected by the -fmessage-length
15239         option, and that control characters will be escaped.
15240         (#pragma GCC error): Document this pragma.
15241         (#pragma GCC warning): Likewise.
15242         * doc/invoke.texi (-fmessage-length): Document this option's
15243         effect on the #warning and #error preprocessor directives and
15244         the deprecated attribute.
15246 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15248         * tree.c (decl_value_expr_lookup): Revert latest change.
15249         (decl_value_expr_insert): Likewise.
15251 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15253         * gimplify.c (nonlocal_vlas): Delete.
15254         (nonlocal_vla_vars): Likewise.
15255         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15256         referenced VLAs.
15257         (gimplify_body): Do not create and destroy nonlocal_vlas.
15258         * tree-nested.c: Include diagnostic.h.
15259         (use_pointer_in_frame): Tweak.
15260         (lookup_field_for_decl): Add assertion and declare the transformation.
15261         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15262         internal error when the reference is in a wrong context.  Do not
15263         create a debug decl by default.
15264         (note_nonlocal_block_vlas): Delete.
15265         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15266         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15267         create a debug decl by default.
15268         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15269         call is in a wrong context.
15270         (fixup_vla_decls): New function.
15271         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15272         debug variables were created.
15273         * tree.c (decl_value_expr_lookup): Add checking assertion.
15274         (decl_value_expr_insert): Likewise.
15276 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15278         PR middle-end/82479
15279         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15280         * tree-scalar-evolution.c (interpret_expr): Likewise.
15281         (expression_expensive_p): Likewise.
15282         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15283         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15284         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15285         (ssa_defined_by_minus_one_stmt_p): New.
15287 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15289         PR middle-end/64946
15290         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15291         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15292         * dojump.c (do_jump): Likewise.
15293         * expr.c (expand_expr_real_2): Check operand type's sign.
15294         * fold-const.c (const_unop): Handle ABSU_EXPR.
15295         (fold_abs_const): Likewise.
15296         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15297         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15298         (strip_sign_op_1): Likesise.
15299         * match.pd: Add new pattern to generate ABSU_EXPR.
15300         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15301         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15302         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15303         * tree-inline.c (estimate_operator_cost): Likewise.
15304         * tree-pretty-print.c (dump_generic_node): Likewise.
15305         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15306         * tree.def (ABSU_EXPR): New.
15308 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15310         PR middle-end/86095
15311         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15312         documented as preserved for backward compatibility only.
15313         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15315         PR rtl-optimization/86108
15316         * bb-reorder.c (create_forwarder_block): Renamed to ...
15317         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15318         jump from new landing pad to the second part.
15319         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15320         Adjust callers.
15322 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15324         PR middle-end/85878
15325         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15326         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15327         Only call store_expr for halves if the mode is the same.
15329         PR middle-end/86123
15330         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15331         Fix up comment formatting.
15333 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15335         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15336         (typed_splay_tree::closure,
15337         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15338         (typed_splay_tree::typed_splay_tree,
15339         typed_splay_tree::operator =): Declared private.
15340         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15341         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15342         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15343         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15344         typed_splay_tree::splay_tree_splay,
15345         typed_splay_tree::splay_tree_foreach_helper,
15346         typed_splay_tree::splay_tree_insert,
15347         typed_splay_tree::splay_tree_remove,
15348         typed_splay_tree::splay_tree_lookup,
15349         typed_splay_tree::splay_tree_predecessor,
15350         typed_splay_tree::splay_tree_successor,
15351         typed_splay_tree::splay_tree_min,
15352         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15353         (typed_splay_tree::root, typed_splay_tree::comp,
15354         typed_splay_tree::delete_key,
15355         typed_splay_tree::delete_value): New data members.
15356         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15357         typed_splay_tree::remove.
15359 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15361         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15362         -mginv and -mno-ginv to the assembler.
15363         * config/mips/mips.opt (-mcrc): New option.
15364         (-mginv): Likewise.
15365         * doc/invoke.text (-mcrc): Document.
15366         (-mginv): Likewise.
15368 2018-06-15  Nick Clifton  <nickc@redhat.com>
15370         PR 84195
15371         * tree.c (escaped_string): New class.  Converts an unescaped
15372         string into its escaped equivalent.
15373         (warn_deprecated_use): Use the new class to convert the
15374         deprecation message, if present.
15375         (test_escaped_strings): New self test.
15376         (test_c_tests): Add test_escaped_strings.
15377         * doc/extend.texi (deprecated): Add a note that the
15378         deprecation message is affected by the -fmessage-length
15379         option, and that control characters will be escaped.
15380         (#pragma GCC error): Document this pragma.
15381         (#pragma GCC warning): Likewise.
15382         * doc/invoke.texi (-fmessage-length): Document this option's
15383         effect on the #warning and #error preprocessor directives and
15384         the deprecated attribute.
15386 2018-06-15  Richard Biener  <rguenther@suse.de>
15388         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15389         here, also noting vector size used.
15390         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15391         size used in MSG_OPTIMIZED_LOCATIONS dump.
15392         (pass_slp_vectorize::execute): Adjust.
15394 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15396         PR target/85968
15397         * config/arc/arc.c (arc_return_address_register): Fix
15398         if-condition.
15400 2018-06-15  Richard Biener  <rguenther@suse.de>
15402         PR middle-end/86159
15403         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15404         leave useless conversion stripping to force_gimple_operand_gsi.
15405         (gimplify_build2): Likewise.
15406         (gimplify_build1): Likewise.
15408 2018-06-15  Richard Biener  <rguenther@suse.de>
15410         PR middle-end/86076
15411         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15412         before adjusting their block.
15414 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15416         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15417         multilibs for *-*-rtems*.
15418         * config/riscv/t-rtems: New file.
15420 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15422         PR middle-end/86122
15423         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15424         unsigned_type_for returns NULL.
15426         PR target/85945
15427         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15428         subregs of multi-word pseudos unless the float mode has word size.
15430 2018-06-14  Richard Biener  <rguenther@suse.de>
15432         PR middle-end/86139
15433         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15434         duplicate and harmful type_hash_canon.
15435         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15437 2018-06-14  Richard Biener  <rguenther@suse.de>
15439         PR ipa/86124
15440         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15441         NULL cgraph_node.
15443 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15445         * config/rtems.h (STDINT_LONG32): Define.
15447 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15448             Prachi Godbole  <prachi.godbole@imgtec.com>
15450         * config/mips/mips-cpus.def: Define P6600.
15451         * config/mips/mips-tables.opt: Regenerate.
15452         * config/mips/mips.c (mips_ucbranch_type): New enum.
15453         (mips_rtx_cost_data): Add support for P6600.
15454         (mips_issue_rate): Likewise.
15455         (mips_multipass_dfa_lookahead): Likewise.
15456         (mips_avoid_hazard): Likewise.
15457         (mips_reorg_process_insns): Likewise.
15458         (mips_classify_branch_p6600): New function.
15459         * config/mips/mips.h (TUNE_P6600): New define.
15460         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15461         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15462         * config/mips/mips.md: Include p6600.md.
15463         (processor): Add p6600.
15464         * config/mips/p6600.md: New file.
15465         * doc/invoke.texi: Add p6600 to supported architectures.
15467 2018-06-13  Martin Sebor  <msebor@redhat.com>
15469         PR tree-optimization/86114
15470         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15471         of integer types.
15472         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15474 2018-06-13  Richard Biener  <rguenther@suse.de>
15476         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15477         Properly set vector type of the intermediate stmt.
15478         * tree-vect-stmts.c (vectorizable_operation): The destination
15479         var always has vectype_out type.
15481 2018-06-13  Jeff Law  <law@redhat.com>
15483         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15484         integer 0 for argument to print_rtl_with_bb.
15485         (rl78_reorg): Likewise.
15487 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15489         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15490         from rtx to rtx_insn *.
15491         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15492         "label".
15493         (add_sched_insns_for_speculation): Likewise for local "target",
15494         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15495         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15496         from rtx_insn ** to rtx_code_label **.
15497         (reorg_emit_nops): Likewise.
15498         (c6x_reorg): Likewise for local "call_labels".
15499         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15500         rtx to rtx_insn *.
15501         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
15502         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
15503         the loops over LABEL_REFS.
15504         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
15505         braf_label.
15506         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15507         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
15508         (split_branches): Strengthen local "olabel" from rtx to
15509         rtx_insn *, adding a safe_as_a cast.
15510         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
15511         to "rtx_insn *".
15512         (add_insn_after): Likewise for first two params.
15513         (add_insn_before): Likewise.
15514         (remove_insn): Likewise for param.
15515         (emit_pattern_before_noloc): Likewise for second and third params.
15516         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
15517         (emit_call_insn_before_noloc): Likewise.
15518         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
15519         to "rtx_insn *".
15520         (emit_barrier_before): Likewise.
15521         (emit_label_before): Strengthen "label" param from "rtx" to
15522         "rtx_code_label *".  Strengthen "before" param from "rtx" to
15523         "rtx_insn *".
15524         (emit_insn_after_1): Strengthen "after" param from "rtx" to
15525         "rtx_insn *".
15526         (emit_pattern_after_noloc): Likewise.
15527         (emit_insn_after_noloc): Likewise.
15528         (emit_jump_insn_after_noloc): Likewise.
15529         (emit_call_insn_after_noloc): Likewise.
15530         (emit_debug_insn_after_noloc): Likewise.
15531         (emit_barrier_after): Likewise.
15532         (emit_label_after): Likewise for both params.
15533         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
15534         "loc" param from "int" to "location_t".
15535         (emit_insn_after_setloc): Likewise.
15536         (emit_jump_insn_after_setloc): Likewise.
15537         (emit_call_insn_after_setloc): Likewise.
15538         (emit_debug_insn_after_setloc): Likewise.
15539         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
15540         "loc" param from "int" to "location_t".
15541         (emit_pattern_before): Convert NULL_RTX to NULL.
15542         (emit_insn_before_setloc): Convert "loc" param from "int" to
15543         "location_t".
15544         (emit_jump_insn_before_setloc): Likewise.
15545         (emit_call_insn_before_setloc): Likewise.
15546         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
15547         rtx_insn *.  Convert "loc" param from "int" to "location_t".
15548         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
15549         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
15550         Convert 3rd param from "int" to "location_t".
15551         (emit_barrier_before, emit_barrier_after, next_real_insn):
15552         Strengthen param from rtx to rtx_insn *.
15553         (emit_label_before): Strengthen 1st param from "rtx" to
15554         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
15555         "rtx_insn *".
15556         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
15557         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
15558         Strengthen 2nd param from "rtx" to "rtx_insn *".
15559         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
15560         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
15561         Likewise. Convert 3rd param from "int" to "location_t".
15562         (emit_label_after): Strengthen 1st param from "rtx" to
15563         "rtx_code_label *".
15564         (next_real_insn, remove_insn): Strengthen param from "rtx" to
15565         "rtx_insn *".
15566         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
15567         from "rtx" to "rtx_insn *".
15569 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
15571         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
15572         bodies streamed in with -Q.
15573         * dumpfile.c (dump_files): Add lto-stream-out dump file.
15574         * dumpfile.h (tree_dump_index): Add lto_stream_out.
15575         * gimple-streamer-out.c: Include gimple-pretty-print.h
15576         (output_bb): Dump stmts streamed.
15577         * lto-section-out.c: Include print-tree.h
15578         (lto_begin_section): Dump sections created.
15579         (lto_output_decl_index): Dump decl encoded.
15580         * lto-streamer-out.c: Include print-tree.h
15581         (create_output_block): Dump output block created.
15582         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
15583         (output_function): Dump function output.
15584         (output_constructor): Dump constructor streamed.
15585         (write_global_stream): Output indexes encoded.
15586         (produce_asm_for_decls): Dump streams encoded.
15587         * lto-streamer.c (streamer_dump_file): New global var.
15588         * lto-streamer.h (streamer_dump_file): Declare.
15589         * passes.c (ipa_write_summaries): Initialize streamer dump.
15590         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
15591         in.
15593 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
15595         PR target/86048
15596         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
15597         offsets for register save directives.  Emit a second batch of save
15598         directives, if need be, when the function accesses prior frames.
15600 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15602         * config/arc/fpu.md (fmasf4): Force operand to register.
15603         (fnmasf4): Likewise.
15605 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15607         * config/arc/arc-protos.h (arc_pad_return): Remove.
15608         * config/arc/arc.c (machine_function): Remove force_short_suffix
15609         and size_reason.
15610         (arc_print_operand): Adjust printing of '&'.
15611         (arc_verify_short): Remove conditional printing of short suffix.
15612         (arc_final_prescan_insn): Remove reference to size_reason.
15613         (pad_return): New function.
15614         (arc_reorg): Call pad_return.
15615         (arc_pad_return): Remove.
15616         (arc_init_machine_status): Remove reference to force_short_suffix.
15617         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
15618         (attr length): When attribute iscompact is true force to 2
15619         regardless; in the case of maybe check if we want to force the
15620         instruction to have 4 bytes length.
15621         (nopv): Change it to generate 4 byte long nop as well.
15622         (blockage): New pattern.
15623         (simple_return): Remove call to arc_pad_return.
15624         (p_return_i): Likewise.
15626 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15628         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
15630 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15632         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
15633         ARC cores.
15635 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15637         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
15638         for ARC700 and ARCv2.
15640 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
15642         PR target/86076
15643         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
15644         operands[2] instead of operands[1].
15647 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15649         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
15650         case, check whether the outer register overlaps an unallocatable
15651         register, not just whether it fits the required class.
15653 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15655         * poly-int.h (can_div_trunc_p): Add new overload in which all values
15656         are poly_ints.
15657         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
15658         (memrefs_conflict_p): Likewise.
15659         (init_alias_analysis): Likewise.
15660         * cfgexpand.c (expand_debug_expr): Likewise.
15661         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
15662         * cse.c (fold_rtx): Likewise.
15663         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
15664         * expr.c (emit_block_move_hints): Likewise.
15665         (clear_storage_hints, push_block, emit_push_insn): Likewise.
15666         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
15667         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
15668         (emit_group_store): Likewise.
15669         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
15670         to read the PRE/POST_MODIFY increment.
15671         * calls.c (store_one_arg): Use strip_offset.
15672         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
15673         poly_int_rtx_p.
15674         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
15675         by a VEC_SELECT.
15676         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
15677         (simplify_binary_operation_1): Extend CONST_INT handling to
15678         poly_int_rtx_p.
15679         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
15680         than a HOST_WIDE_INT.
15681         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
15682         poly_int64.
15683         (adjust_mems, add_stores): Update accodingly.
15684         (vt_canonicalize_addr): Track polynomial offsets.
15685         (emit_note_insn_var_location): Likewise.
15686         (vt_add_function_parameter): Likewise.
15687         (vt_initialize): Likewise.
15689 2018-06-12  Jeff Law  <law@redhat.com>
15691         * config.gcc (alpha*-*-freebsd*): Remove.
15692         * config/alpha/freebsd.h: Remove.
15694 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
15696         PR other/69968
15697         * spellcheck-tree.c (levenshtein_distance): Rename to...
15698         (get_edit_distance): ...this, and update for underlying renaming.
15699         * spellcheck-tree.h (levenshtein_distance): Rename to...
15700         (get_edit_distance): ...this.
15701         * spellcheck.c (levenshtein_distance): Rename to...
15702         (get_edit_distance): ...this.  Convert from Levenshtein distance
15703         to Damerau-Levenshtein distance by supporting transpositions of
15704         adjacent characters.  Rename "v1" to "v_next" and "v0" to
15705         "v_one_ago".
15706         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
15707         (selftest::test_edit_distance_unit_test_oneway): ...this, and
15708         update for underlying renaming.
15709         (selftest::levenshtein_distance_unit_test): Rename to...
15710         (selftest::test_get_edit_distance_unit): ...this, and update for
15711         underlying renaming.
15712         (selftest::test_find_closest_string): Add example from PR 69968
15713         where transposition helps
15714         (selftest::test_metric_conditions): Update for renaming.
15715         (selftest::test_metric_conditions): Likewise.
15716         (selftest::spellcheck_c_tests): Likewise.
15717         * spellcheck.h (levenshtein_distance): Rename both overloads to...
15718         (get_edit_distance): ...this.
15719         (best_match::consider): Update for renaming.
15721 2018-06-12  Martin Sebor  <msebor@redhat.com>
15723         PR tree-optimization/85259
15724         * builtins.c (compute_objsize): Handle constant offsets.
15725         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
15726         true iff a warning has been issued.
15727         * gimple.h (gimple_nonartificial_location): New function.
15728         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
15729         gimple_nonartificial_location and handle -Wno-system-headers.
15730         (handle_builtin_stxncpy): Same.
15732 2018-06-12  Martin Sebor  <msebor@redhat.com>
15734         PR c/85931
15735         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
15737 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
15739         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
15740         BUILTIN_VEC_XST entries for pointer to double and long long.
15742 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
15744         PR target/85990
15745         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15746         Update comments.
15747         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15748         Likewise.
15750 2018-06-12  Martin Liska  <mliska@suse.cz>
15752         * doc/options.texi: Document IntegerRange.
15754 2018-06-12  Martin Liska  <mliska@suse.cz>
15756         * config/i386/i386.opt: Make MPX-related options as Deprecated.
15757         * opt-functions.awk: Handle Deprecated flag.
15758         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
15759         and report error.
15760         (read_cmdline_option): Report warning for a deprecated option.
15761         * opts.h (struct cl_option): Add new field cl_deprecated.
15762         (CL_ERR_DEPRECATED): New.
15764 2018-06-12  Martin Liska  <mliska@suse.cz>
15766         * doc/options.texi: Document Deprecated option flag.
15768 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15770         * config/arc/arc-arch.h (arc_extras): New enum.
15771         (arc_cpu_t):Add field extra.
15772         (arc_cpu_types): Consider the extras.
15773         * config/arc/arc-cpus.def: Add extras info.
15774         * config/arc/arc-opts.h (processor_type): Consider extra field.
15775         * config/arc/arc.c (arc_override_options): Handle extra field.
15777 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15779         * config/arc/arc-arch.h: Update ARC_OPTX macro.
15780         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
15781         field.
15782         * config/arc/arc.c (arc_init): Update pic warning.
15783         (irq_range): Update irq range parsing warnings.
15784         (arc_override_options): Update various warning messages.
15785         (arc_handle_aux_attribute): Likewise.
15787 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
15789         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
15791 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15793         * doc/sourcebuild.texi: Document usage of line number 0 in verify
15794         compiler messages directives.
15796 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
15798         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
15799         * config/mips/mips-tables.opt: Regenerate.
15800         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
15801         mips64r6.
15802         * doc/invoke.texi: Document -march=i6500.
15804 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
15806         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
15807         (i6400_gpmul): Add cpu_unit.
15808         (i6400_gpdiv): Likewise.
15809         (i6400_msa_add_d): Update reservations.
15810         (i6400_msa_int_add) Likewise.
15811         (i6400_msa_short_logic3) Likewise.
15812         (i6400_msa_short_logic2) Likewise.
15813         (i6400_msa_short_logic) Likewise.
15814         (i6400_msa_move) Likewise.
15815         (i6400_msa_cmp) Likewise.
15816         (i6400_msa_short_float2) Likewise.
15817         (i6400_msa_div_d) Likewise.
15818         (i6400_msa_long_logic1) Likewise.
15819         (i6400_msa_long_logic2) Likewise.
15820         (i6400_msa_mult) Likewise.
15821         (i6400_msa_long_float2) Likewise.
15822         (i6400_msa_long_float4) Likewise.
15823         (i6400_msa_long_float5) Likewise.
15824         (i6400_msa_long_float8) Likewise.
15825         (i6400_fpu_fadd): Include frint type.
15826         (i6400_fpu_store): New define_insn_reservation.
15827         (i6400_fpu_load): Likewise.
15828         (i6400_fpu_move): Likewise.
15829         (i6400_fpu_fcmp): Likewise.
15830         (i6400_fpu_fmadd): Likewise.
15831         (i6400_int_mult): Include imul3nc type and update reservation.
15832         (i6400_int_div): Include idiv3 type and update reservation.
15833         (i6400_int_load): Update to check type not move_type.
15834         (i6400_int_store): Likewise.
15835         (i6400_int_prefetch): Set zero latency.
15837 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
15839         * gcc.c: Document new %@{...} sequence.
15840         (LINK_COMMAND_SPEC): Use it for the -L switches.
15841         (cpp_unique_options): Use it for the -I switches.
15842         (at_file_argbuf): New global variable.
15843         (in_at_file): Likewise.
15844         (alloc_args): Create at_file_argbuf.
15845         (clear_args): Truncate at_file_argbuf.
15846         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
15847         (open_at_file): New function.
15848         (close_at_file): Likewise.
15849         (create_at_file): Delete.
15850         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
15851         <'o'>: Likewise.
15852         <'@'>: New case.
15853         (validate_switches_from_spec): Deal with %@{...} sequence.
15854         (validate_switches): Likewise.
15855         (driver::finalize): Call clear_args.
15857 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
15859         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
15861 2018-06-11  Martin Sebor  <msebor@redhat.com>
15863         * doc/invoke.texi (-Wall): List -Wc++17-compat.
15864         (Wno-class-memaccess): Add @opindex.
15865         (Wno-templates, Wno-multiple-inheritance): Same.
15866         (Wno-virtual-inheritance, Wno-namespaces): Same.
15867         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
15868         (Wno-format-overflow, Wno-format-truncation): Same.
15869         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
15870         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
15871         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
15872         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
15873         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
15874         (Wno-misspelled-isr): Same.
15876 2018-06-11  Martin Sebor  <msebor@redhat.com>
15878         * PR tree-optimization/86083
15879         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
15881 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
15883         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
15885 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
15887         PR target/85755
15888         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
15889         on the correct operand.
15890         (*movdi_internal64): Ditto.
15892 2018-06-11  Martin Liska  <mliska@suse.cz>
15894         PR tree-optimization/86089
15895         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
15897 2018-06-11  Julia Koval  <julia.koval@intel.com>
15899         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
15900         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
15901         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
15903 2018-06-11  Olivier Hainque  <hainque@adacore.com>
15905         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
15906         for Ada with strict dwarf2.
15908 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
15910         PR target/85755
15911         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
15912         addresses.
15914 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
15916         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
15918 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
15920         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
15921         TARGET_ELF.
15923 2018-06-08  Martin Liska  <mliska@suse.cz>
15925         * tree-cfg.h (debug_function): Fix argument type to match
15926         implementation.
15928 2018-06-08  Martin Liska  <mliska@suse.cz>
15930         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
15931         Remove usage of MPX-related (and removed) fields.
15932         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
15934 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
15936         * cfg.c (debug): Use TDF_NONE rather than 0.
15937         * cfghooks.c (debug): Likewise.
15938         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
15939         (struct dump_option_value_info): Convert to...
15940         (struct kv_pair): ...this template type.
15941         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
15942         rather than 0.
15943         (optinfo_verbosity_options): Likewise.
15944         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
15945         OPTGROUP_NONE.
15946         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
15947         than int for "optgroup_flags" param.
15948         (dump_generic_expr_loc): Use dump_flags_t rather than int for
15949         "dump_kind" param.
15950         (dump_dec): Likewise.
15951         (dump_finish): Use TDF_NONE rather than 0.
15952         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
15953         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
15954         than 0.  Update for change to option_ptr.
15955         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
15956         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
15957         0.  Update for changes to optinfo_verbosity_options and
15958         optgroup_options.
15959         (opt_info_switch_p): Convert optgroup_flags from int to
15960         optgroup_flags_t.
15961         (dump_basic_block): Use dump_flags_t rather than int
15962         for "dump_kind" param.
15963         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
15964         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
15965         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
15966         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
15967         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
15968         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
15969         TDF_NONE): Convert from macros to...
15970         (enum dump_flag): ...this new enum.
15971         (dump_flags_t): Update to use enum.
15972         (operator|, operator&, operator~, operator|=, operator&=):
15973         Implement for dump_flags_t.
15974         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
15975         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
15976         Convert from macros to...
15977         (enum optgroup_flag): ...this new enum.
15978         (optgroup_flags_t): New typedef.
15979         (operator|, operator|=): Implement for optgroup_flags_t.
15980         (struct dump_file_info): Convert field "alt_flags" to
15981         dump_flags_t.  Convert field "optgroup_flags" to
15982         optgroup_flags_t.
15983         (dump_basic_block): Use dump_flags_t rather than int for param.
15984         (dump_generic_expr_loc): Likewise.
15985         (dump_dec): Likewise.
15986         (dump_register): Convert param "optgroup_flags" to
15987         optgroup_flags_t.
15988         (opt_info_enable_passes): Likewise.
15989         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
15990         than 0.
15991         * gimple-pretty-print.c (debug): Likewise.
15992         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
15993         (merged_store_group::apply_stores): Likewise.
15994         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
15995         * gimple.c (verify_gimple_pp): Likewise.
15996         * graphite-poly.c (print_pbb_body): Likewise.
15997         * passes.c (pass_manager::register_one_dump_file): Convert
15998         local "optgroup_flags" to optgroup_flags_t.
15999         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16000         (debug): Likewise.
16001         (debug_body): Likewise.
16002         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16003         to optgroup_flags_t.
16004         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16005         than 0.
16006         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16007         (convert_mult_to_fma): Likewise.
16008         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16009         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16010         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16011         "dump_kind" to dump_flags_t.
16013 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16015         * config/rs6000/rs6000.c (min, max): Delete.
16017 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16019         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16020         -mabi=spe and -mabi=no-spe.
16022 2018-06-08  Martin Liska  <mliska@suse.cz>
16024         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16025         where we expect an existing summary.
16027 2018-06-08  Martin Liska  <mliska@suse.cz>
16029         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16030         * ipa-inline.h (estimate_edge_growth): Likewise.
16032 2018-06-08  Martin Liska  <mliska@suse.cz>
16034         * cgraph.c (function_version_hasher::hash): Use
16035         cgraph_node::get_uid ().
16036         (function_version_hasher::equal):
16037         * cgraph.h (cgraph_node::get_uid): New method.
16038         * ipa-inline.c (update_caller_keys): Use
16039         cgraph_node::get_uid ().
16040         (update_callee_keys): Likewise.
16041         * ipa-utils.c (searchc): Likewise.
16042         (ipa_reduced_postorder): Likewise.
16043         * lto-cgraph.c (input_node): Likewise.
16044         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16045         * symbol-summary.h (symtab_insertion): Likewise.
16046         (symtab_removal): Likewise.
16047         (symtab_duplication): Likewise.
16048         * tree-pretty-print.c (dump_function_header): Likewise.
16049         * tree-sra.c (convert_callers_for_node): Likewise.
16051 2018-06-08  Martin Liska  <mliska@suse.cz>
16053         * cgraph.c (symbol_table::create_edge): Always assign a new
16054         unique number.
16055         (symbol_table::free_edge): Do not recycle numbers.
16056         * cgraph.h (cgraph_edge::get): New method.
16057         * symbol-summary.h (symtab_removal): Use it.
16058         (symtab_duplication): Likewise.
16059         (call_summary::hashable_uid): Remove.
16061 2018-06-08  Martin Liska  <mliska@suse.cz>
16063         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16064         (initialize_growth_caches): Remove.
16065         (free_growth_caches): Likewise.
16066         (do_estimate_edge_time): Use edge_growth_cache.
16067         (do_estimate_edge_size): Likewise.
16068         (do_estimate_edge_hints): Likewise.
16069         * ipa-inline.c (reset_edge_caches): Likewise.
16070         (recursive_inlining): Likewise.
16071         (inline_small_functions): Likewise.
16072         * ipa-inline.h (initialize_growth_caches): Remove.
16073         (estimate_edge_size): Likewise.
16074         (estimate_edge_time): Likewise.
16075         (estimate_edge_hints): Likewise.
16076         (reset_edge_growth_cache): Likewise.
16077         * symbol-summary.h (call_summary::remove): New method.
16079 2018-06-08  Martin Liska  <mliska@suse.cz>
16081         * ipa-cp.c (class edge_clone_summary): New summary.
16082         (grow_edge_clone_vectors): Remove.
16083         (ipcp_edge_duplication_hook): Remove.
16084         (class edge_clone_summary_t): New call_summary class.
16085         (ipcp_edge_removal_hook): Remove.
16086         (edge_clone_summary_t::duplicate): New function.
16087         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16088         (create_specialized_node): Likewise.
16089         (ipcp_driver): Initialize edge_clone_summaries and do not
16090         register hooks.
16092 2018-06-08  Martin Liska  <mliska@suse.cz>
16094         * symbol-summary.h (get): New function.
16095         (call_summary::m_initialize_when_cloning): New class member.
16097 2018-06-08  Martin Liska  <mliska@suse.cz>
16099         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16100         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16101         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16102         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16103         (remove_cgraph_node_from_order): Use the removed_nodes set.
16104         (do_per_function_toporder): Likwise.
16105         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16106         instead of summary_uid.
16107         (symtab_removal): Likewise.
16108         (symtab_duplication): Likewise.
16110 2018-06-08  Martin Liska  <mliska@suse.cz>
16112         * ipa-cp.c (ipcp_store_bits_results): Use
16113         ipcp_transformation_sum.
16114         (ipcp_store_vr_results): Likewise.
16115         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16116         to ...
16117         (ipcp_transformation_initialize): ... this.
16118         (ipa_set_node_agg_value_chain):
16119         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16120         (write_ipcp_transformation_info): Likewise.
16121         (read_ipcp_transformation_info): Likewise.
16122         (ipcp_update_bits): Likewise.
16123         (ipcp_update_vr): Likewise.
16124         (ipcp_transform_function): Likewise.
16125         * ipa-prop.h: Rename ipcp_transformation_summary to
16126         ipcp_transformation.
16127         (class ipcp_transformation_t): New function summary.
16128         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16129         (ipa_get_agg_replacements_for_node): Likewise.
16131 2018-06-08  Martin Liska  <mliska@suse.cz>
16133         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16134         of struct.
16135         (class funct_state_summary_t): New function_summary class.
16136         (has_function_state): Remove.
16137         (get_function_state): Likewise.
16138         (set_function_state): Likewise.
16139         (add_new_function): Likewise.
16140         (funct_state_summary_t::insert): New function.
16141         (duplicate_node_data): Remove.
16142         (remove_node_data): Remove.
16143         (funct_state_summary_t::duplicate): New function.
16144         (register_hooks): Create new funct_state_summaries.
16145         (pure_const_generate_summary): Use it.
16146         (pure_const_write_summary): Likewise.
16147         (pure_const_read_summary): Likewise.
16148         (propagate_pure_const): Likewise.
16149         (propagate_nothrow): Likewise.
16150         (dump_malloc_lattice): Likewise.
16151         (propagate_malloc): Likewise.
16152         (execute): Do not register hooks, just remove summary
16153         instead.
16154         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16155         constructor.
16157 2018-06-08  Martin Liska  <mliska@suse.cz>
16159         * ipa-reference.c (remove_node_data): Remove.
16160         (duplicate_node_data): Likewise.
16161         (class ipa_ref_var_info_summary_t): New class.
16162         (class ipa_ref_opt_summary_t): Likewise.
16163         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16164         (get_reference_optimization_summary): Use
16165         ipa_ref_opt_sum_summaries.
16166         (set_reference_vars_info): Remove.
16167         (set_reference_optimization_summary): Likewise.
16168         (ipa_init): Create summaries.
16169         (init_function_info): Use function summary.
16170         (ipa_ref_opt_summary_t::duplicate): New function.
16171         (ipa_ref_opt_summary_t::remove): New function.
16172         (get_read_write_all_from_node): Fix GNU coding style.
16173         (propagate): Use function summary.
16174         (write_node_summary_p): Fix GNU coding style.
16175         (stream_out_bitmap): Likewise.
16176         (ipa_reference_read_optimization_summary): Use function summary.
16177         (ipa_reference_c_finalize): Do not release hooks.
16179 2018-06-08  Martin Liska  <mliska@suse.cz>
16181         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16182         (analyze_function_body): Extract multiple calls of get_create.
16183         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16184         * ipa-inline.c (recursive_inlining): Use ::get method.
16185         * ipa-inline.h (estimate_edge_growth): Likewise.
16187 2018-06-08  Martin Liska  <mliska@suse.cz>
16189         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16190         HSA_INVALID.
16191         (hsa_function_summary::hsa_function_summary): Use the new enum
16192         value.
16193         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16194         * hsa-gen.c (hsa_get_host_function): Likewise.
16195         (get_brig_function_name): Likewise.
16196         * ipa-hsa.c (process_hsa_functions): Likewise.
16197         (ipa_hsa_write_summary): Likewise.
16198         * symbol-summary.h (symtab_duplication): Use ::get function/
16199         (get): New function.
16201 2018-06-08  Martin Liska  <mliska@suse.cz>
16203         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16204         of get.
16205         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16206         (hsa_register_kernel): Likewise.
16207         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16208         * hsa-gen.c (hsa_get_host_function): Likewise.
16209         (get_brig_function_name): Likewise.
16210         (generate_hsa): Likewise.
16211         (pass_gen_hsail::execute): Likewise.
16212         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16213         (devirtualization_time_bonus): Likewise.
16214         (ipcp_propagate_stage): Likewise.
16215         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16216         (edge_set_predicate): Likewise.
16217         (evaluate_conditions_for_known_args): Likewise.
16218         (evaluate_properties_for_edge): Likewise.
16219         (ipa_fn_summary::reset): Likewise.
16220         (ipa_fn_summary_t::duplicate): Likewise.
16221         (dump_ipa_call_summary): Likewise.
16222         (ipa_dump_fn_summary): Likewise.
16223         (analyze_function_body): Likewise.
16224         (compute_fn_summary): Likewise.
16225         (estimate_edge_devirt_benefit): Likewise.
16226         (estimate_edge_size_and_time): Likewise.
16227         (estimate_calls_size_and_time): Likewise.
16228         (estimate_node_size_and_time): Likewise.
16229         (inline_update_callee_summaries): Likewise.
16230         (remap_edge_change_prob): Likewise.
16231         (remap_edge_summaries): Likewise.
16232         (ipa_merge_fn_summary_after_inlining): Likewise.
16233         (ipa_update_overall_fn_summary): Likewise.
16234         (read_ipa_call_summary): Likewise.
16235         (inline_read_section): Likewise.
16236         (write_ipa_call_summary): Likewise.
16237         (ipa_fn_summary_write): Likewise.
16238         (ipa_free_fn_summary): Likewise.
16239         * ipa-hsa.c (process_hsa_functions): Likewise.
16240         (ipa_hsa_write_summary): Likewise.
16241         (ipa_hsa_read_section): Likewise.
16242         * ipa-icf.c (sem_function::merge): Likewise.
16243         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16244         (do_estimate_edge_time): Likewise.
16245         (estimate_size_after_inlining): Likewise.
16246         (estimate_growth): Likewise.
16247         (growth_likely_positive): Likewise.
16248         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16249         (inline_call): Likewise.
16250         * ipa-inline.c (caller_growth_limits): Likewise.
16251         (can_inline_edge_p): Likewise.
16252         (can_inline_edge_by_limits_p): Likewise.
16253         (compute_uninlined_call_time): Likewise.
16254         (compute_inlined_call_time): Likewise.
16255         (want_inline_small_function_p): Likewise.
16256         (edge_badness): Likewise.
16257         (update_caller_keys): Likewise.
16258         (update_callee_keys): Likewise.
16259         (recursive_inlining): Likewise.
16260         (inline_small_functions): Likewise.
16261         (inline_to_all_callers_1): Likewise.
16262         (dump_overall_stats): Likewise.
16263         (early_inline_small_functions): Likewise.
16264         (early_inliner): Likewise.
16265         * ipa-inline.h (estimate_edge_growth): Likewise.
16266         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16267         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16268         * ipa-prop.h (IPA_NODE_REF): Likewise.
16269         (IPA_EDGE_REF): Likewise.
16270         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16271         (propagate_malloc): Likewise.
16272         * ipa-split.c (execute_split_functions): Likewise.
16273         * symbol-summary.h: Rename get to get_create.
16274         (get): Likewise.
16275         (get_create): Likewise.
16276         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16278 2018-06-08  Martin Liska  <mliska@suse.cz>
16280         * symbol-summary.h (release): Move definition out of class
16281         declaration.
16282         (symtab_removal): Likewise.
16283         (symtab_duplication): Likewise.
16285 2018-06-08  Martin Liska  <mliska@suse.cz>
16287         * symbol-summary.h (function_summary): Move constructor
16288         implementation out of class declaration.
16289         (release): Likewise.
16290         (symtab_insertion): Likewise.
16291         (symtab_removal): Likewise.
16292         (symtab_duplication): Likewise.
16293         (get): Likewise.
16295 2018-06-08  Martin Liska  <mliska@suse.cz>
16297         * Makefile.in: Remove support for MPX (macros, related functions,
16298         fields in cgraph_node, ...).
16299         * builtin-types.def (BT_BND): Likewise.
16300         (BT_FN_BND_CONST_PTR): Likewise.
16301         (BT_FN_CONST_PTR_BND): Likewise.
16302         (BT_FN_VOID_PTR_BND): Likewise.
16303         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16304         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16305         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16306         (expand_builtin_mempcpy_with_bounds): Likewise.
16307         (expand_builtin_memset_with_bounds): Likewise.
16308         (expand_builtin_memset_args): Likewise.
16309         (std_expand_builtin_va_start): Likewise.
16310         (expand_builtin): Likewise.
16311         (expand_builtin_with_bounds): Likewise.
16312         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16313         (DEF_LIB_BUILTIN_CHKP): Likewise.
16314         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16315         (DEF_CHKP_BUILTIN): Likewise.
16316         (BUILT_IN_MEMCPY): Likewise.
16317         (BUILT_IN_MEMMOVE): Likewise.
16318         (BUILT_IN_MEMPCPY): Likewise.
16319         (BUILT_IN_MEMSET): Likewise.
16320         (BUILT_IN_STPCPY): Likewise.
16321         (BUILT_IN_STRCAT): Likewise.
16322         (BUILT_IN_STRCHR): Likewise.
16323         (BUILT_IN_STRCPY): Likewise.
16324         (BUILT_IN_STRLEN): Likewise.
16325         (BUILT_IN_MEMCPY_CHK): Likewise.
16326         (BUILT_IN_MEMMOVE_CHK): Likewise.
16327         (BUILT_IN_MEMPCPY_CHK): Likewise.
16328         (BUILT_IN_MEMSET_CHK): Likewise.
16329         (BUILT_IN_STPCPY_CHK): Likewise.
16330         (BUILT_IN_STRCAT_CHK): Likewise.
16331         (BUILT_IN_STRCPY_CHK): Likewise.
16332         * calls.c (store_bounds): Likewise.
16333         (emit_call_1): Likewise.
16334         (special_function_p): Likewise.
16335         (maybe_warn_nonstring_arg): Likewise.
16336         (initialize_argument_information): Likewise.
16337         (finalize_must_preallocate): Likewise.
16338         (compute_argument_addresses): Likewise.
16339         (expand_call): Likewise.
16340         * cfgexpand.c (expand_call_stmt): Likewise.
16341         (expand_return): Likewise.
16342         (expand_gimple_stmt_1): Likewise.
16343         (pass_expand::execute): Likewise.
16344         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16345         (cgraph_node::remove): Likewise.
16346         (cgraph_node::dump): Likewise.
16347         (cgraph_node::verify_node): Likewise.
16348         * cgraph.h (chkp_function_instrumented_p): Likewise.
16349         (symtab_node::get_alias_target): Likewise.
16350         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16351         (cgraph_local_p): Likewise.
16352         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16353         (cgraph_edge::rebuild_references): Likewise.
16354         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16355         (walk_polymorphic_call_targets): Likewise.
16356         (cgraph_node::expand_thunk): Likewise.
16357         (symbol_table::output_weakrefs): Likewise.
16358         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16359         (ix86_handle_option): Likewise.
16360         * config/i386/constraints.md: Likewise.
16361         * config/i386/i386-builtin-types.def (BND): Likewise.
16362         (VOID): Likewise.
16363         (PVOID): Likewise.
16364         (ULONG): Likewise.
16365         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16366         (BDESC_FIRST): Likewise.
16367         (BDESC): Likewise.
16368         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16369         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16370         * config/i386/i386.c (enum reg_class): Likewise.
16371         (ix86_target_string): Likewise.
16372         (ix86_option_override_internal): Likewise.
16373         (ix86_conditional_register_usage): Likewise.
16374         (ix86_valid_target_attribute_inner_p): Likewise.
16375         (ix86_set_indirect_branch_type): Likewise.
16376         (ix86_set_current_function): Likewise.
16377         (ix86_function_arg_regno_p): Likewise.
16378         (init_cumulative_args): Likewise.
16379         (ix86_function_arg_advance): Likewise.
16380         (ix86_function_arg): Likewise.
16381         (ix86_pass_by_reference): Likewise.
16382         (ix86_function_value_regno_p): Likewise.
16383         (ix86_function_value_1): Likewise.
16384         (ix86_function_value_bounds): Likewise.
16385         (ix86_return_in_memory): Likewise.
16386         (ix86_setup_incoming_vararg_bounds): Likewise.
16387         (ix86_va_start): Likewise.
16388         (indirect_thunk_need_prefix): Likewise.
16389         (print_reg): Likewise.
16390         (ix86_print_operand): Likewise.
16391         (ix86_expand_call): Likewise.
16392         (ix86_output_function_return): Likewise.
16393         (reg_encoded_number): Likewise.
16394         (BDESC_VERIFYS): Likewise.
16395         (ix86_init_mpx_builtins): Likewise.
16396         (ix86_init_builtins): Likewise.
16397         (ix86_emit_cmove): Likewise.
16398         (ix86_emit_move_max): Likewise.
16399         (ix86_expand_builtin): Likewise.
16400         (ix86_builtin_mpx_function): Likewise.
16401         (ix86_get_arg_address_for_bt): Likewise.
16402         (ix86_load_bounds): Likewise.
16403         (ix86_store_bounds): Likewise.
16404         (ix86_load_returned_bounds): Likewise.
16405         (ix86_store_returned_bounds): Likewise.
16406         (ix86_class_likely_spilled_p): Likewise.
16407         (ix86_hard_regno_mode_ok): Likewise.
16408         (x86_order_regs_for_local_alloc): Likewise.
16409         (ix86_mitigate_rop): Likewise.
16410         (ix86_bnd_prefixed_insn_p): Likewise.
16411         (ix86_mpx_bound_mode): Likewise.
16412         (ix86_make_bounds_constant): Likewise.
16413         (ix86_initialize_bounds): Likewise.
16414         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16415         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16416         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16417         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16418         (TARGET_CHKP_BOUND_MODE): Likewise.
16419         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16420         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16421         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16422         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16423         * config/i386/i386.h (TARGET_MPX): Likewise.
16424         (TARGET_MPX_P): Likewise.
16425         (VALID_BND_REG_MODE): Likewise.
16426         (FIRST_BND_REG): Likewise.
16427         (LAST_BND_REG): Likewise.
16428         (enum reg_class): Likewise.
16429         (BND_REG_P): Likewise.
16430         (BND_REGNO_P): Likewise.
16431         (BNDmode): Likewise.
16432         (ADJUST_INSN_LENGTH): Likewise.
16433         * config/i386/i386.md: Likewise.
16434         * config/i386/i386.opt: Likewise.
16435         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16436         (defined): Likewise.
16437         (LINK_MPX): Likewise.
16438         (MPX_SPEC): Likewise.
16439         (LIBMPX_SPEC): Likewise.
16440         (LIBMPXWRAPPERS_SPEC): Likewise.
16441         (CHKP_SPEC): Likewise.
16442         * config/i386/predicates.md: Likewise.
16443         * dbxout.c (dbxout_type): Likewise.
16444         * doc/extend.texi: Likewise.
16445         * doc/invoke.texi: Likewise.
16446         * doc/md.texi: Likewise.
16447         * doc/tm.texi: Likewise.
16448         * doc/tm.texi.in: Likewise.
16449         * dwarf2out.c (is_base_type): Likewise.
16450         (gen_formal_types_die): Likewise.
16451         (gen_subprogram_die): Likewise.
16452         (gen_type_die_with_usage): Likewise.
16453         (gen_decl_die): Likewise.
16454         (dwarf2out_late_global_decl): Likewise.
16455         * expr.c (expand_assignment): Likewise.
16456         (emit_storent_insn): Likewise.
16457         (store_expr_with_bounds): Likewise.
16458         (store_expr): Likewise.
16459         (expand_expr_real_1): Likewise.
16460         * expr.h (store_expr_with_bounds): Likewise.
16461         * function.c (use_register_for_decl): Likewise.
16462         (struct bounds_parm_data): Likewise.
16463         (assign_parms_augmented_arg_list): Likewise.
16464         (assign_parm_find_entry_rtl): Likewise.
16465         (assign_parm_is_stack_parm): Likewise.
16466         (assign_parm_load_bounds): Likewise.
16467         (assign_bounds): Likewise.
16468         (assign_parms): Likewise.
16469         (expand_function_start): Likewise.
16470         * gcc.c (CHKP_SPEC): Likewise.
16471         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16472         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16473         (wrestrict_dom_walker::check_call): Likewise.
16474         * gimple.c (gimple_build_call_from_tree): Likewise.
16475         * gimple.h (enum gf_mask): Likewise.
16476         (gimple_call_with_bounds_p): Likewise.
16477         (gimple_call_set_with_bounds): Likewise.
16478         * gimplify.c (gimplify_init_constructor): Likewise.
16479         * ipa-cp.c (initialize_node_lattices): Likewise.
16480         (propagate_constants_across_call): Likewise.
16481         (find_more_scalar_values_for_callers_subset): Likewise.
16482         * ipa-hsa.c (process_hsa_functions): Likewise.
16483         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16484         * ipa-icf.c (sem_function::merge): Likewise.
16485         * ipa-inline.c (early_inliner): Likewise.
16486         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16487         (warn_function_cold): Likewise.
16488         (propagate_pure_const): Likewise.
16489         * ipa-ref.h (enum GTY): Likewise.
16490         * ipa-split.c (find_retbnd): Likewise.
16491         (consider_split): Likewise.
16492         (split_function): Likewise.
16493         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16494         * ipa.c (walk_polymorphic_call_targets): Likewise.
16495         (symbol_table::remove_unreachable_nodes): Likewise.
16496         (process_references): Likewise.
16497         (cgraph_build_static_cdtor_1): Likewise.
16498         * lto-cgraph.c (lto_output_node): Likewise.
16499         (output_refs): Likewise.
16500         (compute_ltrans_boundary): Likewise.
16501         (input_overwrite_node): Likewise.
16502         (input_node): Likewise.
16503         (input_cgraph_1): Likewise.
16504         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
16505         * passes.c (pass_manager::execute_early_local_passes): Likewise.
16506         (class pass_chkp_instrumentation_passes): Likewise.
16507         (make_pass_chkp_instrumentation_passes): Likewise.
16508         * passes.def: Likewise.
16509         * rtl.h (struct GTY): Likewise.
16510         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
16511         * stor-layout.c (layout_type): Likewise.
16512         * symtab.c: Likewise.
16513         * target.def: Likewise.
16514         * targhooks.c (default_chkp_bound_type): Likewise.
16515         (default_chkp_bound_mode): Likewise.
16516         (default_builtin_chkp_function): Likewise.
16517         (default_chkp_function_value_bounds): Likewise.
16518         (default_chkp_make_bounds_constant): Likewise.
16519         (default_chkp_initialize_bounds): Likewise.
16520         * targhooks.h (default_chkp_bound_type): Likewise.
16521         (default_chkp_bound_mode): Likewise.
16522         (default_builtin_chkp_function): Likewise.
16523         (default_chkp_function_value_bounds): Likewise.
16524         (default_chkp_make_bounds_constant): Likewise.
16525         (default_chkp_initialize_bounds): Likewise.
16526         * toplev.c (compile_file): Likewise.
16527         (process_options): Likewise.
16528         * tree-core.h (DEF_BUILTIN): Likewise.
16529         (DEF_BUILTIN_CHKP): Likewise.
16530         * tree-inline.c (declare_return_variable): Likewise.
16531         (remap_gimple_stmt): Likewise.
16532         (copy_bb): Likewise.
16533         (initialize_inlined_parameters): Likewise.
16534         (expand_call_inline): Likewise.
16535         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
16536         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
16537         (make_pass_ipa_chkp_produce_thunks): Likewise.
16538         (make_pass_chkp): Likewise.
16539         (make_pass_chkp_opt): Likewise.
16540         (make_pass_chkp_instrumentation_passes): Likewise.
16541         * tree-pretty-print.c (dump_generic_node): Likewise.
16542         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
16543         * tree-ssa-dce.c (propagate_necessity): Likewise.
16544         (eliminate_unnecessary_stmts): Likewise.
16545         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
16546         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
16547         * tree-ssa-sccvn.h: Likewise.
16548         * tree-ssa-strlen.c (get_string_length): Likewise.
16549         (valid_builtin_call): Likewise.
16550         (adjust_last_stmt): Likewise.
16551         (handle_builtin_strchr): Likewise.
16552         (handle_builtin_strcpy): Likewise.
16553         (handle_builtin_stxncpy): Likewise.
16554         (handle_builtin_memcpy): Likewise.
16555         (handle_builtin_strcat): Likewise.
16556         (strlen_check_and_optimize_stmt): Likewise.
16557         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
16558         * tree-streamer-in.c: Likewise.
16559         * tree-streamer.c (record_common_node): Likewise.
16560         * tree.c (tree_code_size): Likewise.
16561         (wide_int_to_tree_1): Likewise.
16562         (type_contains_placeholder_1): Likewise.
16563         (build_common_tree_nodes): Likewise.
16564         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
16565         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
16566         (POINTER_BOUNDS_P): Likewise.
16567         (BOUNDED_TYPE_P): Likewise.
16568         (BOUNDED_P): Likewise.
16569         (CALL_WITH_BOUNDS_P): Likewise.
16570         (pointer_bounds_type_node): Likewise.
16571         * value-prof.c (gimple_ic): Likewise.
16572         * var-tracking.c (vt_add_function_parameters): Likewise.
16573         * varasm.c (make_decl_rtl): Likewise.
16574         (assemble_start_function): Likewise.
16575         (output_constant): Likewise.
16576         (maybe_assemble_visibility): Likewise.
16577         * varpool.c (ctor_for_folding): Likewise.
16578         * chkp-builtins.def: Remove.
16579         * ipa-chkp.c: Remove.
16580         * ipa-chkp.h: Remove.
16581         * rtl-chkp.c: Remove.
16582         * rtl-chkp.h: Remove.
16583         * tree-chkp-opt.c: Remove.
16584         * tree-chkp.c: Remove.
16585         * tree-chkp.h: Remove.
16587 2018-06-07  Carl Love  <cel@us.ibm.com>
16589         * config/rs6000/vsx.md (vextract_fp_from_shorth,
16590         vextract_fp_from_shortl): Add BE support.
16592 2018-06-07  Paul Koning  <ni1d@arrl.net>
16594         * compare-elim.c (try_merge_compare): Don't merge compare if
16595         address contains a side effect.
16596         (try_eliminate_compare): Likewise.
16598 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
16600         * config.gcc: Support "tremont".
16601         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
16602         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16603         PROCESSOR_TREMONT.
16604         * config/i386/i386.c (m_TREMONT): Define.
16605         (processor_target_table): Add "tremont".
16606         (PTA_TREMONT): Define.
16607         (ix86_lea_outperforms): Add TARGET_TREMONT.
16608         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
16609         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
16610         and M_INTEL_GOLDMONT_PLUS.
16611         (fold_builtin_cpu): Add "tremont".
16612         (ix86_add_stmt_cost): Add TARGET_TREMONT.
16613         (ix86_option_override_internal): Add "tremont".
16614         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
16615         (processor_type): Add PROCESSOR_TREMONT.
16616         * config/i386/x86-tune.def: Add m_TREMONT.
16617         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
16619 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16621         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
16622         symbol defined for msp430i* devices to be lower case.
16624 2018-06-07  Richard Biener  <rguenther@suse.de>
16626         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
16627         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
16628         Properly wrap signed arithmetic if overflow wraps.
16630 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
16632         PR tree-optimization/69615
16633         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
16634         of a cast from a same precision integral SSA_NAME in a bb dominated
16635         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
16636         cast to utype if rhs2 has already a compatible type.
16638 2018-06-07  Richard Biener  <rguenther@suse.de>
16640         PR tree-optimization/85935
16641         * graphite-scop-detection.c (find_params_in_bb): Analyze
16642         condition operands with respect to the correct loop.  Assert
16643         the analysis doesn't fail.
16645 2018-06-04  Carl Love  <cel@us.ibm.com>
16647         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
16648         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
16649         as it is slightly cheaper.
16650         (first_match_or_eos_index_<mode>):
16651         Calculate index using natural element order.
16652         (first_match_index_<mode>):
16653         Calculate index using natural element order.
16654         (first_match_or_eos_index_<mode>):
16655         Calculate index using natural order.
16656         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
16657         for BE and LE modes.
16658         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
16659         P9V_BUILTIN_VCLZLSBB_V16QI.
16660         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
16661         specific.
16663 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16665         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
16666         indentation and line wrap for many prototypes.  Add missing
16667         @smallexample directives around block of prototypes for vec_xl and
16668         vec_xst.
16670 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
16672         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
16673         track if we pass or return IEEE 128-bit floating point.
16674         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
16675         C++ mangling that is compatible with GCC 8.1.
16676         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
16677         (init_cumulative_args): Note if we pass or return IEEE 128-bit
16678         floating point types.
16679         (rs6000_function_arg_advance_1): Likewise.
16680         (rs6000_mangle_type): Optionally generate mangled names that match
16681         what GCC 8.1 generated for IEEE 128-bit floating point types.
16682         (rs6000_globalize_decl_name): If we have an external function that
16683         passes or returns IEEE 128-bit types, generate a weak reference
16684         from the mangled name used in GCC 8.1 to the current mangled
16685         name.
16686         (rs6000_init_builtins): Make __ibm128 use the long double type if
16687         long double is IBM extended double.  Make __float128 use the long
16688         double type if long double is IEEE 128-bit.
16690         PR target/85657
16691         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
16692         macro for __ibm128 built-in functions.
16693         (PACK_IF): Add __ibm128 pack/unpack functions.
16694         (UNPACK_IF): Likewise.
16695         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
16696         enable long double built-in functions if long double is IEEE
16697         128-bit floating point.
16698         (rs6000_invalid_builtin): Update long double built-in function
16699         error message.
16700         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
16701         functions, adjust the built-in function to use the long double
16702         built-in function if __ibm128 and long double are the same type.
16703         * doc/extend.texi (PowerPC builtins): Update documention for
16704         __builtin_{,un}pack_longdouble.  Add documentation for
16705         __builtin_{,un}pack_ibm128.
16707 2018-06-06  Jim Wilson  <jimw@sifive.com>
16709         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
16710         (struct machine_function): New field interrupt_mode.
16711         (riscv_handle_type_attribute): New function.  Add forward declaration.
16712         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
16713         (riscv_expand_epilogue): Check interrupt_mode field.
16714         (riscv_set_current_function): Check interrupt attribute args and
16715         set interrupt_mode field.
16716         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
16717         (riscv_sret, riscv_uret): New.
16718         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
16719         new arguments to interrupt attribute.
16721 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
16723         PR target/63177
16724         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
16725         Don't handle -mcpu=power8 if -mpower9-vector is also used.
16727 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16729         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
16730         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
16731         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16732         several redundant entries.
16734 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
16736         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
16737         type from "rtx" to "rtx_insn *".
16738         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
16739         for local "call_insn", removing cast.
16740         (ix86_expand_call): Likewise, introducing a "call_insn" local.
16742 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
16744         PR tree-optimization/86066
16745         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
16746         for BIT_INSERT_EXPR stores.
16748 2018-06-06  Richard Biener  <rguenther@suse.de>
16750         PR tree-optimization/86062
16751         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16752         component refs ontop
16753         of to be offsetted base.
16755 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16757         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
16758         to be static and remove check on interrupt attribute name.
16760 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16762         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16763         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
16765 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
16767         PR target/79924
16768         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
16769         second argument.
16770         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
16771         Remove second argument, change how error is called.
16772         (aarch64_layout_arg): Remove second argument from
16773         aarch64_err_no_fpadvsimd call.
16774         (aarch64_init_cumulative_args): Ditto.
16775         (aarch64_gimplify_va_arg_expr): Ditto.
16776         * config/aarch64/aarch64.md (mov<mode>): Ditto.
16778 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
16780         * config/i386/i386.md (simple_return_indirect_internal): New expander.
16781         (*simple_return_indirect_internal<mode>): Rename from
16782         simple_return_indirect_internal.  Use W mode iterator.
16783         (rstorssp): New expander.
16784         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
16785         (clrssbsy): New expander.
16786         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
16788 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16790         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
16791         __typeof__.
16792         (cmse_check_pointed_object): Likewise.
16794 2018-06-05  Martin Liska  <mliska@suse.cz>
16796         PR gcov-profile/47618
16797         * doc/invoke.texi: Document how -fprofile-dir format
16798         is extended.
16800 2018-06-05  Richard Biener  <rguenther@suse.de>
16802         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
16803         removal pretend DOM info isn't available so we do not update
16804         it and only remove edges, not dominated blocks.  Actually free
16805         DOM info in case we removed something.  Remove unreachable blocks.
16806         (mfb_keep_latches): Work with either DOM info or marked backedges.
16807         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
16808         first.  Mark backedges if DOM info isn't available.
16809         (Re-)compute DOM info after cleanup_control_flow_pre.
16811 2018-06-05  Richard Biener  <rguenther@suse.de>
16813         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
16814         (locus_discrim_hasher::hash): Adjust.
16815         (locus_discrim_hasher::equal): Likewise.
16816         (next_discriminator_for_locus): Work on line directly.
16817         (same_line_p): Pass in expanded locus1 as well.
16818         (assign_discriminators): Avoid redundant location expansions.
16820 2018-06-05  Richard Biener  <rguenther@suse.de>
16822         PR tree-optimization/86046
16823         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
16824         if required after clearing TREE_ADDRESSABLE.
16826 2018-06-05  Richard Biener  <rguenther@suse.de>
16828         PR tree-optimization/86047
16829         * tree-ssa-loop.c (for_each_index): Glob handling of all
16830         decls and constants and really handle all of them.
16832 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16834         PR target/81497
16835         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
16836         qualifier_void_pointer and qualifier_const_void_pointer.
16837         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
16838         (arm_init_builtins): Handle the above.
16839         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
16840         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
16841         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
16842         void intrinsics.
16844 2018-06-05  Martin Liska  <mliska@suse.cz>
16846         * auto-profile.c (read_autofdo_file): Do not use
16847         gcov_ctr_summary struct.
16848         (afdo_callsite_hot_enough_for_early_inline): Likewise.
16849         * coverage.c (struct counts_entry): Likewise.
16850         (read_counts_file): Read just single summary entry.
16851         (get_coverage_counts): Use gcov_summary struct.
16852         * coverage.h (get_coverage_counts): Likewise.
16853         * gcov-dump.c (dump_working_sets): Likewise.
16854         (tag_summary): Dump just single summary.
16855         * gcov-io.c (gcov_write_summary): Write just histogram
16856         summary.
16857         (gcov_read_summary): Read just single summary.
16858         (compute_working_sets): Use gcov_summary struct.
16859         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
16860         of GCOV_COUNTERS_SUMMABLE.
16861         (GCOV_COUNTERS_SUMMABLE): Remove.
16862         (GCOV_FIRST_VALUE_COUNTER): Replace with
16863         GCOV_COUNTER_V_INTERVAL.
16864         (struct gcov_ctr_summary): Remove.
16865         (struct gcov_summary): Directly use fields of former
16866         gcov_ctr_summary.
16867         (compute_working_sets): Use gcov_summary struct.
16868         * gcov.c (read_count_file): Do not use ctrs fields.
16869         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
16870         struct.
16871         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
16872         struct.
16873         * profile.c: Likewise.
16874         * profile.h: Likewise.
16876 2018-06-05  Martin Liska  <mliska@suse.cz>
16878         PR gcov-profile/84846
16879         * gcov.c (output_lines): Print working directory only
16880         in intermediate format.
16882 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
16884         * config/s390/s390-builtin-types.def: Add void function type.
16885         * config/s390/s390-builtins.def: Use the function type for the
16886         tbeginc builtin.
16888 2018-06-04  Jim Wilson  <jimw@sifive.com>
16890         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
16891         to int.
16892         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
16893         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
16894         handle EH_RETURN_DATA_REGNO registers properly.
16895         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
16896         (riscv_expand_epilogue): Update comment.  Change argument name and
16897         type.  Update code to use new name and type.  Pass new args to
16898         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
16899         EXCEPTION_RETURN.
16900         * config/riscv/riscv.md (NORMAL_RETURN): New.
16901         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
16902         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16903         (eh_return): Call gen_eh_return_internal and emit barrier.
16904         (eh_return_internal): Call riscv_expand_epilogue.
16906 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
16908         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
16909         bit_insertion field and declare can_be_merged_into method.
16910         (merged_store_group::can_be_merged_into): New method.
16911         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
16912         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
16913         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
16915 2018-06-04  Richard Biener  <rguenther@suse.de>
16917         PR tree-optimization/85955
16918         * builtins.c (fold_builtin_sincos): Convert pointers to
16919         destination to appropriate type before dereferencing.
16921 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
16923         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
16925 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
16927         * expr.c (expand_expr_real_1): Force the operand into memory if
16928         its TYPE_MODE is BLKmode and if there is no integer mode for
16929         the number of bits being extracted.
16931 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16933         PR target/85832
16934         PR target/86036
16935         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16936         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
16938 2018-06-04  Richard Biener  <rguenther@suse.de>
16940         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
16941         (cleanup_tree_cfg_noloop): ... single caller.  Do
16942         start_recording_case_labels later.
16944 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
16946         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
16947         to _IMMINTRIN_H_INCLUDED.
16948         * config/i386/pconfigintrin.h: Ditto.
16949         * config/i386/waitpkgintrin.h: Ditto.
16950         * config/i386/immintrin.h: Add includes for sgxintrin.h,
16951         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
16952         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
16953         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
16954         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
16955         waitpkgintrin.h and cldemoteintrin.h.
16957 2018-06-04  Richard Biener  <rguenther@suse.de>
16959         PR tree-optimization/86038
16960         * tracer.c (find_best_successor): Check probability for
16961         being initialized, bail out if not.
16963 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
16965         PR target/86003
16966         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
16967         of bits to ignore when comparing architectures.
16969 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16971         PR tree-optimization/69615
16972         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
16973         maximum or minimum of the type, try to merge it also as if
16974         range1 is + [-, x - 1] or + [x + 1, -].
16976         PR c++/86025
16977         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
16979 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
16981         PR tree-optimization/86034
16982         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
16983         the unsigned bitfield type in a bit insertion sequence if it does not
16984         have a larger precision than the bitfield size.
16985         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
16987 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
16989         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
16991 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
16993         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
16994         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
16995         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
16996         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
16998 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17000         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17001         Disable -fdelete-null-pointer-checks for ELF toolchain.
17003 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17004             Kito Cheng  <kito.cheng@gmail.com>
17006         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17007         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17008         (nds32*-*-*): Add glibc and uclibc conditions.
17009         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17010         (TARGET_EXCEPT_UNWIND_INFO): Define.
17011         * config/nds32/elf.h: New file.
17012         * config/nds32/linux.h: New file.
17013         * config/nds32/nds32-elf.opt: New file.
17014         * config/nds32/nds32-linux.opt: New file.
17015         * config/nds32/nds32-fp-as-gp.c
17016         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17017         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17018         TARGET_LINUX_ABI.
17019         (nds32_asm_file_end): Ditto.
17020         (nds32_print_operand): Ditto.
17021         (nds32_insert_attributes): Ditto.
17022         (nds32_init_libfuncs): New function.
17023         (TARGET_HAVE_TLS): Define.
17024         (TARGET_INIT_LIBFUNCS): Define.
17025         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17026         spec content.
17027         (TARGET_ELF): Apply different mcmodel setting.
17028         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17029         been migrated into elf.h and linux.h files.
17030         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17031         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17032         (mcmodel): The content has been migrated into nds32-elf.opt and
17033         nds32-linux.opt files.
17034         * config/nds32/t-elf: New file.
17035         * config/nds32/t-linux: New file.
17037 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17038             Shiva Chen  <shiva0217@gmail.com>
17040         * config/nds32/constants.md (unspec_volatile_element): Add
17041         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17042         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17043         optimization.
17044         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17045         (make_pass_nds32_fp_as_gp): Declare.
17046         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17047         optmization pass.
17048         (nds32_asm_function_end_prologue): Remove unused asm output.
17049         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17050         (nds32_asm_file_start): Output necessary fp_as_gp information.
17051         (nds32_option_override): Adjust register usage.
17052         (nds32_expand_prologue): Consider fp_as_gp situation.
17053         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17054         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17055         (epilogue): Ditto.
17056         (return): Ditto.
17057         (simple_return): Ditto.
17058         (omit_fp_begin): Output special directive for fp_as_gp.
17059         (omit_fp_end): Output special directive for fp_as_gp.
17060         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17061         mforbid-fp-as-gp): New options.
17063 2018-06-01  Mark Wielaard  <mark@klomp.org>
17065         * dwarf2out.c (dwarf2out_finish): Remove generation of
17066         DW_AT_loclists_base.
17068 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17070         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17071         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17072         (struct merged_store_group): Add bit_insertion field.
17073         (dump_char_array): Use standard hexadecimal format.
17074         (merged_store_group::merged_store_group): Set bit_insertion to false.
17075         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17076         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17077         also print the mask in the dump file.
17078         (pass_store_merging::gate): Minor tweak.
17079         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17080         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17081         stores with INTEGER_CST stores.
17082         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17083         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17084         and use it throughout.  Generate bit insertion sequences if need be.
17085         (pass_store_merging::process_store): Remove redundant condition.
17086         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17088 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17090         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17091         the 128-bit floating point types.  Fix function comment.
17093 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17095         * config/aarch64/aarch64-simd.md
17096         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17097         mnemonics.
17098         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17099         mnemonics.
17101 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17103         PR tree-optimization/85989
17104         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17105         variable.
17106         (backprop::intersect_uses): Check it when deciding whether this
17107         is a backedge reference.
17108         (backprop::process_block): Add each phi to m_visited_phis
17109         after visiting it, then clear it at the end.
17111 2018-06-01  Richard Biener  <rguenther@suse.de>
17113         * tree-vectorizer.h (vect_dr_stmt): New function.
17114         (vect_get_load_cost): Adjust.
17115         (vect_get_store_cost): Likewise.
17116         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17117         Use vect_dr_stmt instead of DR_SMTT.
17118         (vect_record_base_alignments): Likewise.
17119         (vect_calculate_target_alignment): Likewise.
17120         (vect_compute_data_ref_alignment): Likewise and make static.
17121         (vect_update_misalignment_for_peel): Likewise.
17122         (vect_verify_datarefs_alignment): Likewise.
17123         (vector_alignment_reachable_p): Likewise.
17124         (vect_get_data_access_cost): Likewise.  Pass down
17125         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17126         (vect_get_peeling_costs_all_drs): Likewise.
17127         (vect_peeling_hash_get_lowest_cost): Likewise.
17128         (vect_enhance_data_refs_alignment): Likewise.
17129         (vect_find_same_alignment_drs): Likewise.
17130         (vect_analyze_data_refs_alignment): Likewise.
17131         (vect_analyze_group_access_1): Likewise.
17132         (vect_analyze_group_access): Likewise.
17133         (vect_analyze_data_ref_access): Likewise.
17134         (vect_analyze_data_ref_accesses): Likewise.
17135         (vect_vfa_segment_size): Likewise.
17136         (vect_small_gap_p): Likewise.
17137         (vectorizable_with_step_bound_p): Likewise.
17138         (vect_prune_runtime_alias_test_list): Likewise.
17139         (vect_analyze_data_refs): Likewise.
17140         (vect_supportable_dr_alignment): Likewise.
17141         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17142         (vect_gen_prolog_loop_niters): Likewise.
17143         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17144         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17145         modify DR_STMT.
17146         (vect_recog_mask_conversion_pattern): Likewise.
17147         (vect_try_gather_scatter_pattern): Likewise.
17148         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17149         to vect_get_store_cost.
17150         (vect_get_store_cost): Get stmt_info instead of DR.
17151         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17152         (vect_get_load_cost): Get stmt_info instead of DR.
17154 2018-06-01  Richard Biener  <rguenther@suse.de>
17156         PR middle-end/86017
17157         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17158         vars wrapped in MEM_REFs.
17160 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17162         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17163         Fix subreg tests so that we only return a choice between
17164         GENERAL_REGS and FP_REGS if the original classes included both.
17166 2018-06-01  Richard Biener  <rguenther@suse.de>
17168         PR ipa/85960
17169         * tree-ssa-structalias.c (get_function_part_constraint):
17170         Handle NULL fi->decl.
17171         (find_func_aliases_for_call): Properly handle indirect
17172         fi from direct call.
17173         (find_func_clobbers): Likewise.
17174         (ipa_pta_execute): Likewise.
17175         (create_variable_info_for): For functions that are ifunc_resolver
17176         resolve to a varinfo that contains the result of the resolver call.
17177         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17178         aliases.
17180 2018-05-31  Michael Collison  <michael.collison@arm.com>
17182         * config/aarch64/aarch64.md:
17183         (*fix_to_zero_extenddfdi2): New pattern.
17184         * gcc.target/aarch64/fix_extend1.c: New testcase.
17186 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17188         PR middle-end/78809
17189         PR middle-end/83026
17190         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17191         and BUILT_IN_STRNCMP_EQ.
17192         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17193         BUILT_IN_STRNCMP_EQ.
17194         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17195         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17196         (gimple_fold_builtin): Likewise.
17197         * tree-ssa-strlen.c (compute_string_length): New function.
17198         (determine_min_obsize): New function.
17199         (handle_builtin_string_cmp): New function to handle calls to
17200         string compare functions.
17201         (strlen_optimize_stmt): Add handling to builtin string compare
17202         calls.
17203         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17204         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17205         * tree.c (build_common_builtin_nodes): Add new defines of
17206         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17208 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17210         PR target/85984
17211         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17212         functions with naked attribute.
17214 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17216         * config/i386/sse.md (avx_vec_concat<mode>):
17217         Substitute concat_tg_mode mode attribute with xtg_mode.
17218         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17219         (concat_tg_mode): Remove mode attribute.
17221 2018-05-31  Martin Sebor  <msebor@redhat.com>
17223         PR c/82063
17224         * calls.c (alloc_max_size): Correct a logic error/typo.
17225         Treat excessive arguments as infinite.  Warn for invalid arguments.
17226         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17228 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17230         PR target/85829
17231         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17232         and movx for Haswell.
17234 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17235             Cesar Philippidis  <cesar@codesourcery.com>
17237         PR middle-end/85879
17238         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17239         when emitting error on private/firstprivate reductions.
17240         * omp-low.c (lower_omp_target): Avoid reference-type processing
17241         on pointers for firstprivate clause.
17243 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17245         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17246         (st1x2): Likewise.
17247         (st1x3): Likewise.
17248         * config/aarch64/aarch64-simd.md
17249         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17250         (aarch64_ld1_x3_<mode>): Likewise
17251         (aarch64_st1x2<VALLDIF:mode>): Likewise
17252         (aarch64_st1_x2_<mode>): Likewise
17253         (aarch64_st1x3<VALLDIF:mode>): Likewise
17254         (aarch64_st1_x3_<mode>): Likewise
17255         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17256         (vld1_s8_x3): Likewise.
17257         (vld1_u16_x3): Likewise.
17258         (vld1_s16_x3): Likewise.
17259         (vld1_u32_x3): Likewise.
17260         (vld1_s32_x3): Likewise.
17261         (vld1_u64_x3): Likewise.
17262         (vld1_s64_x3): Likewise.
17263         (vld1_f16_x3): Likewise.
17264         (vld1_f32_x3): Likewise.
17265         (vld1_f64_x3): Likewise.
17266         (vld1_p8_x3): Likewise.
17267         (vld1_p16_x3): Likewise.
17268         (vld1_p64_x3): Likewise.
17269         (vld1q_u8_x3): Likewise.
17270         (vld1q_s8_x3): Likewise.
17271         (vld1q_u16_x3): Likewise.
17272         (vld1q_s16_x3): Likewise.
17273         (vld1q_u32_x3): Likewise.
17274         (vld1q_s32_x3): Likewise.
17275         (vld1q_u64_x3): Likewise.
17276         (vld1q_s64_x3): Likewise.
17277         (vld1q_f16_x3): Likewise.
17278         (vld1q_f32_x3): Likewise.
17279         (vld1q_f64_x3): Likewise.
17280         (vld1q_p8_x3): Likewise.
17281         (vld1q_p16_x3): Likewise.
17282         (vld1q_p64_x3): Likewise.
17283         (vst1_s64_x2): Likewise.
17284         (vst1_u64_x2): Likewise.
17285         (vst1_f64_x2): Likewise.
17286         (vst1_s8_x2): Likewise.
17287         (vst1_p8_x2): Likewise.
17288         (vst1_s16_x2): Likewise.
17289         (vst1_p16_x2): Likewise.
17290         (vst1_s32_x2): Likewise.
17291         (vst1_u8_x2): Likewise.
17292         (vst1_u16_x2): Likewise.
17293         (vst1_u32_x2): Likewise.
17294         (vst1_f16_x2): Likewise.
17295         (vst1_f32_x2): Likewise.
17296         (vst1_p64_x2): Likewise.
17297         (vst1q_s8_x2): Likewise.
17298         (vst1q_p8_x2): Likewise.
17299         (vst1q_s16_x2): Likewise.
17300         (vst1q_p16_x2): Likewise.
17301         (vst1q_s32_x2): Likewise.
17302         (vst1q_s64_x2): Likewise.
17303         (vst1q_u8_x2): Likewise.
17304         (vst1q_u16_x2): Likewise.
17305         (vst1q_u32_x2): Likewise.
17306         (vst1q_u64_x2): Likewise.
17307         (vst1q_f16_x2): Likewise.
17308         (vst1q_f32_x2): Likewise.
17309         (vst1q_f64_x2): Likewise.
17310         (vst1q_p64_x2): Likewise.
17311         (vst1_s64_x3): Likewise.
17312         (vst1_u64_x3): Likewise.
17313         (vst1_f64_x3): Likewise.
17314         (vst1_s8_x3): Likewise.
17315         (vst1_p8_x3): Likewise.
17316         (vst1_s16_x3): Likewise.
17317         (vst1_p16_x3): Likewise.
17318         (vst1_s32_x3): Likewise.
17319         (vst1_u8_x3): Likewise.
17320         (vst1_u16_x3): Likewise.
17321         (vst1_u32_x3): Likewise.
17322         (vst1_f16_x3): Likewise.
17323         (vst1_f32_x3): Likewise.
17324         (vst1_p64_x3): Likewise.
17325         (vst1q_s8_x3): Likewise.
17326         (vst1q_p8_x3): Likewise.
17327         (vst1q_s16_x3): Likewise.
17328         (vst1q_p16_x3): Likewise.
17329         (vst1q_s32_x3): Likewise.
17330         (vst1q_s64_x3): Likewise.
17331         (vst1q_u8_x3): Likewise.
17332         (vst1q_u16_x3): Likewise.
17333         (vst1q_u32_x3): Likewise.
17334         (vst1q_u64_x3): Likewise.
17335         (vst1q_f16_x3): Likewise.
17336         (vst1q_f32_x3): Likewise.
17337         (vst1q_f64_x3): Likewise.
17338         (vst1q_p64_x3): Likewise.
17340 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17342         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17343         user_label_prefix to name.
17345         * tree-core.h: Update comment about the format of NAME string
17346         passed to handler in attribute_spec.
17348         * config/msp430/msp430.md: Remove erroneous subreg expression from
17349         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17350         zero_extend{q,h}isi2.
17352 2018-05-30  Borislav Petkov  <bp@suse.de>
17354         * doc/extend.texi: Document some architecture specific
17355         constraints and sort entries.
17357 2018-05-30  Martin Sebor  <msebor@redhat.com>
17359         PR middle-end/85369
17360         * builtins.c (expand_builtin_stpcpy_1): New function.
17361         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17362         only if the former succeeds.
17364 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17366         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17367         in saphira.
17369 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17371         * doc/invoke.texi (-flinker-output): Document
17373 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17375         * passes.c (ipa_write_summaries): Only modify statements if body
17376         is in memory.
17377         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17378         incrementally linking.
17379         (ipa_passes): Likewise.
17380         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17381         pass down resolution info.
17382         * common.opt (flag_incremental_link): Update info.
17383         * gcc.c (plugin specs): Turn flinker-output=* to
17384         -plugin-opt=-linker-output-known
17385         * toplev.c (compile_file): Also cut compilation when doing incremental
17386         link.
17387         * flag-types. (enum lto_partition_model): Add
17388         LTO_LINKER_OUTPUT_NOLTOREL.
17389         (invoke.texi): Add -flinker-output docs.
17390         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17391         link same way as WPA; do not stream in dead initializers.
17393         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17394         darf2out_register_external_decl): Support incremental link.
17396 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17398         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17400 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17402         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17403         it down to simple_object_copy_lto_debug_sections.
17404         (run_gcc): Determine incremental LTO link time and configure
17405         lto1 into non-wpa mode, disable renaming of debug sections.
17407 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17409         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17410         descriptions of various incorrectly documented functions.
17412 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17414         Revert:
17415         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17416         address check not strict.
17418 2018-05-30  Richard Biener  <rguenther@suse.de>
17420         PR tree-optimization/85964
17421         * tracer.c (better_p): Drop initialized count check, we only
17422         call the function with initialized counts now.
17423         (find_best_successor): Do find a best edge if one
17424         has uninitialized count.
17425         (find_best_predecessor): Likewise.  Do BB frequency check only
17426         if count is initialized.
17428 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17430         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17431         (aarch64_ldrstr_offset_compare): New.
17432         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17433         load/store orderings.
17434         (aarch64_gen_adjusted_ldpstp): Likewise.
17436 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17438         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17439         Check for subset of GENERAL_REGS and FP_REGS.
17440         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17441         r=w alternative.
17443 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17445         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17446         and wi::to_poly_offset.  Add the current offset and then check
17447         whether the sum fits, rather than using an unchecked addition of
17448         a checked term.  Check for a shwi rather than a uhwi.
17449         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17450         (store_constructor): Use poly_int_tree_p.
17451         (expand_expr_real_1): Likewise.
17452         * function.c (assign_temp): Likewise.
17453         * fold-const.c (const_binop): Use poly_int_tree_p and
17454         wi::to_poly_offset.
17455         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17456         division.
17457         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17458         to_poly_offset for MEM offsets.
17459         * ipa-icf.c (sem_variable::equals): Likewise.
17460         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17461         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17462         wi::to_poly_offset for BIT_FIELD_REF offsets.
17463         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17464         wi::to_poly_offset.
17465         * var-tracking.c (emit_note_insn_var_location): Use
17466         tree_to_poly_uint64.
17468 2018-05-29  Jim Wilson  <jimw@sifive.com>
17470         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17472 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17474         PR target/85950
17475         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17476         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17477         sequence.
17478         (sse4_1_round<mode>2): Use nonimmediate_operand
17479         for operand 1 predicate.
17481 2018-05-29  Martin Sebor  <msebor@redhat.com>
17482             Richard Biener  <rguenther@suse.de>
17484         PR testsuite/85888
17485         * calls.c (get_size_range): Call determine_value_range instead
17486         of get_value_range..
17487         * tree-vrp.h (determine_value_range): Declared new function.
17488         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17490 2018-05-29  Richard Biener  <rguenther@suse.de>
17492         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17493         sure to use non-pattern stmts for get_earlier_stmt arguments.
17494         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17495         called on pattern stmts.
17496         (get_later_stmt): Likewise.
17498 2018-05-29  Martin Liska  <mliska@suse.cz>
17500         PR gcov-profile/85759
17501         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
17502         env variables.
17504 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
17506         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
17507         VEC_UNPACK_*_EXPR.
17508         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
17509         VEC_PACK_*_EXPR.
17511         PR target/85918
17512         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17513         VEC_PACK_FLOAT_EXPR): New tree codes.
17514         * tree-pretty-print.c (op_code_prio): Handle
17515         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17516         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17517         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17518         * tree-inline.c (estimate_operator_cost): Likewise.
17519         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
17520         * fold-const.c (const_binop): Likewise.
17521         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
17522         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17523         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17524         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
17525         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17526         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17527         * expr.c (expand_expr_real_2): Likewise.
17528         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
17529         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
17530         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
17531         optabs.
17532         * optabs.c (expand_widen_pattern_expr): For
17533         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
17534         sign from result type rather than operand's type.
17535         (expand_binop_directly): For vec_packu_float_optab and
17536         vec_packs_float_optab allow result type to be different from operand's
17537         type.
17538         * optabs-tree.c (optab_for_tree_code): Handle
17539         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17540         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
17541         * tree-vect-generic.c (expand_vector_operations_1):  Handle
17542         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17543         VEC_PACK_FLOAT_EXPR.
17544         * tree-vect-stmts.c (supportable_widening_operation): Handle
17545         FIX_TRUNC_EXPR.
17546         (supportable_narrowing_operation): Handle FLOAT_EXPR.
17547         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
17548         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
17549         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
17550         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
17551         mode attributes.
17552         (vec_pack<floatprefix>_float_<mode>): New expander.
17553         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
17554         attributes.
17555         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
17556         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
17557         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
17558         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
17559         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
17560         Document.
17561         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
17562         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
17563         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17564         VEC_PACK_FLOAT_EXPR): Document.
17566 2018-05-29  Richard Biener  <rguenther@suse.de>
17568         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
17569         member.
17570         (stmt_vec_info_vec): Make pointer.
17571         (init_stmt_vec_info_vec): Remove.
17572         (free_stmt_vec_info_vec): Likewise.
17573         (set_stmt_vec_info_vec): New function.
17574         (free_stmt_vec_infos): Likewise.
17575         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
17576         (set_vinfo_for_stmt): Likewise.
17577         (get_earlier_stmt): Likewise.
17578         (get_later_stmt): Likewise.
17579         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
17580         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
17581         (vec_info::~vec_info): Free stmt_vec_infos.
17582         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
17583         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
17584         (pass_slp_vectorize::execute): Likewise.
17585         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
17586         (free_stmt_vec_info_vec): Likewise.
17587         (set_stmt_vec_info_vec): New function.
17588         (free_stmt_vec_infos): Likewise.
17589         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
17590         the global stmt_vec_info_vec.
17591         * tree-parloops.c (gather_scalar_reductions): Use
17592         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
17593         vector.
17595 2018-05-29  Richard Biener  <rguenther@suse.de>
17597         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
17599 2018-05-29  Martin Liska  <mliska@suse.cz>
17600             David Malcolm  <dmalcolm@redhat.com>
17602         * vec.c (test_reverse): New.
17603         (vec_c_tests): Add new test.
17604         * vec.h (vl_ptr>::reverse): New function.
17606 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
17608         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
17610         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
17611         and later.
17613 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17615         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
17617 2018-05-28  Richard Biener  <rguenther@suse.de>
17619         PR tree-optimization/85933
17620         * tree-vect-data-refs.c (vect_record_base_alignments): Only
17621         look at stmts marked as vectorizable.
17623 2018-05-28  Richard Biener  <rguenther@suse.de>
17625         PR tree-optimization/85934
17626         * tree-vect-generic.c (expand_vector_operations_1): Hoist
17627         vector boolean check before scalar optimization.
17629 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
17631         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
17632         for armv5te.
17634 2018-05-28  Mark Wielaard  <mark@klomp.org>
17636         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
17637         if it is an expression containing a minus sign.
17639 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
17641         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
17643 2018-05-27  Paul Koning  <ni1d@arrl.net>
17645         * config/pdp11/pdp11.md (truncsihi2): Remove.
17647 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
17648             Chung-Ju Wu  <jasonwucj@gmail.com>
17650         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
17651         implementation.
17652         (unaligned_store_dw): Ditto.
17653         * config/nds32/nds32-memory-manipulation.c
17654         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17655         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
17656         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
17657         (emit_setmem_word_loop): Rename to ...
17658         (emit_setmem_doubleword_loop): ... this.
17659         (nds32_gen_dup_4_byte_to_word_value): New function.
17660         (nds32_gen_dup_8_byte_to_double_word_value): New function.
17661         (nds32_expand_setmem_loop): Refine implementation.
17662         (nds32_expand_setmem_loop_v3m): Ditto.
17663         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
17664         pattern.
17666 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17668         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
17670 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17672         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
17673         (nds32_init_machine_status): Initialize machine->attr_naked_p and
17674         machine->attr_no_prologue_p.
17675         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
17676         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
17677         (nds32_expand_epilogue): Consider attr_naked_p.
17678         (nds32_expand_epilogue_v3pop): Likewise.
17679         (nds32_can_use_return_insn): Likewise.
17680         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
17681         attr_no_prologue_p fields.
17682         * config/nds32/nds32.opt (mret-in-naked-func): New option.
17684 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
17686         PR target/85918
17687         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
17688         attributes.
17689         * config/i386/sse.md
17690         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
17691         Rename to ...
17692         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
17693         ... this.
17694         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
17695         Rename to ...
17696         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
17697         ... this.
17698         (*<floatsuffix>floatv2div2sf2): Rename to ...
17699         (*float<floatunssuffix>v2div2sf2): ... this.
17700         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
17701         (float<floatunssuffix>v2div2sf2_mask): ... this.
17702         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
17703         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
17704         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
17705         to ...
17706         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
17707         ... this.
17708         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17709         Rename to ...
17710         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17711         ... this.
17712         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17713         Rename to ...
17714         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17715         ... this.
17716         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
17717         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
17718         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
17719         gen_ufix_truncv8dfv8si2.
17720         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
17721         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
17722         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
17723         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
17724         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
17725         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
17726         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
17727         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
17729 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17731         PR target/85900
17732         PR target/85345
17733         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
17735 2018-05-25  Jim Wilson  <jimw@sifive.com>
17737         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
17738         * config/riscv/riscv.c (struct machine_function): Add
17739         interrupt_handler_p and attribute_checked_p fields.
17740         (riscv_attribute_table): Add interrupt.
17741         (riscv_interrupt_type_p): New.
17742         (riscv_save_reg_p): Save extra regs for interrupt handler.
17743         (riscv_use_save_libcall): Return false  for interrupt handler.
17744         (riscv_first_stack_step): Add forward declaration.
17745         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
17746         for interrupt handler with large frame.  Use it for saved reg list.
17747         (riscv_expand_prologue): Move flag_stack_usage_info support to
17748         eliminate duplication.
17749         (riscv_expand_epilogue): Generate mret for interrupt handler.
17750         (riscv_epilogue_uses): New.
17751         (riscv_can_use_return_insn): Return false for interrupt handler.
17752         (riscv_function_ok_for_sibcall): Likewise.
17753         (riscv_set_current_function): Add interrupt handler support.
17754         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
17755         * config/riscv/riscv.md (UNSPECV_MRET): New.
17756         (GP_REGNUM): New.
17757         (riscv_frflags, riscv_fsflags): Use tab after opcode.
17758         (riscv_mret): New.
17759         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
17761 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
17763         PR tree-optimization/85712
17764         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
17765         this candidate has already been replaced in-situ by a copy.
17767 2018-05-25  Jason Merrill  <jason@redhat.com>
17769         PR c++/80485 - inline function non-zero address.
17770         * symtab.c (nonzero_address): Check DECL_COMDAT.
17772 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
17774         PR target/83628
17775         * config/alpha/alpha.md (ashlsi3): New insn pattern.
17776         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
17777         extension of SImode operation.  Use const123_operand predicate.
17778         (*saddsi_1): Remove.
17779         (*saddl_se_1): Ditto.
17780         (*ssubsi_1): Ditto.
17781         (*ssubl_se_1): Ditto.
17782         * config/alpha/predicates.md (const123_operand): New predicate.
17783         * config/alpha/constraints.md (P): Use IN_RANGE.
17785 2018-05-25  Richard Biener  <rguenther@suse.de>
17787         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
17788         defaulted to true.
17789         (ref_maybe_used_by_stmt_p): Likewise.
17790         (stmt_may_clobber_ref_p): Likewise.
17791         (stmt_may_clobber_ref_p_1): Likewise.
17792         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
17793         and pass it along.
17794         (ref_maybe_used_by_stmt_p): Likewise.
17795         (stmt_may_clobber_ref_p): Likewise.
17796         (stmt_may_clobber_ref_p_1): Likewise.
17797         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
17798         the alias oracle to disambiguate DRs with stmts DR analysis
17799         couldn't handle.
17800         (vect_analyze_data_refs): Do not give up on not analyzable
17801         DRs for BB vectorization.  Remove code truncating the dataref
17802         vector.
17804 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
17806         PR target/85832
17807         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17808         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
17809         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
17811 2018-05-25  Richard Biener  <rguenther@suse.de>
17813         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
17814         function, combining stmt data ref gathering and fatal analysis
17815         parts.
17816         (vect_analyze_data_refs): Remove now redudnant code and simplify.
17817         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
17818         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
17819         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
17820         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
17822 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17824         PR tree-optimization/85720
17825         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
17826         SCC if all partitions are builtins.
17827         (version_loop_by_alias_check): New parameter.  Generate cancelable
17828         runtime alias check if all partitions are builtins.
17829         (distribute_loop): Update call to above function.
17831 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17833         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
17834         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
17835         (parm_default_def_partition_arg): Ditto.
17836         (set_parm_default_def_partition): Ditto.
17837         (get_parm_default_def_partitions): Ditto and make it static.
17838         (get_undefined_value_partitions): Ditto and make it static.
17839         (remove_ssa_form): Refactor call to init_var_map here.
17840         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
17841         computation for loop region.
17842         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
17843         (register_default_def): Delete.
17844         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
17845         (parm_default_def_partition_arg): Ditto.
17846         (set_parm_default_def_partition): Ditto.
17847         (get_parm_default_def_partitions): Ditto and make it static.
17848         (get_undefined_value_partitions): Ditto and make it static.
17849         (coalesce_with_default, coalesce_with_default): Update comment.
17850         (create_coalesce_list_for_region): New func factored out from
17851         create_outofssa_var_map.
17852         (populate_coalesce_list_for_outofssa): New func factored out from
17853         create_outofssa_var_map and coalesce_ssa_name.
17854         (create_outofssa_var_map): Delete.
17855         (coalesce_ssa_name): Refactor to support live range computation.
17856         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
17857         (get_parm_default_def_partitions): Delete.
17858         (get_undefined_value_partitions): Ditto.
17859         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
17860         computation for loop region.
17861         (new_tree_live_info, loe_visit_block): Ditto.
17862         (live_worklist, set_var_live_on_entry): Ditto.
17863         (calculate_live_on_exit, verify_live_on_entry): Ditto.
17864         * tree-ssa-live.h (struct _var_map): New fields.
17865         (init_var_map): Change decl.
17866         (region_contains_p): New.
17868 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17870         * tree-ssa-live.h (live_merge_and_clear): Delete.
17872 2018-05-25  Richard Biener  <rguenther@suse.de>
17874         PR c++/85912
17875         * tree-dump.c (dequeue_and_dump): Remove access to removed
17876         operand 2 of a SWITCH_EXPR.
17878 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17880         * doc/sourcebuild.texi (vect_double_cond_arith): Include
17881         multiplication and division.
17882         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
17883         (cond_udiv@var{m}, cond_umod@var{m}): Document.
17884         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
17885         (cond_udiv_optab, cond_umod_optab): New optabs.
17886         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
17887         (IFN_COND_RDIV): New internal functions.
17888         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
17889         TRUNC_MOD_EXPR and RDIV_EXPR.
17890         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
17891         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
17892         New unspecs.
17893         (SVE_INT_BINARY): Include mult.
17894         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
17895         (optab, sve_int_op): Handle mult.
17896         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
17897         UNSPEC_COND_DIV.
17898         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
17899         for SVE_INT_BINARY_SD.
17901 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17903         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
17904         (optab, sve_int_op): Handle div and udiv.
17905         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
17906         for SVE_INT_BINARY_SD.
17907         (*<optab><mode>3): New insn for the same.
17909 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17911         * tree-vect-patterns.c: Include predict.h.
17912         (vect_recog_divmod_pattern): Restrict check for division support
17913         to when optimizing for size.
17915 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17917         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
17918         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
17919         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
17920         (gimple_match_op::set_op): Likewise.
17921         (gimple_resimplify4): Declare.
17922         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
17923         (expr::gen_transform): Likewise.
17924         (decision_tree::gen): Generate a simplification routine for 4 operands.
17925         * gimple-match-head.c (gimple_simplify): Add an overload for
17926         4 operands.  In the top-level function, handle up to 4 call
17927         arguments and call gimple_resimplify4.
17928         (gimple_resimplify4): New function.
17929         (build_call_internal): Pass a fourth operand.
17930         (maybe_push_to_seq): Likewise.
17931         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
17932         Fold VEC_COND_EXPRs of an operation and a default value into
17933         an IFN_COND_* function if possible.
17934         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
17935         New unspecs.
17936         (SVE_COND_FP_BINARY): Include them.
17937         (optab, sve_fp_op): Handle them.
17938         (SVE_INT_BINARY_REV): New code iterator.
17939         (SVE_COND_FP_BINARY_REV): New int iterator.
17940         (commutative): New int attribute.
17941         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
17942         Declare.
17943         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
17944         function.
17945         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
17946         (*cond_<optab><mode>): New patterns for reversed operands.
17948 2018-05-25  Richard Biener  <rguenther@suse.de>
17950         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
17951         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
17952         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
17953         (STMT_VINFO_GROUPED_ACCESS): Adjust.
17954         * tree-vect-data-refs.c (everywhere): Adjust users.
17955         * tree-vect-loop.c (everywhere): Likewise.
17956         * tree-vect-slp.c (everywhere): Likewise.
17957         * tree-vect-stmts.c (everywhere): Likewise.
17958         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
17960 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17962         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
17963         Rename to...
17964         (gcc_cv_as_section_exclude): ... this.
17965         Try Solaris as #exclude syntax.
17966         * configure: Regenerate.
17967         * config.in: Regenerate.
17968         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
17969         SECTION_EXCLUDE.
17970         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
17971         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
17973         * varasm.c (default_elf_asm_named_section): Don't check if
17974         HAVE_GAS_SECTION_EXCLUDE is defined.
17976 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17978         * doc/md.texi: Update the documentation of the cond_* optabs
17979         to mention the new final operand.  Fix GET_MODE_NUNITS call.
17980         Describe the scalar case too.
17981         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
17982         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
17983         instead of 2.
17984         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
17985         (get_conditional_internal_fn): Update comment.
17986         * tree-vect-loop.c (vectorizable_reduction): Pass the original
17987         accumulator value as a final argument to conditional functions.
17988         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
17989         a define_expand and add an "else" operand.  Assert for now that
17990         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
17991         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
17992         (*cond_<optab><mode>): New patterns.
17993         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
17994         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17995         (UNSPEC_COND_EOR): Delete.
17996         (optab): Remove associated mappings.
17997         (SVE_INT_BINARY): New code iterator.
17998         (sve_int_op): Remove int attribute and add "minus" to the code
17999         attribute.
18000         (SVE_COND_INT_OP): Delete.
18001         (SVE_COND_FP_OP): Rename to...
18002         (SVE_COND_FP_BINARY): ...this.
18004 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18006         * optabs.c (can_reuse_operands_p): New function.
18007         (maybe_legitimize_operands): Try to reuse the results for
18008         earlier operands.
18010 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18012         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18013         Add {q} suffix to insn mnemonic.
18015 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18017         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18018         (msp430_warn_func_return): New.
18020 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18022         * fold-const.c (tree_nonzero_bits): New function.
18023         * fold-const.h (tree_nonzero_bits): Likewise.
18024         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18025         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18027 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18029         PR target/85900
18030         PR target/85345
18031         * varasm.c (assemble_alias): Check ifunc_resolver only on
18032         FUNCTION_DECL.
18034 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18036         PR target/85903
18037         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18038         when memory input operand is handled.
18040 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18042         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18043         global.
18044         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18046 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18048         * match.pd: Delay FMA folds until after vectorization.
18050 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18052         PR target/83009
18053         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18054         address check not strict.
18056 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18058         * gimple-match.h (gimple_match_op): New class.
18059         (mprts_hook): Replace parameters with a gimple_match_op *.
18060         (maybe_build_generic_op): Likewise.
18061         (gimple_simplified_result_is_gimple_val): Replace parameters with
18062         a const gimple_match_op *.
18063         (gimple_simplify): Replace code_helper * and tree * parameters with
18064         a gimple_match_op * parameter.
18065         (gimple_resimplify1): Replace code_helper *, tree and tree *
18066         parameters with a gimple_match_op * parameter.
18067         (gimple_resimplify2): Likewise.
18068         (gimple_resimplify3): Likewise.
18069         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18070         parameters with a gimple_match_op * parameter.
18071         * gimple-match-head.c (gimple_simplify): Change prototypes of
18072         auto-generated functions to take a gimple_match_op * instead of
18073         separate code_helper * and tree * parameters.  Make the same
18074         change in the top-level overload and update calls to the
18075         gimple_resimplify routines.  Update calls to the auto-generated
18076         functions and to maybe_push_res_to_seq in the publicly-facing
18077         operation-specific gimple_simplify overloads.
18078         (gimple_match_op::MAX_NUM_OPS): Define.
18079         (gimple_resimplify1): Replace rcode and ops with a single res_op
18080         parameter.  Update call to gimple_simplify.
18081         (gimple_resimplify2): Likewise.
18082         (gimple_resimplify3): Likewise.
18083         (mprts_hook): Replace parameters with a gimple_match_op *.
18084         (maybe_build_generic_op): Likewise.
18085         (build_call_internal): Replace type, nargs and ops with
18086         a gimple_match_op *.
18087         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18088         with a single gimple_match_op *.  Update calls to mprts_hook,
18089         build_call_internal and gimple_simplified_result_is_gimple_val.
18090         Factor out code that is common to the tree_code and combined_fn cases.
18091         * genmatch.c (expr::gen_transform): Replace tem_code and
18092         tem_ops with a gimple_match_op called tem_op.  Update calls
18093         to the gimple_resimplify functions and maybe_push_res_to_seq.
18094         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18095         res_ops.  Update call to the gimple_resimplify functions.
18096         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18097         (decision_tree::gen): Make the functions take a gimple_match_op *
18098         called res_op instead of separate res_code and res_ops parameters.
18099         Update call accordingly.
18100         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18101         and ops with a single res_op parameter.  Update calls to
18102         maybe_build_generic_op and maybe_push_res_to_seq.
18103         (fold_stmt_1): Update calls to gimple_simplify and
18104         replace_stmt_with_simplification.
18105         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18106         and gimple_simplified_result_is_gimple_val.
18107         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18108         gimple_simplify.
18109         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18110         with a gimple_match_op *.
18111         (vn_nary_build_or_lookup): Likewise.  Update call to
18112         vn_nary_build_or_lookup_1.
18113         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18114         gimple_match_op *.  Update calls to the gimple_resimplify routines
18115         and to gimple_simplified_result_is_gimple_val.
18116         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18117         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18118         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18119         (visit_nary_op): Likewise.
18120         (visit_reference_op_load): Likewise.
18122 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18124         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18125         modifier for printing the step amount.
18127 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18129         PR target/78849
18130         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18131         types.
18133 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18135         * doc/sourcebuild.texi (Endianness): New subsubsection.
18137 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18139         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18140         <prefetch_dynamic_strides>: New const bool field.
18141         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18142         prefetch_dynamic_strides.
18143         (exynosm1_prefetch_tune): Likewise.
18144         (thunderxt88_prefetch_tune): Likewise.
18145         (thunderx_prefetch_tune): Likewise.
18146         (thunderx2t99_prefetch_tune): Likewise.
18147         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18148         false.
18149         (aarch64_override_options_internal): Update to set
18150         PARAM_PREFETCH_DYNAMIC_STRIDES.
18151         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18152         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18153         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18154         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18155         prefetch-dynamic-strides setting.
18157 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18159         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18160         <minimum_stride>: New const int field.
18161         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18162         minimum_stride field defaulting to -1.
18163         (exynosm1_prefetch_tune): Likewise.
18164         (thunderxt88_prefetch_tune): Likewise.
18165         (thunderx_prefetch_tune): Likewise.
18166         (thunderx2t99_prefetch_tune): Likewise.
18167         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18168         <default_opt_level>: Set to 3.
18169         (aarch64_override_options_internal): Update to set
18170         PARAM_PREFETCH_MINIMUM_STRIDE.
18171         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18172         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18173         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18174         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18175         stride is constant and is below the minimum stride threshold.
18177 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18179         * config/arm/arm-cpus.in (mode26): Delete.
18180         (armv4): Delete mode26 reference.
18181         * config/arm/arm.c (arm_configure_build_target): Delete use of
18182         isa_bit_mode26.
18184 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18186         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18187         New insn pattern.
18188         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18189         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18190         for non-SSE modes.
18191         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18192         (floatunsdidf2): Ditto.
18194 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18196         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18197         (fixuns_trunc<mode>si2_avx512f): Ditto.
18198         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18199         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18200         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18202 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18204         PR rtl-optimization/79985
18205         * df-scan.c (df_insn_refs_collect): Remove special case for
18206         global registers and asm statements.
18208 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18210         * extend.texi (Global Register Variables): Rewrite the bullet list.
18211         Note that the register is available for allocation. Note that access
18212         via inline asm must use constraints. Add note about async-signal
18213         handlers. Remove paragraph about automagic register selection.
18215 2018-05-23  Richard Biener  <rguenther@suse.de>
18217         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18218         of fixed offset from memset VN.
18220 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18222         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18223         first_interp field.
18224         (alloc_cand_and_find_basis): Initialize first_interp field.
18225         (slsr_process_mul): Modify first_interp field.
18226         (slsr_process_add): Likewise.
18227         (slsr_process_cast): Modify first_interp field for each new
18228         interpretation.
18229         (slsr_process_copy): Likewise.
18230         (dump_candidate): Dump first_interp field.
18231         (replace_mult_candidate): Process all interpretations, not just
18232         subsequent ones.
18233         (replace_rhs_if_not_dup): Likewise.
18234         (replace_one_candidate): Likewise.
18236 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18238         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18239         Add new boolean.
18240         (aarch64_needs_frame_chain): New function.
18241         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18243 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18245         PR target/84882
18246         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18247         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18248         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18249         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18250         as true for strict-align.
18251         (aarch64_can_inline_p): Perform checks even when callee has no
18252         attributes to check for strict alignment.
18253         * doc/extend.texi (AArch64 Function Attributes): Document
18254         no-strict-align.
18255         * doc/invoke.texi: (AArch64 Options): Likewise.
18257 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18259         PR tree-optimization/85853
18260         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18261         the handling of the root of the node to...
18262         (vect_slp_analyze_node_operations_1): ...this new function,
18263         and run the whole thing with the child nodes' def types
18264         set according to their SLP node's def type.
18266 2018-05-23  Richard Biener  <rguenther@suse.de>
18268         PR middle-end/85874
18269         * tree-data-ref.c (create_runtime_alias_checks): Defer
18270         and ignore overflow warnings.
18272 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18274         PR tree-optimization/85822
18275         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18276         constants.
18278 2018-05-23  Richard Biener  <rguenther@suse.de>
18280         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18281         memset constants via native_interpret_expr.
18283 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18285         PR target/85345
18286         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18287         attribute.
18288         (cgraph_node::create_alias): Likewise.
18289         (cgraph_node::get_availability): Check ifunc_resolver instead
18290         of looking up ifunc attribute.
18291         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18292         * varasm.c (do_assemble_alias): Likewise.
18293         (assemble_alias): Likewise.
18294         (default_binds_local_p_3): Likewise.
18295         * cgraph.h (cgraph_node): Add ifunc_resolver.
18296         (cgraph_node::only_called_directly_or_aliased_p): Return false
18297         for IFUNC resolver.
18298         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18299         attribute.
18300         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18301         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18302         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18303         instead of looking up ifunc attribute.
18305 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18307         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18309 2018-05-22  Martin Sebor  <msebor@redhat.com>
18311         PR middle-end/85359
18312         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18313         only when expasion succeeds.
18314         (expand_builtin_strcmp): Same.
18315         (expand_builtin_strncmp): Same.
18317 2018-05-22  Martin Sebor  <msebor@redhat.com>
18319         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18321 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18322             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18324         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18325         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18326         code.
18327         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18328         Add check for legitimate_address.
18329         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18330         (aarch64_swap_ldrstr_operands): New.
18331         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18332         Define prototype.
18334 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18335             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18337         * config/aarch64/aarch64.md: New patterns to generate stp
18338         and ldp.
18339         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18340         single words and double words.
18341         (load_pair_sw, load_pair_dw): Likewise.
18342         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18343         Delete.
18344         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18345         Delete.
18346         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18347         for different mode ldpstp and add peephole for merged zero stores.
18348         Likewise for loads.
18349         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18350         Add size check.
18351         (aarch64_gen_store_pair): Rename calls to match new patterns.
18352         (aarch64_gen_load_pair): Rename calls to match new patterns.
18353         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18354         (load_pair<DREG:mode><DREG2:mode>): ... This.
18355         (store_pair<mode>): Rename to...
18356         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18357         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18358         New mode iterators.
18359         (V_INT_EQUIV): Handle SImode.
18360         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18361         New predicate.
18363 2018-05-22  Martin Sebor  <msebor@redhat.com>
18365         PR c/85623
18366         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18367         or ajust the presumed bound on an operation to avoid unnecessary
18368         warnings.
18370 2018-05-22  Martin Sebor  <msebor@redhat.com>
18372         PR tree-optimization/85826
18373         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18374         assuming that a DECL necesarily has a constant size.
18376 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18378         PR middle-end/85862
18379         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18381 2018-05-22  Richard Biener  <rguenther@suse.de>
18383         PR tree-optimization/85834
18384         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18385         non-constant and non-zero memset arguments.
18387 2018-05-22  Martin Liska  <mliska@suse.cz>
18389         PR ipa/85607
18390         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18392 2018-05-22  Richard Biener  <rguenther@suse.de>
18394         PR tree-optimization/85863
18395         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18396         comparisons when vectype is specified.
18397         (vectorizable_condition): Do not specify vectype for
18398         vect_is_simple_cond when SLP vectorizing.
18400 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18402         PR target/85657
18403         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18404         define __ibm128 as long double.
18405         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18406         as a distinct type when IEEE 128-bit support is enabled.
18407         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18408         128-bit types to use the correct functions.
18409         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18410         convert between 128-bit floating point types that have different
18411         modes but the same representation, instead of using gen_lowpart to
18412         makean alias.
18413         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18414         KFmode.
18415         (IFKF_reg): New attributes to give the register constraints for
18416         IFmode and KFmode.
18417         (extend<mode>tf2_internal): New insns to mark an explicit
18418         conversion between 128-bit floating point types that have a
18419         different mode but share the same representation.
18421 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18423         PR tree-optimization/85814
18424         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18425         a null return from get_strinfo when unsharing the next
18426         strinfo in the chain.
18428 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18430         PR gcc/84923
18431         * varasm.c (weak_finish): Clean up weak_decls.
18433 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18435         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18436         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18437         UNSPEC_UADALP values.
18438         * config/aarch64/iterators.md (ABAL): New int iterator.
18439         (ABDL2): Likewise.
18440         (ADALP): Likewise.
18441         (sur): Add mappings for the above.
18442         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18443         New define_insn.
18444         (aarch64_<sur>abal<mode>_4): Likewise.
18445         (aarch64_<sur>adalp<mode>_3): Likewise.
18446         (<sur>sadv16qi): New define_expand.
18448 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18450         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18451         (*movdf_internal): Ditto.
18452         (*rcpsf2_sse): Ditto.
18453         (*rsqrtsf2_sse): Ditto.
18454         (*sqrt<mode>2_sse): Ditto.
18456 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18458         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18459         eor3q<mode>4.
18460         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18461         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18462         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18463         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18464         vbcaxq_s64): New.
18465         * config/aarch64/arm_neon.h: Likewise.
18466         * config/aarch64/iterators.md (VQ_I): New.
18468 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18470         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18471         arc*-*-linux*.
18472         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18473         MULTILIB_DIRNAMES
18475 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18477         * config/nds32/constraints.md (S): New constraint.
18478         * config/nds32/nds32.md (call_internal): Use constraint S.
18479         (call_value_internal): Likewise.
18480         (sibcall_internal): Likewise.
18481         (sibcall_value_internal): Likewise.
18483 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18484             Chung-Ju Wu  <jasonwucj@gmail.com>
18486         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18487         into consideration.
18489 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18490             Chung-Ju Wu  <jasonwucj@gmail.com>
18492         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18493         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18494         (nds32_rtx_costs_impl): Simplify.
18495         (nds32_address_cost_impl): Simplify.
18496         (nds32_init_rtx_costs): New function.
18497         (nds32_rtx_costs_speed_prefer): Likewise.
18498         (nds32_rtx_costs_size_prefer): Likewise.
18499         (nds32_address_cost_speed_prefer): Likewise.
18500         (nds32_address_cost_speed_fwprop): Likewise.
18501         (nds32_address_cost_size_prefer): Likewise.
18502         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
18503         * config/nds32/nds32.c (nds32_option_override): Use
18504         nds32_init_rtx_costs function.
18506 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18508         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
18509         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
18510         (TARGET_PIPELINE_N8): Likewise.
18511         (TARGET_PIPELINE_N10): Likewise.
18512         (TARGET_PIPELINE_N13): Likewise.
18513         (TARGET_PIPELINE_GRAYWOLF): Likewise.
18515 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
18517         * config/nds32/nds32-fpu.md: Update copyright year.
18519 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18521         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18523 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18525         * config/nds32/nds32.c
18526         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
18527         * config/nds32/nds32.opt (minline-asm-r15): New option.
18529 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18531         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
18532         MASK_HW_ABS.
18533         * config/nds32/nds32.md (abssi2): New pattern.
18535 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
18537         * config/i386/i386.md (rex64namesuffix): New mode attribute.
18538         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
18539         Merge insn pattern from sse_cvtsi2ss<round_name> and
18540         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
18541         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
18542         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
18543         using SWI48 mode iterator.
18544         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
18545         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
18546         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
18547         pattern from sse_cvttss2si<round_saeonly_name>
18548         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
18549         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
18550         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
18551         using SWI48 mode iterator.
18552         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
18553         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
18554         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
18555         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
18556         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
18557         using SWI48 mode iterator.
18558         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
18559         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
18560         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
18561         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
18562         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
18563         SWI48 mode iterator.
18564         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
18565         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
18566         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
18567         pattern from sse_cvttsd2si<round_saeonly_name>
18568         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
18570 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18572         * config/nds32/nds32-md-auxiliary.c
18573         (nds32_valid_smw_lwm_base_p): Refine.
18574         (nds32_output_smw_single_word): Refine.
18575         (nds32_output_smw_double_word): New.
18576         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
18578 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18580         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
18581         (nds32_output_stack_pop): Refine.
18582         (nds32_expand_unaligned_load): Refine.
18583         (nds32_expand_unaligned_store): Refine.
18585 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18586             Chung-Ju Wu  <jasonwucj@gmail.com>
18588         * config/nds32/constants.md: Add TP_REGNUM constant.
18589         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
18590         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
18591         UNSPEC_ADD32.
18592         * config/nds32/nds32-doubleword.md: Consider flag_pic.
18593         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
18594         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
18595         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
18596         and PIC code generation.
18597         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
18598         code generation.
18599         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
18600         optimization.
18601         * config/nds32/nds32.md: Support TLS and PIC.
18602         * config/nds32/nds32.c: Support TLS and PIC.
18603         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
18604         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
18605         predicate.
18607 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18609         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
18610         mode with E_ prefix.
18612 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18613             Chung-Ju Wu  <jasonwucj@gmail.com>
18615         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
18616         * config/nds32/nds32-md-auxiliary.c
18617         (symbolic_reference_mentioned_p): New.
18618         (nds32_legitimize_ict_address): New.
18619         (nds32_expand_ict_move): New.
18620         (nds32_indirect_call_referenced_p): New.
18621         (nds32_symbol_binds_local_p): Delete.
18622         (nds32_long_call_p): Modify.
18623         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18624         * config/nds32/nds32-protos.h
18625         (symbolic_reference_mentioned_p): Declare.
18626         (nds32_legitimize_ict_address): Declare.
18627         (nds32_expand_ict_move): Declare.
18628         (nds32_indirect_call_referenced_p): Declare.
18629         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
18630         (nds32_relax_group): Use nds32_ict_const_p as condition.
18631         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
18632         (nds32_asm_file_start): Output ict_model directive in asm code.
18633         (nds32_legitimate_address_p): Consider indirect call.
18634         (nds32_print_operand): Consider indirect call.
18635         (nds32_print_operand_address): Consider indirect call.
18636         (nds32_insert_attributes): Handle "indirect_call" attribute.
18637         (TARGET_LEGITIMATE_ADDRESS_P): Define.
18638         (TARGET_LEGITIMATE_CONSTANT_P): Define.
18639         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
18640         (TARGET_DELEGITIMIZE_ADDRESS): Define.
18641         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
18642         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
18643         (TARGET_ICT_MODEL_SMALL): Define.
18644         (TARGET_ICT_MODEL_LARGE): Define.
18645         * config/nds32/nds32.md (movsi): Consider ict model.
18646         (call, call_value): Consider ict model.
18647         (sibcall, sibcall_value): Consider ict model.
18648         * config/nds32/nds32.opt (mict-model): New option.
18649         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
18650         model.
18652 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
18653             Monk Chiang  <sh.chiang04@gmail.com>
18654             Jim Wilson <jimw@sifive.com>
18656         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
18657         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
18658         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
18659         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
18660         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
18661         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
18662         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
18663         compute save_libcall_adjustment properly.
18664         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
18665         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
18666         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
18667         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
18668         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
18669         (ABI_SPEC): Handle mabi=ilp32e.
18670         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
18671         (RVE): Add RVE mask.
18672         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
18673         <-march>: Add rv32e as an example.
18675 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
18677         PR c++/82899
18678         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
18679         (intra_create_variable_infos): Handle C++ constructors.
18681 2018-05-18  Martin Liska  <mliska@suse.cz>
18683         * passes.def: Remove a redundant pass.
18685 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
18687         PR bootstrap/85838
18688         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
18690 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18692         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
18693         (ARMv4): Update.
18694         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
18695         (ARMv6m): Update.
18696         (armv2, armv2a, armv3, armv3m): Delete architectures.
18697         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
18698         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
18699         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
18700         Delete cpus.
18701         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
18702         (*mulsidi3adddi): Likewise.
18703         (mulsidi3): Likewise.
18704         (*mulsidi3_nov6): Likewise.
18705         (umulsidi3): Likewise.
18706         (umulsidi3_nov6): Likewise.
18707         (umaddsidi4): Likewise.
18708         (*umulsidi3adddi): Likewise.
18709         (smulsi3_highpart): Likewise.
18710         (*smulsi3_highpart_nov6): Likewise.
18711         (umulsi3_highpart): Likewise.
18712         (*umulsi3_highpart_nov6): Likewise.
18713         * config/arm/arm.h (arm_arch3m): Delete.
18714         * config/arm/arm.c (arm_arch3m): Delete.
18715         (arm_option_override_internal): Update armv3-related comment.
18716         (arm_configure_build_target): Delete use of isa_bit_mode32.
18717         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
18718         (arm_rtx_costs_internal): Delete check of arm_arch3m.
18719         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
18720         (mulsa3): Likewise.
18721         (mulusa3): Likewise.
18722         * config/arm/arm-protos.h (arm_arch3m): Delete.
18723         * config/arm/arm-tables.opt: Regenerate.
18724         * config/arm/arm-tune.md: Likewise.
18725         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
18726         deleted architectures.
18728 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18730         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
18731         (armv5t, armv5te): New features.
18732         (ARMv5, ARMv5e): Delete fgroups.
18733         (ARMv5t, ARMv5te): Adjust for above changes.
18734         (ARMv6m): Likewise.
18735         (armv5, armv5e): Delete arches.
18736         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
18737         arm_arch5.
18738         (*call_reg_arm): Likewise.
18739         (*call_value_reg_armv5): Likewise.
18740         (*call_value_reg_arm): Likewise.
18741         (*call_symbol): Likewise.
18742         (*call_value_symbol): Likewise.
18743         (*sibcall_insn): Likewise.
18744         (*sibcall_value_insn): Likewise.
18745         (clzsi2): Likewise.
18746         (prefetch): Likewise.
18747         (define_split and define_peephole2 dependent on arm_arch5):
18748         Likewise.
18749         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
18750         arm_arch5e.
18751         (TARGET_ARM_QBIT): Likewise.
18752         (TARGET_DSP_MULTIPLY): Likewise.
18753         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
18754         (arm_arch5, arm_arch5e): Delete.
18755         (arm_arch5t, arm_arch5te): Declare.
18756         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
18757         (arm_arch5t): Declare.
18758         (arm_option_reconfigure_globals): Update for the above.
18759         (arm_options_perform_arch_sanity_checks): Update comment, replace
18760         use of arm_arch5 with arm_arch5t.
18761         (use_return_insn): Likewise.
18762         (arm_emit_call_insn): Likewise.
18763         (output_return_instruction): Likewise.
18764         (arm_final_prescan_insn): Likewise.
18765         (arm_coproc_builtin_available): Likewise.
18766         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
18767         arm_arch5e with arm_arch5t and arm_arch5te.
18768         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
18769         (arm_arch5t, arm_arch5te): Declare.
18770         * config/arm/arm-tables.opt: Regenerate.
18771         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
18772         * config/arm/t-multilib: Likewise.
18773         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
18774         instead of arm_arch5.
18775         (*call_reg_thumb1): Likewise.
18776         (*call_value_reg_thumb1_v5): Likewise.
18777         (*call_value_reg_thumb1): Likewise.
18778         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
18779         unreachable path.
18780         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
18782 2018-05-18  Martin Liska  <mliska@suse.cz>
18784         PR gcov-profile/84846
18785         * doc/gcov.texi: Document -t option of gcov tool.
18787 2018-05-18  Martin Liska  <mliska@suse.cz>
18789         PR gcov-profile/84846
18790         * gcov.c (print_usage): Add new -t option.
18791         (process_args): Handle the option.
18792         (generate_results): Use stdout as output when requested by
18793         the option.
18795 2018-05-18  Martin Liska  <mliska@suse.cz>
18797         PR gcov-profile/84846
18798         * coverage.c (coverage_init): Write PWD to .gcno file.
18799         * doc/gcov.texi: Document how working directory is printed.
18800         * gcov-dump.c (dump_gcov_file): Print PWD.
18801         * gcov.c (output_intermediate_file): Likewise.
18802         (read_graph_file): Read PWD string.
18803         (output_lines): Print PWD.
18805 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18807         PR middle-end/85817
18808         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
18809         for retval and return false if all args to phi are zero.
18811 2018-05-18  Richard Biener  <rguenther@suse.de>
18813         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
18814         method.
18815         (evrp_dom_walker::before_dom_children): Call it.
18817 2018-05-18  Richard Biener  <rguenther@suse.de>
18819         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
18820         results when processing array refs with variable index.
18822 2018-05-18  Toon Moene  <toon@moene.org>
18824         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
18825         directly after that of -floop-interchange. Indicate that both
18826         options are enabled by default when specifying -O3.
18828 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18830         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
18831         iterator.  Delete separate integer-mode vec_set<mode> expander.
18832         (aarch64_simd_vec_setv2di): Delete.
18833         (vec_setv2di): Delete.
18834         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
18835         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
18836         the "w, r" alternative.
18838 2018-05-18  Martin Liska  <mliska@suse.cz>
18840         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
18841         * tree-pass.h (make_pass_lower_switch_O0): New function.
18842         * tree-switch-conversion.c (node_has_low_bound): Remove.
18843         (node_has_high_bound): Likewise.
18844         (node_is_bounded): Likewise.
18845         (class pass_lower_switch): Make it a template type and create
18846         two instances.
18847         (pass_lower_switch::execute): Add template argument.
18848         (make_pass_lower_switch): New function.
18849         (make_pass_lower_switch_O0): New function.
18850         (do_jump_if_equal): Remove.
18851         (emit_case_nodes): Simplify to just handle all 3 cases and leave
18852         all the hard work to tree optimization passes.
18854 2018-05-18  Martin Liska  <mliska@suse.cz>
18856         * dbgcnt.c (limit_low): Renamed from limit.
18857         (limit_high): New variable.
18858         (dbg_cnt_is_enabled): Check for upper limit.
18859         (dbg_cnt): Adjust dumping.
18860         (dbg_cnt_set_limit_by_index): Add new argument for high
18861         value.
18862         (dbg_cnt_set_limit_by_name): Likewise.
18863         (dbg_cnt_process_single_pair): Parse new format.
18864         (dbg_cnt_process_opt): Use strtok.
18865         (dbg_cnt_list_all_counters): Remove 'value' and add
18866         'limit_high'.
18867         * doc/invoke.texi: Document changes.
18869 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
18871         * doc/sourcebuild.texi (scalar_all_fma): Document.
18872         * tree.def (FMA_EXPR): Delete.
18873         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
18874         * internal-fn.c (ternary_direct): New macro.
18875         (expand_ternary_optab_fn): Likewise.
18876         (direct_ternary_optab_supported_p): Likewise.
18877         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
18878         * builtins.c (fold_builtin_fma): Delete.
18879         (fold_builtin_3): Don't call it.
18880         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
18881         * expr.c (expand_expr_real_2): Likewise.
18882         * fold-const.c (operand_equal_p): Likewise.
18883         (fold_ternary_loc): Likewise.
18884         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
18885         * gimple.c (DEFTREECODE): Likewise.
18886         * gimplify.c (gimplify_expr): Likewise.
18887         * optabs-tree.c (optab_for_tree_code): Likewise.
18888         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18889         * tree-eh.c (operation_could_trap_p): Likewise.
18890         (stmt_could_throw_1_p): Likewise.
18891         * tree-inline.c (estimate_operator_cost): Likewise.
18892         * tree-pretty-print.c (dump_generic_node): Likewise.
18893         (op_code_prio): Likewise.
18894         * tree-ssa-loop-im.c (stmt_cost): Likewise.
18895         * tree-ssa-operands.c (get_expr_operands): Likewise.
18896         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
18897         * fold-const-call.h (fold_fma): Delete.
18898         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
18899         CFN_FNMA and CFN_FNMS.
18900         (fold_fma): Delete.
18901         * genmatch.c (combined_fn): New enum.
18902         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
18903         (commutative_op): New function.
18904         (commutate): Use it.  Handle more than 2 operands.
18905         (dt_operand::gen_gimple_expr): Use commutative_op.
18906         (parser::parse_expr): Allow :c to be used with non-binary
18907         operators if the commutative operand is known.
18908         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
18909         CFN_FMS, CFN_FNMA and CFN_FNMS.
18910         (backprop::process_assign_use): Remove FMA_EXPR handling.
18911         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
18912         (gen_hsa_fma): New function.
18913         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
18914         IFN_FNMA and IFN_FNMS.
18915         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
18916         * gimple-fold.h (follow_all_ssa_edges): Declare.
18917         * gimple-fold.c (follow_all_ssa_edges): New function.
18918         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
18919         gimple_build interface and use follow_all_ssa_edges to fold the result.
18920         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
18921         instead of checking for optabs directly.
18922         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
18923         rather than FMA_EXPRs.
18924         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
18925         call to IFN_FMA instead of an FMA_EXPR.
18927 2018-05-17  Jim Wilson  <jimw@sifive.com>
18929         * expr.c (do_tablejump): When converting index to Pmode, if we have a
18930         sign extended promoted subreg, and the range does not have the sign bit
18931         set, then do a sign extend.
18933         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
18934         test, check for sign extended subreg and/or constant operands, and
18935         do a sign extend in that case.
18937 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
18939         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
18940         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
18941         Add untyped.
18942         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
18943         Change logics_shift_reg to logics_shift_imm.
18944         (thunderx2t99_fp_loadpair_basic): Delete.
18945         (thunderx2t99_fp_storepair_basic): Delete.
18946         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
18947         (thunderx2t99_asimd_polynomial): Delete.
18948         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
18949         and neon_fp_mul_d_scalar_q.
18950         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
18951         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
18952         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
18953         (thunderx2t99_asimd_lut): Add missing tbl types.
18954         (thunderx2t99_asimd_ext): Delete.
18955         (thunderx2t99_asimd_load1_1_mult): Delete.
18956         (thunderx2t99_asimd_load1_2_mult): Delete.
18957         (thunderx2t99_asimd_load1_ldp): New.
18958         (thunderx2t99_asimd_load1): New.
18959         (thunderx2t99_asimd_load2): Add missing *load2* types.
18960         (thunderx2t99_asimd_load3): New.
18961         (thunderx2t99_asimd_load4): New.
18962         (thunderx2t99_asimd_store1_1_mult): Delete.
18963         (thunderx2t99_asimd_store1_2_mult): Delete.
18964         (thunderx2t99_asimd_store2_mult): Delete.
18965         (thunderx2t99_asimd_store2_onelane): Delete.
18966         (thunderx2t99_asimd_store_stp): New.
18967         (thunderx2t99_asimd_store1): New.
18968         (thunderx2t99_asimd_store2): New.
18969         (thunderx2t99_asimd_store3): New.
18970         (thunderx2t99_asimd_store4): New.
18972 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
18974         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
18975         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
18977 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
18978             Segher Boessenkool  <segher@kernel.crashing.org>
18980         PR target/85698
18981         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
18982         operand.
18984 2018-05-17  Richard Biener  <rguenther@suse.de>
18986         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
18987         for pruning loop and prune defs feeding only already visited PHIs.
18989 2018-05-17  Richard Biener  <rguenther@suse.de>
18991         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
18993 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
18994             Richard Biener  <rguenther@suse.de>
18996         PR tree-optimization/85793
18997         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
18998         for VMAT_ELEMENTWISE.
19000 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19002         * internal-fn.h (lookup_internal_fn): Declare
19003         * internal-fn.c (lookup_internal_fn): New function.
19004         * gimple.c (gimple_build_call_from_tree): Handle calls to
19005         internal functions.
19006         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19007         internal function names.
19008         * tree-pretty-print.c (dump_generic_node): Likewise.
19009         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19011 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19013         * gimple-fold.h (gimple_build): Make the function forms take
19014         combined_fn rather than built_in_function.
19015         (gimple_simplify): Likewise.
19016         * gimple-match-head.c (gimple_simplify): Likewise.
19017         * gimple-fold.c (gimple_build): Likewise.
19018         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19019         rather than gimple_build_call_internal.
19020         (get_initial_defs_for_reduction): Likewise.
19021         (vect_create_epilog_for_reduction): Likewise.
19022         (vectorizable_live_operation): Likewise.
19024 2018-05-17  Martin Liska  <mliska@suse.cz>
19026         * gimple-ssa-sprintf.c (format_directive): Do not use
19027         space in between 'G_' and '('.
19029 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19031         PR target/85323
19032         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19033         even if the mask is not all ones.
19035         PR target/85323
19036         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19037         vector.
19038         (ix86_gimple_fold_builtin): Likewise.
19040         PR target/85323
19041         * config/i386/i386.c: Include tree-vector-builder.h.
19042         (ix86_vector_shift_count): New function.
19043         (ix86_fold_builtin): Fold shift builtins by scalar count.
19044         (ix86_gimple_fold_builtin): Likewise.
19046         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19047         _mm512_setzero): New intrinsics.
19049 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19050             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19052         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19053         code generation for cases where splatting a value is not useful.
19054         * simplify-rtx.c (simplify_ternary_operation): Simplify
19055         vec_merge across a vec_duplicate and a paradoxical subreg forming
19056         a vector mode to a vec_concat.
19058 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19060         * config.gcc: Support "goldmont-plus".
19061         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19062         "goldmont-plus".
19063         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19064         PROCESSOR_GOLDMONT_PLUS.
19065         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19066         (processor_target_table): Add "goldmont-plus".
19067         (PTA_GOLDMONT_PLUS): Define.
19068         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19069         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19070         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19071         (fold_builtin_cpu): Add "goldmont-plus".
19072         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19073         (ix86_option_override_internal): Add "goldmont-plus".
19074         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19075         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19076         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19077         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19079 2018-05-17  Richard Biener  <rguenther@suse.de>
19081         PR tree-optimization/85757
19082         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19083         remove defs that only feed that PHI from further processing.
19085 2018-05-16  Jim Wilson  <jimw@sifive.com>
19087         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19088         asterisk to name.
19089         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19091 2018-05-16  Mark Wielaard  <mark@klomp.org>
19093         * dwarf2out.c (count_index_strings): New function.
19094         (output_indirect_strings): Call count_index_strings and generate
19095         header for dwarf_version >= 5.
19097 2018-05-16  Mark Wielaard  <mark@klomp.org>
19099         * dwarf2out.c (dwarf_FORM): New function.
19100         (set_indirect_string): Use dwarf_FORM.
19101         (reset_indirect_string): Likewise.
19102         (size_of_die): Likewise.
19103         (value_format): Likewise.
19104         (output_die): Likewise.
19105         (add_skeleton_AT_string): Likewise.
19106         (output_macinfo_op): Likewise.
19107         (index_string): Likewise.
19108         (output_index_string_offset): Likewise.
19109         (output_index_string): Likewise.
19110         (count_index_strings): Likewise.
19112 2018-05-16  Carl Love  <cel@us.ibm.com>
19114         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19115         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19117 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19119         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19120         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19122 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19124         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19125         (fnma<mode>4): Likewise.
19126         (fms<mode>4): Likewise.
19127         (fnms<mode>4): Likewise.
19128         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19129         (aarch64_fnma<mode>4): Likewise.
19130         (aarch64_fms<mode>4): Likewise.
19131         (aarch64_fnms<mode>4): Likewise.
19132         (aarch64_fnmadd<mode>4): Likewise.
19134 2018-05-16  Jason Merrill  <jason@redhat.com>
19136         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19138 2018-05-16  Richard Biener  <rguenther@suse.de>
19140         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19141         (dump_stmt_cost): Declare.
19142         (add_stmt_cost): Dump cost we add.
19143         (add_stmt_costs): New function.
19144         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19145         No longer exported.
19146         (vect_analyze_stmt): Adjust prototype.
19147         (vectorizable_condition): Likewise.
19148         (vectorizable_live_operation): Likewise.
19149         (vectorizable_reduction): Likewise.
19150         (vectorizable_induction): Likewise.
19151         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19152         cost vector to pass to vectorizable_ and record afterwards.
19153         (vect_model_reduction_cost): Take cost vector argument and adjust.
19154         (vect_model_induction_cost): Likewise.
19155         (vectorizable_reduction): Likewise.
19156         (vectorizable_induction): Likewise.
19157         (vectorizable_live_operation): Likewise.
19158         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19159         SLP_TREE_NUMBER_OF_VEC_STMTS.
19160         (vect_analyze_slp_cost_1): Remove.
19161         (vect_analyze_slp_cost): Likewise.
19162         (vect_slp_analyze_node_operations): Take visited args and
19163         a target cost vector.  Avoid processing already visited stmt sets.
19164         (vect_slp_analyze_operations): Use a local cost vector to gather
19165         costs and register those of non-discarded instances.
19166         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19167         (vect_schedule_slp_instance): Remove copying of
19168         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19169         zero.
19170         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19171         adding cost.  Record cost entry location.
19172         (vect_prologue_cost_for_slp_op): Function to compute cost of
19173         a constant or invariant generated for SLP vect in the prologue,
19174         split out from vect_analyze_slp_cost_1.
19175         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19176         (vect_model_promotion_demotion_cost): Likewise.
19177         (vect_model_store_cost): Likewise, make static.
19178         (vect_model_load_cost): Likewise.
19179         (vectorizable_bswap): Add cost vector arg and adjust.
19180         (vectorizable_call): Likewise.
19181         (vectorizable_simd_clone_call): Likewise.
19182         (vectorizable_conversion): Likewise.
19183         (vectorizable_assignment): Likewise.
19184         (vectorizable_shift): Likewise.
19185         (vectorizable_operation): Likewise.
19186         (vectorizable_store): Likewise.
19187         (vectorizable_load): Likewise.
19188         (vectorizable_condition): Likewise.
19189         (vectorizable_comparison): Likewise.
19190         (can_vectorize_live_stmts): Likewise.
19191         (vect_analyze_stmt): Likewise.
19192         (vect_transform_stmt): Adjust calls to vectorizable_*.
19193         * tree-vectorizer.c: Include gimple-pretty-print.h.
19194         (dump_stmt_cost): New function.
19196 2018-05-16  Richard Biener  <rguenther@suse.de>
19198         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19199         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19200         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19201         (check_name): New callback.
19202         (dse_classify_store): Track cycles via a visited bitmap of PHI
19203         defs and simplify handling of in-loop and across loop dead stores
19204         and properly fail for loop-variant refs.  Handle byte-tracking with
19205         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19206         limiting the walk.
19208 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19210         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19211         (vect_get_mask_type_for_stmt): Likewise.
19212         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19213         split out from...
19214         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19215         to determine the statement's vector type and the vector type that
19216         should be used for calculating nunits.  Deal with cases in which
19217         the type has to be deferred.
19218         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19219         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19220         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19221         (vect_determine_vf_for_stmt): New functions, split out from...
19222         (vect_determine_vectorization_factor): ...here.
19223         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19224         (vect_get_mask_type_for_stmt): New functions, split out from
19225         vect_determine_vectorization_factor.
19227 2018-05-16  Richard Biener  <rguenther@suse.de>
19229         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19230         verify the [VEC_]COND_EXPR embedded comparison.
19232 2018-05-15  Martin Sebor  <msebor@redhat.com>
19234         PR tree-optimization/85753
19235         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19236         RECORD_TYPE in addition to ARRAY_TYPE.
19238 2018-05-15  Martin Sebor  <msebor@redhat.com>
19240         PR middle-end/85643
19241         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19243 2018-05-15  Richard Biener  <rguenther@suse.de>
19245         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19246         add by_clobber_p one.  Change algorithm to collect all defs
19247         representing uses we need to walk and try reducing them to
19248         a single one before failing.
19249         (dse_dom_walker::dse_optimize_stmt): Adjust.
19251 2018-05-13  Mark Wielaard  <mark@klomp.org>
19253         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19254         (size_of_loc_descr): Likewise.
19255         (output_loc_operands): Likewise.
19256         (output_loc_operands_raw): Likewise.
19257         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19258         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19259         (hash_loc_operands): Likewise.
19260         (compare_loc_operands): Likewise.
19262 2018-05-14  Mark Wielaard  <mark@klomp.org>
19264         * dwarf2out.c (count_index_addrs): New function.
19265         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19267 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19269         PR tree-optimization/83648
19270         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19271         return value as malloc candidate.
19273 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19275         PR ipa/85734
19276         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19277         param as true in call to suggest_attribute.
19279 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19281         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19282         -mreadonly-in-sdata.
19284 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19286         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19287         New pattern.
19288         (aarch64_crypto_aesd_fused): Likewise.
19290 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19292         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19293         (movsi_aarch64): Likewise.
19294         (load_pairsi): Likewise.
19295         (load_pairdi): Likewise.
19296         (store_pairsi): Likewise.
19297         (store_pairdi): Likewise.
19298         (load_pairsf): Likewise.
19299         (load_pairdf): Likewise.
19300         (store_pairsf): Likewise.
19301         (store_pairdf): Likewise.
19302         (zero_extend): Likewise.
19303         (trunc): Swap alternatives.
19304         (fcvt_target): Add '?' to prefer w over r.
19306 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19308         PR target/85756
19309         * config/i386/i386.md: Disallow non-commutative arithmetics in
19310         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19311         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19312         in the peephole2 before it.
19314 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19316         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19317         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19318         (ix86_handle_option): Handle -mcldemote.
19319         * config.gcc: New header.
19320         * config/i386/cldemoteintrin.h: New file.
19321         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19322         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19323         -mcldemote.
19324         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19325         OPTION_MASK_ISA_CLDEMOTE.
19326         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19327         (ix86_valid_target_attribute_inner_p): Ditto.
19328         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19329         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19330         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19331         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19332         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19333         (cldemote): New.
19334         * config/i386/i386.opt: Add -mcldemote.
19335         * config/i386/x86intrin.h: New header.
19336         * doc/invoke.texi: Add -mcldemote.
19338 2018-05-14  Richard Biener  <rguenther@suse.de>
19340         * doc/match-and-simplify.texi: Adjust :s documentation.
19342 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19344         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19345         intended memcpy size.
19346         (REORDER_45): Likewise.
19348 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19350         * sort.cc: New file.
19351         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19352         * vec.c (qsort_chk): Use gcc_qsort.
19353         * Makefile.in (OBJS-libcommon): Add sort.o.
19354         (build/sort.o): New target.  Use it...
19355         (BUILD_RTL): ... here, and...
19356         (build/gencfn-macros): ... here, and...
19357         (build/genmatch): ... here.
19359 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19360             Chung-Ju Wu  <jasonwucj@gmail.com>
19362         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19363         * config/nds32/nds32-graywolf.md: New file.
19364         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19365         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19366         pipeline.
19367         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19368         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19369         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19370         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19371         * config/nds32/pipelines.md: Include n15 settings.
19373 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19374             Chung-Ju Wu  <jasonwucj@gmail.com>
19376         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19377         * config/nds32/nds32-n13.md: New file.
19378         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19379         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19380         pipeline.
19381         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19382         * config/nds32/nds32.md (pipeline_model): Add n13.
19383         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19384         * config/nds32/pipelines.md: Include n13 settings.
19386 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19387             Chung-Ju Wu  <jasonwucj@gmail.com>
19389         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19390         * config/nds32/nds32-n10.md: New file.
19391         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19392         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19393         pipeline.
19394         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19395         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19396         * config/nds32/nds32.md (pipeline_model): Add n10.
19397         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19398         * config/nds32/pipelines.md: Include n10 settings.
19400 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19401             Kito Cheng  <kito.cheng@gmail.com>
19402             Chung-Ju Wu  <jasonwucj@gmail.com>
19404         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19405         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19406         Add enum values for DSP extension instructions.
19407         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19408         New constraints.
19409         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19410         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19411         New code iterators.
19412         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19413         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19414         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19415         * config/nds32/nds32-intrinsic.md: Likewise.
19416         * config/nds32/nds32_intrinsic.h: Likewise.
19417         * config/nds32/nds32-md-auxiliary.c: Likewise.
19418         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19419         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19420         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19421         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19422         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19423         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19424         TYPE_DMAC in switch statement.
19425         * config/nds32/nds32.c: New checking and implementation for DSP
19426         extension instructions.
19427         * config/nds32/nds32.h: Likewise.
19428         * config/nds32/nds32.md: Likewise.
19429         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19430         * config/nds32/predicates.md: Implement new predicates for DSP
19431         extension.
19433 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19435         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19436         Reformat alternatives and attributes so it is easier to identify
19437         which constraints/attributes go with which instruction.
19438         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19439         (mov<mode>_softfloat32, FMOVE64): Likewise.
19440         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19441         (mov<mode>_softfloat64, FMOVE64): Likewise.
19443 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19445         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19446         subsection.
19447         (Basic PowerPC Built-in Functions): The new name of the
19448         subsection previously known as "PowerPC Built-in Functions".
19449         (Basic PowerPC Built-in Functions Available on all Configurations):
19450         New subsubsection.
19451         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19452         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19453         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19454         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19456 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19458         PR ipa/85655
19459         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19460         single const.
19462 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19464         PR target/85733
19465         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19467 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19469         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19470         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19471         (ix86_handle_option): Handle -mwaitpkg.
19472         * config.gcc: New header.
19473         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19474         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19475         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19476         function type.
19477         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19478         OPTION_MASK_ISA_WAITPKG.
19479         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19480         (ix86_option_override_internal): Add PTA_WAITPKG.
19481         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19482         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19483         IX86_BUILTIN_TPAUSE.
19484         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19485         __builtin_ia32_umwait and __builtin_ia32_tpause.
19486         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19487         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19488         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19489         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19490         UNSPECV_TPAUSE): New.
19491         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19492         * config/i386/i386.opt: Add -mwaitpkg.
19493         * config/i386/waitpkgintrin.h: New file.
19494         * config/i386/x86intrin.h: New header.
19495         * doc/invoke.texi: Add -mwaitpkg.
19497 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19499         PR target/85606
19500         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
19501         equivalent.
19502         (cortex-m0): Use armv6s-m isa.
19503         (cortex-m0plus): Likewise.
19504         (cortex-m1): Likewise.
19505         (cortex-m0.small-multiply): Likewise.
19506         (cortex-m0plus.small-multiply): Likewise.
19507         (cortex-m1.small-multiply): Likewise.
19509 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
19510             Jakub Jelinek  <jakub@redhat.com>
19512         PR tree-optimization/85692
19513         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
19514         source permute as well.
19516 2018-05-11  Martin Liska  <mliska@suse.cz>
19518         PR sanitizer/85556
19519         * doc/extend.texi: Document LLVM style format for no_sanitize
19520         attribute.
19522 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
19524         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
19525         mode_supports_vsx_dform_quad to mode_supports_dq_form.
19526         (mode_supports_vsx_dform_quad): Likewise.
19527         (mode_supports_vmx_dform): Move these functions to be next to the
19528         other mode_supports functions.
19529         (mode_supports_dq_form): Likewise.
19530         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
19531         mode_supports_dq_form.
19532         (reg_offset_addressing_ok_p): Likewise.
19533         (offsettable_ok_by_alignment): Likewise.
19534         (rs6000_legitimate_offset_address_p): Likewise.
19535         (legitimate_lo_sum_address_p): Likewise.
19536         (rs6000_legitimize_address): Likewise.
19537         (rs6000_legitimize_reload_address): Likewise.
19538         (rs6000_secondary_reload_inner): Likewise.
19539         (rs6000_preferred_reload_class): Likewise.
19540         (rs6000_output_move_128bit): Likewise.
19542 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19544         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
19545         Generate SImode target register for null target.
19546         <case IX86_BUILTIN_XGETBV>: Ditto.
19547         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
19548         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
19550 2018-05-10  Carl Love  <cel@us.ibm.com>
19552         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19553         dcbtt and dcbtstt if operands[2] is 0.
19555 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19557         PR target/85693
19558         * config/i386/sse.md (usadv64qi): New expander.
19560 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
19562         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
19563         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
19564         -maltivec=be support.
19565         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
19566         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
19567         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
19568         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
19569         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
19570         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
19571         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
19572         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
19573         altivec_vsumsws): Adjust.
19574         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
19575         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
19576         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
19577         support.
19578         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
19579         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
19580         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
19581         (altivec_lve<VI_char>x): Delete expand.
19582         (*altivec_lve<VI_char>x_internal): Rename to...
19583         (altivec_lve<VI_char>x): ... this.
19584         (altivec_lvxl_<mode>): Delete expand.
19585         (*altivec_lvxl_<mode>_internal): Rename to ...
19586         (altivec_lvxl_<mode>): ... this.
19587         (altivec_stvxl_<mode>): Delete expand.
19588         (*altivec_stvxl_<mode>_internal): Rename to ...
19589         (altivec_stvxl_<mode>): ... this.
19590         (altivec_stve<VI_char>x): Delete expand.
19591         (*altivec_stve<VI_char>x_internal): Rename to ...
19592         (altivec_stve<VI_char>x): ... this.
19593         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
19594         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
19595         reduc_plus_scal_<mode>): Adjust.
19596         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
19597         comment.
19598         (rs6000_cpu_cpp_builtins): Adjust.
19599         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
19600         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
19601         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
19602         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
19603         -maltivec=be support.
19604         (rs6000_split_vec_extract_var): Adjust.
19605         (rs6000_split_v4si_init): Adjust.
19606         (swap_selector_for_mode): Delete.
19607         (altivec_expand_lvx_be, altivec_expand_stvx_be,
19608         altivec_expand_stvex_be): Delete.
19609         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
19610         -maltivec=be support.
19611         (rs6000_gimple_fold_builtin): Ditto.
19612         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
19613         Adjust.
19614         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
19615         (TARGET_DIRECT_MOVE_64BIT): Adjust.
19616         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
19617         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
19618         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
19619         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
19620         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
19621         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
19622         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
19623         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
19624         anonymous split): Adjust.
19625         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
19626         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
19628 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
19630         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
19631         when --with-gxx-include-dir is also specified.
19632         * configure: Regenerate.
19634 2018-05-09  Jim Wilson  <jimw@sifive.com>
19636         PR target/84797
19637         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
19638         * config/riscv/t-withmultilib: New.
19639         * config/riscv/withmultilib.h: New.
19640         * doc/install.texi: Document RISC-V --with-multilib-list support.
19642 2018-05-09  Richard Biener  <rguenther@suse.de>
19644         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19645         vector.
19646         (vect_bb_vectorization_profitable_p): Adjust.  Compute
19647         actual scalar cost using the cost vector and the add_stmt_cost
19648         machinery.
19650 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19652         PR rtl-optimization/85645
19653         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
19654         in the REG_CFA_REGISTER note for LR, don't leave it empty.
19656 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19658         PR rtl-optimization/85645
19659         * shrink-wrap.c (spread_components): Return a boolean saying if
19660         anything was changed.
19661         (try_shrink_wrapping_separate): Iterate spread_components until
19662         nothing changes anymore.
19664 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19666         PR rtl-optimization/85645
19667         * regrename.c (build_def_use): Also kill the chains that include the
19668         destination of a REG_CFA_REGISTER note.
19670 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19672         PR rtl-optimization/85645
19673         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
19674         insn that has a REG_CFA_REGISTER note.
19676 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
19678         * cfgexpand.c (expand_clobber): New function.
19679         (expand_gimple_stmt_1): Use it.
19680         * tree-vect-stmts.c (vect_clobber_variable): New function,
19681         split out from...
19682         (vectorizable_simd_clone_call): ...here.
19683         (vectorizable_store): Emit a clobber either side of an
19684         IFN_STORE_LANES sequence.
19685         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
19687 2018-05-09  Tom de Vries  <tom@codesourcery.com>
19689         PR target/85626
19690         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
19691         (define_insn "trap_if_false"): Add exit after trap.
19693 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
19695         PR rtl-optimization/85638
19696         * bb-reorder.c: Include common/common-target.h.
19697         (create_forwarder_block): New function extracted from...
19698         (fix_up_crossing_landing_pad): ...here.  Rename into...
19699         (dw2_fix_up_crossing_landing_pad): ...this.
19700         (sjlj_fix_up_crossing_landing_pad): New function.
19701         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
19702         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
19703         from both partitions and exit the loop after one iteration.
19705 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19707         Revert:
19708         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19709         subsection.
19710         (Basic PowerPC Built-in Functions): The new name of the
19711         subsection previously known as "PowerPC Built-in Functions".
19712         (Basic PowerPC Built-in Functions Available on all Configurations):
19713         New subsubsection.
19714         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19715         subsubsection.
19716         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19717         subsubsection.
19718         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19719         subsubsection.
19720         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19721         subsubsection.
19723 2018-05-08  Jim Wilson  <jimw@sifive.com>
19725         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
19726         (LD_EMUL_SUFFIX): New.
19727         (LINK_SPEC): Use it.
19729 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19731         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19732         subsection.
19733         (Basic PowerPC Built-in Functions): The new name of the
19734         subsection previously known as "PowerPC Built-in Functions".
19735         (Basic PowerPC Built-in Functions Available on all Configurations):
19736         New subsubsection.
19737         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19738         subsubsection.
19739         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19740         subsubsection.
19741         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19742         subsubsection.
19743         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19744         subsubsection.
19746 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19748         PR target/85683
19749         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
19750         after cmpelim optimization.
19752 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
19754         * config.gcc: Support "goldmont".
19755         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
19756         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19757         PROCESSOR_GOLDMONT.
19758         * config/i386/i386.c (m_GOLDMONT): Define.
19759         (processor_target_table): Add "goldmont".
19760         (PTA_GOLDMONT): Define.
19761         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
19762         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
19763         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
19764         (fold_builtin_cpu): Add "goldmont".
19765         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
19766         (ix86_option_override_internal): Add "goldmont".
19767         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
19768         (processor_type): Add PROCESSOR_GOLDMONT.
19769         * config/i386/i386.md: Add CPU "glm".
19770         * config/i386/glm.md: New file.
19771         * config/i386/x86-tune.def: Add m_GOLDMONT.
19772         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
19774 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19776         PR target/85572
19777         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
19778         E_V4DImode.
19779         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
19780         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
19781         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
19783         PR target/85317
19784         * config/i386/i386.c (ix86_fold_builtin): Handle
19785         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
19787         PR target/85480
19788         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
19789         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
19791 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
19793         PR target/85658
19794         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
19795         (check_arch): Likewise.
19796         (check_fpu): Return the result rather than printing it.
19797         (end arch): Fix operator precedence.
19798         (end cpu): Likewise.
19799         (END): Print the result from check_fpu.
19801 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19802             Alan Hayward  <alan.hayward@arm.com>
19803             David Sherwood  <david.sherwood@arm.com>
19805         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
19806         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
19807         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
19808         (*fcmuo<mode>_and): New patterns.
19810 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19812         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
19813         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
19814         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
19815         (cmp_op, sve_imm_con): New code attributes.
19816         (SVE_COND_INT_CMP, imm_con): Delete.
19817         (cmp_op): Remove above unspecs from int attribute.
19818         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
19819         to...
19820         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
19821         comparison-specific unspecs.
19822         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
19823         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
19824         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
19825         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
19826         (*vec_fcm<cmp_op><mode>): Rename to...
19827         (*fcm<cmp_op><mode>): ...this and adjust likewise.
19828         (*vec_fcmuo<mode>): Rename to...
19829         (*fcmuo<mode>): ...this and adjust likewise.
19830         (*pred_fcm<cmp_op><mode>): New pattern.
19831         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
19832         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
19833         functions.
19834         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
19835         and UNORDERED.
19836         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
19837         (aarch64_emit_sve_predicated_cond): New function.
19838         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
19839         (aarch64_emit_unspec_cond_or): Replace with...
19840         (aarch64_emit_sve_or_conds): ...this new function.  Use
19841         aarch64_emit_sve_ptrue_op for the individual comparisons and
19842         aarch64_emit_binop to OR them together.
19843         (aarch64_emit_inverted_unspec_cond): Replace with...
19844         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
19845         aarch64_emit_sve_ptrue_op for the comparison and
19846         aarch64_emit_unop to invert the result.
19847         (aarch64_expand_sve_vec_cmp_float): Update after the above
19848         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
19850 2018-05-07  Nathan Sidwell  <nathan@acm.org>
19852         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
19853         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
19854         (Backwards Compatibility): Likewise.
19856 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19858         PR bootstrap/85681
19859         Revert:
19860         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19862         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19863         <prefetch_dynamic_strides>: New const bool field.
19864         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19865         prefetch_dynamic_strides.
19866         (exynosm1_prefetch_tune): Likewise.
19867         (thunderxt88_prefetch_tune): Likewise.
19868         (thunderx_prefetch_tune): Likewise.
19869         (thunderx2t99_prefetch_tune): Likewise.
19870         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19871         to false.
19872         (aarch64_override_options_internal): Update to set
19873         PARAM_PREFETCH_DYNAMIC_STRIDES.
19874         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19875         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19876         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19877         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19878         prefetch-dynamic-strides setting.
19880         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19882         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19883         <minimum_stride>: New const int field.
19884         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19885         minimum_stride field.
19886         (exynosm1_prefetch_tune): Likewise.
19887         (thunderxt88_prefetch_tune): Likewise.
19888         (thunderx_prefetch_tune): Likewise.
19889         (thunderx2t99_prefetch_tune): Likewise.
19890         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19891         (aarch64_override_options_internal): Update to set
19892         PARAM_PREFETCH_MINIMUM_STRIDE.
19893         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19894         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19895         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19896         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19897         stride is constant and is below the minimum stride threshold.
19899 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19901         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
19902         to 512.
19904 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19906         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19907         <prefetch_dynamic_strides>: New const bool field.
19908         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19909         prefetch_dynamic_strides.
19910         (exynosm1_prefetch_tune): Likewise.
19911         (thunderxt88_prefetch_tune): Likewise.
19912         (thunderx_prefetch_tune): Likewise.
19913         (thunderx2t99_prefetch_tune): Likewise.
19914         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19915         to false.
19916         (aarch64_override_options_internal): Update to set
19917         PARAM_PREFETCH_DYNAMIC_STRIDES.
19918         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19919         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19920         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19921         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19922         prefetch-dynamic-strides setting.
19924 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19926         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19927         <minimum_stride>: New const int field.
19928         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19929         minimum_stride field.
19930         (exynosm1_prefetch_tune): Likewise.
19931         (thunderxt88_prefetch_tune): Likewise.
19932         (thunderx_prefetch_tune): Likewise.
19933         (thunderx2t99_prefetch_tune): Likewise.
19934         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19935         (aarch64_override_options_internal): Update to set
19936         PARAM_PREFETCH_MINIMUM_STRIDE.
19937         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19938         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19939         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19940         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19941         stride is constant and is below the minimum stride threshold.
19943 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
19945         PR c++/85659
19946         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
19947         the type is addressable.  Don't force op into register if it has
19948         BLKmode.
19950 2018-05-05  Roland McGrath  <mcgrathr@google.com>
19952         PR other/77609
19953         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
19954         any section for which we don't know a specific type it should have,
19955         regardless of name.  Previously this was done only for the exact
19956         names ".init_array", ".fini_array", and ".preinit_array".
19957         (default_elf_asm_named_section): Add comment about
19958         relationship with default_section_type_flags and SECTION_NOTYPE.
19959         (get_section): Don't consider it a type conflict if one side has
19960         SECTION_NOTYPE and the other doesn't, as long as neither has the
19961         SECTION_BSS et al used in the default_section_type_flags logic.
19963 2018-05-05  Tom de Vries  <tom@codesourcery.com>
19965         PR target/85653
19966         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
19967         (workaround_barsyncs): New function.
19968         (nvptx_reorg): Use workaround_barsyncs.
19969         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19970         (define_expand "nvptx_membar_cta"): New define_expand.
19971         (define_insn "*nvptx_membar_cta"): New insn.
19973 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
19975         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
19976         To improve optimization opportunities.
19977         * builtin-types.def: The new needed builtin types for the above.
19979 2018-05-04  Richard Biener  <rguenther@suse.de>
19981         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
19982         * gimple-ssa-store-merging.c
19983         (imm_store_chain_info::output_merged_store): Remove redundant create,
19984         release split_store vector contents on failure.
19985         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
19986         scalar stmt vector on cache hit.
19988 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
19990         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
19991         Xilinx FP support.
19992         * config.gcc (powerpc-xilinx-eabi*): Remove.
19993         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
19994         support.
19995         (fusion_addis_mem_combo_load): Ditto.
19996         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
19997         FP support.
19998         (rs6000_cpu_cpp_builtins): Ditto.
19999         * config/rs6000/rs6000-linux.c
20000         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20001         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20002         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20003         support.
20004         (rs6000_setup_reg_addr_masks): Ditto.
20005         (rs6000_init_hard_regno_mode_ok): Ditto.
20006         (rs6000_option_override_internal): Ditto.
20007         (legitimate_lo_sum_address_p): Ditto.
20008         (rs6000_legitimize_address): Ditto.
20009         (rs6000_legitimize_reload_address): Ditto.
20010         (rs6000_legitimate_address_p): Ditto.
20011         (abi_v4_pass_in_fpr): Ditto.
20012         (setup_incoming_varargs): Ditto.
20013         (rs6000_gimplify_va_arg): Ditto.
20014         (rs6000_split_multireg_move): Ditto.
20015         (rs6000_savres_strategy): Ditto.
20016         (rs6000_emit_prologue_components): Ditto.
20017         (rs6000_emit_epilogue_components): Ditto.
20018         (rs6000_emit_prologue): Ditto.
20019         (rs6000_emit_epilogue): Ditto.
20020         (rs6000_elf_file_end): Ditto.
20021         (rs6000_function_value): Ditto.
20022         (rs6000_libcall_value): Ditto.
20023         * config/rs6000/rs6000.h: Ditto.
20024         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20025         (TARGET_MINMAX): ... this.  New.
20026         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20027         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20028         (*movsi_internal1_single): Delete.
20029         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20030         mfpu=, mxilinx-fpu): Delete.
20031         * config/rs6000/singlefp.h: Delete.
20032         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20033         * config/rs6000/t-rs6000: Ditto.
20034         * config/rs6000/t-xilinx: Delete.
20035         * config/rs6000/titan.md: Adjust for fp_type removal.
20036         * config/rs6000/vsx.md: Remove Xilinx FP support.
20037         (VStype_simple): Delete.
20038         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20039         * config/rs6000/xfpu.h: Delete.
20040         * config/rs6000/xfpu.md: Delete.
20041         * config/rs6000/xilinx.h: Delete.
20042         * config/rs6000/xilinx.opt: Delete.
20043         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20044         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20046 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20048         PR libgomp/85639
20049         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20050         if ignore == 0.
20052 2018-05-04  Richard Biener  <rguenther@suse.de>
20054         PR middle-end/85627
20055         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20056         (expand_complex_div_wide): Likewise.
20057         (expand_complex_operations_1): Likewise.
20058         (expand_complex_libcall): Preserve EH info of the original stmt.
20059         (tree_lower_complex): Handle removed blocks.
20060         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20061         on complex multiplication and division libcall builtins.
20063 2018-05-04  Richard Biener  <rguenther@suse.de>
20065         PR middle-end/85574
20066         * fold-const.c (negate_expr_p): Restrict negation of operand
20067         zero of a division to when we know that can happen without
20068         overflow.
20069         (fold_negate_expr_1): Likewise.
20071 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20073         PR libstdc++/85466
20074         * real.h (real_nextafter): Declare.
20075         * real.c (real_nextafter): New function.
20076         * fold-const-call.c (fold_const_nextafter): New function.
20077         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20078         CASE_CFN_NEXTTOWARD.
20079         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20080         even when arg1_mode is different from arg0_mode.
20082 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20084         * doc/extend.texi (Deprecated Features): Remove
20085         -ffriend-injection.
20086         (Backwards Compatibility): Likewise.
20087         * doc/invoke.texi (C++ Language Options): Likewise.
20088         (C++ Dialect Options): Likewise.
20090 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20092         PR target/85530
20093         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20094         _mm512_mask_mullox_epi64): New intrinsics.
20096 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20098         PR testsuite/85106
20099         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20100         dump files): Add offload-tree.
20102 2018-05-03  Richard Biener  <rguenther@suse.de>
20104         PR tree-optimization/85615
20105         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20106         to loops not nested in BBs loop father to avoid creating multi-entry
20107         loops.
20109 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20111         PR tree-optimization/70291
20112         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20113         arguments.  Change return type to tree.  Emit libcall as a new
20114         statement rather than replacing existing one when inplace_p is true.
20115         (expand_complex_multiplication_components): New function.
20116         (expand_complex_multiplication): Expand floating-point complex
20117         multiplication using the above.
20118         (expand_complex_division): Rename inner_type parameter to type.
20119         Update expand_complex_libcall call-site.
20120         (expand_complex_operations_1): Update expand_complex_multiplication
20121         and expand_complex_division call-sites.
20123 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20125         PR target/85582
20126         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20127         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20128         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20129         the highest significant bit of the shift count mask is clear.  In
20130         check whether and[sq]i3 is needed verify that all significant bits
20131         of the shift count other than the highest are set.
20133 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20135         PR libgomp/82428
20136         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20137         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20138         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20139         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20140         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20141         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20142         __builtin_goacc_parlevel_size.
20144 2018-05-02  Richard Biener  <rguenther@suse.de>
20146         PR tree-optimization/85597
20147         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20148         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20149         directly.
20151 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20153         PR testsuite/85106
20154         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20155         dump files): Add ltrans-tree.
20157 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20159         PR testsuite/85106
20160         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20161         dump files): Add wpa-ipa.
20163 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20165         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20166         powerpc*-*-linux*paired* target.
20167         * config/rs6000/750cl.h: Delete.
20168         * config/rs6000/paired.h: Delete.
20169         * config/rs6000/paired.md: Delete.
20170         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20171         float support.
20172         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20173         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20174         comment.  Remove paired float support.
20175         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20176         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20177         VECTOR_PAIRED.
20178         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20179         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20180         declarations.
20181         * config/rs6000/rs6000.c: Remove paired float support.
20182         (paired_expand_vector_init, paired_expand_vector_move,
20183         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20184         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20185         paired_expand_builtin, paired_expand_predicate_builtin,
20186         paired_init_builtins): Delete.
20187         * config/rs6000/rs6000.h: Remove paired float support.
20188         * config/rs6000/rs6000.md: Remove paired float support.
20189         (move_from_CR_ov_bit): Delete.
20190         * config/rs6000/rs6000.opt (mpaired): Delete.
20191         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20192         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20194 2018-05-02  Richard Biener  <rguenther@suse.de>
20196         PR middle-end/85567
20197         * gimplify.c (gimplify_save_expr): When in SSA form allow
20198         SAVE_EXPRs to compute to SSA vars.
20200 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20202         PR target/85582
20203         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20204         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20205         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20206         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20208 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20210         PR tree-optimization/85586
20211         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20212         exit early for statements in the same group if the accesses are
20213         not strided.
20215 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20217         PR lto/85451
20218         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20219         error message.
20221 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20223         PR tree-optimization/85143
20224         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20226 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20228         PR lto/85451
20229         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20230         not found" error message.
20232 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20234         PR other/83786
20235         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20236         * vec.c (test_ordered_remove_if): New function.
20237         (vec_c_tests): Call test_ordered_remove_if.
20238         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20239         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20240         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20241         VEC_ORDERED_REMOVE_IF.
20243 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20245         PR tree-optimization/82665
20246         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20247         pointer subtraction where arguments come from a memchr call.
20249 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20251         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20252         --push-state --as-needed and --pop-state instead of --as-needed and
20253         --no-as-needed if ld supports it.
20254         * configure: Regenerated.
20256         PR web/85578
20257         * doc/install.texi2html: Replace _002d with - and _002a with * in
20258         generated html files using sed.
20260 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20262         PR c++/85523
20263         * gcc-rich-location.c (blank_line_before_p): New function.
20264         (use_new_line): New function.
20265         (gcc_rich_location::add_fixit_insert_formatted): New function.
20266         * gcc-rich-location.h
20267         (gcc_rich_location::add_fixit_insert_formatted): New function.
20269 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20271         * selftest.c (assert_streq): Rename "expected" and "actual" to
20272         "val1" and "val2".  Extend NULL-handling to cover both inputs
20273         symmetrically, while still requiring both to be non-NULL for a pass.
20274         * selftest.h (assert_streq): Rename "expected" and "actual" to
20275         "val1" and "val2".
20276         (ASSERT_EQ): Likewise.
20277         (ASSERT_EQ_AT): Likewise.
20278         (ASSERT_KNOWN_EQ): Likewise.
20279         (ASSERT_KNOWN_EQ_AT): Likewise.
20280         (ASSERT_NE): Likewise.
20281         (ASSERT_MAYBE_NE): Likewise.
20282         (ASSERT_MAYBE_NE_AT): Likewise.
20283         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20284         the assertion to pass.
20285         (ASSERT_STREQ_AT): Likewise.
20287 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20289         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20290         interaction with -pie.
20292 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20294         * selftest.h: Fix alphabetization of per-source-file selftest
20295         declarations.
20297 2018-04-30  Jason Merrill  <jason@redhat.com>
20299         PR c++/61982 - dead stores to destroyed objects.
20300         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20301         of clobber.
20303 2018-04-30  Jason Merrill  <jason@redhat.com>
20305         * tree.c (build_clobber): New.
20306         * tree.h: Declare it.
20307         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20309 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20311         * diagnostic-show-locus.c (layout::layout): Update for
20312         location_get_source_line returning a char_span.
20313         (struct char_span): Move to input.h.
20314         (struct correction): Update for fields in char_span becoming
20315         private.
20316         (struct source_line): Update for location_get_source_line
20317         returning a char_span.
20318         (layout::print_line): Likewise.
20319         * edit-context.c (edited_file::print_content): Likewise.
20320         (edited_file::print_diff_hunk): Likewise.
20321         (edited_file::print_run_of_changed_lines): Likewise.
20322         (edited_file::get_num_lines): Likewise.
20323         (edited_line::edited_line): Likewise.
20324         * final.c (asm_show_source): Likewise.
20325         * input.c (location_get_source_line): Convert return type
20326         from const char * to char_span, losing the final "line_len"
20327         param.
20328         (dump_location_info): Update for the above.
20329         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20330         when handling the literal within the line.
20331         (test_reading_source_line): Update for location_get_source_line
20332         returning a char_span.
20333         * input.h (class char_span): Move here from
20334         diagnostic-show-locus.c, converting from a struct to a class.
20335         Make data members private.
20336         (char_span::operator bool): New.
20337         (char_span::length): New.
20338         (char_span::get_buffer): New.
20339         (char_span::operator[]): New.
20340         (char_span::subspan): Make const.
20341         (char_span::xstrdup): New.
20342         (location_get_source_line): Convert return type from const char *
20343         to char_span, losing the final "line_size" param.
20345 2018-04-30  Jan Hubicka  <jh@suse.cz>
20347         * lto-wrapper.c (ltrans_priorities): New static var.
20348         (cmp_priority): New.
20349         (run_gcc): Read priorities and if doing parallel build order
20350         the Makefile by them.
20352 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20354         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20356 2018-04-30  Richard Biener  <rguenther@suse.de>
20358         * tree-cfg.c (verify_address): Remove base argument, add
20359         flag whether to check TREE_ADDRESSABLE and do that.
20360         (verify_expr): Remove.
20361         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20362         (verify_gimple_assign_single): Likewise.
20363         (verify_gimple_switch): Likewise.
20364         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20365         piece from verify_expr.
20366         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20368 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20370         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20372 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20374         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20375         (small_data_pattern): Likewise.
20376         (arc_rewrite_small_data): Likewise.
20377         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20378         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20379         (get_symbol_alignment): New function.
20380         (legitimate_small_data_address_p): Likewise.
20381         (legitimate_scaled_address): Update, call
20382         legitimate_small_data_address_p.
20383         (output_sdata): New static variable.
20384         (arc_print_operand): Update how we handle small data operands.
20385         (arc_print_operand_address): Likewise.
20386         (arc_legitimate_address_p): Update, use
20387         legitimate_small_data_address_p.
20388         (arc_rewrite_small_data_p): Remove.
20389         (arc_rewrite_small_data_1): Likewise.
20390         (arc_rewrite_small_data): Likewise.
20391         (small_data_pattern): Likewise.
20392         (compact_sda_memory_operand): Update to use
20393         legitimate_small_data_address_p and get_symbol_alignment.
20394         (prepare_move_operands): Don't rewite sdata pattern.
20395         (prepare_extend_operands): Remove.
20396         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20397         pattern.
20398         (zero_extendqisi2): Likewise.
20399         (zero_extendhisi2): Likewise.
20400         (extendqihi2): Likewise.
20401         (extendqisi2): Likewise.
20402         (extendhisi2): Likewise.
20403         (addsi3): Likewise.
20404         (subsi3): Likewise.
20405         (andsi3): Likewise.
20406         * config/arc/constraints.md (Usd): Change it to memory constraint.
20408 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20410         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20411         as source of std instructions.
20412         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20413         allow 6-bit constants as source for store instructions.
20414         (movdi_insn): Update instruction pattern to allow 6-bit constants
20415         as source for store instructions.
20417 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20419         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20421 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20422             Sandra Loosemore <sandra@codesourcery.com>
20424         * dumpfile.c (dump_open): Allow '-' for stdout.
20425         * doc/invoke.texi (Developer Options): Document dump filename
20426         determination early.  Document stdin/stdout selection.
20428 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20430         Microblaze Target: PIC data text relative
20432         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20433         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20434         Add declaration.
20435         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20436         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20437         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20438         New addressing mode for data-text relative position indepenedent code.
20439         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20440         'ADDRESS_SYMBOLIC_TXT_REL'.
20441         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20442         (microblaze_legitimate_pic_operand): Exclude function calls from
20443         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20444         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20445         addresses cases.
20446         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20447         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20448         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20449         for 'address + offset'.
20450         (microblaze_expand_prologue): Add new function prologue call for
20451         'r20' assignation.
20452         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20453         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20454         table in case of TARGET_PIC_DATA_TEXT_REL.
20455         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20456         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20457         Add new macros 'UNSPEC_TEXT',
20458         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20459         + exclude function calls from 'UNSPEC_PLT' in case of data text
20460         relative mode.
20461         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20462         new target hook for generating address diff vector tables in case of
20463         flag_pic.
20464         * doc/tm.texi : Regenerate.
20465         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20466         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20467         of addr diff vector generation.
20468         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20469         target hook definition.
20470         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20471         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20472         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20473         option for data text relative.
20475 2018-04-30  Richard Biener  <rguenther@suse.de>
20477         * tree-chrec.h (evolution_function_is_constant_p): Remove
20478         redundant check.
20479         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20481 2018-04-30  Richard Biener  <rguenther@suse.de>
20483         PR bootstrap/85571
20484         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20486 2018-04-30  Richard Biener  <rguenther@suse.de>
20488         PR tree-optimization/28364
20489         PR tree-optimization/85275
20490         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20491         copying first exit test.
20493 2018-04-28  Mark Wielaard  <mark@klomp.org>
20495         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20496         dwarf_version >= 5.
20497         (dwarf_AT): Handle DW_AT_addr_base.
20498         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20500 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
20502         PR target/84431
20503         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
20504         (*ashl<dwi>3_doubleword_mask_1): Ditto.
20505         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
20506         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
20508 2018-04-28  Richard Biener  <rguenther@suse.de>
20510         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
20511         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
20512         to reflect use.  Only add interesting stmts.
20514 2018-04-27  Martin Jambor  <mjambor@suse.cz>
20516         PR ipa/85549
20517         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
20518         the jump function allows for passing through aggregate values.
20520 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
20522         * input.h (in_system_header_at): Convert from macro to inline
20523         function.
20524         (from_macro_expansion_at): Likewise.
20525         (from_macro_definition_at): Likewise.
20527 2018-04-27  Jeff Law  <law@redhat.com>
20529         * config.gcc: Mark tile* targets as deprecated/obsolete.
20531 2018-04-27  Richard Biener  <rguenther@suse.de>
20533         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
20534         fix for ILP32.
20536 2018-04-27  Richard Biener  <rguenther@suse.de>
20538         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
20540 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
20542         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
20543         with Yd constraint. Set "preferred_for_speed" attribute from
20544         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
20545         with Yd constraint.
20546         (*movdi_internal): Ditto.
20547         (movti_interunit splitters): Remove
20548         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
20549         (movdi_interunit splitters): Ditto.
20550         * config/i386/constraints.md (Ye): Remove.
20551         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
20553 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20555         PR target/85512
20556         * config/aarch64/constraints.md (Usg): Limit to 31.
20557         (Usj): Limit to 63.
20559 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
20561         PR tree-optimization/85529
20562         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
20563         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
20564         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
20565         zero extension or masking of the MSB bit.
20566         (optimize_range_tests): Add FIRST_BB argument, pass it through
20567         to optimize_range_tests_var_bound.
20568         (maybe_optimize_range_tests, reassociate_bb): Adjust
20569         optimize_range_tests callers.
20571 2018-04-26  Richard Biener  <rguenther@suse.de>
20572             Jakub Jelinek  <jakub@redhat.com>
20574         * cgraph.h (symbol_table): Just declare debug method here.
20575         * symtab.c (symbol_table::debug): Define.
20577 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
20579         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
20581 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
20583         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
20584         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
20585         (*movdi_internal): Substitute Yi and Yj constraint with x
20586         and Ym and Yn constraint with y constraint.  Update "isa"
20587         attribute and set "preferred_for_speed" attribute from
20588         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
20589         (*movsi_internal): Ditto.
20590         (*movdf_internal): Ditto.
20591         (*movsf_internal): Ditto.
20592         (*zero_extendsidi2): Ditto.
20593         * config/i386/sse.md (vec_set<mode>_0): Ditto.
20594         (sse2_loadld): Ditto.
20595         (*vec_extract<ssevecmodelower>_0): Ditto.
20596         (*vec_extractv4si_0_zext_sse4): Ditto.
20597         (vec_concatv2di): Ditto.
20598         (*vec_dup<mode>): Ditto.
20599         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
20600         * config/i386/constraints.md (Yi): Remove.
20601         (Yj): Remove.
20602         (Ym): Remove.
20603         (Yn): Remove.
20605 2018-04-26  Nathan Sidwell  <nathan@acm.org>
20607         * dumpfile.c (dump_open): New.
20608         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
20609         (dump_finish): Detect stdio/stderr by value not name.
20611 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
20613         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
20615 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20617         PR target/84952
20618         * config/nvptx/nvptx.c (verify_neutering_jumps)
20619         (verify_neutering_labels): New function
20620         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
20622 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20624         PR target/84025
20625         * config/nvptx/nvptx.c (needs_neutering_p): New function.
20626         (nvptx_single): Use needs_neutering_p to skip over insns that do not
20627         need neutering.
20629 2018-04-26  Richard Biener <rguenther@suse.de>
20630             Tom de Vries  <tom@codesourcery.com>
20632         PR lto/85422
20633         * lto-streamer-out.c (output_function): Fixup loops if required to match
20634         discovery done in the reader.
20636 2018-04-26  Richard Biener  <rguenther@suse.de>
20638         PR tree-optimization/85116
20639         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
20640         have a loop exit from the single latch predecessor.  Remove
20641         case of header with just condition.
20642         (ch_base::copy_headers): Exclude infinite loops from any
20643         processing.
20644         (pass_ch::execute): Record exits.
20646 2018-04-26  Richard Biener  <rguenther@suse.de>
20648         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
20649         prologue cost vector and pass it to vect_get_load_cost.
20650         (vect_get_peeling_costs_all_drs): Likewise.
20651         (vect_peeling_hash_get_lowest_cost): Likewise.
20652         (vect_enhance_data_refs_alignment): Likewise.
20654 2018-04-26  Richard Biener  <rguenther@suse.de>
20656         PR middle-end/85450
20657         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
20658         checking of integer<->pointer conversions.
20659         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
20660         sign-/zero-extending pointer types.
20661         (expand_omp_for_static_chunk): Likewise.
20663 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
20664             Jean Lee  <xiaoyur347@gmail.com>
20666         * config/mips/mips.c (mips_asan_shadow_offset): New function.
20667         (TARGET_ASAN_SHADOW_OFFSET): Define.
20668         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
20669         true for -fsanitize=address.
20671 2018-04-25  Mark Wielaard  <mark@klomp.org>
20673         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
20674         shorter ones.
20676 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20678         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
20679         than "alu", remove explicit "memory" and "imm_disp" attributes.
20680         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
20682         PR middle-end/85414
20683         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
20684         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
20685         gen_lowpart_no_emit.
20687 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
20689         PR target/85473
20690         * config/i386/i386.c (ix86_expand_builtin): Change memory
20691         operand to XI, extend p0 to Pmode.
20692         * config/i386/i386.md: Change unspec volatile and operand
20693         1 mode to XI, change operand 0 mode to P.
20695 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20697         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
20698         GET_MODE_MASK before any checking.
20699         (nds32_can_use_bset_p): Likewise.
20700         (nds32_can_use_btgl_p): Likewise.
20702 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20704         * config/nds32/nds32-doubleword.md: New define_split pattern for
20705         illegal register number.
20707 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20709         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
20711 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20713         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
20715 2018-04-25  Richard Biener  <rguenther@suse.de>
20717         * lto-streamer.h (LTO_major_version): Bump to 8.
20719 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20721         * BASE-VER: Set to 9.0.0.
20723 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
20725         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
20726         in __abskf2 and __powikf2.
20728 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20730         PR target/85512
20731         * config/aarch64/constraints.md (Usg, Usj): New constraints.
20732         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
20733         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
20734         Use the above on operand 2.  Reindent.
20735         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
20737 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20739         PR target/85485
20740         * common/config/i386/i386-common.c (ix86_handle_option): Don't
20741         handle OPT_mcet.
20742         * config/i386/i386.opt (mcet): Removed.
20743         * doc/install.texi: Remove -mcet documentation.
20744         * doc/invoke.texi: Likewise.
20746 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20748         PR target/85485
20749         * doc/install.texi: Remove -mcet from bootstrap-cet.
20751 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20753         PR target/85511
20754         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
20755         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
20756         if TARGET_64BIT.
20758         PR target/85503
20759         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
20760         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
20761         containing a CONST_VECTOR.
20763 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
20765         * doc/install.texi: Update newlib dependency for nvptx.
20767 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20769         PR target/85508
20770         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
20771         instead of INTVAL when shifting x left.
20773 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
20775         PR tree-optimization/85478
20776         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
20777         vect_grouped_store_supported for single element vectors.
20779 2018-04-24  Richard Biener  <rguenther@suse.de>
20781         PR target/85491
20782         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
20783         load cost increase to the case of non-constant step.
20785 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20787         PR target/84828
20788         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
20789         destination if any_malformed_asm.
20791 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
20793         PR middle-end/85496
20794         * expr.c (store_field): In the bitfield case, if the value comes from
20795         a function call and is returned in registers by means of a PARALLEL,
20796         do not change the mode of the temporary unless BLKmode and VOIDmode.
20798 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
20800         PR rtl-optimization/85423
20801         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
20802         dependencies to debug insns when the previous insn is non-debug.
20804 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20806         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
20807         enums into a single definition.
20808         (fls): Fix predicates and printing.
20809         (seti): Likewise.
20811 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20813         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
20814         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
20815         and short u6 immediate.
20816         (check_if_valid_sleep_operand): Remove.
20817         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
20819 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20821         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
20822         flag_always_save_lp condition.
20823         * config/nds32/nds32.opt (malways-save-lp): New option.
20825 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20827         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
20828         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
20829         * config/nds32/nds32.h
20830         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
20831         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
20833 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20835         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
20836         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
20838 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20839             Chung-Ju Wu  <jasonwucj@gmail.com>
20841         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
20842         Declare.
20843         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
20844         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
20846 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20848         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
20850 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20852         * config/nds32/nds32-protos.h (nds32_data_alignment,
20853         nds32_local_alignment): Declare.
20854         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
20855         nds32_local_alignment): New functions.
20856         (TARGET_CONSTANT_ALIGNMENT): Define.
20857         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
20859 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20861         * config/nds32/nds32.c
20862         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
20863         (TARGET_MODES_TIEABLE_P): Likewise.
20865 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20867         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
20868         level Ofast and Og.
20870 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
20871             Chung-Ju Wu  <jasonwucj@gmail.com>
20873         * config/nds32/constants.md (unspec_volatile_element): Add enum values
20874         for unaligned access.
20875         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20876         unaligned access.
20877         * config/nds32/nds32-intrinsic.md: Likewise.
20878         * config/nds32/nds32_intrinsic.h: Likewise.
20879         * config/nds32/nds32.h (nds32_builtins): Likewise.
20880         * config/nds32/nds32.opt (munaligned-access): New option.
20881         * config/nds32/nds32.c (nds32_asm_file_start): Display
20882         flag_unaligned_access status.
20884 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20886         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
20887         -mno-relax is present.
20888         * config/riscv/linux.h (LINK_SPEC): Ditto.
20890 2018-04-20  Martin Sebor  <msebor@redhat.com>
20892         PR c/85365
20893         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
20894         for null pointers.
20895         (gimple_fold_builtin_stxcpy_chk): Same.
20896         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
20898 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
20900         PR target/85456
20901         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
20902         __powikf2 when long double is IEEE 128-bit.
20904 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20906         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
20907         step to make sure stack always aligned.
20909 2018-04-20  Carl Love  <cel@us.ibm.com>
20911         PR target/83402
20912         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
20913         size check for arg0.
20915 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
20916             Tom de Vries  <tom@codesourcery.com>
20918         PR target/85445
20919         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
20920         Emit insns for calls too.
20921         (nvptx_find_par): Always look for worker-level predecessor insn.
20922         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
20923         calls.
20924         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
20925         (nvptx_process_pars): Propagate frames for calls.
20927 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
20929         PR target/85469
20930         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
20931         Removed.
20932         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
20933         (ix86_handle_option): Don't handle OPT_mibt.
20934         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
20935         __SHSTK__.
20936         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
20937         has_ibt and ibt.
20938         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
20939         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
20940         (ix86_target_macros): Define __CET__ with flag_cf_protection
20941         for -fcf-protection.
20942         * config/i386/i386.c (isa2_opts): Remove -mibt.
20943         * config/i386/i386.h (TARGET_IBT): Removed.
20944         (TARGET_IBT_P): Likewise.
20945         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
20946         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
20947         * config/i386/i386.opt (mcet): Update help message.
20948         (mshstk): Likewise.
20949         (mibt): Removed.
20950         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
20951         -mcet as an alias for -mshstk.
20953 2018-04-20  Richard Biener <rguenther@suse.de>
20955         PR middle-end/85475
20956         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
20957         complexity by forcing a single use of the multiply operand.
20959 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20961         ipa/85449
20962         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
20963         recursion dependency to only apply to non-clones.
20965 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20967         ipa/85447
20968         * ipa-cp.c (create_specialized_node): Check that clones of
20969         self-recursive edges exist during IPA-CP.
20971 2018-04-19  Toon Moene  <toon@moene.org>
20973         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
20974         by -O3.
20976 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
20978         PR tree-optimization/85467
20979         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
20980         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
20981         VECTOR_CST element to type.
20983 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20985         PR target/85397
20986         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
20987         * config/i386/i386.md (builtin_setjmp_setup): Removed.
20988         (builtin_longjmp): Likewise.
20989         (save_stack_nonlocal): New pattern.
20990         (restore_stack_nonlocal): Likewise.
20992 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20994         PR target/85404
20995         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20996         Replace ASM_OUTPUT_LABEL with fprintf.
20998 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21000         PR target/85417
21001         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21002         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21003         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21004         define __IBT__ and __SHSTK__ for -fcf-protection.
21005         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21006         TARGET_IBT.
21007         (ix86_trampoline_init): Likewise.
21008         (x86_output_mi_thunk): Likewise.
21009         (ix86_notrack_prefixed_insn_p): Likewise.
21010         (ix86_option_override_internal): Don't disallow -fcf-protection.
21011         * config/i386/i386.md (rdssp<mode>): Also enable for
21012         -fcf-protection.
21013         (incssp<mode>): Likewise.
21014         (nop_endbr): Likewise.
21015         * config/i386/i386.opt (mcet): Change help message to built-in
21016         functions only.
21017         (mibt): Likewise.
21018         (mshstk): Likewise.
21019         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21020         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21021         enable CET built-in functions.
21023 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21025         * common/config/i386/i386-common.c
21026         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21027         OPTION_MASK_ISA_MOVDIRI_UNSET,
21028         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21029         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21030         * config.gcc (movdirintrin.h): New header.
21031         * config/i386/cpuid.h (bit_MOVDIRI,
21032         bit_MOVDIR64B): New bits.
21033         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21034         and -mmvodir64b.
21035         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21036         (VOID, PVOID, PCVOID)): New function types.
21037         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21038         __builtin_ia32_directstoreu_u64,
21039         __builtin_ia32_movdir64b): New builtins.
21040         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21041         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21042         and -mmovdiri.
21043         (ix86_valid_target_attribute_inner_p): Ditto.
21044         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21045         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21046         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21047         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21048         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21049         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21050         (movdiri<mode>, movdir64b_<mode>): New.
21051         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21052         * config/i386/immintrin.h: Include movdirintrin.h.
21053         * config/i386/movdirintrin.h: New file.
21054         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21056 2018-04-19  Richard Biener  <rguenther@suse.de>
21058         PR middle-end/85455
21059         * cfg.c (clear_bb_flags): When loop state says we have
21060         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21062 2018-04-19  Richard Biener  <rguenther@suse.de>
21064         PR tree-optimization/84737
21065         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21066         copying restrict info.
21067         (vect_setup_realignment): Use it.
21068         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21069         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21070         the first DR to all generated stores.
21071         (vectorizable_load): Likewise for loads.
21073 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21075         PR tree-optimization/85446
21076         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21077         the integral and pointer types to have the same precision.
21079         * doc/install.texi: Document --disable-cet being the default and
21080         --enable-cet=auto.
21082 2018-04-18  Martin Liska  <mliska@suse.cz>
21084         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21085         style.
21087 2018-04-18  Martin Liska  <mliska@suse.cz>
21089         Revert
21090         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21092         PR ipa/83983
21093         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21094         arguments if they are comparable.
21096 2018-04-18  Martin Liska  <mliska@suse.cz>
21098         Revert
21099         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21101         PR lto/84805
21102         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21103         incomplete types.
21105 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21107         PR target/85388
21108         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21109         ENDBR after calling __morestack.
21111 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21113         PR jit/85384
21114         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21115         by using gcc_base_ver to generate a gcc_driver_version, and use
21116         it when generating GCC_DRIVER_NAME.
21117         * configure: Regenerate.
21119 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21121         PR target/81084
21122         * config.gcc: Obsolete powerpc*-*-*spe*.
21124 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21126         PR debug/84637
21127         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21128         (stabstr_D): Change type of unum from unsigned int to
21129         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21130         type.
21132 2018-04-17  Jim Wilson  <jimw@sifive.com>
21134         PR 84856
21135         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21136         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21137         Set arg_pointer_offset after using pretend_args_size.
21139 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21141         PR rtl-optimization/85431
21142         * dse.c (record_store): Ignore zero width stores.
21144         PR sanitizer/85230
21145         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21146         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21147         __builtin_stack_restore rather than after it.
21148         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21149         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21150         argument instead of virtual_dynamic_stack_rtx.
21152 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21154         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21155         New prototype.
21156         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21157         Add note to error message to explain internal mapping of overloaded
21158         built-in function name to non-overloaded built-in function name.
21159         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21160         function.
21162 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21164         PR target/85424
21165         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21166         where the inputs overlap with the output.
21168 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21170         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21171         (=v, v) alternative and explicit "memory" attribute.
21172         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21173         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21174         attributes.
21175         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21176         "sselog1" type instead of "sselog".
21177         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21178         "sselog".  Remove explicit "memory" attribute.
21179         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21180         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21181         attributes.
21182         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21183         "sselog1" type instead of "sselog".  Remove explicit "memory"
21184         attribute.
21185         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21186         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21187         attribute.
21188         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21189         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21190         attributes.
21191         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21192         "sselog1" type instead of "sselog".  Remove explicit "memory"
21193         attribute.
21194         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21195         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21196         attribute.
21198         PR target/85430
21199         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21201         PR middle-end/85414
21202         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21203         on a SUBREG.
21205 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21207         PR ipa/85421
21208         * ipa-cp.c (create_specialized_node): Call
21209         expand_all_artificial_thunks if necessary.
21211 2018-04-17  Martin Liska  <mliska@suse.cz>
21213         PR lto/85405
21214         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21215         in message, remote space in between '_G' and '('.
21217 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21219         PR target/85281
21220         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21221         avx512f_vmcmp<mode>3<round_saeonly_name>,
21222         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21223         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21224         avx512f_rndscale<mode><round_saeonly_name>,
21225         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21226         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21227         Use %<iptr>2 instead of %2 for -masm=intel.
21228         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21229         avx512f_vcvttss2usi<round_saeonly_name>,
21230         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21231         -masm=intel.
21232         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21233         avx512f_vcvttsd2usi<round_saeonly_name>,
21234         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21235         Use %q1 instead of %1 for -masm=intel.
21236         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21237         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21238         of %3 for -masm=intel.
21239         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21240         -masm=intel.
21241         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21242         -masm=intel.
21243         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21244         -masm=intel.
21245         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21246         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21247         %g1.
21248         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21249         -masm=intel.
21250         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21251         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21252         %g1 and one with %0 and %1.
21253         (avx512er_vmrcp28<mode><round_saeonly_name>,
21254         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21255         %1 for -masm=intel.
21256         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21257         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21258         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21259         of %0 and %{%4%} for -masm=intel.
21260         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21261         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21262         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21263         order of %0 and %{%5%}%{z%} for -masm=intel.
21265 2018-04-17  Jan Hubicka  <jh@suse.cz>
21267         PR lto/85405
21268         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21270 2018-04-17  Martin Liska  <mliska@suse.cz>
21272         PR ipa/85329
21273         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21274         for target_clone error message.  Make default implementation
21275         clone to be a local declaration.
21276         (separate_attrs): Add new argument and check for an empty
21277         string.
21278         (expand_target_clones): Handle it.
21279         (ipa_target_clone): Make redirection just for target_clones
21280         functions.
21282 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21283             Tom de Vries  <tom@codesourcery.com>
21285         PR middle-end/84955
21286         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21287         tiled basic blocks without omp continue statements.
21289 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21291         PR target/83660
21292         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21293         vec_extract expression as having side effects to make sure it gets
21294         a cleanup point.
21296 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21298         PR target/85403
21299         * config/i386/i386.c (get_builtin_code_for_version): Check
21300         error_mark_node.
21302 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21304         PR target/84331
21305         * config.gcc: Support "skylake".
21306         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21307         PROCESSOR_SKYLAKE.
21308         * config/i386/i386.c (m_SKYLAKE): Define.
21309         (processor_target_table): Add "skylake".
21310         (ix86_option_override_internal): Add "skylake".
21311         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21312         PROCESSOR_CANNONLAKE.
21313         (get_builtin_code_for_version): Fix priority for
21314         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21315         PROCESSOR_SKYLAKE-AVX512.
21316         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21317         (processor_type): Add PROCESSOR_SKYLAKE.
21319 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21320             Jason Merrill  <jason@redhat.com>
21322         PR c++/85112
21323         * convert.c (convert_to_integer_1): Use direct recursion for
21324         enumeral types and types with a precision less than the number
21325         of bits in their mode.
21327 2018-04-16  Julia Koval  <julia.koval@intel.com>
21329         PR target/84413
21330         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21331         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21333 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21335         PR target/85293
21336         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21337         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21338         and -mno-direct-move.
21340 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21342         PR target/83402
21343         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21344         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21345         Ensure negative shifts result in {0}.
21347 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21349         PR rtl-optimization/79916
21350         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21351         regs (if any) to define how to gnerate SD moves when LRA is in
21352         progress.
21354 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21356         PR rtl-optimization/85393
21357         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21358         * except.c (expand_dw2_landing_pad_for_region): Make static.
21359         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21360         a label and unconditional jump to old_bb, rather than
21361         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21362         basic block.
21364         PR rtl-optimization/85376
21365         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21366         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21367         instead of a specific value.
21369 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21370             Bin Cheng  <bin.cheng@arm.com>
21372         PR tree-optimization/82965
21373         PR tree-optimization/83991
21374         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21375         by_profile_only parameter.
21376         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21377         information if the loop was predicted to iterate too many times.
21378         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21380 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21382         PR lto/71991
21383         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21384         always inline.
21386 2018-04-13  Martin Liska  <mliska@suse.cz>
21387             Jakub Jelinek  <jakub@redhat.com>
21389         PR middle-end/81657
21390         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21391         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21392         * builtins.c (expand_builtin_memory_copy_args): Use
21393         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21394         handle dest_addr == pc_rtx.
21396 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21398         PR target/85291
21399         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21400         asked to not generate direct moves.
21401         (fix_trunc<mode>si2_stfiwx): Similar.
21402         (fix_trunc<mode>si2_internal): Similar.
21404 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21406         PR debug/83157
21407         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21408         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21409         lookup if dest in some wider mode is known to be const0_rtx and
21410         if so, record permanent equivalence for it to be ZERO_EXTEND of
21411         the narrower mode destination.
21413 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21415         * lto-streamer-out.c (output_function): Revert 259346.
21416         * omp-expand.c (expand_oacc_for): Likewise.
21418 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21420         PR rtl-optimization/85354
21421         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21422         * sel-sched.c (sel_global_init): ... here.
21424 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21426         PR target/85238
21427         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21428         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21429         mode for PE-COFF targets.
21430         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21431         (i386_pe_asm_lto_end): Likewise.
21432         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21433         (TARGET_ASM_LTO_END): Likewise.
21434         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21435         (i386_pe_asm_lto_start): New function.
21436         (i386_pe_asm_lto_end): Likewise.
21438 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21439             Richard Biener  <rguenther@suse.de>
21441         PR middle-end/84955
21442         * lto-streamer-out.c (output_function): Fix CFG loop state before
21443         streaming out.
21444         * omp-expand.c (expand_oacc_for): Handle calls to internal
21445         functions like regular functions.
21447 2018-04-12  Richard Biener  <rguenther@suse.de>
21449         PR lto/85371
21450         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21451         for the early LTO debug to properly generate references to it
21452         during DIE emission.  Do not re-use that for the skeleton for
21453         split-dwarf.
21454         (dwarf2out_early_finish): Likewise.
21456 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21458         PR target/85328
21459         * config/i386/sse.md
21460         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21461         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21462         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21463         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21464         and output is a reg, avoid creating invalid lowpart subreg, but
21465         instead split into a 512-bit move.  Don't split if not AVX512VL,
21466         input is xmm16+ reg and output is a mem.
21467         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21468         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21469         xmm16+ reg and output is a mem.
21471 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21473         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21474         also for flag_dwarf2_cfi_asm.
21476 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21478         PR rtl-optimization/85342
21479         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21480         a bool scalar var inside of the loop instead.  Don't try to update
21481         recog_data.operand after failed apply_change_group.
21483 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21485         PR target/85296
21486         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21487         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21488         array with flexible array member as array without given dimension.
21489         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21490         argument for undefined param to true.
21492 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21494         PR target/85321
21495         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21496         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21497         from PowerPC section.
21498         * config/rs6000/sysv4.opt (mcall-): Improve help text.
21499         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21500         help text that is too long.
21501         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
21502         help text that is too long.
21503         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
21504         help text that is too long.
21506 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
21508         * config/alpha/alpha.md (stack_probe_internal): Rename
21509         from "probe_stack".  Update all callers.
21511 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21513         PR rtl-optimization/84566
21514         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
21515         sched_macro_fuse_insns.
21517 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21519         PR target/84301
21520         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
21521         (compute_block_dependences): ... from here.
21523 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21525         PR tree-optimization/85331
21526         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
21527         from int to HOST_WIDE_INT.
21529 2018-04-11  Martin Jambor  <mjambor@suse.cz>
21531         PR ipa/84149
21532         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
21533         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
21534         not the same as the source val.
21535         (cgraph_edge_brings_value_p): New parameter.
21536         (gather_edges_for_value): Pass destination value to
21537         cgraph_edge_brings_value_p.
21538         (perhaps_add_new_callers): Likewise.
21539         (get_info_about_necessary_edges): Likewise and exclude values brought
21540         only by self-recursive edges.
21541         (create_specialized_node): Redirect only clones of self-calling edges.
21542         (+self_recursive_pass_through_p): New function.
21543         (find_more_scalar_values_for_callers_subset): Use it.
21544         (find_aggregate_values_for_callers_subset): Likewise.
21545         (known_aggs_to_agg_replacement_list): Removed.
21546         (decide_whether_version_node): Re-calculate known constants for all
21547         remaining context clones.
21549 2018-04-11  Richard Biener  <rguenther@suse.de>
21551         PR lto/85339
21552         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
21553         from early DWARF output.
21554         (dwarf2out_early_finish): Output line info unconditionally into
21555         early DWARF and add reference to it.
21557 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21559         PR target/85281
21560         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
21561         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
21562         other than V2DFmode using iptr mode attribute.
21563         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
21565 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21567         PR rtl-optimization/84659
21568         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
21570 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21572         PR debug/85302
21573         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
21574         SIZEP is NULL.
21575         (output_loc_list): Pass address of a dummy size variable even in the
21576         locview handling loop.
21577         (index_location_lists): Add comment on why skip_loc_list_entry can't
21578         call size_of_locs.
21580 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21582         PR target/85261
21583         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
21584         into register.
21586 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
21588         PR target/85321
21589         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21590         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
21591         and -mstring-compare-inline-limit.
21593 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21595         PR target/85287
21596         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
21597         for stack clash protection in a register whenever we need it to be in
21598         a register.
21600 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21602         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
21603         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
21605 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21607         PR target/85321
21608         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
21609         the help text.
21610         (mlong-double-): Ditto.
21611         * config/rs6000/sysv4.opt (msdata=): Ditto.
21612         (mtls-size=): Ditto.
21614 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21616         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
21617         erroneous entries for
21618         "vector int vec_ldl (int, long int *)", and
21619         "vector unsigned int vec_ldl (int, unsigned long int *)".
21620         Add comments and entries for
21621         "vector bool char vec_ldl (int, bool char *)",
21622         "vector bool short vec_ldl (int, bool short *)",
21623         "vector bool int vec_ldl (int, bool int *)",
21624         "vector bool long long vec_ldl (int, bool long long *)",
21625         "vector pixel vec_ldl (int, pixel *)",
21626         "vector long long vec_ldl (int, long long *)",
21627         "vector unsigned long long vec_ldl (int, unsigned long long *)".
21628         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
21629         type tree bool_long_long_type_node and correct definition of
21630         bool_V2DI_type_node to make reference to this new type tree.
21631         (rs6000_mangle_type): Replace erroneous reference to
21632         bool_long_type_node with bool_long_long_type_node.
21633         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
21634         comments to emphasize sign distinctions for char and int types and
21635         replace RS6000_BTI_bool_long constant with
21636         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
21637         use of RS6000_BTI_pixel.
21638         (bool_long_type_node): Remove this macro definition.
21639         (bool_long_long_type_node): New macro definition
21641 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21643         PR rtl-optimization/85300
21644         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
21645         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
21646         simplify_unary_operation fails.
21648 2018-04-10  Martin Liska  <mliska@suse.cz>
21650         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
21651         cgraph_edge and ipa_ref.
21653 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21655         PR target/85177
21656         PR target/85255
21657         * config/i386/sse.md
21658         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
21659         computation of the VEC_MERGE selector from mask.
21660         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
21661         Fix decoding of the VEC_MERGE selector into mask.
21663 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21665         PR tree-optimization/85286
21666         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
21668 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21670         * final.c (final_1): Set insn_last_address as well as
21671         insn_current_address.
21673 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21675         PR target/85173
21676         * explow.c (emit_stack_probe): Call validize_mem on memory location
21677         before passing it to gen_probe_stack.  Create address operand and
21678         legitimize it for the probe_stack_address case.
21680 2018-04-09  Jan Hubicka  <jh@suse.cz>
21682         PR lto/85078
21683         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
21684         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
21685         * tree.c (free_lang_data_in_type): Fix handling of binfos;
21686         walk basetypes.
21687         (free_lang_data): Rebuild type inheritance graph.
21689 2018-04-09  Martin Sebor  <msebor@redhat.com>
21691         * invoke.texi (-finline-small-functions): Mention other optimization
21692         options.
21693         (-findirect-inlining, -fpartial-inlining): Same.
21694         (-finline-functions-called-once): Same.
21695         (-freorder-blocks-and-partition): Same.
21697 2018-04-09  Jan Hubicka  <jh@suse.cz>
21699         PR rtl/84058
21700         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
21701         jumps; choose last target that matches the criteria (i.e.
21702         no partition changes for non-crossing jumps).
21703         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
21704         support for redirecting crossing jumps to non-crossing.
21706 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
21708         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
21709         also for naked functions.
21711 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
21713         * config/arc/arc.md (add_shift): New pattern.
21714         (add_shift2): Likewise.
21715         (sub_shift): Likewise.
21716         (sub_shift_cmp0_noout): Likewise.
21717         (compare_si_ashiftsi): Likewise.
21718         (xbfu_cmp0_noout): New combine pattern.
21719         (xbfu_cmp0"): Likewise.
21720         (movsi_set_cc_insn): Place the predicable variant first.
21721         (commutative_binary_cmp0_noout): Remove clobber.
21722         (commutative_binary_cmp0): New pattern.
21723         (noncommutative_binary_cmp0): Likewise.
21724         (noncommutative_binary_cmp0_noout): Likewise.
21725         (noncommutative_binary_comparison_result_used): Removed.
21726         (rsub_cmp0): New pattern.
21727         (rsub_cmp0_noout): Likewise.
21728         (extzvsi): Changed, keep only meaningful variants.
21729         (SQH, SEZ): New iterators.
21730         (SQH_postfix): New mode attribute.
21731         (SEZ_prefix): New code attribute.
21732         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
21733         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
21734         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
21735         of numerical value.
21736         (noncommutative_operator): Check the availability of barrel
21737         shifter option.
21739 2018-04-09  Richard Biener  <rguenther@suse.de>
21741         PR tree-optimization/85284
21742         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
21743         Only use the niter constraining form of simple_iv when the exit
21744         is always executed.
21746 2018-04-09  Tom de Vries  <tom@codesourcery.com>
21748         PR target/84041
21749         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21750         (define_expand "*memory_barrier"): New define_expand.
21751         (define_insn "memory_barrier"): New insn.
21753 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21755         PR rtl-optimization/80463
21756         PR rtl-optimization/83972
21757         PR rtl-optimization/83480
21759         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
21760         correct producer for the insn.
21761         (tidy_control_flow): Fixup seqnos in case of debug insns.
21763 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21765         PR rtl-optimization/83913
21767         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
21768         different sched-times when merging exprs.
21770 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21772         PR rtl-optimization/83962
21774         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
21775         tidy_fallthru_edge and tidy_control_flow.
21777 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21779         PR rtl-optimization/83530
21781         * sel-sched.c (force_next_insn): New global variable.
21782         (remove_insn_for_debug): When force_next_insn is true, also leave only
21783         next insn in the ready list.
21784         (sel_sched_region): When the region wasn't scheduled, make another pass
21785         over it with force_next_insn set to 1.
21787 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21789         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
21790         into tm_file.
21791         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21792         for interrupt control.
21793         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21794         functions for interrupt control.
21795         * config/nds32/nds32-intrinsic.md: Likewise.
21796         * config/nds32/nds32_intrinsic.h: Likewise.
21797         * config/nds32/nds32.h (nds32_builtins): Likewise.
21799 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21801         * config/nds32/nds32.c (nds32_init_machine_status,
21802         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
21803         strict_aligned_p field.
21804         (nds32_expand_to_rtl_hook): New function.
21805         (TARGET_EXPAND_TO_RTL_HOOK): Define.
21806         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
21808 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21809             Chung-Ju Wu  <jasonwucj@gmail.com>
21811         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
21812         * config/nds32/nds32-n7.md: New file.
21813         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
21814         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
21815         pipeline.
21816         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
21817         * config/nds32/nds32.md (pipeline_model): Add n7.
21818         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21819         * config/nds32/pipelines.md: Include n7 settings.
21821 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21822             Chung-Ju Wu  <jasonwucj@gmail.com>
21824         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
21825         * config/nds32/nds32-e8.md: New file.
21826         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
21827         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
21828         pipeline.
21829         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
21830         * config/nds32/nds32.md (pipeline_model): Add e8.
21831         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
21832         * config/nds32/pipelines.md: Include e8 settings.
21834 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21835             Chung-Ju Wu  <jasonwucj@gmail.com>
21837         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
21838         * config/nds32/nds32-n8.md: New file.
21839         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
21840         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
21841         pipeline.
21842         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
21843         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
21844         * config/nds32/nds32.md (pipeline_model): Add n8.
21845         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
21846         * config/nds32/pipelines.md: Include n8 settings.
21848 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21849             Chung-Ju Wu  <jasonwucj@gmail.com>
21851         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
21852         * config/nds32/nds32-n9-2r1w.md: New file.
21853         * config/nds32/nds32-n9-3r2w.md: New file.
21854         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
21855         nds32_register_ports): New or modify for cpu n9.
21856         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
21857         pipeline.
21858         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
21859         * config/nds32/nds32-utils.c: New file.
21860         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
21861         TARGET_MUL_SLOW): Define.
21862         * config/nds32/nds32.md (pipeline_model): New attribute.
21863         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
21864         New options that support cpu n9.
21865         * config/nds32/pipelines.md: Include n9 settings.
21866         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
21868 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21870         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
21871         information if necessary.
21872         (output_cond_branch_compare_zero): Likewise.
21873         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
21874         (nds32_target_alignment): Refine for alignment.
21875         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
21876         (FUNCTION_BOUNDARY): Modify.
21877         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
21878         align case.
21879         * config/nds32/nds32.opt (malways-align, malign-functions): New.
21881 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21883         * config/nds32/constants.md (unspec_volatile_element): Add values for
21884         TLB operation and data prefetch.
21885         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21886         functions for TLB operation and data prefetch.
21887         * config/nds32/nds32-intrinsic.md: Likewise.
21888         * config/nds32/nds32_intrinsic.h: Likewise.
21889         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
21890         (nds32_print_operand): Likewise.
21891         * config/nds32/nds32.h (nds32_builtins): Likewise.
21893 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21894         Andrew Pinski <pinsika@gcc.gnu.org>
21896         PR middle-end/82976
21897         * match.pd: Use constant_boolean_node of correct type instead of
21898         boolean_true_node or boolean_false_node for simplifying
21899         pointer comparisons to zero.
21901 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21903         PR tree-optimization/80021
21904         * tree.c (verify_type_variant): Make error call in verify_variant_match
21905         translatable and remove final full stop.
21907 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21909         * config/nds32/constants.md (unspec_volatile_element): Add
21910         UNSPEC_VOLATILE_EH_RETURN.
21911         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
21912         nds32_output_stack_pop): Support dwarf exception handling process.
21913         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
21914         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
21915         exception handling process.
21916         (nds32_compute_stack_frame): Likewise.
21917         (nds32_return_addr_rtx): Likewise.
21918         (nds32_initial_elimination_offset): Likewise.
21919         (nds32_expand_prologue): Likewise.
21920         (nds32_expand_epilogue): Likewise.
21921         (nds32_dynamic_chain_address): New function.
21922         * config/nds32/nds32.h (machine_function): Add fields for dwarf
21923         exception handling.
21924         (DYNAMIC_CHAIN_ADDRESS): Define.
21925         (EH_RETURN_DATA_REGNO): Define.
21926         (EH_RETURN_STACKADJ_RTX): Define.
21927         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
21928         patterns for dwarf exception handling.
21930 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21932         * config/nds32/nds32.h: Clean up obsolete macros.
21934 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21936         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21937         Add enum values for particular instructions.
21938         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21939         particular intrinsic functions.
21940         * config/nds32/nds32-intrinsic.md: Likewise.
21941         * config/nds32/nds32_intrinsic.h: Likewise.
21942         * config/nds32/nds32.h (nds32_builtins): Likewise.
21943         * config/nds32/nds32.md (type): Add pbsad and pbsada.
21944         (btst, ave): New patterns for particular instructions.
21946 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21948         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21949         Add enum values for atomic load/store and memory sync.
21950         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
21951         and memory sync.
21952         * config/nds32/nds32-intrinsic.md: Likewise.
21953         * config/nds32/nds32_intrinsic.h: Likewise.
21954         * config/nds32/nds32.h (nds32_builtins): Likewise.
21956 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21958         PR tree-optimization/85257
21959         * fold-const.c (native_encode_vector): If not all elts could fit
21960         and off is -1, return 0 rather than offset.
21961         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
21962         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
21963         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
21964         adjust buffer in native_interpret_expr call.
21966 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21968         * config/nds32/constants.md (unspec_volatile_element): Add cache
21969         control enum values.
21970         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
21971         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
21972         * config/nds32/nds32.c (nds32_cctl_names): New.
21973         (nds32_print_operand): Handle cache control register names.
21974         * config/nds32/nds32.h (nds32_builtins): New enum values.
21975         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
21976         macros.
21977         * config/nds32/nds32.md (type): Add mmu.
21978         * config/nds32/pipelines.md (simple_insn): Add mmu.
21980 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21982         * config/nds32/nds32.md (type): Remove call.
21983         * config/nds32/pipelines.md (simple_insn): Likewise.
21985 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21987         * config/nds32/constants.md (unspec_volatile_element): Add
21988         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
21989         UNSPEC_VOLATILE_FMFCFG.
21990         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
21991         description for fmfcfg and fmfcsr.
21992         (bdesc_1arg): Add fmtcsr.
21993         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
21994         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
21995         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
21996         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
21997         unspec_fmfcfg): New patterns.
21998         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
21999         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22000         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22001         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22002         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22003         __nds32__fmfcfg): Define.
22005 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22007         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22008         intrinsic register names.
22009         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22010         intrinsic register enum values and macros.
22012 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22014         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22015         for load/store addressing form.
22016         (nds32_print_operand_address): Likewise.
22018 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22020         PR target/85196
22021         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22022         based on LABEL_REF.  Remove useless assertion.
22023         (pic_address_needs_scratch): Fix formatting.
22024         (sparc_legitimize_pic_address): Minor tweaks.
22025         (sparc_delegitimize_address): Adjust assertion accordingly.
22026         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22027         into symbolic_operand.
22028         (movsi_high_pic_label_ref): Likewise.
22029         (movsi_lo_sum_pic_label_ref): Likewise.
22030         (movdi_pic_label_ref): Likewise.
22031         (movdi_high_pic_label_ref): Likewise.
22032         (movdi_lo_sum_pic_label_ref): Likewise.
22034 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22036         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22037         custom LIB_SPEC setup.
22039 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22040             Kito Cheng  <kito.cheng@gmail.com>
22042         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22043         * config/riscv/freebsd.h: New.
22045 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22047         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22048         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22049         file.
22051 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22052             Kito Cheng  <kito.cheng@gmail.com>
22054         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22055         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22056         * config/nds32/nds32-protos.h (nds32_output_call,
22057         nds32_output_return): Declare.
22058         * config/nds32/nds32.md: Refine all the call and return patterns.
22060 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22062         PR debug/85252
22063         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22064         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22066         PR rtl-optimization/84872
22067         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22068         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22069         EDGE_CROSSING edge.
22071 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22073         * expr.c (copy_blkmode_to_reg): Revert 254862.
22074         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22076 2018-04-06  Richard Biener  <rguenther@suse.de>
22078         PR middle-end/85244
22079         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22080         after seeing a component reference with an adjacent field.  Treat
22081         refs to arrays at struct end of external decls similar to
22082         refs to unconstrained commons.
22084 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22086         PR sanitizer/85213
22087         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22088         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22089         recursive calls.
22090         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22091         save_p here.
22093 2018-04-06  Richard Biener  <rguenther@suse.de>
22095         PR middle-end/85180
22096         * alias.c (find_base_term): New wrapper around find_base_term
22097         unwinding CSELIB_VAL_PTR changes.
22098         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22099         recursion.
22101 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22103         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22104         instructions.
22105         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22106         constant definitions.
22107         ("nop"): lr 0,0 -> nopr r0
22108         ("nop_lr0", "nop_lr1"): New insn definitions.
22110 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22112         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22113         NDS32_V3PUSH_AVAILABLE_P macro.
22115 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22116             Chung-Ju Wu  <jasonwucj@gmail.com>
22118         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22119         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22120         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22121         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22122         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22123         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22124         * config/nds32/constraints.md: New constraints and checking for hard
22125         float configuration.
22126         * config/nds32/iterators.md: New mode iterator and attribute for hard
22127         float configuration.
22128         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22129         patterns.
22130         * config/nds32/nds32-fpu.md: New file.
22131         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22132         deal with hard float code generation.
22133         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22134         ARCH_V3S.
22135         (abi_type, float_reg_number): New enum type.
22136         * config/nds32/nds32-predicates.c: New predicates for hard float.
22137         * config/nds32/nds32-protos.h: Declare functions for hard float.
22138         * config/nds32/nds32.c: Implementation for hard float configuration.
22139         * config/nds32/nds32.h: Definitions for hard float configuration.
22140         * config/nds32/nds32.md: Include hard float machine description and
22141         modify patterns for hard float configuration.
22142         * config/nds32/nds32.opt: New options for hard float configuration.
22143         * config/nds32/predicates.md: New predicates for hard float
22144         configuration.
22146 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22148         * common/config/nds32/nds32-common.c
22149         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22151 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22153         PR middle-end/85195
22154         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22155         CONSTRUCTOR_ELT (ctor, ...)->value.
22157 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22159         PR target/85193
22160         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22162 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22164         PR target/85204
22165         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22166         cond jump.
22168 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22169             Kito Cheng  <kito.cheng@gmail.com>
22171         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22172         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22173         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22174         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22176 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22177             Kito Cheng  <kito.cheng@gmail.com>
22179         * config/nds32/constraints.md (Ufe): New memory constraint.
22180         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22181         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22182         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22183         operands.
22184         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22185         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22187 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22189         * config/nds32/nds32.md: Use optimize_size in the condition for
22190         alu-shift instructions.
22192 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22194         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22196 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22198         * config/nds32/nds32.md (negsi2): Refine pattern.
22200 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22201             Chung-Ju Wu  <jasonwucj@gmail.com>
22203         * config/nds32/iterators.md (shift_rotate): New code iterator.
22204         (shift): New code attribute.
22205         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22206         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22207         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22208         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22209         bit-wise operations.
22210         (andsi3, *andsi3): Ditto.
22211         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22212         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22213         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22214         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22215         nds32_ior_operand, nds32_xor_operand): New predicates.
22217 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22219         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22220         (addsi3, subsi3): ... this.
22222 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22224         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22226 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22228         * config/nds32/nds32.md: Adjust indention.
22230 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22232         * config/nds32/nds32.md (feature): New attribute.
22234 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22236         * config/nds32/nds32.md (subtype): New attribute.
22238 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22240         PR target/85203
22241         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22242         expansion to perform a bitwise AND of the argument followed by a
22243         boolean negation of the result.
22245 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22247         PR rtl-optimization/84878
22248         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22249         the basic block.  Assert the use reference is not artificial and that
22250         it has an associated insn.
22252 2018-04-04  Michael Matz  <matz@suse.de>
22254         * builtins.c (compute_objsize): Pass correct operand
22255         to array_at_struct_end_p.
22257 2018-04-04  Richard Biener  <rguenther@suse.de>
22259         PR lto/85176
22260         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22261         from contexts for DINFO_LEVEL_TERSE and below.
22263 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22265         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22266         resiter_operand condition.
22267         * config/nds32/nds32.md (*move<mode>): Ditto.
22269 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22270             Monk Chiang  <sh.chiang04@gmail.com>
22272         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22274 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22276         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22278 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22279             Kito Cheng  <kito.cheng@gmail.com>
22281         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22282         nds32_cond_code_str, output_cond_branch,
22283         output_cond_branch_compare_zero, nds32_expand_cbranch,
22284         nds32_expand_cstore, nds32_expand_movcc,
22285         nds32_output_cbranchsi4_equality_zero,
22286         nds32_output_cbranchsi4_equality_reg,
22287         nds32_output_cbranchsi4_equality_reg_or_const_int,
22288         nds32_output_cbranchsi4_greater_less_zero: New functions.
22289         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22290         nds32_expand_cstore, nds32_expand_movcc,
22291         nds32_output_cbranchsi4_equality_zero,
22292         nds32_output_cbranchsi4_equality_reg,
22293         nds32_output_cbranchsi4_equality_reg_or_const_int,
22294         nds32_output_cbranchsi4_greater_less_zero): Declare.
22295         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22296         nds32_rimm11s_operand): New predicates.
22297         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22298         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22299         patterns.
22301 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22303         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22304         * config/nds32/nds32.md: Ditto.
22305         * config/nds32/pipelines.md: Ditto.
22307 2018-04-04  Richard Biener  <rguenther@suse.de>
22309         PR tree-optimization/85168
22310         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22311         propagating abnormals.
22313 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22315         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22317 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22318             Kito Cheng  <kito.cheng@gmail.com>
22320         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22321         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22322         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22323         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22324         * config/nds32/nds32.md (sibcall_internal): New.
22325         (sibcall_register): Remove.
22326         (sibcall_immediate): Remove.
22327         (sibcall_value_internal): New.
22328         (sibcall_value_register): Remove.
22329         (sibcall_value_immediate): Remove.
22330         * config/nds32/predicates.md (nds32_general_register_operand): New.
22331         (nds32_call_address_operand): New.
22333 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22335         PR rtl-optimization/85167
22336         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22337         bb_defs if *split_p, instead preinitialize it to NULL.
22339         PR tree-optimization/85156
22340         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22341         evaluating the argument multiple times.
22343 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22345         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22346         than vector.
22347         (_mm_cvtpd_ps): Likewise.
22348         (_mm_cvttpd_epi32): Likewise.
22349         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22350         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22351         vector, pixel, and bool following altivec.h include.
22353 2018-04-03  Martin Sebor  <msebor@redhat.com>
22355         * doc/extend.texi (Common Function Attributes): Clarify.
22356         (const attribute): Likewise.
22357         (pure attribute): Likewise.
22359 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22361         PR target/85169
22362         * config/i386/i386.c (ix86_expand_vector_set): Use
22363         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22365 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22367         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22368         instructions when changing rounding bits to preserve precision bits
22369         in the x87 control word.
22371 2018-04-03  Martin Liska  <mliska@suse.cz>
22373         PR tree-optimization/82491
22374         * rtl.h (strip_offset_and_add): Replace += suboffset with
22375         poly_uint64 () + suboffset.
22377 2018-03-29  Martin Liska  <mliska@suse.cz>
22378             Martin Jambor  <mjambor@suse.cz>
22380         PR ipa/84947
22381         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22382         param_type is not an integral or pointer type.
22384 2018-04-03  Richard Biener  <rguenther@suse.de>
22386         * sese.h (recompute_all_dominators): Remove.
22388 2018-04-02  Martin Sebor  <msebor@redhat.com>
22390         * doc/invoke.texi (-Wrestrict): Fix typos.
22392 2018-04-02  Jim Wilson  <jimw@sifive.com>
22394         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22395         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22396         (<optab>di3, <optab>si3_extend): Likewise.
22397         (<optab>si3_mask, <optab>si3_mask_1): New.
22398         (<optab>di3_mask, <optab>di3_mask_1): New.
22399         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22400         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22401         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22403 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22405         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22406         example.
22408 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22410         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22411         (nds32_canonicalize_comparison): New function.
22413 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22414             Kito Cheng  <kito.cheng@gmail.com>
22415             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22417         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22418         * config/nds32/constants.md (unspec_volatile_element): Add
22419         UNSPEC_VOLATILE_RELAX_GROUP.
22420         * config/nds32/nds32-relax-opt.c: New file.
22421         * config/nds32/nds32-predicates.c
22422         (nds32_symbol_load_store_p): New function.
22423         * config/nds32/nds32-protos.h
22424         (nds32_symbol_load_store_p): Declare function.
22425         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22426         * config/nds32/nds32.c
22427         (nds32_register_pass): New function to register pass.
22428         (nds32_register_passes): New function to register passes.
22429         * config/nds32/nds32.md (relax_group): New pattern.
22430         * config/nds32/nds32.opt (mrelax-hint): New option.
22431         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22433 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22435         * config/nds32/t-nds32: Modify files dependency.
22437 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22439         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22440         (PROFILE_HOOK): Define its implementation.
22442 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22444         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22445         type and 32-bit size.
22447 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22449         PR middle-end/85090
22450         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22451         (V_128_256): New mode iterator.
22452         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22453         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22454         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22455         of V.
22456         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22457         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22459 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22461         PR target/83315
22462         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22463         NaN inputs correctly.
22465 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22467         PR target/80546
22468         * config/rs6000/vsx.md (??r): New mode attribute.
22469         (*vsx_mov<mode>_64bit): Use it.
22470         (*vsx_mov<mode>_32bit): Likewise.
22472 2018-03-30  Martin Sebor  <msebor@redhat.com>
22474         PR tree-optimization/84818
22475         * builtins.c (check_access): Use warning_n.
22477 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22479         PR target/83822
22480         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22481         condition.
22482         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22483         condition.
22485 2018-03-30  Julia Koval  <julia.koval@intel.com>
22487         PR target/84413
22488         * x86-tune.def (movx, partial_reg_dependency): Enable for
22489         m_SKYLAKE_AVX512.
22491 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22493         PR inline-asm/84985
22494         * lra-constraints.c (process_alt_operands): Move setting
22495         this_alternative_matches below.
22497 2018-03-29  Martin Liska  <mliska@suse.cz>
22499         PR lto/84995.
22500         * doc/invoke.texi: Document how LTO works with debug info.
22501         Describe auto-load support of binutils.  Mention 'x86-64'
22502         as valid option value of -march option.
22504 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
22506         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
22508         PR c/85094
22509         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
22510         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
22511         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
22512         checking.
22514 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22516         PR target/84912
22517         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
22518         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
22519         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
22520         for RS6000_BTM_POWERPC64.
22521         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
22522         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
22523         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
22524         definition.
22525         (DIVDE): Use it.
22526         (DIVDEU): Likewise.
22528 2018-03-28  Carl Love  <cel@us.ibm.com>
22530         Revert
22531         2017-09-27  Carl Love  <cel@us.ibm.com>
22533         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
22534         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
22535         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
22536         fctiw instruction.
22538 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22540         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
22541         instead of __vector bool.
22542         (_mm_max_pu8): Likewise.
22543         (_mm_min_pi16): Likewise.
22545 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22547         PR target/84912
22548         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
22549         (DIVWEUO): Likewise.
22550         (DIVDEO): Likewise.
22551         (DIVDEUO): Likewise.
22552         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
22553         DIVWEUO and DIVDEUO.
22554         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
22555         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
22556         (div_extend): Likewise.
22557         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
22558         builtin function.
22559         (__builtin_divweuo): Likewise.
22560         (__builtin_divdeo): Likewise.
22561         (__builtin_divdeuo): Likewise.
22563 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22565         PR target/85095
22566         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
22567         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
22569         PR tree-optimization/82004
22570         * gimple-match-head.c (optimize_pow_to_exp): New function.
22571         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
22572         Don't fold to exp if optimize_pow_to_exp is false.
22574 2018-03-28  Martin Liska  <mliska@suse.cz>
22576         PR other/84819
22577         * calls.c (initialize_argument_information): Fix trailing space.
22578         * common.opt: Fix typo and provide better explanation for
22579         -fsanitize-coverage option.
22580         * config/i386/i386.opt: Fix typo.
22582 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22583             Martin Liska  <mliska@suse.cz>
22585         PR sanitizer/85081
22586         * gimplify.c (asan_poison_variable): Don't do the check for
22587         gimplify_omp_ctxp here.
22588         (gimplify_decl_expr): Do it here.
22589         (gimplify_target_expr): Likewise.
22591 2018-03-28  Martin Liska  <mliska@suse.cz>
22593         PR target/84988
22594         * config/i386/i386.c (ix86_function_arg_advance): Do not call
22595         chkp_type_bounds_count if MPX is not enabled.
22597 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
22599         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
22601 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
22603         PR target/84914
22604         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
22605         function to create the function decl for complex long double
22606         multiply and divide for -mabi=ieeelongdouble.
22607         (init_float128_ieee): Call it.
22609 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
22611         PR target/85044
22612         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
22613         -fcf-protection=branch -mibt.
22614         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
22616 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22618         PR target/81863
22619         * config/arm/arm.c (arm_valid_symbolic_address): Handle
22620         arm_word_relocations.
22622 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
22624         PR target/85056
22625         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
22626         extern array declarations.
22628 2018-03-27  Richard Biener  <rguenther@suse.de>
22630         PR middle-end/84067
22631         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
22632         explicit single_use checks.
22634 2018-03-27  Richard Biener  <rguenther@suse.de>
22636         PR tree-optimization/85082
22637         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
22638         Valueize the VUSE.
22640 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22642         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
22643         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
22644         Turn on fasynchronous-unwind-tables and funwind-tables.
22646 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
22648         PR target/85073
22649         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
22650         (*bmi_blsr_<mode>_ccz): Ditto.
22652 2018-03-26  Tom de Vries  <tom@codesourcery.com>
22654         PR tree-optimization/85063
22655         * omp-general.c (offloading_function_p): New function.  Factor out
22656         of ...
22657         * omp-offload.c (pass_omp_target_link::gate): ... here.
22658         * omp-general.h (offloading_function_p): Declare.
22659         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
22660         with attribute omp declare target for offloading functions.
22662 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
22664         PR tree-optimization/84005
22665         * tree-data-ref.h (get_base_for_alignment): Declare.
22666         * tree-data-ref.c (get_base_for_alignment_1): New function.
22667         (get_base_for_alignment): Likewise.
22668         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
22669         get_base_for_alignment to find a suitable base object, instead
22670         of always using drb->base_address.
22672 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22674         PR inline-asm/85022
22675         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
22676         known size by default.
22678 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
22680         PR inline-asm/85030
22681         * lra-constraints.c (process_alt_operands): Don't match BLKmode
22682         and non BLKmode operands.
22684 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22686         PR target/85026
22687         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
22688         Clean up attributes.
22690 2018-03-23  Richard Biener  <rguenther@suse.de>
22692         PR debug/85020
22693         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
22694         we are going to emit early debug for LTO.
22696 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22698         PR inline-asm/85034
22699         * function.c (match_asm_constraints_1): Don't optimize if input
22700         doesn't satisfy general_operand predicate for output's mode.
22702         PR inline-asm/85022
22703         * alias.c (write_dependence_p): Don't require for x_canonicalized
22704         non-VOIDmode if x has VOIDmode.
22706         PR sanitizer/85029
22707         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
22708         just don't try to optimize it rather than assert it never happens.
22710 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22712         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
22713         macro expansions for definition of ST_INTERNAL_<mode> and
22714         LD_INTERNAL_<mode> builtins.
22715         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
22716         Remove prototype.
22717         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
22718         function.
22719         (altivec_expand_st_builtin): Likewise.
22720         (altivec_expand_builtin): Remove calls to deleted functions.
22721         (rs6000_address_for_altivec): Delete this function.
22722         * config/rs6000/vector.md: Remove expands for
22723         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
22725 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
22727         PR target/84826
22728         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
22729         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
22730         re-computing once computed.
22731         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
22732         (arm_init_machine_status): Initialize
22733         machine->static_chain_stack_bytes.
22735 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22737         PR target/84760
22738         * doc/extend.texi: Add four new prototypes for vec_ld.
22739         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
22740         definitions for more logical presentation.
22741         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
22742         entries for V1TI variants of __builtin_altivec_ld builtin.
22743         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
22744         handling of V1TI variant of LVX icode pattern.
22745         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
22746         (rs6000_gimple_fold_builtin): Likewise.
22747         (altivec_init_builtins): Add code to define
22748         __builtin_altivec_lvx_v1ti function.
22750 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22752         PR inline-asm/84941
22753         * function.c (match_asm_constraints_1): Don't do the optimization
22754         if input isn't a REG, SUBREG, MEM or constant.
22756 2018-03-22  Tom de Vries  <tom@codesourcery.com>
22758         PR tree-optimization/84956
22759         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
22760         bb_has_abnormal_pred.
22762 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22764         PR sanitizer/85018
22765         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
22766         DECL_INITIAL (decl) to decl at the end.
22767         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
22768         adjust the comment.
22770 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
22772         * doc/extend.texi (__builtin_tgmath): Document when complex
22773         integer types are treated as _Complex _Float64.
22775 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22777         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
22779 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22781         PR tree-optimization/84960
22782         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
22783         if it is ENTRY block, move them into single succ of ENTRY in that case.
22785 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
22787         PR tree-optimization/84811
22788         * poly-int.h (poly_span_traits): Remove the T3 parameter and
22789         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
22790         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
22791         (known_subrange_p): Update accordingly.  Cast each value involved
22792         in the size comparison, rather than casting the result of the
22793         subtraction.
22795 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22797         PR tree-optimization/84982
22798         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
22799         by flipping the least significant bit rather than all bits from
22800         bitpos to bitpos + bitsize - 1.
22802 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22804         * doc/extend.texi (Deprecated Features): Remove mention of
22805         long-deleted deprecations.
22807 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22809         PR jit/84288
22810         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
22811         * configure: Regenerate.
22813 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22815         PR tree-optimization/83126
22816         * tree-parloops.c (num_phis): New function.
22817         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
22819 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22821         * doc/extend.texi (Deprecated Features): Update deprecated flags,
22822         mention anon-struct/union members and trailing attributes.
22824 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
22826         PR tree-optimization/84969
22827         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
22828         builtin memset partitions if they set different rhs values.
22830 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22832         PR rtl-optimization/84989
22833         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
22834         VEC_DUPLICATE with scalar result mode.
22836 2018-03-21  Martin Liska  <mliska@suse.cz>
22838         PR ipa/84963
22839         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
22840         not intended return statement.
22842 2018-03-21  Martin Liska  <mliska@suse.cz>
22844         PR target/84988
22845         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
22846         (chkp_find_bound_slots_1): Limit number of iterations.
22848 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
22850         PR target/84838
22851         * Minor grammar fixes for x86 options.
22853 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22855         PR debug/84875
22856         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
22857         holding REG_CFA_RESTORE notes, instead turn them into a USE.
22859 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
22861         PR target/83789
22862         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
22863         (altivec_lvx_<mode>_1op): Likewise.
22864         (altivec_stvx_<mode>_2op): Likewise.
22865         (altivec_stvx_<mode>_1op): Likewise.
22866         (altivec_lvx_<VM2:mode>): New define_expand.
22867         (altivec_stvx_<VM2:mode>): Likewise.
22868         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
22869         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22870         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
22871         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22872         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
22873         (rs6000_gen_lvx): Likewise.
22874         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
22875         (altivec_expand_stv_builtin): Likewise.
22876         (altivec_expand_builtin): Likewise.
22877         * config/rs6000/vector.md: Likewise.
22879 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22881         PR target/82518
22882         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
22883         BYTES_BIG_ENDIAN.
22885 2018-03-20  Richard Biener  <rguenther@suse.de>
22887         PR target/84986
22888         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
22889         sign-conversions as zero, fall back to standard scalar_stmt
22890         cost for the rest.
22892 2018-03-20  Martin Liska  <mliska@suse.cz>
22894         PR ipa/84825
22895         * predict.c (rebuild_frequencies): Handle case when we have
22896         PROFILE_ABSENT, but flag_guess_branch_prob is false.
22898 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22900         PR target/84990
22901         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
22902         flag_section_anchors.
22903         * varasm.c (use_blocks_for_decl_p): Remove hack for
22904         dw2_force_const_mem.
22906         PR target/84845
22907         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
22908         to ...
22909         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
22910         be created, use lowpart_subreg of operands[0] rather than operands[0]
22911         itself.
22912         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
22913         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
22914         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
22915         and n constraint instead of aarch64_shift_imm_di and Usd.
22916         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
22917         (*aarch64_<optab>_reg_minus<mode>3): ... this.
22919 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
22921         PR target/82989
22922         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
22923         to favor GPR over NEON registers.
22924         (<shift>di3_neon): Likewise.
22926 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22928         PR target/84952
22929         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
22930         (nvptx_process_pars): Emit bar.sync asap and alap.
22932 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22934         PR target/84954
22935         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
22936         seen_label if seen_label is already set.
22938 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22940         PR target/84945
22941         * config/i386/i386.c (fold_builtin_cpu): For features above 31
22942         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
22943         Use 1U instead of 1.  Formatting fixes.
22945         PR c/84953
22946         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
22947         instead of TREE_TYPE (s1) for the return value.
22949 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
22951         PR tree-optimization/84946
22952         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
22953         bitsize + bitsize in poly_uint64 rather than poly_int64.
22955         PR sanitizer/78651
22956         * dwarf2asm.c: Include fold-const.c.
22957         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
22958         of decl rather than decl itself.
22960         PR rtl-optimization/84643
22961         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
22963 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
22965         PR sanitizer/78651
22966         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
22967         calling assemble_variable.
22969 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
22971         PR target/81647
22972         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
22973         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
22975 2018-03-19  Jim Wilson  <jimw@sifive.com>
22977         PR bootstrap/84856
22978         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
22979         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
22980         (riscv_first_stack_step): Likewise.
22981         (riscv_option_override): Use STACK_BOUNDARY instead of
22982         MIN_STACK_BOUNDARY.
22983         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
22984         MIN_STACK_BOUNDARY.
22985         (BIGGEST_ALIGNMENT): Set to 128.
22986         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
22987         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
22988         STACK_BOUNDARY.
22990 2018-03-19  Richard Biener  <rguenther@suse.de>
22992         PR tree-optimization/84933
22993         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
22994         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
22996 2018-03-19  Richard Biener  <rguenther@suse.de>
22998         PR tree-optimization/84859
22999         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23000         (cond_if_else_store_replacement): Perform sinking operation on
23001         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23002         Generalize what a BB with a single eligible store is.
23004 2018-03-19  Richard Biener  <rguenther@suse.de>
23006         PR tree-optimization/84929
23007         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23008         chrec_is_positive against non-chrec arg.
23010 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23012         PR target/84711
23013         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23015 2018-03-18  Martin Liska  <mliska@suse.cz>
23017         PR rtl-optimization/84635
23018         * regrename.c (build_def_use): Use matches_mode only when
23019         matches >= 0.
23021 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23023         PR tree-optimization/84913
23024         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23025         vectorize chains of COND_EXPRs.
23027 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23029         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23031 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23033         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23035 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23037         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23039 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23040             Kito Cheng  <kito.cheng@gmail.com>
23042         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23043         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23044         (nds32_adjust_reg_alloc_order): New function.
23045         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23047 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23049         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23050         nds32_print_operand, nds32_print_operand_address): Use
23051         HOST_WIDE_INT_PRINT_DEC instead.
23053 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23055         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23057 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23059         PR target/84902
23060         * config/i386/i386.c (initial_ix86_tune_features,
23061         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23062         unsigned long long.
23063         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23064         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23065         rather than 1u << ix86_tune.  Formatting fix.
23066         (ix86_option_override_internal): Change ix86_arch_mask from
23067         unsigned int to unsigned HOST_WIDE_INT, initialize to
23068         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23069         (ix86_function_specific_restore): Likewise.
23071 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23073         PR target/84899
23074         * postreload.c (reload_combine_recognize_pattern): Perform
23075         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23076         truncate_int_for_mode the result for the destination's mode.
23078         PR c/84909
23079         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23080         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23081         Likewise.
23082         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23083         Likewise.
23085 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23087         PR target/84876
23088         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23089         regno_allocno_class_array and sorted_pseudos.
23090         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23091         insns where regno is used.
23093 2018-03-16  Martin Liska  <mliska@suse.cz>
23095         PR ipa/84833
23096         * multiple_target.c (create_dispatcher_calls): Redirect
23097         reference in the symbol table.
23099 2018-03-16  Martin Liska  <mliska@suse.cz>
23101         PR ipa/84722
23102         * multiple_target.c (create_dispatcher_calls): Redirect also
23103         an alias.
23105 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23107         PR c++/79937
23108         PR c++/82410
23109         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23110         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23111         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23113 2018-03-16  Julia Koval  <julia.koval@intel.com>
23115         * doc/invoke.texi (Skylake Server): Add CLWB.
23116         Cannonlake): Remove CLWB.
23118 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23120         PR tree-optimization/84841
23121         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23122         1 << 3.
23123         (FLOAT_ONE_CONST_TYPE): Define.
23124         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23125         (sort_by_operand_rank): Put entries with higher constant_type last
23126         rather than first to match comments.
23128 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23130         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23131         split predicate.
23133 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23135         PR c++/79085
23136         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23137         check and use address of target always.
23139 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23141         PR target/84574
23142         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23143         (indirect_thunk_bnd_needed): Likewise.
23144         (indirect_thunks_used): Likewise.
23145         (indirect_thunks_bnd_used): Likewise.
23146         (indirect_return_needed): New.
23147         (indirect_return_bnd_needed): Likewise.
23148         (output_indirect_thunk_function): Add a bool argument for
23149         function return.
23150         (output_indirect_thunk_function): Don't generate alias for
23151         function return thunk.
23152         (ix86_code_end): Call output_indirect_thunk_function to generate
23153         function return thunks.
23154         (ix86_output_function_return): Set indirect_return_bnd_needed
23155         and indirect_return_needed instead of indirect_thunk_bnd_needed
23156         and indirect_thunk_needed.
23158 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23160         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23161         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23162         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23164 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23165             Paul Hua <paul.hua.gm@gmail.com>
23167         PR c/84852
23168         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23170 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23172         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23173         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23174         resp. SFmode cases.
23176 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23178         PR target/84711
23179         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23180         instead of GET_MODE_SIZE when comparing Units.
23182 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23184         PR target/68256
23185         * varasm.c (hash_section): Return an unchangeble hash value
23186         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23187         Return !aarch64_can_use_per_function_literal_pools_p ().
23189 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23191         PR target/84860
23192         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23193         rather than address of cmode as last argument to prepare_cmp_insn.
23195 2018-03-15  Julia Koval  <julia.koval@intel.com>
23197         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23198         F_AVX512VNNI, F_AVX512BITALG): New.
23200 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23202         PR target/83451
23203         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23204         insn for floating-point loads and stores.
23206 2018-03-14  Carl Love  <cel@us.ibm.com>
23208         * config/rs6000/rs6000-c.c: Add macro definitions for
23209         ALTIVEC_BUILTIN_VEC_PERMXOR.
23210         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23211         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23212         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23213         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23214         UNSPEC_VPERMXOR.
23215         * config/doc/extend.texi: Add prototypes for vec_permxor.
23217 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23219         PR c/84852
23220         * diagnostic-show-locus.c (class layout_point): Convert m_line
23221         from int to linenum_type.
23222         (line_span::comparator): Use linenum "compare" function when
23223         comparing line numbers.
23224         (test_line_span): New function.
23225         (layout_range::contains_point): Convert param "row" from int to
23226         linenum_type.
23227         (layout_range::intersects_line_p): Likewise.
23228         (layout::will_show_line_p): Likewise.
23229         (layout::print_source_line): Likewise.
23230         (layout::should_print_annotation_line_p): Likewise.
23231         (layout::print_annotation_line): Likewise.
23232         (layout::print_leading_fixits): Likewise.
23233         (layout::annotation_line_showed_range_p): Likewise.
23234         (struct line_corrections): Likewise for field m_row.
23235         (line_corrections::line_corrections): Likewise for param "row".
23236         (layout::print_trailing_fixits): Likewise.
23237         (layout::get_state_at_point): Likewise.
23238         (layout::get_x_bound_for_row): Likewise.
23239         (layout::print_line): Likewise.
23240         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23241         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23242         * input.c (selftest::test_linenum_comparisons): New function.
23243         (selftest::input_c_tests): Call it.
23244         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23245         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23246         * selftest.h (ASSERT_GT): New macro.
23247         (ASSERT_GT_AT): New macro.
23248         (ASSERT_LT): New macro.
23249         (ASSERT_LT_AT): New macro.
23251 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23253         PR rtl-optimization/84780
23254         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23256 2018-03-14  Martin Liska  <mliska@suse.cz>
23258         * tree.c (record_node_allocation_statistics): Use
23259         get_stats_node_kind.
23260         (get_stats_node_kind): New function extracted from
23261         record_node_allocation_statistics.
23262         (free_node): Use get_stats_node_kind.
23264 2018-03-14  Richard Biener  <rguenther@suse.de>
23266         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23267         that the value-set of ANTIC_IN doesn't grow.
23269         Revert
23270         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23271         member.
23272         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23273         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23275 2018-03-14  Julia Koval  <julia.koval@intel.com>
23277         * config.gcc (icelake-client, icelake-server): New.
23278         (icelake): Remove.
23279         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23280         (initial_ix86_arch_features): Ditto.
23281         (PTA_SKYLAKE): Add SGX.
23282         (PTA_ICELAKE): Remove.
23283         (PTA_ICELAKE_CLIENT): New.
23284         (PTA_ICELAKE_SERVER): New.
23285         (ix86_option_override_internal): Split up icelake on icelake client and
23286         icelake server.
23287         (get_builtin_code_for_version): Ditto.
23288         (fold_builtin_cpu): Ditto.
23289         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23290         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23291         * config/i386/i386.h (processor_type): Ditto.
23292         * doc/invoke.texi: Ditto.
23294 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23296         PR sanitizer/83392
23297         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23298         INTEGER_CST offset, add it together with bitpos / 8 and
23299         sign extend based on POINTER_SIZE.
23301         PR target/84844
23302         Revert
23303         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23305         PR target/78090
23306         * config/i386/constraints.md (Yc): New register constraint.
23307         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23308         Use Yc constraint for alternative 2 of operand 0.  Remove
23309         preferred_for_speed attribute.
23311 2018-03-14  Richard Biener  <rguenther@suse.de>
23313         PR tree-optimization/84830
23314         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23315         with the old one to avoid oscillations.
23317 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23319         PR target/83712
23320         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23321         pseudos.
23322         (assign_by_spills): Return a flag of reload assignment failure.
23323         Do not process the reload assignment failures.  Do not spill other
23324         reload pseudos if they has the same reg class.  Update n if
23325         necessary.
23326         (lra_assign): Add a return arg.  Set up from the result of
23327         assign_by_spills call.
23328         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23329         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23330         usage_insns if it is not NULL.
23331         (spill_hard_reg_in_range): New function.
23332         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23333         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23334         function prototypes.
23335         (lra_assign): Change prototype.
23336         * lra.c (lra): Add code to deal with fails by splitting hard reg
23337         live ranges.
23339 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23341         * config/riscv/riscv.opt (mrelax): New option.
23342         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23343         "norelax" when riscv_mrelax is disabled.
23344         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23346 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23348         PR target/84743
23349         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23350         reassociation for int modes.
23352 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23354         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23355         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23356         for big-endian.
23357         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23358         * config/aarch64/aarch64-sve.md
23359         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23360         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23361         (*extend<mode><Vwide>2): Rename to...
23362         (aarch64_sve_extend<mode><Vwide>2): ...this.
23363         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23364         renaming the old pattern to...
23365         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23366         unsigned packs.
23367         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23368         define_expand, renaming the old pattern to...
23369         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23370         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23371         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23372         account when deciding which SVE instruction the optab should use.
23373         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23375 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23377         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23378         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23379         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23380         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23381         (tlsdesc_small_<mode>): Turn a define_expand and use
23382         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23383         (tlsdesc_small_advsimd_<mode>): ...this.
23384         (tlsdesc_small_sve_<mode>): New pattern.
23386 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23388         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23389         (UNSPEC_UMUL_HIGHPART): New constants.
23390         (MUL_HIGHPART): New int iteraor.
23391         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23392         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23393         define_expand.
23394         (*<su>mul<mode>3_highpart): New define_insn.
23396 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23398         PR lto/84805
23399         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23400         incomplete types.
23402 2018-03-13  Martin Liska  <mliska@suse.cz>
23404         PR ipa/84658.
23405         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23406         vector.
23407         (sem_item_optimizer::~sem_item_optimizer): Release it.
23408         (sem_item_optimizer::merge_classes): Register variable aliases.
23409         (sem_item_optimizer::fixup_pt_set): New function.
23410         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23411         * ipa-icf.h: Declare new variables and functions.
23413 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23415         PR middle-end/84834
23416         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23417         integer_pow2p@2 and test integer_pow2p in condition.
23418         (A < 0 ? C : 0): Similarly for @1.
23420         PR middle-end/84831
23421         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23422         characters starting at p contain '\0' character, don't look beyond
23423         that.
23425         PR target/84827
23426         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23427         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23429         PR target/84828
23430         * reg-stack.c (change_stack): Change update_end var from int to
23431         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23432         also call set_block_for_insn on the newly added insns and rescan.
23434         PR target/84786
23435         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23436         on the last operand.
23438         PR c++/84704
23439         * tree.c (stabilize_reference_1): Return save_expr (e) for
23440         STATEMENT_LIST even if it doesn't have side-effects.
23442 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23444         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23446 2018-03-12  Renlin Li  <renlin.li@arm.com>
23448         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23449         aarch64_output_scalar_simd_mov_immediate.
23451 2018-03-12  Martin Sebor  <msebor@redhat.com>
23453         PR tree-optimization/83456
23454         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23455         for perfectly overlapping calls to memcpy.
23456         (gimple_fold_builtin_memory_chk): Same.
23457         (gimple_fold_builtin_strcpy): Handle no-warning.
23458         (gimple_fold_builtin_stxcpy_chk): Same.
23459         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23461 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23463         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23464         parameter.  Use it for SFmode.
23465         (rs6000_function_arg_advance_1): Adjust.
23466         (rs6000_function_arg): Adjust.
23467         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23469 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23471         PR rtl-optimization/84169
23472         PR rtl-optimization/84780
23473         * combine.c (can_combine_p): Check for a 2-insn combination whether
23474         the destination register is used between the two insns, too.
23476 2018-03-12  Richard Biener  <rguenther@suse.de>
23478         PR tree-optimization/84803
23479         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23480         for refs DR analysis didn't process.
23482 2018-03-12  Richard Biener  <rguenther@suse.de>
23484         PR tree-optimization/84777
23485         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23486         force-vectorize loops ignore whether we are optimizing for size.
23488 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23490         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23491         (TARGET_MD_ASM_ADJUST): Define.
23493 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23494             Kito Cheng  <kito.cheng@gmail.com>
23495             Chung-Ju Wu  <jasonwucj@gmail.com>
23497         * config/nds32/nds32.c (nds32_compute_stack_frame,
23498         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23499         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23500         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
23501         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
23502         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
23503         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
23504         * config/nds32/nds32.md (prologue, epilogue): Use macro
23505         NDS32_V3PUSH_AVAILABLE_P to do checking.
23507 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
23509         PR debug/58150
23510         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
23511         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
23512         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
23513         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
23514         addition of most attributes on !orig_type_die or the attribute not
23515         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
23517 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23518             Chung-Ju Wu  <jasonwucj@gmail.com>
23520         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
23521         __NDS32_VH__ macro.
23522         * config/nds32/nds32.opt (mvh): New option.
23524 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23525             Chung-Ju Wu  <jasonwucj@gmail.com>
23527         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
23528         function.
23529         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
23530         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
23531         definition.
23533 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23534             Chung-Ju Wu  <jasonwucj@gmail.com>
23536         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
23537         function.
23538         * config/nds32/nds32-multiple.md (strlensi): New pattern.
23539         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
23541 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
23542             Kito Cheng  <kito.cheng@gmail.com>
23543             Chung-Ju Wu  <jasonwucj@gmail.com>
23545         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
23546         UNSPEC_FFMISM and UNSPEC_FLMISM.
23547         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
23548         for ffb, ffmism and flmism.
23549         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
23550         (unspec_ffmism): Ditto.
23551         (unspec_flmism): Ditto.
23552         (nds32_expand_builtin_impl): Check if string extension is available.
23553         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
23554         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
23556 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
23558         Reverting patch:
23559         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23561         PR target/83712
23562         * lra-assigns.c (assign_by_spills): Return a flag of reload
23563         assignment failure.  Do not process the reload assignment
23564         failures.  Do not spill other reload pseudos if they has the same
23565         reg class.
23566         (lra_assign): Add a return arg.  Set up from the result of
23567         assign_by_spills call.
23568         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23569         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23570         usage_insns if it is not NULL.
23571         (spill_hard_reg_in_range): New function.
23572         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23573         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23574         function prototypes.
23575         (lra_assign): Change prototype.
23576         * lra.c (lra): Add code to deal with fails by splitting hard reg
23577         live ranges.
23579 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
23581         PR target/84807
23582         * config/i386/i386.opt: Replace Enforcment with Enforcement.
23584 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
23586         PR debug/84620
23587         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
23588         (dw_val_node): Add val_symbolic_view.
23589         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
23590         (symview_upper_bound): New.
23591         (new_line_info_table): Initialize symviews_since_reset.
23592         (dwarf2out_source_line): Count symviews_since_reset and set
23593         symview_upper_bound.
23594         (dw_val_equal_p): Handle symview.
23595         (add_AT_symview): New.
23596         (print_dw_val): Handle symview.
23597         (attr_checksum, attr_checksum_ordered): Likewise.
23598         (same_dw_val_p, size_of_die): Likewise.
23599         (value_format, output_die): Likewise.
23600         (add_high_low_attributes): Use add_AT_symview for entry_view.
23601         (dwarf2out_finish): Reset symview_upper_bound, clear
23602         zero_view_p.
23604 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
23606         PR target/83969
23607         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
23608         Add strict argument and use it.
23609         (rs6000_split_multireg_move): Update for new strict argument.
23610         (mem_operand_gpr): Disallow all non-offsettable addresses.
23611         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
23613 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
23615         PR target/84772
23616         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
23617         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
23618         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
23620         PR c++/84767
23621         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
23622         decl, use remap_type if we want to use the type.
23624 2018-03-09  Martin Sebor  <msebor@redhat.com>
23626         PR tree-optimization/84526
23627         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23628         Remove dead code.
23629         (builtin_access::generic_overlap): Be prepared to handle non-array
23630         base objects.
23632 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
23634         PR rtl-optimization/84682
23635         * lra-constraints.c (process_address_1): Check is_address flag
23636         for address constraints.
23637         (process_alt_operands): Likewise.
23638         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
23639         preprocess_constraints.
23640         * recog.h (preprocess_constraints): Add oploc parameter.
23641         Adjust callers.
23642         * recog.c (preprocess_constraints): Test address_operand for
23643         CT_ADDRESS constraints.
23645 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23647         PR target/83712
23648         * lra-assigns.c (assign_by_spills): Return a flag of reload
23649         assignment failure.  Do not process the reload assignment
23650         failures.  Do not spill other reload pseudos if they has the same
23651         reg class.
23652         (lra_assign): Add a return arg.  Set up from the result of
23653         assign_by_spills call.
23654         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23655         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23656         usage_insns if it is not NULL.
23657         (spill_hard_reg_in_range): New function.
23658         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23659         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23660         function prototypes.
23661         (lra_assign): Change prototype.
23662         * lra.c (lra): Add code to deal with fails by splitting hard reg
23663         live ranges.
23665 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23667         PR target/83193
23668         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
23669         Accept complain bool parameter.  Only emit errors if it is true.
23670         (arm_parse_cpu_option_name): Likewise.
23671         (arm_target_thumb_only): Adjust callers of the above.
23672         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
23673         prototype to take a default true bool parameter.
23674         (arm_parse_arch_option_name): Likewise.
23676 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
23677             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
23679         PR jit/64089
23680         PR jit/84288
23681         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
23682         * configure: Regenerate.
23683         * configure.ac ("linker --version-script option"): New.
23684         ("linker soname option"): New.
23686 2018-03-09  Richard Biener  <rguenther@suse.de>
23688         PR tree-optimization/84775
23689         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
23690         immediate uses of predicate stmts and mark them modified.
23692         Revert
23693         PR tree-optimization/84178
23694         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23695         to caller.
23696         (version_loop_for_if_conversion): Delay update_ssa call.
23697         (tree_if_conversion): Delay update_ssa until after predicate
23698         insertion.
23700 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
23702         PR target/84763
23703         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
23704         when the function accesses prior frames.
23706 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23708         PR debug/84456
23709         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
23710         gen_llsym, otherwise call maybe_gen_llsym.
23712         PR inline-asm/84742
23713         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
23714         has ',' character inside of it.
23716 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23718         PR target/84748
23719         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
23720         as clobbering CC_REGNUM.
23722 2018-03-08  Richard Biener  <rguenther@suse.de>
23724         PR middle-end/84552
23725         * tree-scalar-evolution.c: Include tree-into-ssa.h.
23726         (follow_copies_to_constant): Do not follow SSA names registered
23727         for update.
23729 2018-03-08  Richard Biener  <rguenther@suse.de>
23731         PR tree-optimization/84178
23732         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23733         to caller.
23734         (version_loop_for_if_conversion): Delay update_ssa call.
23735         (tree_if_conversion): Delay update_ssa until after predicate
23736         insertion.
23738 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
23740         PR tree-optimization/84178
23741         * tree-if-conv.c (release_bb_predicate): Remove the
23742         the assertion that the stmts have NULL use_ops.
23743         Discard the statements, asserting that they haven't
23744         yet been added to a BB.
23746 2018-03-08  Richard Biener  <rguenther@suse.de>
23748         PR tree-optimization/84746
23749         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
23750         (phi_translate): Pass in destination ANTIC_OUT set.
23751         (phi_translate_1): Likewise.  For a simplified result lookup
23752         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
23753         (phi_translate_set): Adjust.
23754         (do_pre_regular_insertion): Likewise.
23755         (do_pre_partial_partial_insertion): Likewise.
23757 2018-03-08  Martin Liska  <mliska@suse.cz>
23759         PR gcov-profile/84735
23760         * doc/gcov.texi: Document usage of profile files.
23761         * gcov-io.h: Document changes in the format.
23763 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
23765         PR debug/84404
23766         PR debug/84408
23767         * dwarf2out.c (struct dw_line_info_table): Update comments for
23768         view == -1.
23769         (FORCE_RESET_NEXT_VIEW): New.
23770         (FORCE_RESETTING_VIEW_P): New.
23771         (RESETTING_VIEW_P): Check for -1 too.
23772         (ZERO_VIEW_P): Likewise.
23773         (new_line_info_table): Force-reset next view.
23774         (dwarf2out_begin_function): Likewise.
23775         (dwarf2out_source_line): Simplify zero_view_p initialization.
23776         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
23777         view directly.  Omit view when omitting .loc at line 0.
23779 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23781         PR tree-optimization/84740
23782         * tree-switch-conversion.c (process_switch): Call build_constructors
23783         only if info.phi_count is non-zero.
23785         PR tree-optimization/84739
23786         * tree-tailcall.c (find_tail_calls): Check call arguments against
23787         DECL_ARGUMENTS (current_function_decl) rather than
23788         DECL_ARGUMENTS (func) when checking for tail recursion.
23790 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23792         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
23793         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
23794         Volker Reichelt's entry and add entries for people that perform
23795         GCC fuzzy testing and report numerous bugs.
23797 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
23799         PR target/82411
23800         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
23801         readonly data in sdata, if that is disabled.
23802         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
23803         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
23804         -mreadonly-in-sdata option.
23806 2018-03-07  Martin Sebor  <msebor@redhat.com>
23808         PR tree-optimization/84468
23809         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
23810         basic block when looking for nul assignment.
23812 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
23814         PR target/84277
23815         * except.h (output_function_exception_table): Adjust prototype.
23816         * except.c (output_function_exception_table): Remove FNNAME parameter
23817         and add SECTION parameter.  Ouput one part of the table at a time.
23818         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
23819         the first part of the exception table and emit unwind directives.
23820         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
23821         (i386_pe_seh_cold_init): Likewise.
23822         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
23823         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
23824         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
23825         (ix86_output_call_insn): Emit a nop in one more case for SEH.
23826         * config/i386/winnt.c: Include except.h.
23827         (struct seh_frame_state): Add reg_offset, after_prologue and
23828         in_cold_section fields.
23829         (i386_pe_seh_end_prologue): Set seh->after_prologue.
23830         (i386_pe_seh_cold_init): New function.
23831         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
23832         to seh->in_cold_section.
23833         (seh_emit_push): Record the offset of the push.
23834         (seh_emit_save): Record the offet of the save.
23835         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
23836         Test seh->after_prologue to disregard the epilogue.
23837         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
23838         (i386_pe_end_cold_function): New function.
23840 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23842         PR fortran/84565
23843         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
23844         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
23846         PR c++/84704
23847         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
23848         on tmp_var.
23849         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
23850         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
23852         PR middle-end/84723
23853         * multiple_target.c: Include tree-inline.h and intl.h.
23854         (expand_target_clones): Diagnose and fail if node->definition and
23855         !tree_versionable_function_p (node->decl).
23857 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
23859         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
23860         sprint_ul.
23861         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
23862         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
23863         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
23865 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23867         PR target/84710
23868         * combine.c (try_combine): Use reg_or_subregno instead of handling
23869         just paradoxical SUBREGs and REGs.
23871 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
23873         * config/arc/arc.c (arc_finalize_pic): Remove function.
23874         (arc_must_save_register): We use single base PIC register, remove
23875         checks to save/restore the PIC register.
23876         (arc_expand_prologue): Likewise.
23877         * config/arc/arc-protos.h (arc_set_default_type_attributes):
23878         Remove.
23879         (arc_verify_short): Likewise.
23880         (arc_attr_type): Likewise.
23881         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
23882         (walk_stores): Likewise.
23883         (arc_address_cost): Make it static.
23884         (arc_verify_short): Likewise.
23885         (branch_dest): Likewise.
23886         (arc_attr_type): Likewise.
23887         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
23888         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
23889         (arc_final_prescan_insn): Remove inserting the nops due to
23890         hardware hazards.  It is done in reorg step.
23891         (insn_length_variant_t): Remove.
23892         (insn_length_parameters_t): Likewise.
23893         (arc_insn_length_parameters): Likewise.
23894         (arc_get_insn_variants): Likewise.
23895         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
23897 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23899         PR inline-asm/84683
23900         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
23901         assertion failure.
23903         PR tree-optimization/84687
23904         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
23905         on new_node->decl.
23906         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
23908 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23910         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
23911         Rename to ppc_speculation_barrier.
23912         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
23913         __builtin_ppc_speculation_barrier.
23915 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23917         PR target/84700
23918         * combine.c (combine_simplify_rtx): Don't try to simplify if
23919         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
23920         are equal to x.
23922 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
23924         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
23925         to 32 bytes when compiling for POWER9.
23927 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23929         PR target/84564
23930         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
23931         regparm >= 3 with no arg reg available also for calls with
23932         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
23934         PR target/84524
23935         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
23936         orig,vex.
23937         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
23939 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
23941         PR target/84264
23942         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
23944 2018-03-05  Richard Biener  <rguenther@suse.de>
23946         PR tree-optimization/84486
23947         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
23948         When inserting a __builtin_assume_aligned call set the LHS
23949         SSA name alignment info accordingly.
23951 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
23953         PR tree-optimization/84114
23954         * config/aarch64/aarch64.c (aarch64_reassociation_width)
23955         Avoid reassociation of FLOAT_MODE addition.
23957 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
23959         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
23960         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
23961         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
23962         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
23963         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
23964         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
23965         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
23966         and -mwbnoinvd.
23967         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
23968         __builtin_ia32_wbinvd): New builtins.
23969         (SPECIAL_ARGS2): New.
23970         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
23971         (SPECIAL_ARGS2): New.
23972         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
23973         (ix86_valid_target_attribute_inner_p): Ditto.
23974         (ix86_init_mmx_sse_builtins): Add special_args2.
23975         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
23976         TARGET_WBNOINVD_P): New.
23977         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
23978         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
23979         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
23980         * config/i386/immintrin.h (_wbinvd): New intrinsic.
23981         * config/i386/pconfigintrin.h: New file.
23982         * config/i386/wbnoinvdintrin.h: Ditto.
23983         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
23984         wbnoinvdintrin.h.
23985         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
23987 2018-03-05  Richard Biener  <rguenther@suse.de>
23989         PR tree-optimization/84670
23990         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23991         member.
23992         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23993         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23994         (compute_antic_aux): Only assert the number of values in ANTIC_IN
23995         doesn't grow if all successors (recursively) were visited at least
23996         once.
23998 2018-03-05  Richard Biener  <rguenther@suse.de>
24000         PR tree-optimization/84650
24001         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24002         if executed in the loop pipeline.
24004 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24006         * doc/configfiles.texi (Configuration Files): Move info about
24007         conditionalizing $target-protos.h to...
24008         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24009         differs from $target-protos.h.
24011 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24012             Chung-Ju Wu  <jasonwucj@gmail.com>
24014         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24015         * config/nds32/nds32-multiple.md (setmemsi): Define.
24016         * config/nds32/nds32-memory-manipulation.c
24017         (nds32_gen_dup_4_byte_to_word_value): New.
24018         (emit_setmem_word_loop): New.
24019         (emit_setmem_byte_loop): New.
24020         (nds32_expand_setmem_loop): New.
24021         (nds32_expand_setmem_loop_v3m): New.
24022         (nds32_expand_setmem_unroll): New.
24023         (nds32_expand_setmem): New.
24025 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24026             Chung-Ju Wu  <jasonwucj@gmail.com>
24028         * config/nds32/nds32-memory-manipulation.c
24029         (nds32_emit_load_store): New.
24030         (nds32_emit_post_inc_load_store): New.
24031         (nds32_emit_mem_move): New.
24032         (nds32_emit_mem_move_block): New.
24033         (nds32_expand_movmemsi_loop_unknown_size): New.
24034         (nds32_expand_movmemsi_loop_known_size): New.
24035         (nds32_expand_movmemsi_loop): New.
24036         (nds32_expand_movmemsi_unroll): New.
24037         (nds32_expand_movmemqi): Rename ...
24038         (nds32_expand_movmemsi): ... to this.
24039         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24040         (movmemsi): ... to this.
24041         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24042         (nds32_expand_movmemsi): ... to this.
24044 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24045             Monk Chiang  <sh.chiang04@gmail.com>
24046             Chung-Ju Wu  <jasonwucj@gmail.com>
24048         * config/nds32/nds32-protos.h
24049         (nds32_expand_load_multiple): New arguments.
24050         (nds32_expand_store_multiple): Ditto.
24051         (nds32_valid_multiple_load_store): Rename ...
24052         (nds32_valid_multiple_load_store_p): ... to this.
24053         * config/nds32/nds32-memory-manipulation.c
24054         (nds32_expand_load_multiple): Refine implementation.
24055         (nds32_expand_store_multiple): Ditto.
24056         * config/nds32/nds32-multiple.md
24057         (load_multiple): Update nds32_expand_load_multiple interface.
24058         (store_multiple): Update nds32_expand_store_multiple interface.
24059         * config/nds32/nds32-predicates.c
24060         (nds32_valid_multiple_load_store): Rename ...
24061         (nds32_valid_multiple_load_store_p): ... to this and refine
24062         implementation.
24063         * config/nds32/predicates.md
24064         (nds32_load_multiple_and_update_address_operation): New predicate.
24065         (nds32_store_multiple_and_update_address_operation): New predicate.
24067 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24068             Chung-Ju Wu  <jasonwucj@gmail.com>
24070         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24071         (combo): New attribute.
24072         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24074 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24076         * config/nds32/nds32.opt: Change -mcmodel= default value.
24078 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24079             Monk Chiang  <sh.chiang04@gmail.com>
24080             Chung-Ju Wu  <jasonwucj@gmail.com>
24082         * config/nds32/constants.md (unspec_element): New enum.
24083         * config/nds32/constraints.md (Umw): New constraint.
24084         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24085         * config/nds32/nds32-intrinsic.md: Likewise.
24086         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24087         (nds32_valid_smw_lwm_base_p): New.
24088         (nds32_output_smw_single_word): New.
24089         (nds32_output_lmw_single_word): New.
24090         (nds32_expand_unaligned_load): New.
24091         (nds32_expand_unaligned_store): New.
24092         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24093         (nds32_output_smw_single_word): Declare.
24094         (nds32_output_lmw_single_word): Declare.
24095         (nds32_expand_unaligned_load): Declare.
24096         (nds32_expand_unaligned_store): Declare.
24097         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24098         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24099         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24100         NDS32_BUILTIN_UASTORE_DW.
24101         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24102         predicate.
24104 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24105             Kito Cheng  <kito.cheng@gmail.com>
24106             Chung-Ju Wu  <jasonwucj@gmail.com>
24108         * config/nds32/nds32-intrinsic.c
24109         (nds32_expand_builtin_null_ftype_reg): Delete.
24110         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24111         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24112         (nds32_read_argument): New.
24113         (nds32_legitimize_target): Ditto.
24114         (nds32_legitimize_argument): Ditto.
24115         (nds32_check_constant_argument): Ditto.
24116         (nds32_expand_unop_builtin): Ditto.
24117         (nds32_expand_unopimm_builtin): Ditto.
24118         (nds32_expand_binop_builtin): Ditto.
24119         (nds32_builtin_decl_impl): Ditto.
24120         (builtin_description): Ditto.
24121         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24122         (nds32_init_builtins_impl): Ditto.
24123         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24124         (nds32_builtin_decl): New.
24125         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24126         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24128 2018-03-02  Jeff Law  <law@redhat.com>
24130         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24131         (redundant_insn, fill_simple_delay_slots): Likewise.
24132         (fill_slots_from_thread): Likewise.
24133         * resource.c (mark_referenced_resources): Likewise.
24134         (mark_set_resources, find_dead_or_set_registers): Likewise.
24136 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24138         * substring-locations.h (format_warning_va): Formatting fix for
24139         ATTRIBUTE_GCC_DIAG.
24140         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24141         argument.
24142         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24143         * substring-locations.c: Include intl.h.
24144         (format_warning_va): Turned into small wrapper around
24145         format_warning_n_va, renamed to ...
24146         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24147         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24148         use ngettext.
24149         (format_warning_at_substring_n): New function.
24150         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24151         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24152         format_warning_at_substring with just a shorter name instead of
24153         const function pointer.
24154         (fmtwarn_n): New function.
24155         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24156         appropriate, get rid of all the fmtstr temporaries, move conditionals
24157         with G_() wrapped string literals directly into fmtwarn arguments,
24158         cast dir.len to (int), formatting fixes.
24160 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24162         * doc/invoke.texi: Remove "Cilk Plus" references.
24164 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24165             Richard Biener  <rguenther@suse.de>
24167         PR ipa/84628
24168         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24169         for error or warning attributes if CALL_FROM_THUNK_P is set.
24170         Formatting fixes.
24172 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24174         PR target/56540
24175         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24176         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24178         PR target/56540
24179         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24180         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24182         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24183         instead of -1U in last predictors element's probability member.
24185 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24187         PR ipa/83983
24188         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24189         arguments if they are comparable.
24191 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24193         PR tree-optimization/84634
24194         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24195         masks and masked_loop_p with a single loop_masks, making sure it's
24196         null for bb vectorization.
24198 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24200         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24201         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24202         loop->force_vectorize to check whether there is no alias.
24204 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24206         PR target/84614
24207         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24208         prototypes.
24209         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24210         comments.
24211         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24212         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24213         instead of a loop around prev_real_insn.
24214         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24215         prev_real_insn.
24217         PR inline-asm/84625
24218         * config/i386/i386.c (ix86_print_operand): Use conditional
24219         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24220         zero vector.
24222 2018-03-02  Richard Biener  <rguenther@suse.de>
24224         PR tree-optimization/84427
24225         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24226         (bitmap_set_subtract_values): Rewrite to handle multiple
24227         exprs per value.
24228         (clean): Likewise.
24229         (prune_clobbered_mems): Likewise.
24230         (phi_translate): Take edge instead of pred/phiblock.
24231         (phi_translate_1): Likewise.
24232         (phi_translate_set): Likewise.  Insert all translated
24233         exprs for a value into the set, keeping possibly multiple
24234         expressions per value.
24235         (compute_antic_aux): Adjust for phi_translate changes.
24236         When intersecting union the expressions and prune those
24237         not in the final value set, keeping possibly multiple
24238         expressions per value.  Do not use value-insertion
24239         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24240         all expressions.  Add verification that the value-sets
24241         only shrink during iteration.
24242         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24243         (do_pre_regular_insertion): Likewise.
24244         (do_pre_partial_partial_insertion): Likewise.
24246 2018-03-02  Richard Biener  <rguenther@suse.de>
24248         PR target/82005
24249         * config/darwin.c (saved_debug_info_level): New static global.
24250         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24251         (darwin_asm_lto_end): Restore debug info generation settings.
24253 2018-03-01  Martin Liska  <mliska@suse.cz>
24255         PR sanitizer/82484
24256         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24257         volatile arguments.
24259 2018-03-01  Richard Biener  <rguenther@suse.de>
24261         PR debug/84645
24262         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24263         type annotation with LTO when debug was disabled at compile-time.
24265 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24267         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24268         XINT with INTVAL.
24269         (mips_final_postscan_insn): Likewise.
24271 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24273         PR rtl-optimization/84528
24274         * alias.c (init_alias_target): Add commentary.
24275         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24276         a unique base value if the frame pointer is not eliminated
24277         to the stack pointer.
24279 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24281         PR rtl-optimization/83327
24282         * lra-int.h (hard_regs_spilled_into): Declare.
24283         * lra.c (hard_regs_spilled_into): Define.
24284         (init_reg_info): Init hard_regs_spilled_into.
24285         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24286         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24287         (process_bb_lives): Handle hard_regs_spilled_into.
24288         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24289         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24291 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24293         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24294         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24295         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24296         * config/rs6000/aix72.h: New file.
24298 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24300         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24301         instead of warning_at with conditional singular and plural messages
24302         where possible.
24304         PR target/52991
24305         * stor-layout.c (update_alignment_for_field): For
24306         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24307         && !DECL_PACKED (field), do the alignment update, just use
24308         only desired_align instead of MAX (type_align, desired_align)
24309         as the alignment.
24310         (place_field): Don't do known_align < desired_align handling
24311         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24312         is non-NULL, instead do it after rli->prev_field handling and
24313         only if not within a bitfield word.  For DECL_PACKED (field)
24314         use type_align of BITS_PER_UNIT.
24316 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24318         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24319         superfluous parentheses and trailing spaces.
24321 2018-02-28  Richard Biener  <rguenther@suse.de>
24323         PR tree-optimization/84584
24324         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24325         SCoPs with fake exit edge.
24327 2018-02-28  Martin Liska  <mliska@suse.cz>
24329         PR testsuite/84597
24330         * timevar.c (timer::print): Fix format to properly print 100%
24331         values.
24333 2018-02-28  Richard Biener  <rguenther@suse.de>
24335         PR middle-end/84607
24336         * genmatch.c (capture_info::walk_match): Do not mark
24337         captured expressions without operands as expr_p given
24338         they act more like predicates and should be subject to
24339         "lost tail" side-effect preserving.
24341 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24343         PR rtl-optimization/81611
24344         * auto-inc-dec.c (attempt_change): Move dead note from
24345         mem_insn if it's the next use of regno
24346         (find_address): Take address use of reg holding
24347         non-incremented value.  Add parm to limit search to the named
24348         reg only.
24349         (merge_in_block): Attempt to use a mem insn that is the next
24350         use of the original regno.
24352 2018-02-27  Martin Sebor  <msebor@redhat.com>
24354         PR c++/83871
24355         * doc/invoke.texi (-Wmissing-attributes): New option.
24356         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24358 2018-02-27  Martin Sebor  <msebor@redhat.com>
24360         PR translation/84207
24361         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24362         n argument to unsigned HOST_WIDE_INT.
24363         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24364         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24365         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24366         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24368 2018-02-27  Richard Biener  <rguenther@suse.de>
24370         PR tree-optimization/84512
24371         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24372         Do not use the estimate returned from record_stmt_cost for
24373         the scalar iteration cost but sum properly using add_stmt_cost.
24375 2018-02-27  Richard Biener  <rguenther@suse.de>
24377         PR tree-optimization/84466
24378         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24379         Adjust last change to less strictly validate use operands.
24381 2018-02-27  Martin Liska  <mliska@suse.cz>
24383         PR gcov-profile/84548
24384         * gcov.c (process_file): Allow partial overlap and consider it
24385         also as group functions.
24386         (output_lines): Properly calculate range of lines for a group.
24388 2018-02-27  Martin Liska  <mliska@suse.cz>
24390         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24391         'ggc' suffixes.  Change first column width.
24392         (timer::print): Fix formatting of the column.
24394 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24396         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24397         preserve inline entry blocks for the sake of debug inline
24398         entry point markers alone.
24399         (remove_unused_locals): Suggest in comments a better place to
24400         force the preservation of inline entry blocks that are
24401         otherwise unused, but do not preserve them.
24403 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24405         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24407 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24409         PR target/84039
24410         * config/i386/constraints.md (Bs): Replace
24411         ix86_indirect_branch_register with
24412         TARGET_INDIRECT_BRANCH_REGISTER.
24413         (Bw): Likewise.
24414         * config/i386/i386.md (indirect_jump): Likewise.
24415         (tablejump): Likewise.
24416         (*sibcall_memory): Likewise.
24417         (*sibcall_value_memory): Likewise.
24418         Peepholes of indirect call and jump via memory: Likewise.
24419         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24420         (*sibcall_value_GOT_32): Likewise.
24421         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24422         (GOT_memory_operand): Likewise.
24423         (call_insn_operand): Likewise.
24424         (sibcall_insn_operand): Likewise.
24425         (GOT32_symbol_operand): Likewise.
24426         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24428 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24430         PR rtl-optimization/83496
24431         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24432         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24433         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24434         redundant insn, if any.
24435         (relax_delay_slots): Likewise.
24436         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24438 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24440         PR tree-optimization/83965
24441         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24442         that grouped statements are part of a reduction chain.  Return
24443         true if the statement is not marked as a reduction itself but
24444         is part of a group.
24445         (vect_recog_dot_prod_pattern): Don't check whether the statement
24446         is part of a group here.
24447         (vect_recog_sad_pattern): Likewise.
24448         (vect_recog_widen_sum_pattern): Likewise.
24450 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24452         PR debug/84545
24453         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24455 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24457         PR target/84530
24458         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24459         the bool argument.
24460         (ix86_output_indirect_function_return): New prototype.
24461         (ix86_split_simple_return_pop_internal): Likewise.
24462         * config/i386/i386.c (indirect_return_via_cx): New.
24463         (indirect_return_via_cx_bnd): Likewise.
24464         (indirect_thunk_name): Handle return va CX_REG.
24465         (output_indirect_thunk_function): Create alias for
24466         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24467         (ix86_output_indirect_jmp): Remove the bool argument.
24468         (ix86_output_indirect_function_return): New function.
24469         (ix86_split_simple_return_pop_internal): Likewise.
24470         * config/i386/i386.md (*indirect_jump): Don't pass false
24471         to ix86_output_indirect_jmp.
24472         (*tablejump_1): Likewise.
24473         (simple_return_pop_internal): Change it to define_insn_and_split.
24474         Call ix86_split_simple_return_pop_internal to split it for
24475         -mfunction-return=.
24476         (simple_return_indirect_internal): Call
24477         ix86_output_indirect_function_return instead of
24478         ix86_output_indirect_jmp.
24480 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24482         PR bootstrap/84405
24483         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24484         memset and value initialization afterwards.
24486 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24488         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24490 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24492         PR target/84521
24493         * common/config/aarch64/aarch64-common.c
24494         (aarch_option_optimization_table[]): Switch
24495         off fomit-frame-pointer
24497 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24498             Chung-Ju Wu  <jasonwucj@gmail.com>
24500         * config/nds32/nds32-multiple.md (load_multiple): Disallow
24501         volatile memory.
24502         (store_multiple): Ditto.
24504 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24506         * config.gcc: Add --with-cpu support for nds32 target.
24507         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
24508         * config/nds32/nds32.opt: Add -mcpu= option.
24510 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
24512         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
24513         isel=yes): Warn for these deprecated options.
24515 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
24517         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
24518         ISA_2_5_MASKS_EMBEDDED.
24520 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
24522         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
24523         p->max as pointers rather than using iterative_hash_expr.
24525 2018-02-23  Carl Love  <cel@us.ibm.com>
24527         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
24528         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
24529         BU_P8V_OVERLOAD_2.
24530         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
24531         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
24532         P8V_BUILTIN_VEC_VUNSIGNED2.
24534 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
24536         PR target/81572
24537         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
24538         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
24539         LRA_UNKNOWN_ALT.
24540         * lra-constraints.c (curr_insn_transform): Set up
24541         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
24542         LRA_UNKNOWN_ALT.
24543         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
24544         * lra-eliminations.c (spill_pseudos): Ditto.
24545         (process_insn_for_elimination): Ditto.
24546         * lra-lives.c (reg_early_clobber_p): Use the new macros.
24547         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
24548         LRA_NON_CLOBBERED_ALT.
24550 2018-02-22  Martin Sebor  <msebor@redhat.com>
24552         PR tree-optimization/84480
24553         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
24554         to maybe_diag_stxncpy_trunc.  Call it.
24555         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
24556         from gimple_fold_builtin_strcpy.  Print inlining stack.
24557         (handle_builtin_stxncpy): Print inlining stack.
24558         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
24560 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
24562         PR target/84176
24563         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
24564         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
24565         and -fcheck-pointer-bounds are used together.
24566         (indirect_thunk_prefix): New enum.
24567         (indirect_thunk_need_prefix): New function.
24568         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
24569         "_nt" instead of "_bnd" for NOTRACK prefix.
24570         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
24571         (output_indirect_thunk_function): Likewise.
24572         (): Likewise.
24573         (ix86_code_end): Update output_indirect_thunk_function calls.
24574         (ix86_output_indirect_branch_via_reg): Replace
24575         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
24576         (ix86_output_indirect_branch_via_push): Likewise.
24577         (ix86_output_function_return): Likewise.
24578         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
24579         incompatible with -fcf-protection=branch and
24580         -fcheck-pointer-bounds.
24582 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24584         PR target/83335
24585         * config/aarch64/aarch64.c (aarch64_print_address_internal):
24586         Change gcc_assert call to output_operand_lossage.
24588 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24590         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
24592 2018-02-22  DJ Delorie  <dj@redhat.com>
24593             Sebastian Perta  <sebastian.perta@renesas.com>
24594             Oleg Endo  <olegendo@gcc.gnu.org>
24596         * config/rx/rx.c (rx_rtx_costs): New function.
24597         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
24599 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24601         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
24603 2018-02-22  Martin Liska  <mliska@suse.cz>
24605         PR driver/83193
24606         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
24607         Add "native" as a possible value.
24609 2018-02-22  Martin Liska  <mliska@suse.cz>
24611         PR driver/83193
24612         * config/i386/i386.c (ix86_option_override_internal):
24613         Add "native" as a possible value for -march and -mtune.
24615 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
24617         PR target/84502
24618         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
24619         to all type variants.
24621         PR tree-optimization/84503
24622         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
24623         width as info->bitpos + info->bitsize - start.
24624         (merged_store_group::merge_overlapping): Simplify width computation.
24625         (check_no_overlap): New function.
24626         (imm_store_chain_info::try_coalesce_bswap): Compute expected
24627         start + width and last_order of the group, fail if check_no_overlap
24628         fails.
24629         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
24630         to group if check_no_overlap fails.
24632 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24634         * config/rs6000/altivec.md: Delete contraint arguments to
24635         define_expand, define_split, and define_peephole2, and in
24636         define_insn_and_split if always unused.
24637         * config/rs6000/darwin.md: Ditto.
24638         * config/rs6000/dfp.md: Ditto.
24639         * config/rs6000/rs6000.md: Ditto.
24640         * config/rs6000/sync.md: Ditto.
24641         * config/rs6000/vector.md: Ditto.
24642         * config/rs6000/vsx.md: Ditto.
24644 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24646         * config/rs6000/altivec.md: Write output control strings as braced
24647         blocks instead of double-quoted strings.
24648         * config/rs6000/darwin.md: Ditto.
24649         * config/rs6000/rs6000.md: Ditto.
24650         * config/rs6000/vector.md: Ditto.
24651         * config/rs6000/vsx.md: Ditto.
24653 2018-02-21  Jason Merrill  <jason@redhat.com>
24655         PR c++/84314 - ICE with templates and fastcall attribute.
24656         * attribs.c (build_type_attribute_qual_variant): Remove assert.
24658 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24660         * ipa-cp.c (determine_versionability): Fix comment typos.
24662 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24664         PR c/84229
24665         * ipa-cp.c (determine_versionability): Do not version functions caling
24666         va_arg_pack.
24668 2018-02-21  Martin Liska  <mliska@suse.cz>
24670         PR driver/83193
24671         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
24672         Add "native" as a possible value.
24673         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
24674         the macro when native cpu detection is available.
24676 2018-02-21  Martin Liska  <mliska@suse.cz>
24678         PR driver/83193
24679         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
24680         Add "native" as a possible value.
24681         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
24682         when native cpu detection is available.
24684 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
24685             Martin Sebor  <msebor@redhat.com>
24687         PR tree-optimization/84478
24688         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
24689         false.
24690         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
24691         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
24692         support which is conservatively correct, for 2 only stay conservative
24693         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
24694         argument to the 2 argument get_range_strlen, adjust 6 arg
24695         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
24696         false.
24697         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
24698         (gimple_fold_builtin_strlen): Pass true as last argument to
24699         get_range_strlen.
24701 2018-02-20  Martin Sebor  <msebor@redhat.com>
24703         PR middle-end/84095
24704         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
24705         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
24706         (builtin_memref::builtin_memref): Factor out parts into
24707         set_base_and_offset and call it.
24709 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24711         PR middle-end/84406
24712         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
24713         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
24714         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
24715         search at the associated MODE_INT.
24717 2018-02-20  Jeff Law  <law@redhat.com>
24719         PR middle-end/82123
24720         PR tree-optimization/81592
24721         PR middle-end/79257
24722         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
24723         for range data rather than using global data.
24724         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
24725         range data rather than using global data.
24726         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
24727         pass it to children as needed.
24728         (struct directive::fmtresult): Similarly.
24729         (struct directive::set_width): Similarly.
24730         (struct directive::set_precision): Similarly.
24731         (format_integer, format_directive, parse_directive): Similarly.
24732         (format_none): Accept unnamed vr_values parameter.
24733         (format_percent, format_floating, format_character): Similarly.
24734         (format_string, format_plain): Similarly.
24735         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
24736         the EVRP range analyzer for range data rather than using global data.
24737         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
24738         gimple-ssa-evrp-analyze.h
24739         (class sprintf_dom_walker): Add after_dom_children member function.
24740         Add evrp_range_analyzer member.
24741         (sprintf_dom_walker::before_dom_children): Call into the EVRP
24742         range analyzer as needed.
24743         (sprintf_dom_walker::after_dom_children): New member function.
24744         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
24745         if not optimizing.
24746         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
24747         (evrp_range_analyzer::pop_to_marker): Likewise.
24749 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24751         PR tree-optimization/84419
24752         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
24753         with the required type if its current type is compatible but
24754         different.
24756 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
24758         PR middle-end/82004
24759         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
24760         after vectorization.
24762 2018-02-20  Martin Liska  <mliska@suse.cz>
24764         PR driver/83193
24765         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
24766         possible values if we don't have a hint.
24768 2018-02-20  Martin Liska  <mliska@suse.cz>
24770         PR c/84310
24771         PR target/79747
24772         * final.c (shorten_branches): Build align_tab array with one
24773         more element.
24774         * opts.c (finish_options): Add alignment option limit check.
24775         (MAX_CODE_ALIGN): Likewise.
24776         (MAX_CODE_ALIGN_VALUE): Likewise.
24777         * doc/invoke.texi: Document maximum allowed option value for
24778         all -falign-* options.
24780 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
24782         PR target/84146
24783         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
24784         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
24785         * var-tracking.c (emit_note_insn_var_location): Remove all references
24786         to NOTE_INSN_CALL_ARG_LOCATION.
24787         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
24788         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
24789         Use copy_rtx_if_shared.
24790         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
24791         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
24792         (dwarf2out_var_location): Remove handling of
24793         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
24794         on call_insn.
24795         * final.c (final_scan_insn): Remove all references to
24796         NOTE_INSN_CALL_ARG_LOCATION.
24797         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
24798         before dumping final insns.
24799         * except.c (emit_note_eh_region_end): Remove all references to
24800         NOTE_INSN_CALL_ARG_LOCATION.
24801         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
24802         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
24803         * config/arc/arc.c (hwloop_optimize): Likewise.
24804         * config/arm/arm.c (create_fix_barrier): Likewise.
24805         * config/s390/s390.c (s390_chunkify_start): Likewise.
24806         * config/sh/sh.c (find_barrier): Likewise.
24807         * config/i386/i386.c (rest_of_insert_endbranch,
24808         ix86_seh_fixup_eh_fallthru): Likewise.
24809         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
24810         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
24811         * config/frv/frv.c (frv_function_prologue): Likewise.
24812         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
24813         reg note.
24814         (note_outside_basic_block_p): Remove all references to
24815         NOTE_INSN_CALL_ARG_LOCATION.
24816         * gengtype.c (adjust_field_rtx_def): Likewise.
24817         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
24818         Likewise.
24819         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
24820         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
24822         PR c++/84444
24823         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
24824         is ADDR_EXPR.
24826         PR tree-optimization/84452
24827         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
24828         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
24829         is NULL.
24831 2018-02-19  Martin Liska  <mliska@suse.cz>
24833         PR sanitizer/82183
24834         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
24836 2018-02-19  Martin Liska  <mliska@suse.cz>
24837             Richard Sandiford  <richard.sandiford@linaro.org>
24839         PR tree-optimization/82491
24840         * gimple-fold.c (get_base_constructor): Make earlier bail out
24841         to prevent ubsan.
24843 2018-02-19  Carl Love  <cel@us.ibm.com>
24845         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
24846         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
24847         BU_P8V_OVERLOAD_1.
24848         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
24849         P8V_BUILTIN_VEC_NEG.
24851 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
24853         * config/rl78/rl78.md (movdf): New define expand.
24855 2018-02-19  Martin Liska  <mliska@suse.cz>
24857         PR other/80589
24858         * doc/invoke.texi: Fix typo.
24859         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
24861 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
24863         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
24864         handle rs6000_single_float and rs6000_double_float specially for
24865         e500 family CPUs.
24867 2018-02-16  Jeff Law  <law@redhat.com>
24869         * config/rx/rx.c (add_pop_cfi_notes): New function.;
24870         (pop_regs): Use it.
24872 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24874         PR ipa/84425
24875         * ipa-inline.c (inline_small_functions): Fix a typo.
24877 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24879         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
24881 2018-02-16  Carl Love  <cel@us.ibm.com>
24883         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
24884         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
24885         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
24886         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
24887         expansion to P8V_BUILTIN_VEC_FLOAT2.
24889 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
24891         PR rtl-optimization/70023
24892         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
24893         src_regno into account.
24895 2018-02-16  Carl Love  <cel@us.ibm.com>
24897         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
24898         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
24899         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
24900         * config/rs6000/rs6000.c: Remove case statements for
24901         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
24902         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
24903         and P9V_BUILTIN_VEC_VINSERT4B.
24904         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
24905         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
24906         * config/rs6000/vsx.md:
24907         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
24908         vec_insert4b.
24910 2018-02-16  Carl Love  <cel@us.ibm.com>
24912         * config/rs6000/altivec.h: Add builtin names vec_extract4b
24913         vec_insert4b.
24914         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
24915         definitions.
24916         * config/rs6000/rs6000-c.c: Add the definitions for
24917         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
24918         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
24919         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
24920         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
24921         definition for insert4b and define insn *insert3b_internal.
24922         * doc/extend.texi: Add documentation for vec_extract4b.
24924 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24926         * doc/extend.texi (Backwards Compatibility): Mention friend
24927         injection.  Note for-scope is deprecated.
24928         * doc/invoke.texi (-ffriend-injection): Deprecate.
24930 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
24932         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
24933         that moved to I2, also allow destinations that are a paradoxical
24934         subreg (instead of a normal reg).
24936 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
24938         PR target/83831
24939         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
24940         to QImode.
24942 2018-02-16  Richard Biener  <rguenther@suse.de>
24944         PR tree-optimization/84037
24945         PR tree-optimization/84016
24946         PR target/82862
24947         * config/i386/i386.c (ix86_builtin_vectorization_cost):
24948         Adjust vec_construct for the fact we need additional higher latency
24949         128bit inserts for AVX256 and AVX512 vector builds.
24950         (ix86_add_stmt_cost): Scale vector construction cost for
24951         elementwise loads.
24953 2018-02-16  Richard Biener  <rguenther@suse.de>
24955         PR tree-optimization/84417
24956         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
24957         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
24958         (non_rewritable_lvalue_p): Likewise, use poly-ints.
24960 2018-02-16  Martin Liska  <mliska@suse.cz>
24962         PR sanitizer/84307
24963         * internal-fn.def (ASAN_CHECK): Set proper flags.
24964         (ASAN_MARK): Likewise.
24966 2018-02-16  Julia Koval  <julia.koval@intel.com>
24968         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
24969         from PTA_CANNONLAKE.
24971 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24973         PR target/84272
24974         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
24975         Use ++iter rather than iter++ for std::list iterators.
24976         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
24977         defer deleting them until all nodes in the forest are processed.  Do
24978         free even leaf nodes.  Change to_process into auto_vec.
24980         PR bootstrap/84405
24981         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
24982         * vec.h (vec_default_construct): Use memset instead of placement new
24983         if BROKEN_VALUE_INITIALIZATION is defined.
24984         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
24985         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
24986         is defined.
24988         PR rtl-optimization/83723
24989         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
24990         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
24991         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
24992         recursive calls.
24993         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
24994         callers.
24995         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
24997 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
24999         PR rtl-optimization/81443
25000         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25001         from inner REGs to paradoxical SUBREGs.
25003 2018-02-16  Richard Biener  <rguenther@suse.de>
25005         PR tree-optimization/84399
25006         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25007         For operands we can analyze at their definition make sure we can
25008         analyze them at each use as well.
25010 2018-02-16  Richard Biener  <rguenther@suse.de>
25012         PR tree-optimization/84190
25013         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25014         volatile accesses if the decl isn't volatile.
25016 2018-02-15  Jason Merrill  <jason@redhat.com>
25018         PR c++/84314 - ICE with templates and fastcall attribute.
25019         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25020         TYPE_CANONICAL on an existing type.
25022 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25024         PR tree-optimization/84383
25025         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25026         dstoff nor call operand_equal_p if dstbase is NULL.
25028         PR tree-optimization/84334
25029         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25030         also a CONSTANT_CLASS_P, punt.
25032 2018-02-14  Jim Wilson  <jimw@sifive.com>
25034         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25035         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25036         to where locals are set.  Add TARGET_RVC support.
25037         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25039 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25041         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25043 2018-02-14  Martin Sebor  <msebor@redhat.com>
25045         PR tree-optimization/83698
25046         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25047         arrays constrain the offset range to their bounds.
25048         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25049         (builtin_access::overlap): Avoid setting the size of overlap if it's
25050         already been set.
25051         (maybe_diag_overlap): Also consider arrays when deciding what values
25052         of offsets to include in diagnostics.
25054 2018-02-14  Martin Sebor  <msebor@redhat.com>
25056         PR c/84108
25057         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25058         that correspond to the kind of a declaration.
25060 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25062         PR target/83984
25063         * config/pa/pa.md: Load address of PIC label using the linkage table
25064         if the label is nonlocal.
25066 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25068         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25069         warning message if user requests -maltivec=be.
25070         * doc/invoke.texi: Document deprecation of -maltivec=be.
25072 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25074         PR target/84220
25075         * config/rs6000/rs6000-c.c: Update definitions for
25076         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25077         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25079 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25081         PR target/84239
25082         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25083         add _get_ssp intrinsics. Remove argument from
25084         __builtin_ia32_rdssp[d|q].
25085         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25086         * config/i386/i386-builtin.def: Remove argument from
25087         __builtin_ia32_rdssp[d|q].
25088         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25089         ix86_expand_special_args_builtin for _rdssp[d|q].
25090         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25091         Clear register before usage.
25092         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25093         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25095 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25097         PR tree-optimization/84357
25098         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25099         operand 1 of an ARRAY_REF too.
25101 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25103         PR target/83831
25104         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25105         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25106         declarations.
25107         (set_of_reg): New struct.
25108         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25109         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25110         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25111         functions.
25112         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25113         Split into bitclr, bitset, bitinvert patterns if appropriate.
25114         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25115         use rx_fuse_in_memory_bitop.
25116         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25117         to named insn, correct maximum insn length.
25119 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25121         PR target/79242
25122         * machmode.def: Define a complex mode for PARTIAL_INT.
25123         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25124         MODE_PARTIAL_INT.
25125         * doc/rtl.texi: Document CSPImode.
25126         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25127         handling.
25128         (msp430_hard_regno_nregs_with_padding): Likewise.
25130 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25132         PR target/84279
25133         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25135 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25137         PR rtl-optimization/84169
25138         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25139         we generated a parallel as new i3 and we split that to new i2 and i3
25140         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25141         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25142         those to i2, not i1.  Partially rewrite this scan code.
25144 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25146         PR c/82210
25147         * stor-layout.c (place_field): For variable length fields, adjust
25148         offset_align afterwards not just based on the field's alignment,
25149         but also on the size.
25151         PR middle-end/84309
25152         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25153         of exps and logs in the use_exp2 case.
25155 2018-02-13  Jeff Law  <law@redhat.com>
25157         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25158         entry for "vector".
25160         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25161         ARGS as unused.
25163 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25165         PR debug/84342
25166         PR debug/84319
25167         * common.opt (gas-loc-support, gas-locview-support): New.
25168         (ginline-points, ginternal-reset-location-views): New.
25169         * doc/invoke.texi: Document them.  Use @itemx where intended.
25170         (gvariable-location-views): Adjust.
25171         * target.def (reset_location_view): New.
25172         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25173         (TARGET_RESET_LOCATION_VIEW): New.
25174         * doc/tm.texi: Rebuilt.
25175         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25176         (dwarf2out_default_as_locview_support): New.
25177         (output_asm_line_debug_info): Use option variables.
25178         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25179         (output_loc_list): Likewise.
25180         (add_high_low_attributes): Check option variables.
25181         Don't output entry view attribute in strict mode.
25182         (gen_inlined_subroutine_die): Check option variables.
25183         (dwarf2out_inline_entry): Likewise.
25184         (init_sections_and_labels): Likewise.
25185         (dwarf2out_early_finish): Likewise.
25186         (maybe_reset_location_view): New, from...
25187         (dwarf2out_var_location): ... here.  Call it.
25188         * debug.h (dwarf2out_default_as_loc_support): Declare.
25189         (dwarf2out_default_as_locview_support): Declare.
25190         * hooks.c (hook_int_rtx_insn_0): New.
25191         * hooks.h (hook_int_rtx_insn_0): Declare.
25192         * toplev.c (process_options): Take -gas-loc-support and
25193         -gas-locview-support from dwarf2out.  Enable
25194         -gvariable-location-views by default only with locview
25195         assembler support.  Enable -ginternal-reset-location-views by
25196         default only if the target defines the corresponding hook.
25197         Enable -ginline-points by default if location views are
25198         enabled; force it disabled if statement frontiers are
25199         disabled.
25200         * tree-inline.c (expand_call_inline): Check option variables.
25201         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25203 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25205         PR tree-optimization/84321
25206         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25207         handling.  Also check whether the anti-range contains any values
25208         that satisfy the mask; switch to a VR_RANGE if not.
25210 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25212         PR sanitizer/84340
25213         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25215 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25217         PR c++/83990
25218         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25219         of call statements, also set location of a load to a temporary.
25221 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25223         * config/rl78/rl78.c (add_vector_labels): New function.
25224         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25225         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25226         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25227         which checks that no arguments are passed.
25228         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25229         * doc/extend.texi: Documentation for the new attribute.
25231 2018-02-13  Andreas Schwab  <schwab@suse.de>
25233         * config/riscv/linux.h (CPP_SPEC): Define.
25235 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25237         PR target/84335
25238         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25239         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25240         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25241         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25242         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25243         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25244         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25245         temporarily for AES and PCLMUL builtins.
25247         PR tree-optimization/84339
25248         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25249         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25250         Formatting fixes.
25252         PR middle-end/84309
25253         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25254         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25255         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25256         inline function.
25257         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25258         inline function.
25259         * omp-simd-clone.h: New file.
25260         * omp-simd-clone.c: Include omp-simd-clone.h.
25261         (expand_simd_clones): No longer static.
25262         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25263         cgraph.h and omp-simd-clone.h.
25264         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25265         (vect_recog_widen_shift_pattern): Formatting fix.
25266         (vect_pattern_recog_1): Don't check optab for calls.
25268         PR target/84336
25269         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25270         operands[2] into a REG before using gen_lowpart on it.
25272 2018-02-12  Jeff Law  <law@redhat.com>
25274         PR target/83760
25275         * config/sh/sh.c (find_barrier): Consider a sibling call
25276         a barrier as well.
25278         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25279         successfully back substituting a reg.
25281 2018-02-12  Richard Biener  <rguenther@suse.de>
25283         PR tree-optimization/84037
25284         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25285         parameter, move visited init to caller.
25286         (vect_slp_analyze_operations): Separate cost from validity
25287         check, initialize visited once for all instances.
25288         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25289         for all instances.
25290         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25291         out an assert.
25292         (vect_model_promotion_demotion_cost): Likewise.
25293         (vectorizable_bswap): Guard cost modeling with !slp_node
25294         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25295         SLP stmts.
25296         (vectorizable_call): Likewise.
25297         (vectorizable_conversion): Likewise.
25298         (vectorizable_assignment): Likewise.
25299         (vectorizable_shift): Likewise.
25300         (vectorizable_operation): Likewise.
25301         (vectorizable_store): Likewise.
25302         (vectorizable_load): Likewise.
25303         (vectorizable_condition): Likewise.
25304         (vectorizable_comparison): Likewise.
25306 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25308         PR sanitizer/84307
25309         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25310         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25311         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25313 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25315         PR middle-end/83665
25316         * params.def (inline-min-speedup): Increase from 8 to 15.
25317         (max-inline-insns-auto): Decrease from 40 to 30.
25318         * ipa-split.c (consider_split): Add some buffer for function to
25319         be considered inlining candidate.
25320         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25321         default values.
25323 2018-02-12  Richard Biener  <rguenther@suse.de>
25325         PR tree-optimization/84037
25326         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25327         matched stmts if we cannot swap the non-matched ones.
25329 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25331         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25332         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25333         _mm_maskz_scalef_round_ss): New intrinsics.
25334         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25335         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25336         __builtin_ia32_scalefss_round): Remove.
25337         (__builtin_ia32_scalefsd_mask_round,
25338         __builtin_ia32_scalefss_mask_round): New intrinsics.
25339         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25340         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25341         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25342         "<round_constraint>")): Changed to ...
25343         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25344         "<round_scalar_constraint>")): ... this.
25345         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25346         %0, %1, %2<round_op3>}"): Changed to ...
25347         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25348         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25349         %2<round_scalar_mask_op3>}"): ... this.
25350         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25352 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25354         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25355         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25356         (_mm_maskz_sqrt_round_ss): New intrinsics.
25357         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25358         (__builtin_ia32_sqrtsd_mask_round)
25359         (__builtin_ia32_sqrtss_mask_round): New builtins.
25360         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25361         (__builtin_ia32_sqrtss_round): Remove.
25362         (__builtin_ia32_sqrtsd_mask_round)
25363         (__builtin_ia32_sqrtss_mask_round): New builtins.
25364         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25365         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25366         ((match_operand:VF_128 1 "vector_operand"
25367         "xBm,<round_constraint>")): Changed to ...
25368         ((match_operand:VF_128 1 "vector_operand"
25369         "xBm,<round_scalar_constraint>")): ... this.
25370         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25371         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25372         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25373         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25374         %<iptr>1<round_scalar_mask_op3>}): ... this.
25375         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25376         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25378 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25380         PR target/84266
25381         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25382         Cast vec_cmpeq result to correct type.
25383         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25384         Cast vec_cmpgt result to correct type.
25386 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25388         * final.c (final_scan_insn_1): Renamed from...
25389         (final_scan_insn): ... this.  New wrapper, to recover
25390         seen from the outermost call in recursive ones.
25391         * config/sparc/sparc.c (output_return): Drop seen from call.
25392         (output_sibcall): Likewise.
25393         * config/visium/visium.c (output_branch): Likewise.
25395 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25397         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25398         function label.
25400 2018-02-10  Alan Modra  <amodra@gmail.com>
25402         PR target/84300
25403         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25404         Specify LR as an input.
25406 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25408         PR sanitizer/83987
25409         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25410         remove_member_access_dummy_vars): New functions.
25411         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25412         lower_omp_1, execute_lower_omp): Use them.
25414         PR rtl-optimization/84308
25415         * shrink-wrap.c (spread_components): Release todo vector.
25417 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25419         PR rtl-optimization/57193
25420         * ira-color.c (struct allocno_color_data): Add member
25421         conflict_allocno_hard_prefs.
25422         (update_conflict_allocno_hard_prefs): New.
25423         (bucket_allocno_compare_func): Add a preference based on
25424         conflict_allocno_hard_prefs.
25425         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25426         (color_allocnos): Remove a dead code.  Initiate
25427         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25429 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25431         PR target/84226
25432         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25433         constraint from =wa to wa.  Avoid a subreg on the output operand,
25434         instead use a pseudo and subreg it in a move.
25435         (p9_xxbrd_<mode>): Changed to ...
25436         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25437         (p9_xxbrd_v2df): New expander.
25438         (p9_xxbrw_<mode>): Changed to ...
25439         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25440         (p9_xxbrw_v4sf): New expander.
25442 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25444         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25446 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25448         PR target/83926
25449         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25450         multiply in 32-bit mode.
25451         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25452         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25453         mode.
25455 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25457         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25458         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25459         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25460         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25462 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25464         PR lto/84213
25465         * dwarf2out.c (is_trivial_indirect_ref): New function.
25466         (dwarf2out_late_global_decl): Do not generate a location
25467         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25468         and that are not defined in the current unit.
25470 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25472         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25473         instead of a libcall for UNORDERED.
25475 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25477         PR target/82641
25478         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25479         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25481 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25483         PR target/PR84295
25484         * config/s390/s390.c (s390_set_current_function): Invoke
25485         s390_indirect_branch_settings also if fndecl didn't change.
25487 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25489         * config/rs6000/rs6000.md (blockage): Set length to zero.
25491 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25493         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25495 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25497         PR sanitizer/84285
25498         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25499         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25500         -static-lib*san.
25502         PR debug/84252
25503         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
25504         PARALLEL incoming that failed vt_get_decl_and_offset check.
25506         PR middle-end/84237
25507         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
25508         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
25509         TREE_READONLY bit.
25510         (get_variable_section): For decls in named .bss* sections pass true as
25511         second argument to bss_initializer_p.
25513 2018-02-09  Marek Polacek  <polacek@redhat.com>
25514             Jakub Jelinek  <jakub@redhat.com>
25516         PR c++/83659
25517         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
25518         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
25519         Sync some changes from cxx_fold_indirect_ref.
25521 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25523         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
25524         markers.
25525         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
25526         (BLOCK_INLINE_ENTRY_LABEL): New.
25527         (dwarf2out_var_location): Disregard inline entry markers.
25528         (inline_entry_data): New struct.
25529         (inline_entry_data_hasher): New hashtable type.
25530         (inline_entry_data_hasher::hash): New.
25531         (inline_entry_data_hasher::equal): New.
25532         (inline_entry_data_table): New variable.
25533         (add_high_low_attributes): Add DW_AT_entry_pc and
25534         DW_AT_GNU_entry_view attributes if a pending entry is found
25535         in inline_entry_data_table.  Add old entry_pc attribute only
25536         if debug nonbinding markers are disabled.
25537         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
25538         markers are enabled.
25539         (block_within_block_p, dwarf2out_inline_entry): New.
25540         (dwarf2out_finish): Check that no entries remained in
25541         inline_entry_data_table.
25542         * final.c (reemit_insn_block_notes): Handle inline entry notes.
25543         (final_scan_insn, notice_source_line): Likewise.
25544         (rest_of_clean_state): Skip inline entry markers.
25545         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
25546         markers.
25547         * gimple.c (gimple_build_debug_inline_entry): New.
25548         * gimple.h (enum gimple_debug_subcode): Add
25549         GIMPLE_DEBUG_INLINE_ENTRY.
25550         (gimple_build_debug_inline_entry): Declare.
25551         (gimple_debug_inline_entry_p): New.
25552         (gimple_debug_nonbind_marker_p): Adjust.
25553         * insn-notes.def (INLINE_ENTRY): New.
25554         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
25555         inline entry marker notes.
25556         (print_insn): Likewise.
25557         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
25558         (INSN_DEBUG_MARKER_KIND): Likewise.
25559         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
25560         * tree-inline.c (expand_call_inline): Build and insert
25561         debug_inline_entry stmt.
25562         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
25563         inline entry blocks early, if nonbind markers are enabled.
25564         (dump_scope_block): Dump fragment info.
25565         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
25566         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
25567         (gimple_build_debug_inline_entry): New.
25568         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
25569         Enable/disable inline entry points too.
25570         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
25571         (DEBUG_INSN): Describe inline entry markers.
25573         * common.opt (gvariable-location-views): New.
25574         (gvariable-location-views=incompat5): New.
25575         * config.in: Rebuilt.
25576         * configure: Rebuilt.
25577         * configure.ac: Test assembler for view support.
25578         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
25579         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
25580         * dwarf2out.c (var_loc_view): New typedef.
25581         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
25582         (dwarf2out_locviews_in_attribute): New.
25583         (dwarf2out_locviews_in_loclist): New.
25584         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
25585         (enum dw_line_info_opcode): Add LI_adv_address.
25586         (struct dw_line_info_table): Add view.
25587         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
25588         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
25589         (zero_view_p): New variable.
25590         (ZERO_VIEW_P): New macro.
25591         (output_asm_line_debug_info): New.
25592         (struct var_loc_node): Add view.
25593         (add_AT_view_list, AT_loc_list): New.
25594         (add_var_loc_to_decl): Add view param.  Test it against last.
25595         (new_loc_list): Add view params.  Record them.
25596         (AT_loc_list_ptr): Handle loc and view lists.
25597         (view_list_to_loc_list_val_node): New.
25598         (print_dw_val): Handle dw_val_class_view_list.
25599         (size_of_die): Likewise.
25600         (value_format): Likewise.
25601         (loc_list_has_views): New.
25602         (gen_llsym): Set vl_symbol too.
25603         (maybe_gen_llsym, skip_loc_list_entry): New.
25604         (dwarf2out_maybe_output_loclist_view_pair): New.
25605         (output_loc_list): Output view list or entries too.
25606         (output_view_list_offset): New.
25607         (output_die): Handle dw_val_class_view_list.
25608         (output_dwarf_version): New.
25609         (output_compilation_unit_header): Use it.
25610         (output_skeleton_debug_sections): Likewise.
25611         (output_rnglists, output_line_info): Likewise.
25612         (output_pubnames, output_aranges): Update version comments.
25613         (output_one_line_info_table): Output view numbers in asm comments.
25614         (dw_loc_list): Determine current endview, pass it to new_loc_list.
25615         Call maybe_gen_llsym.
25616         (loc_list_from_tree_1): Adjust.
25617         (add_AT_location_description): Create view list attribute if
25618         needed, check it's absent otherwise.
25619         (convert_cfa_to_fb_loc_list): Adjust.
25620         (maybe_emit_file): Call output_asm_line_debug_info for test.
25621         (dwarf2out_var_location): Reset views as needed.  Precompute
25622         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
25623         attribute.  Set view.
25624         (new_line_info_table): Reset next view.
25625         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
25626         (dwarf2out_source_line): Likewise.  Output view resets and labels to
25627         the assembler, or select appropriate line info opcodes.
25628         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
25629         (optimize_string_length): Catch it.  Adjust.
25630         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
25631         dw_val_class_view_list, and remove it if no longer needed.
25632         (hash_loc_list): Hash view numbers.
25633         (loc_list_hasher::equal): Compare them.
25634         (optimize_location_lists): Check whether a view list symbol is
25635         needed, and whether the locview attribute is present, and
25636         whether they match.  Remove the locview attribute if no longer
25637         needed.
25638         (index_location_lists): Call skip_loc_list_entry for test.
25639         (dwarf2out_finish): Call output_asm_line_debug_info for test.
25640         Use output_dwarf_version.
25641         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
25642         (struct dw_val_node): Add val_view_list.
25643         * final.c (SEEN_NEXT_VIEW): New.
25644         (set_next_view_needed): New.
25645         (clear_next_view_needed): New.
25646         (maybe_output_next_view): New.
25647         (final_start_function): Rename to...
25648         (final_start_function_1): ... this.  Take pointer to FIRST,
25649         add SEEN parameter.  Emit param bindings in the initial view.
25650         (final_start_function): Reintroduce SEEN-less interface.
25651         (final): Rename to...
25652         (final_1): ... this.  Take SEEN parameter.  Output final pending
25653         next view at the end.
25654         (final): Reintroduce seen-less interface.
25655         (final_scan_insn): Output pending next view before switching
25656         sections or ending a block.  Mark the next view as needed when
25657         outputting variable locations.  Notify debug backend of section
25658         changes, and of location view changes.
25659         (rest_of_handle_final): Adjust.
25660         * toplev.c (process_options): Autodetect value for debug variable
25661         location views option.  Warn on incompat5 without -gdwarf-5.
25662         * doc/invoke.texi (gvariable-location-views): New.
25663         (gvariable-location-views=incompat5): New.
25664         (gno-variable-location-views): New.
25666 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
25668         PR tree-optimization/84136
25669         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
25670         that the result of find_edge is non-NULL.
25672 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25674         PR target/83008
25675         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
25676         storing integer register in SImode.  Fix cost of 256 and 512
25677         byte aligned SSE register store.
25679 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25681         * config/i386/i386.c (ix86_multiplication_cost): Fix
25682         multiplication cost for TARGET_AVX512DQ.
25684 2018-02-08  Marek Polacek  <polacek@redhat.com>
25686         PR tree-optimization/84238
25687         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
25688         get_range_strlen.
25690 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25692         PR tree-optimization/84265
25693         * tree-vect-stmts.c (vectorizable_store): Don't treat
25694         VMAT_CONTIGUOUS accesses as grouped.
25695         (vectorizable_load): Likewise.
25697 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25699         PR tree-optimization/81635
25700         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
25701         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
25702         (test_round_for_mask): New functions.
25703         (wide_int_cc_tests): Call test_round_for_mask.
25704         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
25705         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
25706         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
25707         range returned by get_range_info.
25709 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25711         PR ipa/81360
25712         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
25713         * symtab.c: Include builtins.h
25714         (symtab_node::output_to_lto_symbol_table_p): Move here
25715         from lto-streamer-out.c:output_symbol_p.
25716         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
25717         (output_symbol_p): Move all logic to symtab.c
25718         (produce_symtab): Update.
25720 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25722         * config/s390/s390-opts.h (enum indirect_branch): Define.
25723         * config/s390/s390-protos.h (s390_return_addr_from_memory)
25724         (s390_indirect_branch_via_thunk)
25725         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
25726         (enum s390_indirect_branch_type): Define.
25727         * config/s390/s390.c (struct s390_frame_layout, struct
25728         machine_function): Remove.
25729         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
25730         (indirect_branch_table_label_no, indirect_branch_table_name):
25731         Define variables.
25732         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
25733         (enum s390_indirect_branch_option): Define.
25734         (s390_return_addr_from_memory): New function.
25735         (s390_handle_string_attribute): New function.
25736         (s390_attribute_table): Add new attribute handler.
25737         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
25738         (s390_indirect_branch_via_thunk): New function.
25739         (s390_indirect_branch_via_inline_thunk): New function.
25740         (s390_function_ok_for_sibcall): When jumping via thunk disallow
25741         sibling call optimization for non z10 compiles.
25742         (s390_emit_call): Force indirect branch target to be a single
25743         register.  Add r1 clobber for non-z10 compiles.
25744         (s390_emit_epilogue): Emit return jump via return_use expander.
25745         (s390_reorg): Handle JUMP_INSNs as execute targets.
25746         (s390_option_override_internal): Perform validity checks for the
25747         new command line options.
25748         (s390_indirect_branch_attrvalue): New function.
25749         (s390_indirect_branch_settings): New function.
25750         (s390_set_current_function): Invoke s390_indirect_branch_settings.
25751         (s390_output_indirect_thunk_function):  New function.
25752         (s390_code_end): Implement target hook.
25753         (s390_case_values_threshold): Implement target hook.
25754         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
25755         macros.
25756         * config/s390/s390.h (struct s390_frame_layout)
25757         (struct machine_function): Move here from s390.c.
25758         (TARGET_INDIRECT_BRANCH_NOBP_RET)
25759         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
25760         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
25761         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
25762         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
25763         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
25764         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
25765         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
25766         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
25767         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
25768         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
25769         (mnemonic attribute): Add values which aren't recognized
25770         automatically.
25771         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
25772         pattern for branch conversion.  Fix mnemonic attribute.
25773         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
25774         indirect branch via thunk if requested.
25775         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
25776         ("*indirect_jump"): Disable for branch conversion using out of
25777         line thunks.
25778         ("indirect_jump_via_thunk<mode>_z10")
25779         ("indirect_jump_via_thunk<mode>")
25780         ("indirect_jump_via_inlinethunk<mode>_z10")
25781         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
25782         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
25783         ("casesi_jump_via_inlinethunk<mode>_z10")
25784         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
25785         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
25786         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
25787         ("*indirect2_jump"): Disable for branch conversion.
25788         ("casesi_jump"): Turn into expander and expand patterns for branch
25789         conversion.
25790         ("return_use"): New expander.
25791         ("*return"): Emit return via thunk and rename it to ...
25792         ("*return<mode>"): ... this one.
25793         * config/s390/s390.opt: Add new options and and enum for the
25794         option values.
25796 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25798         * lra-constraints.c (match_reload): Unconditionally use
25799         gen_lowpart_SUBREG, rather than selecting between that
25800         and equivalent gen_rtx_SUBREG code.
25802 2018-02-08  Richard Biener  <rguenther@suse.de>
25804         PR tree-optimization/84233
25805         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
25806         changed flag instead of boguously re-using phi_inserted.
25808 2018-02-08  Martin Jambor  <mjambor@suse.cz>
25810         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
25811         static local variables.
25813 2018-02-08  Richard Biener  <rguenther@suse.de>
25815         PR tree-optimization/84278
25816         * tree-vect-stmts.c (vectorizable_store): When looking for
25817         smaller vector types to perform grouped strided loads/stores
25818         make sure the mode is supported by the target.
25819         (vectorizable_load): Likewise.
25821 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25823         * config/aarch64/aarch64.c (aarch64_components_for_bb):
25824         Increase LDP/STP opportunities by adding adjacent callee-saves.
25826 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25828         PR rtl-optimization/84068
25829         PR rtl-optimization/83459
25830         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
25832 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
25834         PR tree-optimization/84224
25835         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
25836         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
25837         non-zero arguments.
25839 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
25841         PR target/84113
25842         * config/rs6000/altivec.md (*restore_world): Remove LR use.
25843         * config/rs6000/predicates.md (restore_world_operation): Adjust op
25844         count, remove one USE.
25846 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25848         * doc/install.texi (Configuration): Document the
25849         --with-long-double-format={ibm,ieee} PowerPC configuration
25850         options.
25852         PR target/84154
25853         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
25854         Convert from define_expand to be define_insn_and_split.  Rework
25855         float/double/_Float128 conversions to QI/HI/SImode to work with
25856         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
25857         conversions to QI/HImode types did a store and then a load to
25858         truncate the value.  For conversions to VSX registers, don't split
25859         the insn, instead emit the code directly.  Use the code iterator
25860         any_fix to combine signed and unsigned conversions.
25861         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
25862         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
25863         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25864         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
25865         (fix_<mode>di2_hw): Likewise.
25866         (fixuns_<mode>di2_hw): Likewise.
25867         (fix_<mode>si2_hw): Likewise.
25868         (fixuns_<mode>si2_hw): Likewise.
25869         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
25870         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25871         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
25872         fix<uns>_trunc<SFDF:mode>si2_p8.
25873         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
25874         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
25875         (fix<uns>_<mode>_mem): Likewise.
25876         (fctiw<u>z_<mode>_mem): Likewise.
25877         (fix<uns>_<mode>_mem): Likewise.
25878         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
25879         the register allocator from doing a direct move to the GPRs to do
25880         a store, and instead use the ISA 3.0 store byte/half-word from
25881         vector register instruction.  For IEEE 128-bit floating point,
25882         also optimize stores of 32-bit ints.
25883         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
25885 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
25887         * genextract.c (push_pathstr_operand): New function to support
25888         [a-zA-Z].
25889         (walk_rtx): Call push_pathstr_operand.
25890         (print_path): Support [a-zA-Z].
25892 2018-02-07  Richard Biener  <rguenther@suse.de>
25894         PR tree-optimization/84037
25895         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
25896         (cse_and_gimplify_to_preheader): Declare.
25897         (vect_get_place_in_interleaving_chain): Likewise.
25898         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25899         ivexpr_map.
25900         (_loop_vec_info::~_loop_vec_info): Delete it.
25901         (cse_and_gimplify_to_preheader): New function.
25902         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
25903         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
25904         (vectorizable_load): Likewise.  For grouped stores always base
25905         the IV on the first element.
25906         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
25907         condition before gimplifying.
25909 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25911         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
25912         *DIV_EXPR and *MOD_EXPR.
25914 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
25916         PR target/84248
25917         * config/i386/i386.c (ix86_option_override_internal): Mask out
25918         the CF_SET bit when checking -fcf-protection.
25920 2018-02-07  Tom de Vries  <tom@codesourcery.com>
25922         PR libgomp/84217
25923         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
25924         enough.
25926 2018-02-07  Richard Biener  <rguenther@suse.de>
25928         PR tree-optimization/84204
25929         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
25930         this place.
25932         PR tree-optimization/84205
25933         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
25934         special-case isl_ast_op_zdiv_r.
25936         PR tree-optimization/84223
25937         * graphite-scop-detection.c (gather_bbs::before_dom_children):
25938         Only add conditions from within the region.
25939         (gather_bbs::after_dom_children): Adjust.
25941 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
25943         PR target/84209
25944         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
25945         * config/avr/avr.md: Only post-reload split REG-REG moves if
25946         either register is GENERAL_REG_P.
25948 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25950         PR tree-optimization/84235
25951         * tree-ssa-scopedtables.c
25952         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
25953         if the subtraction is performed in floating point type where NaNs are
25954         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
25955         build 1.  Formatting fix.
25957 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
25959         PR target/84146
25960         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
25961         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
25962         and skip it regardless of bb boundaries.  Use CALL_P macro,
25963         don't test INSN_P (insn) together with CALL_P or JUMP_P check
25964         unnecessarily, formatting fix.
25966 2018-02-06  Michael Collison  <michael.collison@arm.com>
25968         * config/arm/thumb2.md:
25969         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
25970         (*thumb_mov_notscc): Ditto.
25972 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
25974         PR target/84154
25975         * config/rs6000/rs6000.md (su code attribute): Use "u" for
25976         unsigned_fix, not "s".
25978 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25980         * configure.ac (gcc_fn_eh_frame_ro): New function.
25981         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
25982         correct .eh_frame permissions.
25983         * configure: Regenerate.
25985 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
25987         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
25988         irrelevant options.
25990 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25992         * config/rs6000/rs6000.c (rs6000_option_override_internal):
25993         Display warning message for -mno-speculate-indirect-jumps.
25995 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
25997         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
25998         Undocumented.
25999         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26001 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26003         PR tree-optimization/84225
26004         * tree-eh.c (find_trapping_overflow): Only call
26005         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26007 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26009         PR target/84145
26010         * config/i386/i386.c: Reimplement the check of possible options
26011         -mibt/-mshstk conbination. Change error messages.
26012         * doc/invoke.texi: Fix a typo: remove extra '='.
26014 2018-02-06  Marek Polacek  <polacek@redhat.com>
26016         PR tree-optimization/84228
26017         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26019 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26021         PR target/82641
26022         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26023         emitted arch directives.
26024         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26025         __ARM_FEATURE_COPROC before changing architectures.
26027 2018-02-06  Richard Biener  <rguenther@suse.de>
26029         * config/i386/i386.c (print_reg): Fix typo.
26030         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26032 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26034         * configure: Regenerate.
26036 2018-02-05  Martin Sebor  <msebor@redhat.com>
26038         PR tree-optimization/83369
26039         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26040         inlining context.
26042 2018-02-05  Martin Liska  <mliska@suse.cz>
26044         * doc/invoke.texi: Cherry-pick upstream r323995.
26046 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26048         * ira.c (ira_init_register_move_cost): Adjust comment.
26050 2018-02-05  Martin Liska  <mliska@suse.cz>
26052         PR gcov-profile/84137
26053         * doc/gcov.texi: Fix typo in documentation.
26055 2018-02-05  Martin Liska  <mliska@suse.cz>
26057         PR gcov-profile/83879
26058         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26059         using dlopen functionality.
26061 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26063         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26064         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26065         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26066         _mm_maskz_range_round_ss): New intrinsics.
26067         (__builtin_ia32_rangesd128_round)
26068         (__builtin_ia32_rangess128_round): Remove.
26069         (__builtin_ia32_rangesd128_mask_round,
26070         __builtin_ia32_rangess128_mask_round): New builtins.
26071         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26072         __builtin_ia32_rangess128_round): Remove.
26073         (__builtin_ia32_rangesd128_mask_round,
26074         __builtin_ia32_rangess128_mask_round): New builtins.
26075         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26076         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26077         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26078         "<round_saeonly_constraint>")): Changed to ...
26079         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26080         "<round_saeonly_scalar_constraint>")): ... this.
26081         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26082         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26083         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26084         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26085         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26087 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26089         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26090         options.
26091         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26092         Remove all values except native, 8540 and 8548.
26094 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26096         * config/i386/i386.c (ix86_output_function_return): Pass
26097         INVALID_REGNUM, instead of -1, as invalid register number to
26098         indirect_thunk_name and output_indirect_thunk.
26100 2018-02-02  Julia Koval  <julia.koval@intel.com>
26102         * config.gcc: Add -march=icelake.
26103         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26104         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26105         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26106         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26107         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26108         (processor_target_table): Add icelake.
26109         (ix86_option_override_internal): Handle new PTAs.
26110         (get_builtin_code_for_version): Handle icelake.
26111         (M_INTEL_COREI7_ICELAKE): New.
26112         (fold_builtin_cpu): Handle icelake.
26113         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26114         * doc/invoke.texi: Add -march=icelake.
26116 2018-02-02  Julia Koval  <julia.koval@intel.com>
26118         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26119         to wide_int_bitmask.
26120         * wide-int-bitmask.h: New.
26122 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26124         PR target/84066
26125         * config/i386/i386.md: Replace Pmode with word_mode in
26126         builtin_setjmp_setup and builtin_longjmp to support x32.
26128 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26130         PR target/56010
26131         PR target/83743
26132         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26133         #include "opts.h".
26134         (rs6000_supported_cpu_names): New static variable.
26135         (linux_cpu_translation_table): Likewise.
26136         (elf_platform) <cpu>: Define new static variable and use it.
26137         Translate kernel AT_PLATFORM name to canonical name if needed.
26138         Error if platform name is unknown.
26140 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26142         PR target/84089
26143         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26145 2018-02-01  Jeff Law  <law@redhat.com>
26147         PR target/84128
26148         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26149         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26150         the scratch if RELEASE_VIA_POP is false.
26151         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26152         If we have to save a temporary register, decrement SIZE appropriately.
26153         Pass new arguments to release_scratch_register_on_entry.
26154         (ix86_adjust_stack_and_probe): Likewise.
26155         (ix86_emit_probe_stack_range): Pass new arguments to
26156         release_scratch_register_on_entry.
26158 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26160         PR rtl-optimization/84157
26161         * combine.c (change_zero_ext): Use REG_P predicate in
26162         front of HARD_REGISTER_P predicate.
26164 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26166         * config/avr/avr.c (avr_option_override): Move disabling of
26167         -fdelete-null-pointer-checks to...
26168         * common/config/avr/avr-common.c (avr_option_optimization_table):
26169         ...here.
26171 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26173         PR tree-optimization/81635
26174         * tree-data-ref.c (split_constant_offset_1): For types that
26175         wrap on overflow, try to use range info to prove that wrapping
26176         cannot occur.
26178 2018-02-01  Renlin Li  <renlin.li@arm.com>
26180         PR target/83370
26181         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26182         TAILCALL_ADDR_REGS.
26183         (aarch64_register_move_cost): Likewise.
26184         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26185         TAILCALL_ADDR_REGS.
26186         (REG_CLASS_NAMES): Likewise.
26187         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26188         TAILCALL_ADDR_REGS. Remove IP registers.
26189         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26191 2018-02-01  Richard Biener  <rguenther@suse.de>
26193         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26194         for specifying RPO order and allow NULL for that.
26195         * domwalk.c (dom_walker::dom_walker): Likewise.
26196         (dom_walker::walk): Handle NULL RPO order.
26197         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26198         in RPO order.
26199         (rewrite_update_dom_walker): Likewise.
26200         (mark_def_dom_walker): Likewise.
26202 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26204         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26205         (aarch64_maybe_expand_sve_subreg_move): Declare.
26206         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26207         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26208         predicate.
26209         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26210         that are semantically a reverse operation.
26211         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26212         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26213         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26214         functions.
26215         (aarch64_can_change_mode_class): For big-endian, forbid changes
26216         between two SVE modes if they have different element sizes.
26218 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26220         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26221         the TImode handling for big-endian targets.
26223 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26225         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26226         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26227         not just bytes.
26228         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26229         Remove BSWAP handing for big-endian targets and use the form of
26230         LD1RQ appropariate for the mode.
26232 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26234         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26235         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26236         duplicated element.
26238 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26240         PR tearget/83845
26241         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26242         check for operands that need to go through aarch64_sve_reload_be.
26244 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26246         PR tree-optimization/81661
26247         PR tree-optimization/84117
26248         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26249         * tree-eh.c: Include gimplify.h.
26250         (find_trapping_overflow, replace_trapping_overflow,
26251         rewrite_to_non_trapping_overflow): New functions.
26252         * tree-vect-loop.c: Include tree-eh.h.
26253         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26254         * tree-data-ref.c: Include tree-eh.h.
26255         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26257 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26259         PR rtl-optimization/84123
26260         * combine.c (change_zero_ext): Check if hard register satisfies
26261         can_change_dest_mode before calling gen_lowpart_SUBREG.
26263 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26265         PR target/82444
26266         * ira.c (ira_init_register_move_cost): Remove assert.
26268 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26270         PR rtl-optimization/84071
26271         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26272         * doc/tm.texi: Regenerate.
26274 2018-01-31  Richard Biener  <rguenther@suse.de>
26276         PR tree-optimization/84132
26277         * tree-data-ref.c (analyze_miv_subscript): Properly
26278         check whether evolution_function_is_affine_multivariate_p
26279         before calling gcd_of_steps_may_divide_p.
26281 2018-01-31  Julia Koval  <julia.koval@intel.com>
26283         PR target/83618
26284         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26285         * config/i386/i386.md (rdpid_rex64) New.
26286         (rdpid): Make 32bit only.
26288 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26290         PR lto/84105
26291         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26292         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26294 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26296         Revert
26297         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26299         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26301 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26303         PR rtl-optimization/84071
26304         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26305         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26307 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26309         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26310         (arc_attribute_table): Add 'aux' attribute.
26311         (arc_in_small_data_p): Consider aux like variables.
26312         (arc_is_aux_reg_p): New function.
26313         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26314         (arc_get_aux_arg): New function.
26315         (prepare_move_operands): Handle aux-register access.
26316         (arc_handle_aux_attribute): New function.
26317         * doc/extend.texi (ARC Variable attributes): Add subsection.
26319 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26321         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26322         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26323         (arc_attribute_table): Add 'uncached' attribute.
26324         (arc_print_operand): Print '.di' flag for uncached memory
26325         accesses.
26326         (arc_in_small_data_p): Do not consider for small data the uncached
26327         types.
26328         (arc_is_uncached_mem_p): New function.
26329         * config/arc/predicates.md (compact_store_memory_operand): Check
26330         for uncached memory accesses.
26331         (nonvol_nonimm_operand): Likewise.
26332         * doc/extend.texi (ARC Type Attribute): New subsection.
26334 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26336         PR c/84100
26337         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26338         falign-loops=): Add Optimization flag.
26340 2018-01-30  Jeff Law  <law@redhat.com>
26342         PR target/84064
26343         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26344         INT_REGISTERS_SAVED.  Check it prior to calling
26345         get_scratch_register_on_entry.
26346         (ix86_adjust_stack_and_probe): Similarly.
26347         (ix86_emit_probe_stack_range): Similarly.
26348         (ix86_expand_prologue): Corresponding changes.
26350 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26352         PR target/40411
26353         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26354         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26356 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26358         PR target/84112
26359         * lra-constraints.c (curr_insn_transform): Process AND in the
26360         address.
26362 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26364         PR rtl-optimization/83986
26365         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26366         dependence against last_pending_memory_flush in addition to
26367         pending_jump_insns.
26369 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26371         PR tree-optimization/81611
26372         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26373         copies.
26375 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26377         PR target/83758
26378         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26379         a reg rtx.
26381 2018-01-30  Richard Biener  <rguenther@suse.de>
26382             Jakub Jelinek  <jakub@redhat.com>
26384         PR tree-optimization/84111
26385         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26386         inner loops added during recursion, as they don't have up-to-date
26387         SSA form.
26389 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26391         PR ipa/81360
26392         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26393         (can_inline_edge_by_limits_p): ... here.
26394         (can_early_inline_edge_p, check_callers,
26395         update_caller_keys, update_callee_keys, recursive_inlining,
26396         add_new_edges_to_heap, speculation_useful_p,
26397         inline_small_functions,
26398         inline_small_functions, flatten_function,
26399         inline_to_all_callers_1): Update.
26401 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26403         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26404         zeros correctly.
26406 2018-01-30  Richard Biener  <rguenther@suse.de>
26408         PR tree-optimization/83008
26409         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26410         invariant and constant vector uses in stmts when they need
26411         more than one stmt.
26413 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26415         PR bootstrap/84017
26416         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26417         * configure: Regenerate.
26419 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26421         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26422         pattern.
26423         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26424         Use gen_rtx_REG rather than gen_lowpart.
26426 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26428         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26429         rather than 0 when creating partial subregs.
26431 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26433         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26434         of usage.
26436 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26438         PR target/81550
26439         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26440         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26441         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26442         flags.  This restores the settings used before the 2017-07-24.
26443         Turning off pre increment/decrement/modify allows IVOPTS to
26444         optimize DF/SF loops where the index is an int.
26446 2018-01-29  Richard Biener  <rguenther@suse.de>
26447             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26449         PR bootstrap/80867
26450         * tree-vect-stmts.c (vectorizable_call): Don't call
26451         targetm.vectorize_builtin_md_vectorized_function if callee is
26452         NULL.
26454 2018-01-22  Carl Love  <cel@us.ibm.com>
26456         * doc/extend.tex: Fix typo in second arg in
26457         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26459 2018-01-29  Richard Biener  <rguenther@suse.de>
26461         PR tree-optimization/84086
26462         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26463         (flush_ssaname_freelist): When SSA names were released reset
26464         the SCEV hash table.
26466 2018-01-29  Richard Biener  <rguenther@suse.de>
26468         PR tree-optimization/84057
26469         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26470         removed paths when removing edges.
26472 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26474         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26475         -mfunction-return=@var{choice}.
26477 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26479         PR diagnostic/84034
26480         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26481         Handle CR like TAB.
26482         (layout::print_source_line): Likewise.
26483         (test_get_line_width_without_trailing_whitespace): Add test cases.
26485 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26487         PR middle-end/84040
26488         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26489         debug insns.
26491 2018-01-26  Jim Wilson  <jimw@sifive.com>
26493         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26495         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26496         specified.
26498 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26500         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
26501         and CMP + SUB-immediate -> SUBS.
26503 2018-01-26  Martin Sebor  <msebor@redhat.com>
26505         PR tree-optimization/83896
26506         * tree-ssa-strlen.c (get_string_len): Rename...
26507         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
26508         Avoid assuming length is constant.
26509         (handle_char_store): Use HOST_WIDE_INT for string length.
26511 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
26513         PR target/81763
26514         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
26515         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
26517 2018-01-26  Richard Biener  <rguenther@suse.de>
26519         PR rtl-optimization/84003
26520         * dse.c (record_store): Only record redundant stores when
26521         the earlier store aliases at least all accesses the later one does.
26523 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
26525         PR rtl-optimization/83985
26526         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
26527         REG_CFA_RESTORE insns.
26528         (delete_unmarked_insns): Don't ignore separate shrink wrapping
26529         REG_CFA_RESTORE insns here.
26531         PR c/83989
26532         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
26533         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
26535 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26537         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
26538         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
26539         (arc_init): Likewise.
26540         (arc_override_options): Likewise.
26541         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
26542         value.
26543         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
26544         support.
26545         * config/arc/arc.h (TARGET_DBNZ): Define.
26546         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
26547         properly set the tune attribute.
26548         (dbnz): Use TARGET_DBNZ guard.
26549         * config/arc/arc.opt (mtune): Add core3 option.
26551 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26553         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
26554         recognize new pic like addresses.
26555         (arc_delegitimize_address): Clean up.
26557 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26559         * config/arc/arc-arches.def: Option mrf16 valid for all
26560         architectures.
26561         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
26562         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
26563         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
26564         * config/arc/arc-tables.opt: Regenerate.
26565         * config/arc/arc.c (arc_conditional_register_usage): Handle
26566         reduced register file case.
26567         (arc_file_start): Set must have build attributes.
26568         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
26569         mrf16 option value.
26570         * config/arc/arc.opt (mrf16): Add new option.
26571         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
26572         * config/arc/genmultilib.awk: Handle new mrf16 option.
26573         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
26574         * config/arc/t-multilib: Regenerate.
26575         * doc/invoke.texi (ARC Options): Document mrf16 option.
26577 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26579         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
26580         * config/arc/arc.c (arc_handle_secure_attribute): New function.
26581         (arc_attribute_table): Add 'secure_call' attribute.
26582         (arc_print_operand): Print secure call operand.
26583         (arc_function_ok_for_sibcall): Don't optimize tail calls when
26584         secure.
26585         (arc_is_secure_call_p): New function.  * config/arc/arc.md
26586         (call_i): Add support for sjli instruction.
26587         (call_value_i): Likewise.
26588         * config/arc/constraints.md (Csc): New constraint.
26590 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26591             John Eric Martin  <John.Martin@emmicro-us.com>
26593         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
26594         * config/arc/arc.c (_arc_jli_section): New struct.
26595         (arc_jli_section): New type.
26596         (rc_jli_sections): New static variable.
26597         (arc_handle_jli_attribute): New function.
26598         (arc_attribute_table): Add jli_always and jli_fixed attribute.
26599         (arc_file_end): New function.
26600         (TARGET_ASM_FILE_END): Define.
26601         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
26602         (arc_add_jli_section): New function.
26603         (jli_call_scan): Likewise.
26604         (arc_reorg): Call jli_call_scan.
26605         (arc_output_addsi): Remove 'S' from printing asm operand.
26606         (arc_is_jli_call_p): New function.
26607         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
26608         operand.
26609         (movhi_insn): Likewise.
26610         (movsi_insn): Likewise.
26611         (movsi_set_cc_insn): Likewise.
26612         (loadqi_update): Likewise.
26613         (load_zeroextendqisi_update): Likewise.
26614         (load_signextendqisi_update): Likewise.
26615         (loadhi_update): Likewise.
26616         (load_zeroextendhisi_update): Likewise.
26617         (load_signextendhisi_update): Likewise.
26618         (loadsi_update): Likewise.
26619         (loadsf_update): Likewise.
26620         (movsicc_insn): Likewise.
26621         (bset_insn): Likewise.
26622         (bxor_insn): Likewise.
26623         (bclr_insn): Likewise.
26624         (bmsk_insn): Likewise.
26625         (bicsi3_insn): Likewise.
26626         (cmpsi_cc_c_insn): Likewise.
26627         (movsi_ne): Likewise.
26628         (movsi_cond_exec): Likewise.
26629         (clrsbsi2): Likewise.
26630         (norm_f): Likewise.
26631         (normw): Likewise.
26632         (swap): Likewise.
26633         (divaw): Likewise.
26634         (flag): Likewise.
26635         (sr): Likewise.
26636         (kflag): Likewise.
26637         (ffs): Likewise.
26638         (ffs_f): Likewise.
26639         (fls): Likewise.
26640         (call_i): Remove 'S' asm letter, add jli instruction.
26641         (call_value_i): Likewise.
26642         * config/arc/arc.op (mjli-always): New option.
26643         * config/arc/constraints.md (Cji): New constraint.
26644         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
26645         operand.
26646         (subsf3_fpx): Likewise.
26647         (mulsf3_fpx): Likewise.
26648         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
26649         asm operand.
26650         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
26651         function attrbutes.
26652         * doc/invoke.texi (ARC): Document mjli-always option.
26654 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
26656         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
26657         avoid addition with 0 and use incw and decw where possible.
26659 2018-01-26  Richard Biener  <rguenther@suse.de>
26661         PR tree-optimization/81082
26662         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
26663         association if it requires casting to unsigned.
26664         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
26665         from fold_plusminus_mult_expr to catch important cases late when
26666         range info is available.
26668 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26670         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
26671         * configure.ac (hidden_linkonce): New test.
26672         * configure: Regenerate.
26673         * config.in: Regenerate.
26675 2018-01-26  Julia Koval  <julia.koval@intel.com>
26677         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
26678         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
26679         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
26680         _mm_mask_bitshuffle_epi64_mask): Fix type.
26681         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
26682         USI_FTYPE_V4DI_V4DI_USI): Remove.
26683         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
26684         __builtin_ia32_vpshufbitqmb256_mask,
26685         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
26686         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
26687         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
26689 2018-01-26  Alan Modra  <amodra@gmail.com>
26691         PR target/84033
26692         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
26693         UNSPEC_VBPERMQ.  Sort other unspecs.
26695 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
26697         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
26699 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
26701         PR middle-end/83055
26702         * predict.c (drop_profile): Do not push/pop cfun; update also
26703         node->count.
26704         (handle_missing_profiles): Fix logic looking for zero profiles.
26706 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
26708         PR middle-end/83977
26709         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
26710         on functions with #pragma omp declare simd or functions with simd
26711         attribute.
26712         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
26713         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
26714         Remove trailing \n from warning_at calls.
26716 2018-01-25  Tom de Vries  <tom@codesourcery.com>
26718         PR target/84028
26719         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26720         for neutered workers.
26722 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
26724         PR target/68467
26725         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
26726         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
26728 2018-01-24  Jeff Law  <law@redhat.com>
26730         PR target/83994
26731         * i386.c (get_probe_interval): Move to earlier point.
26732         (ix86_compute_frame_layout): If -fstack-clash-protection and
26733         the frame is larger than the probe interval, then use pushes
26734         to save registers rather than reg->mem moves.
26735         (ix86_expand_prologue): Remove conditional for int_registers_saved
26736         assertion.
26738 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
26740         PR target/84014
26741         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
26742         min/max for never referenced object.
26744 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
26746         PR middle-end/83977
26747         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
26748         here.
26749         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
26750         attributes from DECL_ATTRIBUTES (decl) without affecting
26751         DECL_ATTRIBUTES (current_function_decl).
26752         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
26753         functions with non-NULL DECL_ABSTRACT_ORIGIN.
26755 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
26757         PR tree-optimization/83979
26758         * fold-const.c (fold_comparison): Use constant_boolean_node
26759         instead of boolean_{true,false}_node.
26761 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
26763         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
26764         with zero counts.
26766 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26768         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26769         Simplify the clause that sets the length attribute.
26770         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26771         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
26772         clause that sets the length attribute.
26773         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26775 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26777         PR target/83589
26778         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
26779         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
26780         Add strict parameter.
26781         (prevent_branch_around_nothing): Insert dummy insn between branch to
26782         label and label with no ptx insn inbetween.
26783         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
26785 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26787         PR target/81352
26788         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26789         for neutered threads in warp.
26790         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
26792 2018-01-24  Richard Biener  <rguenther@suse.de>
26794         PR tree-optimization/83176
26795         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
26796         operands.
26798 2018-01-24  Richard Biener  <rguenther@suse.de>
26800         PR tree-optimization/82819
26801         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
26802         code generating pluses that are no-ops in the target precision.
26804 2018-01-24  Richard Biener  <rguenther@suse.de>
26806         PR middle-end/84000
26807         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
26809 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26811         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
26812         to merge probabilities.
26813         * predict.c (probably_never_executed): Also mark as cold functions
26814         with global 0 profile and guessed local profile.
26815         * profile-count.c (profile_probability::combine_with_count): New
26816         member function.
26817         * profile-count.h (profile_probability::operator*,
26818         profile_probability::operator*=, profile_probability::operator/,
26819         profile_probability::operator/=): Reduce precision to adjusted
26820         and set value to guessed on contradictory divisions.
26821         (profile_probability::combine_with_freq): Remove.
26822         (profile_probability::combine_wiht_count): Declare.
26823         (profile_count::force_nonzero):: Set to adjusted.
26824         (profile_count::probability_in):: Set quality to adjusted.
26825         * tree-ssa-tail-merge.c (replace_block_by): Use
26826         combine_with_count.
26828 2018-01-23  Andrew Waterman  <andrew@sifive.com>
26829             Jim Wilson  <jimw@sifive.com>
26831         * config/riscv/riscv.c (riscv_stack_boundary): New.
26832         (riscv_option_override): Set riscv_stack_boundary.  Handle
26833         riscv_preferred_stack_boundary_arg.
26834         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
26835         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
26836         (STACK_BOUNDARY): Set to riscv_stack_boundary.
26837         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
26838         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
26839         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
26841 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
26843         PR target/83905
26844         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
26845         of struct ix86_frame.
26846         (ix86_expand_epilogue): Likewise.  Add a local variable for
26847         the reg_save_offset field in struct ix86_frame.
26849 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
26851         PR tree-optimization/82604
26852         * tree-loop-distribution.c (enum partition_kind): New enum item
26853         PKIND_PARTIAL_MEMSET.
26854         (partition_builtin_p): Support above new enum item.
26855         (generate_code_for_partition): Ditto.
26856         (compute_access_range): Differentiate cases that equality can be
26857         proven at all loops, the innermost loops or no loops.
26858         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
26859         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
26860         (finalize_partitions, distribute_loop): Don't fuse partition of
26861         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
26862         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
26863         parloop is enabled.
26865 2018-01-23  Martin Liska  <mliska@suse.cz>
26867         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
26868         order to ignore the predictor.
26869         (PRED_POLYMORPHIC_CALL): Likewise.
26870         (PRED_RECURSIVE_CALL): Likewise.
26872 2018-01-23  Martin Liska  <mliska@suse.cz>
26874         * tree-profile.c (tree_profiling): Print function header to
26875         aware reader which function we are working on.
26876         * value-prof.c (gimple_find_values_to_profile): Do not print
26877         not interesting value histograms.
26879 2018-01-23  Martin Liska  <mliska@suse.cz>
26881         * profile-count.h (enum profile_quality): Add
26882         profile_uninitialized as the first value. Do not number values
26883         as they are zero based.
26884         (profile_count::verify): Update sanity check.
26885         (profile_probability::verify): Likewise.
26887 2018-01-23  Nathan Sidwell  <nathan@acm.org>
26889         * doc/invoke.texi (ffor-scope): Deprecate.
26891 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26893         PR tree-optimization/83510
26894         * domwalk.c (set_all_edges_as_executable): New function.
26895         (dom_walker::dom_walker): Convert bool param
26896         "skip_unreachable_blocks" to enum reachability.  Move setup of
26897         edge flags to set_all_edges_as_executable and only do it when
26898         reachability is REACHABLE_BLOCKS.
26899         * domwalk.h (enum dom_walker::reachability): New enum.
26900         (dom_walker::dom_walker): Convert bool param
26901         "skip_unreachable_blocks" to enum reachability.
26902         (set_all_edges_as_executable): New decl.
26903         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
26904         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
26905         "reachability".
26906         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
26907         but converting true to REACHABLE_BLOCKS.
26908         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
26909         * tree-vrp.c
26910         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
26911         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
26912         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
26913         REACHABLE_BLOCKS.
26914         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
26915         if check_all_array_refs will be called.
26917 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26919         * tree.c (selftest::test_location_wrappers): Add more test
26920         coverage.
26922 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26924         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
26925         (selftest::test_bit_in_range): Likewise.
26927 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26929         PR testsuite/83888
26930         * doc/sourcebuild.texi (vect_float): Say that the selector
26931         only describes the situation when -funsafe-math-optimizations is on.
26932         (vect_float_strict): Document.
26934 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26936         PR tree-optimization/83965
26937         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
26938         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
26939         instead of checking only for a reduction.
26940         (vect_recog_widen_sum_pattern): Likewise.
26942 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26944         * predict.c (probably_never_executed): Only use precise profile info.
26945         (compute_function_frequency): Skip after inlining hack since we now
26946         have quality checking.
26948 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26950         * profile-count.h (profile_probability::very_unlikely,
26951         profile_probability::unlikely, profile_probability::even): Set
26952         precision to guessed.
26954 2018-01-23  Richard Biener  <rguenther@suse.de>
26956         PR tree-optimization/83963
26957         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
26958         Properly terminate dominator walk when crossing the exit edge not
26959         when visiting its source block.
26961 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
26963         PR c++/83918
26964         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
26965         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
26967 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
26969         PR tree-optimization/83957
26970         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
26971         semicolon after for body surrounded by braces.
26973         PR tree-optimization/83081
26974         * profile-count.h (profile_probability::split): New method.
26975         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
26976         Use profile_probability::split.
26977         (do_compare_rtx_and_jump): Fix adjustment of probabilities
26978         when splitting a single conditional jump into 2.
26980 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
26982         PR tree-optimization/69452
26983         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
26984         decl.
26986 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26988         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
26989         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
26990         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
26992 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26994         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
26995         declaration.
26996         * config/rl78/rl78.md (movdi): New define_expand.
26997         * config/rl78/rl78.c (rl78_split_movdi): New function.
26999 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27001         PR target/83862
27002         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27003         no longer used.
27004         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27005         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27006         128-bit to produce an UNSPEC move to get the double word with the
27007         signbit and then a shift directly to do signbit.
27008         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27009         implementation with a new version that just does either a direct
27010         move or a regular move.  Move memory interface to separate insns.
27011         Move insns so they are next to the expander.
27012         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27013         with signbit move.  Split big and little endian case.
27014         (signbit<mode>2_dm_mem_le): Likewise.
27015         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27016         (signbit<mode>2_dm2): Likewise.
27018 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27020         * config/rl78/rl78.md (anddi3): New define_expand.
27022 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27024         * config/rl78/rl78.md (umindi3): New define_expand.
27026 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27028         * config/rl78/rl78.md (smindi3): New define_expand.
27030 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27032         * config/rl78/rl78.md (smaxdi3): New define_expand.
27034 2018-01-22  Carl Love  <cel@us.ibm.com>
27036         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27037         LVX_V1TI): Add macro expansion.
27038         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27039         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27040         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27041         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27042         Change check to determine if the instruction is a byte reversing
27043         entry.  Fix typo in comment.
27044         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27045         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27046         Add def_builtin calls for new builtins.
27047         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27048         Add define_insn expansion.
27050 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27052         * config/rl78/rl78.md (umaxdi3): New define_expand.
27054 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27056         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27057         for non-QImode registers.
27059 2018-01-22  Richard Biener  <rguenther@suse.de>
27061         PR tree-optimization/83963
27062         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27063         including the loop exit block.
27064         (scop_detection::merge_sese): Likewise.
27065         (scop_detection::add_scop): Do it here instead.
27067 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27069         * doc/sourcebuild.texi (arm_softfloat): Document.
27071 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27073         PR gcc/77734
27074         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27075         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27076         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27078 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27079             David Edelsohn  <dje.gcc@gmail.com>
27081         PR target/83946
27082         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27083         Change "crset eq" to "crset 2".
27084         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27085         (*call_indirect_aix<mode>_nospec): Likewise.
27086         (*call_value_indirect_aix<mode>_nospec): Likewise.
27087         (*call_indirect_elfv2<mode>_nospec): Likewise.
27088         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27089         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27090         change assembly output from . to $.
27091         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27092         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27093         (*tablejump<mode>_internal1_nospec): Likewise.
27095 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27097         PR target/80870
27098         * config/sh/sh_optimize_sett_clrt.cc:
27099         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27101 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27103         PR tree-optimization/83940
27104         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27105         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27106         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27107         use it to pass back the definition type.
27108         (vect_check_store_rhs): Likewise rhs_dt_out.
27109         (vect_build_gather_load_calls): Add a mask_dt argument and use
27110         it instead of a call to vect_is_simple_use.
27111         (vectorizable_store): Update calls to vect_check_load_store_mask
27112         and vect_check_store_rhs.  Use the dt returned by the latter instead
27113         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27114         instead of calls to vect_is_simple_use.  Pass the scalar rather
27115         than the vector operand to vect_is_simple_use when handling
27116         second and subsequent copies of an rhs value.
27117         (vectorizable_load): Update calls to vect_check_load_store_mask
27118         and vect_build_gather_load_calls.  Use the cached mask_dt and
27119         gs_info.offset_dt instead of calls to vect_is_simple_use.
27121 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27123         PR middle-end/83945
27124         * tree-emutls.c: Include gimplify.h.
27125         (lower_emutls_2): New function.
27126         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27127         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27128         it before further processing.
27130         PR target/83930
27131         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27132         UINTVAL (trueop1) instead of INTVAL (op1).
27134 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27136         PR debug/81570
27137         PR debug/83728
27138         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27139         INCOMING_FRAME_SP_OFFSET if not defined.
27140         (scan_trace): Add ENTRY argument.  If true and
27141         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27142         emit a note to adjust the CFA offset.
27143         (create_cfi_notes): Adjust scan_trace callers.
27144         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27145         INCOMING_FRAME_SP_OFFSET in the CIE.
27146         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27147         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27148         Likewise.
27149         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27150         * doc/tm.texi: Regenerated.
27152 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27154         PR rtl-optimization/83147
27155         * lra-constraints.c (remove_inheritance_pseudos): Use
27156         lra_substitute_pseudo_within_insn.
27158 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27159             Cesar Philippidis  <cesar@codesourcery.com>
27161         PR target/83920
27162         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27164 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27166         PR target/83790
27167         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27168         spaces for function labels.
27170 2018-01-19  Martin Liska  <mliska@suse.cz>
27172         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27173         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27174         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27175         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27176         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27177         (PRED_CONST_RETURN): Change from 69 to 65.
27178         (PRED_NULL_RETURN): Change from 91 to 71.
27179         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27180         (PRED_LOOP_GUARD): Change from 66 to 73.
27182 2018-01-19  Martin Liska  <mliska@suse.cz>
27184         * predict.c (predict_insn_def): Add new assert.
27185         (struct branch_predictor): Change type to signed integer.
27186         (test_prediction_value_range): Amend test to cover
27187         PROB_UNINITIALIZED.
27188         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27189         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27190         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27191         (PRED_LOOP_IV_COMPARE): Likewise.
27192         * predict.h (PROB_UNINITIALIZED): Define new constant.
27194 2018-01-19  Martin Liska  <mliska@suse.cz>
27196         * predict.c (dump_prediction): Add new format for
27197         analyze_brprob.py script which is enabled with -details
27198         suboption.
27199         * profile-count.h (precise_p): New function.
27201 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27203         PR tree-optimization/83922
27204         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27205         there are no statements that need masking.
27206         (vect_active_double_reduction_p): New function.
27207         (vect_analyze_loop_operations): Use it when handling phis that
27208         are not in the loop header.
27210 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27212         PR tree-optimization/83914
27213         * tree-vect-loop.c (vectorizable_induction): Don't convert
27214         init_expr or apply the peeling adjustment for inductions
27215         that are nested within the vectorized loop.
27217 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27219         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27220         instead of NEG.
27222 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27224         PR sanitizer/81715
27225         PR testsuite/83882
27226         * function.h (gimplify_parameters): Add gimple_seq * argument.
27227         * function.c: Include gimple.h and options.h.
27228         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27229         for the added local temporaries if needed.
27230         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27231         if there are any parameter cleanups, wrap whole body into a
27232         try/finally with the cleanups.
27234 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27236         PR target/82964
27237         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27238         Use GET_MODE_CLASS for scalar floating point.
27240 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27242         PR ipa/82256
27243         patch by PaX Team
27244         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27245         Fix call of call_cgraph_insertion_hooks.
27247 2018-01-18  Martin Sebor  <msebor@redhat.com>
27249         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27251 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27253         PR ipa/83619
27254         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27255         frequencies.
27257 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27259         PR other/70268
27260         * common.opt: (-ffile-prefix-map): New option.
27261         * opts.c (common_handle_option): Defer it.
27262         * opts-global.c (handle_common_deferred_options): Handle it.
27263         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27264         * file-prefix-map.h: New file.
27265         (remap_debug_filename, add_debug_prefix_map): ...here.
27266         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27267         * final.c (debug_prefix_map, add_debug_prefix_map
27268         remap_debug_filename): Move to...
27269         * file-prefix-map.c: New file.
27270         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27271         generalize, get rid of alloca(), use strrchr() instead of strchr().
27272         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27273         Implement in terms of add_prefix_map().
27274         (remap_macro_filename, remap_debug_filename): Implement in term of
27275         remap_filename().
27276         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27277         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27278         * dbxout.c: Include file-prefix-map.h.
27279         * varasm.c: Likewise.
27280         * vmsdbgout.c: Likewise.
27281         * xcoffout.c: Likewise.
27282         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27283         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27284         * doc/invoke.texi (-ffile-prefix-map): Document.
27285         (-fdebug-prefix-map): Update description.
27287 2018-01-18  Martin Liska  <mliska@suse.cz>
27289         * config/i386/i386.c (indirect_thunk_name): Document that also
27290         lfence is emitted.
27291         (output_indirect_thunk): Document why both instructions
27292         (pause and lfence) are generated.
27294 2018-01-18  Richard Biener  <rguenther@suse.de>
27296         PR tree-optimization/83887
27297         * graphite-scop-detection.c
27298         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27299         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27300         (scop_detection::merge_sese): Re-implement with a flood-fill
27301         algorithm that properly finds a SESE region if it exists.
27303 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27305         PR c/61240
27306         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27307         pointer_diff optimizations use view_convert instead of convert.
27309 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27311         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27312         Generate different code for -mno-speculate-indirect-jumps.
27313         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27314         (*call_indirect_aix<mode>): Disable for
27315         -mno-speculate-indirect-jumps.
27316         (*call_indirect_aix<mode>_nospec): New define_insn.
27317         (*call_value_indirect_aix<mode>): Disable for
27318         -mno-speculate-indirect-jumps.
27319         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27320         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27321         -mno-speculate-indirect-jumps.
27322         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27324 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27326         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27327         long double type, set the flags for noting the default long double
27328         type, even if we don't pass or return a long double type.
27330 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27332         PR ipa/83051
27333         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27334         failure.
27336 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27338         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27339         support for merge[hl].
27340         (fold_mergehl_helper): New helper function.
27341         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27342         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27343         (altivec_vmrglw_direct): Add xxmrglw insn.
27345 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27347         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27348         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27350 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27352         PR lto/83121
27353         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27354         call the lto_location_cache before reading the
27355         DECL_SOURCE_LOCATION of the types.
27357 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27358             Richard Sandiford  <richard.sandiford@linaro.org>
27360         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27361         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27362         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27363         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27364         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27365         Add declaration.
27366         * config/aarch64/constraints.md (aarch64_movti_operand):
27367         Limit immediates.
27368         * config/aarch64/predicates.md (Uti): Add new constraint.
27370 2018-01-17  Carl Love  <cel@us.ibm.com>
27372         * config/rs6000/vsx.md (define_expand xl_len_r,
27373         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27374         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27375         lxvll.
27376         (define_expand, define_insn): Move the shift left from  the
27377         define_insn to the define_expand for lxvl and stxvl instructions.
27378         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27379         and XL_LEN_R definitions to PURE.
27381 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27383         * config/i386/i386.c (indirect_thunk_name): Declare regno
27384         as unsigned int.  Compare regno with INVALID_REGNUM.
27385         (output_indirect_thunk): Ditto.
27386         (output_indirect_thunk_function): Ditto.
27387         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27388         in the call to output_indirect_thunk_function.
27390 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27392         PR middle-end/83884
27393         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27394         rather than the size of inner_type to determine the stack slot size
27395         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27397 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27399         PR target/83546
27400         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27401         to PTA_SILVERMONT.
27403 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27405         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27406         endian Linux systems to optionally enable multilibs for selecting
27407         the long double type if the user configured an explicit type.
27408         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27409         have no long double multilibs if not defined.
27410         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27411         warn if the user used -mabi={ieee,ibm}longdouble and we built
27412         multilibs for long double.
27413         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27414         appropriate multilib option.
27415         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27416         multilib options.
27417         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27418         for building long double multilibs.
27419         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27421 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27423         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27424         copies.
27426         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27427         64 bits.
27428         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27429         128 bits.
27431         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27432         variables.
27434         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27435         return value.
27437 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27439         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27440         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27442 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27444         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27445         different rtl trees depending on TARGET_64BIT.
27446         (rs6000_gen_lvx): Likewise.
27448 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27450         * config/visium/visium.md (nop): Tweak comment.
27451         (hazard_nop): Likewise.
27453 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27455         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27456         -mspeculate-indirect-jumps.
27457         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27458         for -mno-speculate-indirect-jumps.
27459         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27460         (*call_value_indirect_elfv2<mode>): Disable for
27461         -mno-speculate-indirect-jumps.
27462         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27463         (indirect_jump): Emit different RTL for
27464         -mno-speculate-indirect-jumps.
27465         (*indirect_jump<mode>): Disable for
27466         -mno-speculate-indirect-jumps.
27467         (*indirect_jump<mode>_nospec): New define_insn.
27468         (tablejump): Emit different RTL for
27469         -mno-speculate-indirect-jumps.
27470         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27471         (tablejumpsi_nospec): New define_expand.
27472         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27473         (tablejumpdi_nospec): New define_expand.
27474         (*tablejump<mode>_internal1): Disable for
27475         -mno-speculate-indirect-jumps.
27476         (*tablejump<mode>_internal1_nospec): New define_insn.
27477         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27478         option.
27480 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27482         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27483         callers updated.
27485 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27486             Richard Biener  <rguenth@suse.de>
27488         PR libgomp/83590
27489         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27490         return early, inline manually is_gimple_sizepos.  Make sure if we
27491         call gimplify_expr we don't end up with a gimple constant.
27492         * tree.c (variably_modified_type_p): Don't return true for
27493         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27494         * gimplify.h (is_gimple_sizepos): Remove.
27496 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27498         PR tree-optimization/83857
27499         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27500         vectorizable_live_operation for pure SLP statements.
27501         (vectorizable_live_operation): Handle PHIs.
27503 2018-01-16  Richard Biener  <rguenther@suse.de>
27505         PR tree-optimization/83867
27506         * tree-vect-stmts.c (vect_transform_stmt): Precompute
27507         nested_in_vect_loop_p since the scalar stmt may get invalidated.
27509 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27511         PR c/83844
27512         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
27513         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
27514         If off is not INTEGER_CST, issue a may not be aligned warning
27515         rather than isn't aligned.  Use isn%'t rather than isn't.
27516         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
27517         into MULT_EXPR.
27518         <case MULT_EXPR>: Improve the case when bottom and one of the
27519         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
27520         operand, in that case check if the other operand is multiple of
27521         bottom divided by the INTEGER_CST operand.
27523 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27525         PR target/83858
27526         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
27527         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
27528         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
27529         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
27530         * config/pa/pa.c (pa_function_arg_advance): Likewise.
27531         (pa_function_arg, pa_arg_partial_bytes): Likewise.
27532         (pa_function_arg_size): New function.
27534 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27536         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
27537         in a separate statement.
27539 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27541         PR tree-optimization/83847
27542         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
27543         group gathers and scatters.
27545 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27547         PR rtl-optimization/86620
27548         * params.def (max-sched-ready-insns): Bump minimum value to 1.
27550         PR rtl-optimization/83213
27551         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
27552         to last if both are JUMP_INSNs.
27554         PR tree-optimization/83843
27555         * gimple-ssa-store-merging.c
27556         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
27557         store_immediate_info for bswap/nop orig_stores.
27559 2018-01-15  Andrew Waterman  <andrew@sifive.com>
27561         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
27562         !TARGET_MUL.
27563         <UDIV>: Increase cost if !TARGET_DIV.
27565 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
27567         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
27568         (define_attr "cr_logical_3op"): New.
27569         (cceq_ior_compare): Adjust.
27570         (cceq_ior_compare_complement): Adjust.
27571         (*cceq_rev_compare): Adjust.
27572         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
27573         (is_cracked_insn): Adjust.
27574         (insn_must_be_first_in_group): Adjust.
27575         * config/rs6000/40x.md: Adjust.
27576         * config/rs6000/440.md: Adjust.
27577         * config/rs6000/476.md: Adjust.
27578         * config/rs6000/601.md: Adjust.
27579         * config/rs6000/603.md: Adjust.
27580         * config/rs6000/6xx.md: Adjust.
27581         * config/rs6000/7450.md: Adjust.
27582         * config/rs6000/7xx.md: Adjust.
27583         * config/rs6000/8540.md: Adjust.
27584         * config/rs6000/cell.md: Adjust.
27585         * config/rs6000/e300c2c3.md: Adjust.
27586         * config/rs6000/e500mc.md: Adjust.
27587         * config/rs6000/e500mc64.md: Adjust.
27588         * config/rs6000/e5500.md: Adjust.
27589         * config/rs6000/e6500.md: Adjust.
27590         * config/rs6000/mpc.md: Adjust.
27591         * config/rs6000/power4.md: Adjust.
27592         * config/rs6000/power5.md: Adjust.
27593         * config/rs6000/power6.md: Adjust.
27594         * config/rs6000/power7.md: Adjust.
27595         * config/rs6000/power8.md: Adjust.
27596         * config/rs6000/power9.md: Adjust.
27597         * config/rs6000/rs64.md: Adjust.
27598         * config/rs6000/titan.md: Adjust.
27600 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27602         * config/i386/predicates.md (indirect_branch_operand): Rewrite
27603         ix86_indirect_branch_register logic.
27605 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27607         * config/i386/constraints.md (Bs): Update
27608         ix86_indirect_branch_register check.  Don't check
27609         ix86_indirect_branch_register with GOT_memory_operand.
27610         (Bw): Likewise.
27611         * config/i386/predicates.md (GOT_memory_operand): Don't check
27612         ix86_indirect_branch_register here.
27613         (GOT32_symbol_operand): Likewise.
27615 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27617         * config/i386/predicates.md (constant_call_address_operand):
27618         Rewrite ix86_indirect_branch_register logic.
27619         (sibcall_insn_operand): Likewise.
27621 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27623         * config/i386/constraints.md (Bs): Replace
27624         ix86_indirect_branch_thunk_register with
27625         ix86_indirect_branch_register.
27626         (Bw): Likewise.
27627         * config/i386/i386.md (indirect_jump): Likewise.
27628         (tablejump): Likewise.
27629         (*sibcall_memory): Likewise.
27630         (*sibcall_value_memory): Likewise.
27631         Peepholes of indirect call and jump via memory: Likewise.
27632         * config/i386/i386.opt: Likewise.
27633         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27634         (GOT_memory_operand): Likewise.
27635         (call_insn_operand): Likewise.
27636         (sibcall_insn_operand): Likewise.
27637         (GOT32_symbol_operand): Likewise.
27639 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27641         PR middle-end/83837
27642         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
27643         type rather than type addr's type points to.
27644         (expand_omp_atomic_mutex): Likewise.
27645         (expand_omp_atomic): Likewise.
27647 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27649         PR target/83839
27650         * config/i386/i386.c (output_indirect_thunk_function): Use
27651         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
27652         for  __x86_return_thunk.
27654 2018-01-15  Richard Biener  <rguenther@suse.de>
27656         PR middle-end/83850
27657         * expmed.c (extract_bit_field_1): Fix typo.
27659 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27661         PR target/83687
27662         * config/arm/iterators.md (VF): New mode iterator.
27663         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
27664         Remove integer-related logic from pattern.
27665         (neon_vabd<mode>_3): Likewise.
27667 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27669         PR middle-end/82694
27670         * common.opt (fstrict-overflow): No longer an alias.
27671         (fwrapv-pointer): New option.
27672         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
27673         also for pointer types based on flag_wrapv_pointer.
27674         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
27675         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
27676         opts->x_flag_wrapv got set.
27677         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
27678         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
27679         POINTER_TYPE_OVERFLOW_UNDEFINED.
27680         * match.pd: Likewise in address comparison pattern.
27681         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
27683 2018-01-15  Richard Biener  <rguenther@suse.de>
27685         PR lto/83804
27686         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
27687         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
27688         Reset type names to their identifier if their TYPE_DECL doesn't
27689         have linkage (and thus is used for ODR and devirt).
27690         (save_debug_info_for_decl): Remove.
27691         (save_debug_info_for_type): Likewise.
27692         (add_tree_to_fld_list): Adjust.
27693         * tree-pretty-print.c (dump_generic_node): Make dumping of
27694         type names more robust.
27696 2018-01-15  Richard Biener  <rguenther@suse.de>
27698         * BASE-VER: Bump to 8.0.1.
27700 2018-01-14  Martin Sebor  <msebor@redhat.com>
27702         PR other/83508
27703         * builtins.c (check_access): Avoid warning when the no-warning bit
27704         is set.
27706 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
27708         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
27709         * ira-color (allocno_hard_regs_compare): Likewise.
27711 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
27713         PR target/83013
27714         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
27715         Use .pushsection/.popsection.
27717 2018-01-14  Martin Sebor  <msebor@redhat.com>
27719         PR c++/81327
27720         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
27722 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
27724         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
27725         entry from extra_headers.
27726         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
27727         extra_headers, make the list bitwise identical to the i?86-*-* one.
27729 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27731         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
27732         -mcmodel=large with -mindirect-branch=thunk,
27733         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
27734         -mfunction-return=thunk-extern.
27735         * doc/invoke.texi: Document -mcmodel=large is incompatible with
27736         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
27737         -mfunction-return=thunk and -mfunction-return=thunk-extern.
27739 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27741         * config/i386/i386.c (print_reg): Print the name of the full
27742         integer register without '%'.
27743         (ix86_print_operand): Handle 'V'.
27744         * doc/extend.texi: Document 'V' modifier.
27746 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27748         * config/i386/constraints.md (Bs): Disallow memory operand for
27749         -mindirect-branch-register.
27750         (Bw): Likewise.
27751         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27752         (GOT_memory_operand): Likewise.
27753         (call_insn_operand): Likewise.
27754         (sibcall_insn_operand): Likewise.
27755         (GOT32_symbol_operand): Likewise.
27756         * config/i386/i386.md (indirect_jump): Call convert_memory_address
27757         for -mindirect-branch-register.
27758         (tablejump): Likewise.
27759         (*sibcall_memory): Likewise.
27760         (*sibcall_value_memory): Likewise.
27761         Disallow peepholes of indirect call and jump via memory for
27762         -mindirect-branch-register.
27763         (*call_pop): Replace m with Bw.
27764         (*call_value_pop): Likewise.
27765         (*sibcall_pop_memory): Replace m with Bs.
27766         * config/i386/i386.opt (mindirect-branch-register): New option.
27767         * doc/invoke.texi: Document -mindirect-branch-register option.
27769 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27771         * config/i386/i386-protos.h (ix86_output_function_return): New.
27772         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
27773         set function_return_type.
27774         (indirect_thunk_name): Add ret_p to indicate thunk for function
27775         return.
27776         (output_indirect_thunk_function): Pass false to
27777         indirect_thunk_name.
27778         (ix86_output_indirect_branch_via_reg): Likewise.
27779         (ix86_output_indirect_branch_via_push): Likewise.
27780         (output_indirect_thunk_function): Create alias for function
27781         return thunk if regno < 0.
27782         (ix86_output_function_return): New function.
27783         (ix86_handle_fndecl_attribute): Handle function_return.
27784         (ix86_attribute_table): Add function_return.
27785         * config/i386/i386.h (machine_function): Add
27786         function_return_type.
27787         * config/i386/i386.md (simple_return_internal): Use
27788         ix86_output_function_return.
27789         (simple_return_internal_long): Likewise.
27790         * config/i386/i386.opt (mfunction-return=): New option.
27791         (indirect_branch): Mention -mfunction-return=.
27792         * doc/extend.texi: Document function_return function attribute.
27793         * doc/invoke.texi: Document -mfunction-return= option.
27795 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27797         * config/i386/i386-opts.h (indirect_branch): New.
27798         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
27799         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
27800         with local indirect jump when converting indirect call and jump.
27801         (ix86_set_indirect_branch_type): New.
27802         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
27803         (indirectlabelno): New.
27804         (indirect_thunk_needed): Likewise.
27805         (indirect_thunk_bnd_needed): Likewise.
27806         (indirect_thunks_used): Likewise.
27807         (indirect_thunks_bnd_used): Likewise.
27808         (INDIRECT_LABEL): Likewise.
27809         (indirect_thunk_name): Likewise.
27810         (output_indirect_thunk): Likewise.
27811         (output_indirect_thunk_function): Likewise.
27812         (ix86_output_indirect_branch_via_reg): Likewise.
27813         (ix86_output_indirect_branch_via_push): Likewise.
27814         (ix86_output_indirect_branch): Likewise.
27815         (ix86_output_indirect_jmp): Likewise.
27816         (ix86_code_end): Call output_indirect_thunk_function if needed.
27817         (ix86_output_call_insn): Call ix86_output_indirect_branch if
27818         needed.
27819         (ix86_handle_fndecl_attribute): Handle indirect_branch.
27820         (ix86_attribute_table): Add indirect_branch.
27821         * config/i386/i386.h (machine_function): Add indirect_branch_type
27822         and has_local_indirect_jump.
27823         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
27824         to true.
27825         (tablejump): Likewise.
27826         (*indirect_jump): Use ix86_output_indirect_jmp.
27827         (*tablejump_1): Likewise.
27828         (simple_return_indirect_internal): Likewise.
27829         * config/i386/i386.opt (mindirect-branch=): New option.
27830         (indirect_branch): New.
27831         (keep): Likewise.
27832         (thunk): Likewise.
27833         (thunk-inline): Likewise.
27834         (thunk-extern): Likewise.
27835         * doc/extend.texi: Document indirect_branch function attribute.
27836         * doc/invoke.texi: Document -mindirect-branch= option.
27838 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
27840         PR ipa/83051
27841         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
27843 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
27845         * ipa-inline.c (want_inline_small_function_p): Return false if
27846         inlining has already failed with CIF_FINAL_ERROR.
27847         (update_caller_keys): Call want_inline_small_function_p before
27848         can_inline_edge_p.
27849         (update_callee_keys): Likewise.
27851 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27853         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
27854         New function.
27855         (rs6000_quadword_masked_address_p): Likewise.
27856         (quad_aligned_load_p): Likewise.
27857         (quad_aligned_store_p): Likewise.
27858         (const_load_sequence_p): Add comment to describe the outer-most loop.
27859         (mimic_memory_attributes_and_flags): New function.
27860         (rs6000_gen_stvx): Likewise.
27861         (replace_swapped_aligned_store): Likewise.
27862         (rs6000_gen_lvx): Likewise.
27863         (replace_swapped_aligned_load): Likewise.
27864         (replace_swapped_load_constant): Capitalize argument name in
27865         comment describing this function.
27866         (rs6000_analyze_swaps): Add a third pass to search for vector loads
27867         and stores that access quad-word aligned addresses and replace
27868         with stvx or lvx instructions when appropriate.
27869         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
27870         New function prototype.
27871         (rs6000_quadword_masked_address_p): Likewise.
27872         (rs6000_gen_lvx): Likewise.
27873         (rs6000_gen_stvx): Likewise.
27874         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
27875         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
27876         when memory address is aligned.
27877         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
27878         this split to select lvx instruction when memory address is aligned.
27879         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
27880         instruction when memory address is aligned.
27881         (*vsx_le_perm_load_v16qi): Likewise.
27882         (four unnamed splitters): Modify to select the stvx instruction
27883         when memory is aligned.
27885 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
27887         * predict.c (determine_unlikely_bbs): Handle correctly BBs
27888         which appears in the queue multiple times.
27890 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27891             Alan Hayward  <alan.hayward@arm.com>
27892             David Sherwood  <david.sherwood@arm.com>
27894         * tree-vectorizer.h (vec_lower_bound): New structure.
27895         (_loop_vec_info): Add check_nonzero and lower_bounds.
27896         (LOOP_VINFO_CHECK_NONZERO): New macro.
27897         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
27898         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
27899         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
27900         fields.  Make seg_len the distance travelled, not including the
27901         access size.
27902         (dr_direction_indicator): Declare.
27903         (dr_zero_step_indicator): Likewise.
27904         (dr_known_forward_stride_p): Likewise.
27905         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
27906         tree-ssanames.h.
27907         (runtime_alias_check_p): Allow runtime alias checks with
27908         variable strides.
27909         (operator ==): Compare access_size and align.
27910         (prune_runtime_alias_test_list): Rework for new distinction between
27911         the access_size and seg_len.
27912         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
27913         segment lengths.
27914         (get_segment_min_max): New function.
27915         (create_intersect_range_checks): Use it.
27916         (dr_step_indicator): New function.
27917         (dr_direction_indicator): Likewise.
27918         (dr_zero_step_indicator): Likewise.
27919         (dr_known_forward_stride_p): Likewise.
27920         * tree-loop-distribution.c (data_ref_segment_size): Return
27921         DR_STEP * (niters - 1).
27922         (compute_alias_check_pairs): Update call to the dr_with_seg_len
27923         constructor.
27924         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
27925         (vect_preserves_scalar_order_p): New function, split out from...
27926         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
27927         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
27928         (vect_vfa_access_size): New function.
27929         (vect_vfa_align): Likewise.
27930         (vect_compile_time_alias): Take access_size_a and access_b arguments.
27931         (dump_lower_bound): New function.
27932         (vect_check_lower_bound): Likewise.
27933         (vect_small_gap_p): Likewise.
27934         (vectorizable_with_step_bound_p): Likewise.
27935         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
27936         depencies if the vectorization factor is 1.  Convert the checks
27937         for nonzero steps into checks on the bounds of DR_STEP.  Try using
27938         a bunds check for variable steps if the minimum required step is
27939         relatively small. Update calls to the dr_with_seg_len
27940         constructor and to vect_compile_time_alias.
27941         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
27942         function.
27943         (vect_loop_versioning): Call it.
27944         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
27945         when retrying.
27946         (vect_estimate_min_profitable_iters): Account for any bounds checks.
27948 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27949             Alan Hayward  <alan.hayward@arm.com>
27950             David Sherwood  <david.sherwood@arm.com>
27952         * doc/sourcebuild.texi (vect_scatter_store): Document.
27953         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
27954         optabs.
27955         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
27956         Document.
27957         * genopinit.c (main): Add supports_vec_scatter_store and
27958         supports_vec_scatter_store_cached to target_optabs.
27959         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
27960         IFN_MASK_SCATTER_STORE.
27961         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
27962         functions.
27963         * internal-fn.h (internal_store_fn_p): Declare.
27964         (internal_fn_stored_value_index): Likewise.
27965         * internal-fn.c (scatter_store_direct): New macro.
27966         (expand_scatter_store_optab_fn): New function.
27967         (direct_scatter_store_optab_supported_p): New macro.
27968         (internal_store_fn_p): New function.
27969         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
27970         IFN_MASK_SCATTER_STORE.
27971         (internal_fn_mask_index): Likewise.
27972         (internal_fn_stored_value_index): New function.
27973         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
27974         for scatter stores.
27975         * optabs-query.h (supports_vec_scatter_store_p): Declare.
27976         * optabs-query.c (supports_vec_scatter_store_p): New function.
27977         * tree-vectorizer.h (vect_get_store_rhs): Declare.
27978         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
27979         true for scatter stores.
27980         (vect_gather_scatter_fn_p): Handle scatter stores too.
27981         (vect_check_gather_scatter): Consider using scatter stores if
27982         supports_vec_scatter_store_p.
27983         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
27984         scatter stores too.
27985         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27986         internal_fn_stored_value_index.
27987         (check_load_store_masking): Handle scatter stores too.
27988         (vect_get_store_rhs): Make public.
27989         (vectorizable_call): Use internal_store_fn_p.
27990         (vectorizable_store): Handle scatter store internal functions.
27991         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
27992         when deciding whether the end of the group has been reached.
27993         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
27994         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
27995         (mask_scatter_store<mode>): New insns.
27997 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27998             Alan Hayward  <alan.hayward@arm.com>
27999             David Sherwood  <david.sherwood@arm.com>
28001         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28002         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28003         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28004         function.
28005         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28006         Use vect_truncate_gather_scatter_offset if we can't treat the
28007         operation as a normal gather load or scatter store.
28008         (get_group_load_store_type): Take the gather_scatter_info
28009         as argument.  Try using a gather load or scatter store for
28010         single-element groups.
28011         (get_load_store_type): Update calls to get_group_load_store_type
28012         and vect_use_strided_gather_scatters_p.
28014 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28015             Alan Hayward  <alan.hayward@arm.com>
28016             David Sherwood  <david.sherwood@arm.com>
28018         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28019         optional tree argument.
28020         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28021         null target hooks.
28022         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28023         but continue to use the current value as a fallback.
28024         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28025         to compare the updates.
28026         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28027         (get_load_store_type): Use it when handling a strided access.
28028         (vect_get_strided_load_store_ops): New function.
28029         (vect_get_data_ptr_increment): Likewise.
28030         (vectorizable_load): Handle strided gather loads.  Always pass
28031         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28033 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28034             Alan Hayward  <alan.hayward@arm.com>
28035             David Sherwood  <david.sherwood@arm.com>
28037         * doc/md.texi (gather_load@var{m}): Document.
28038         (mask_gather_load@var{m}): Likewise.
28039         * genopinit.c (main): Add supports_vec_gather_load and
28040         supports_vec_gather_load_cached to target_optabs.
28041         * optabs-tree.c (init_tree_optimization_optabs): Use
28042         ggc_cleared_alloc to allocate target_optabs.
28043         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28044         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28045         functions.
28046         * internal-fn.h (internal_load_fn_p): Declare.
28047         (internal_gather_scatter_fn_p): Likewise.
28048         (internal_fn_mask_index): Likewise.
28049         (internal_gather_scatter_fn_supported_p): Likewise.
28050         * internal-fn.c (gather_load_direct): New macro.
28051         (expand_gather_load_optab_fn): New function.
28052         (direct_gather_load_optab_supported_p): New macro.
28053         (direct_internal_fn_optab): New function.
28054         (internal_load_fn_p): Likewise.
28055         (internal_gather_scatter_fn_p): Likewise.
28056         (internal_fn_mask_index): Likewise.
28057         (internal_gather_scatter_fn_supported_p): Likewise.
28058         * optabs-query.c (supports_at_least_one_mode_p): New function.
28059         (supports_vec_gather_load_p): Likewise.
28060         * optabs-query.h (supports_vec_gather_load_p): Declare.
28061         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28062         and memory_type field.
28063         (NUM_PATTERNS): Bump to 15.
28064         * tree-vect-data-refs.c: Include internal-fn.h.
28065         (vect_gather_scatter_fn_p): New function.
28066         (vect_describe_gather_scatter_call): Likewise.
28067         (vect_check_gather_scatter): Try using internal functions for
28068         gather loads.  Recognize existing calls to a gather load function.
28069         (vect_analyze_data_refs): Consider using gather loads if
28070         supports_vec_gather_load_p.
28071         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28072         (vect_get_gather_scatter_offset_type): Likewise.
28073         (vect_convert_mask_for_vectype): Likewise.
28074         (vect_add_conversion_to_patterm): Likewise.
28075         (vect_try_gather_scatter_pattern): Likewise.
28076         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28077         (vect_vect_recog_func_ptrs): Add it.
28078         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28079         internal_fn_mask_index and internal_gather_scatter_fn_p.
28080         (check_load_store_masking): Take the gather_scatter_info as an
28081         argument and handle gather loads.
28082         (vect_get_gather_scatter_ops): New function.
28083         (vectorizable_call): Check internal_load_fn_p.
28084         (vectorizable_load): Likewise.  Handle gather load internal
28085         functions.
28086         (vectorizable_store): Update call to check_load_store_masking.
28087         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28088         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28089         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28090         (aarch64_gather_scale_operand_d): New predicates.
28091         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28092         (mask_gather_load<mode>): New insns.
28094 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28095             Alan Hayward  <alan.hayward@arm.com>
28096             David Sherwood  <david.sherwood@arm.com>
28098         * optabs.def (fold_left_plus_optab): New optab.
28099         * doc/md.texi (fold_left_plus_@var{m}): Document.
28100         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28101         * internal-fn.c (fold_left_direct): Define.
28102         (expand_fold_left_optab_fn): Likewise.
28103         (direct_fold_left_optab_supported_p): Likewise.
28104         * fold-const-call.c (fold_const_fold_left): New function.
28105         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28106         * tree-parloops.c (valid_reduction_p): New function.
28107         (gather_scalar_reductions): Use it.
28108         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28109         (vect_finish_replace_stmt): Declare.
28110         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28111         (needs_fold_left_reduction_p): New function, split out from...
28112         (vect_is_simple_reduction): ...here.  Accept reductions that
28113         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28114         (vect_force_simple_reduction): Also store the reduction type in
28115         the assignment's STMT_VINFO_REDUC_TYPE.
28116         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28117         (merge_with_identity): New function.
28118         (vect_expand_fold_left): Likewise.
28119         (vectorize_fold_left_reduction): Likewise.
28120         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28121         scalar phi in place for it.  Check for target support and reject
28122         cases that would reassociate the operation.  Defer the transform
28123         phase to vectorize_fold_left_reduction.
28124         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28125         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28126         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28128 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28130         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28131         call to ifc_temp_var.
28133 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28134             Alan Hayward  <alan.hayward@arm.com>
28135             David Sherwood  <david.sherwood@arm.com>
28137         * target.def (legitimize_address_displacement): Take the original
28138         offset as a poly_int.
28139         * targhooks.h (default_legitimize_address_displacement): Update
28140         accordingly.
28141         * targhooks.c (default_legitimize_address_displacement): Likewise.
28142         * doc/tm.texi: Regenerate.
28143         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28144         as an argument, moving assert of ad->disp == ad->disp_term to...
28145         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28146         Try calling targetm.legitimize_address_displacement before expanding
28147         the address rather than afterwards, and adjust for the new interface.
28148         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28149         Match the new hook interface.  Handle SVE addresses.
28150         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28151         new hook interface.
28153 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28155         * Makefile.in (OBJS): Add early-remat.o.
28156         * target.def (select_early_remat_modes): New hook.
28157         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28158         * doc/tm.texi: Regenerate.
28159         * targhooks.h (default_select_early_remat_modes): Declare.
28160         * targhooks.c (default_select_early_remat_modes): New function.
28161         * timevar.def (TV_EARLY_REMAT): New timevar.
28162         * passes.def (pass_early_remat): New pass.
28163         * tree-pass.h (make_pass_early_remat): Declare.
28164         * early-remat.c: New file.
28165         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28166         function.
28167         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28169 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28170             Alan Hayward  <alan.hayward@arm.com>
28171             David Sherwood  <david.sherwood@arm.com>
28173         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28174         vfm1 with a bound_epilog parameter.
28175         (vect_do_peeling): Update calls accordingly, and move the prologue
28176         call earlier in the function.  Treat the base bound_epilog as 0 for
28177         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28178         this base when peeling for gaps.
28179         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28180         with fully-masked loops.
28181         (vect_estimate_min_profitable_iters): Handle the single peeled
28182         iteration in that case.
28184 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28185             Alan Hayward  <alan.hayward@arm.com>
28186             David Sherwood  <david.sherwood@arm.com>
28188         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28189         single-element interleaving even if the size is not a power of 2.
28190         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28191         accesses for single-element interleaving if the group size is
28192         not a power of 2.
28194 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28195             Alan Hayward  <alan.hayward@arm.com>
28196             David Sherwood  <david.sherwood@arm.com>
28198         * doc/md.texi (fold_extract_last_@var{m}): Document.
28199         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28200         * optabs.def (fold_extract_last_optab): New optab.
28201         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28202         * internal-fn.c (fold_extract_direct): New macro.
28203         (expand_fold_extract_optab_fn): Likewise.
28204         (direct_fold_extract_optab_supported_p): Likewise.
28205         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28206         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28207         EXTRACT_LAST_REDUCTION.
28208         (get_initial_def_for_reduction): Do not create an initial vector
28209         for EXTRACT_LAST_REDUCTION reductions.
28210         (vectorizable_reduction): Leave the scalar phi in place for
28211         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28212         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28213         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28214         transform phase to vectorizable_condition.
28215         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28216         split out from...
28217         (vect_finish_stmt_generation): ...here.
28218         (vect_finish_replace_stmt): New function.
28219         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28220         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28221         pattern.
28222         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28224 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28225             Alan Hayward  <alan.hayward@arm.com>
28226             David Sherwood  <david.sherwood@arm.com>
28228         * doc/md.texi (extract_last_@var{m}): Document.
28229         * optabs.def (extract_last_optab): New optab.
28230         * internal-fn.def (EXTRACT_LAST): New internal function.
28231         * internal-fn.c (cond_unary_direct): New macro.
28232         (expand_cond_unary_optab_fn): Likewise.
28233         (direct_cond_unary_optab_supported_p): Likewise.
28234         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28235         loops using EXTRACT_LAST.
28236         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28237         (extract_last_<mode>): ...this optab.
28238         (vec_extract<mode><Vel>): Update accordingly.
28240 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28241             Alan Hayward  <alan.hayward@arm.com>
28242             David Sherwood  <david.sherwood@arm.com>
28244         * target.def (empty_mask_is_expensive): New hook.
28245         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28246         * doc/tm.texi: Regenerate.
28247         * targhooks.h (default_empty_mask_is_expensive): Declare.
28248         * targhooks.c (default_empty_mask_is_expensive): New function.
28249         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28250         if the target says that empty masks are expensive.
28251         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28252         New function.
28253         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28255 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28256             Alan Hayward  <alan.hayward@arm.com>
28257             David Sherwood  <david.sherwood@arm.com>
28259         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28260         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28261         (vect_use_loop_mask_for_alignment_p): New function.
28262         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28263         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28264         niters_skip argument.  Make sure that the first niters_skip elements
28265         of the first iteration are inactive.
28266         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28267         Update call to vect_set_loop_masks_directly.
28268         (get_misalign_in_elems): New function, split out from...
28269         (vect_gen_prolog_loop_niters): ...here.
28270         (vect_update_init_of_dr): Take a code argument that specifies whether
28271         the adjustment should be added or subtracted.
28272         (vect_update_init_of_drs): Likewise.
28273         (vect_prepare_for_masked_peels): New function.
28274         (vect_do_peeling): Skip prologue peeling if we're using a mask
28275         instead.  Update call to vect_update_inits_of_drs.
28276         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28277         mask_skip_niters.
28278         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28279         alignment.  Do not include the number of peeled iterations in
28280         the minimum threshold in that case.
28281         (vectorizable_induction): Adjust the start value down by
28282         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28283         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28284         Take the number of skipped iterations into account when calculating
28285         the loop bounds.
28286         * tree-vect-stmts.c (vect_gen_while_not): New function.
28288 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28289             Alan Hayward  <alan.hayward@arm.com>
28290             David Sherwood  <david.sherwood@arm.com>
28292         * doc/sourcebuild.texi (vect_fully_masked): Document.
28293         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28294         default value to 0.
28295         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28296         split out from...
28297         (vect_analyze_loop_2): ...here. Don't check the vectorization
28298         factor against the number of loop iterations if the loop is
28299         fully-masked.
28301 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28302             Alan Hayward  <alan.hayward@arm.com>
28303             David Sherwood  <david.sherwood@arm.com>
28305         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28306         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28307         (dump_groups): Update accordingly.
28308         (iv_use::mem_type): New member variable.
28309         (address_p): New function.
28310         (record_use): Add a mem_type argument and initialize the new
28311         mem_type field.
28312         (record_group_use): Add a mem_type argument.  Use address_p.
28313         Remove obsolete null checks of base_object.  Update call to record_use.
28314         (find_interesting_uses_op): Update call to record_group_use.
28315         (find_interesting_uses_cond): Likewise.
28316         (find_interesting_uses_address): Likewise.
28317         (get_mem_type_for_internal_fn): New function.
28318         (find_address_like_use): Likewise.
28319         (find_interesting_uses_stmt): Try find_address_like_use before
28320         calling find_interesting_uses_op.
28321         (addr_offset_valid_p): Use the iv mem_type field as the type
28322         of the addressed memory.
28323         (add_autoinc_candidates): Likewise.
28324         (get_address_cost): Likewise.
28325         (split_small_address_groups_p): Use address_p.
28326         (split_address_groups): Likewise.
28327         (add_iv_candidate_for_use): Likewise.
28328         (autoinc_possible_for_pair): Likewise.
28329         (rewrite_groups): Likewise.
28330         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28331         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28332         (get_alias_ptr_type_for_ptr_address): New function.
28333         (rewrite_use_address): Rewrite address uses in calls that were
28334         identified by find_address_like_use.
28336 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28337             Alan Hayward  <alan.hayward@arm.com>
28338             David Sherwood  <david.sherwood@arm.com>
28340         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28341         TARGET_MEM_REFs.
28342         * gimple-expr.h (is_gimple_addressable: Likewise.
28343         * gimple-expr.c (is_gimple_address): Likewise.
28344         * internal-fn.c (expand_call_mem_ref): New function.
28345         (expand_mask_load_optab_fn): Use it.
28346         (expand_mask_store_optab_fn): Likewise.
28348 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28349             Alan Hayward  <alan.hayward@arm.com>
28350             David Sherwood  <david.sherwood@arm.com>
28352         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28353         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28354         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28355         (cond_umax@var{mode}): Document.
28356         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28357         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28358         (cond_umin_optab, cond_umax_optab): New optabs.
28359         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28360         (COND_IOR, COND_XOR): New internal functions.
28361         * internal-fn.h (get_conditional_internal_fn): Declare.
28362         * internal-fn.c (cond_binary_direct): New macro.
28363         (expand_cond_binary_optab_fn): Likewise.
28364         (direct_cond_binary_optab_supported_p): Likewise.
28365         (get_conditional_internal_fn): New function.
28366         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28367         Cope with reduction statements that are vectorized as calls rather
28368         than assignments.
28369         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28370         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28371         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28372         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28373         (UNSPEC_COND_EOR): New unspecs.
28374         (optab): Add mappings for them.
28375         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28376         (sve_int_op, sve_fp_op): New int attributes.
28378 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28379             Alan Hayward  <alan.hayward@arm.com>
28380             David Sherwood  <david.sherwood@arm.com>
28382         * optabs.def (while_ult_optab): New optab.
28383         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28384         * internal-fn.def (WHILE_ULT): New internal function.
28385         * internal-fn.h (direct_internal_fn_supported_p): New override
28386         that takes two types as argument.
28387         * internal-fn.c (while_direct): New macro.
28388         (expand_while_optab_fn): New function.
28389         (convert_optab_supported_p): Likewise.
28390         (direct_while_optab_supported_p): New macro.
28391         * wide-int.h (wi::udiv_ceil): New function.
28392         * tree-vectorizer.h (rgroup_masks): New structure.
28393         (vec_loop_masks): New typedef.
28394         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28395         and fully_masked_p.
28396         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28397         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28398         (vect_max_vf): New function.
28399         (slpeel_make_loop_iterate_ntimes): Delete.
28400         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28401         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28402         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28403         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28404         internal-fn.h, stor-layout.h and optabs-query.h.
28405         (vect_set_loop_mask): New function.
28406         (add_preheader_seq): Likewise.
28407         (add_header_seq): Likewise.
28408         (interleave_supported_p): Likewise.
28409         (vect_maybe_permute_loop_masks): Likewise.
28410         (vect_set_loop_masks_directly): Likewise.
28411         (vect_set_loop_condition_masked): Likewise.
28412         (vect_set_loop_condition_unmasked): New function, split out from
28413         slpeel_make_loop_iterate_ntimes.
28414         (slpeel_make_loop_iterate_ntimes): Rename to..
28415         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28416         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28417         (vect_do_peeling): Update call accordingly.
28418         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28419         loops.
28420         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28421         mask_compare_type, can_fully_mask_p and fully_masked_p.
28422         (release_vec_loop_masks): New function.
28423         (_loop_vec_info): Use it to free the loop masks.
28424         (can_produce_all_loop_masks_p): New function.
28425         (vect_get_max_nscalars_per_iter): Likewise.
28426         (vect_verify_full_masking): Likewise.
28427         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28428         retries, and free the mask rgroups before retrying.  Check loop-wide
28429         reasons for disallowing fully-masked loops.  Make the final decision
28430         about whether use a fully-masked loop or not.
28431         (vect_estimate_min_profitable_iters): Do not assume that peeling
28432         for the number of iterations will be needed for fully-masked loops.
28433         (vectorizable_reduction): Disable fully-masked loops.
28434         (vectorizable_live_operation): Likewise.
28435         (vect_halve_mask_nunits): New function.
28436         (vect_double_mask_nunits): Likewise.
28437         (vect_record_loop_mask): Likewise.
28438         (vect_get_loop_mask): Likewise.
28439         (vect_transform_loop): Handle the case in which the final loop
28440         iteration might handle a partial vector.  Call vect_set_loop_condition
28441         instead of slpeel_make_loop_iterate_ntimes.
28442         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28443         (check_load_store_masking): New function.
28444         (prepare_load_store_mask): Likewise.
28445         (vectorizable_store): Handle fully-masked loops.
28446         (vectorizable_load): Likewise.
28447         (supportable_widening_operation): Use vect_halve_mask_nunits for
28448         booleans.
28449         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28450         (vect_gen_while): New function.
28451         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28452         (aarch64_uqdec<mode>): New insn.
28454 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28455             Alan Hayward  <alan.hayward@arm.com>
28456             David Sherwood  <david.sherwood@arm.com>
28458         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28459         (reduc_xor_scal_optab): New optabs.
28460         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28461         (reduc_xor_scal_@var{m}): Document.
28462         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28463         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28464         internal functions.
28465         * fold-const-call.c (fold_const_call): Handle them.
28466         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28467         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28468         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28469         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28470         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28471         (UNSPEC_XORV): New unspecs.
28472         (optab): Add entries for them.
28473         (BITWISEV): New int iterator.
28474         (bit_reduc_op): New int attributes.
28476 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28477             Alan Hayward  <alan.hayward@arm.com>
28478             David Sherwood  <david.sherwood@arm.com>
28480         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28481         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28482         * optabs.def (vec_shl_insert_optab): New optab.
28483         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28484         (duplicate_and_interleave): Likewise.
28485         * tree-vect-loop.c: Include internal-fn.h.
28486         (neutral_op_for_slp_reduction): New function, split out from
28487         get_initial_defs_for_reduction.
28488         (get_initial_def_for_reduction): Handle option 2 for variable-length
28489         vectors by loading the neutral value into a vector and then shifting
28490         the initial value into element 0.
28491         (get_initial_defs_for_reduction): Replace the code argument with
28492         the neutral value calculated by neutral_op_for_slp_reduction.
28493         Use gimple_build_vector for constant-length vectors.
28494         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28495         but the first group_size elements have a neutral value.
28496         Use duplicate_and_interleave otherwise.
28497         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28498         Update call to get_initial_defs_for_reduction.  Handle SLP
28499         reductions for variable-length vectors by creating one vector
28500         result for each scalar result, with the elements associated
28501         with other scalar results stubbed out with the neutral value.
28502         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
28503         Require IFN_VEC_SHL_INSERT for double reductions on
28504         variable-length vectors, or SLP reductions that have
28505         a neutral value.  Require can_duplicate_and_interleave_p
28506         support for variable-length unchained SLP reductions if there
28507         is no neutral value, such as for MIN/MAX reductions.  Also require
28508         the number of vector elements to be a multiple of the number of
28509         SLP statements when doing variable-length unchained SLP reductions.
28510         Update call to vect_create_epilog_for_reduction.
28511         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
28512         and remove initial values.
28513         (duplicate_and_interleave): Make public.
28514         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
28515         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
28517 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28518             Alan Hayward  <alan.hayward@arm.com>
28519             David Sherwood  <david.sherwood@arm.com>
28521         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
28522         (can_duplicate_and_interleave_p): New function.
28523         (vect_get_and_check_slp_defs): Take the vector of statements
28524         rather than just the current one.  Remove excess parentheses.
28525         Restriction rejectinon of vect_constant_def and vect_external_def
28526         for variable-length vectors to boolean types, or types for which
28527         can_duplicate_and_interleave_p is false.
28528         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
28529         (duplicate_and_interleave): New function.
28530         (vect_get_constant_vectors): Use gimple_build_vector for
28531         constant-length vectors and suitable variable-length constant
28532         vectors.  Use duplicate_and_interleave for other variable-length
28533         vectors.  Don't defer the update when inserting new statements.
28535 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28536             Alan Hayward  <alan.hayward@arm.com>
28537             David Sherwood  <david.sherwood@arm.com>
28539         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
28540         min_profitable_iters doesn't go negative.
28542 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28543             Alan Hayward  <alan.hayward@arm.com>
28544             David Sherwood  <david.sherwood@arm.com>
28546         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
28547         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
28548         * optabs.def (vec_mask_load_lanes_optab): New optab.
28549         (vec_mask_store_lanes_optab): Likewise.
28550         * internal-fn.def (MASK_LOAD_LANES): New internal function.
28551         (MASK_STORE_LANES): Likewise.
28552         * internal-fn.c (mask_load_lanes_direct): New macro.
28553         (mask_store_lanes_direct): Likewise.
28554         (expand_mask_load_optab_fn): Handle masked operations.
28555         (expand_mask_load_lanes_optab_fn): New macro.
28556         (expand_mask_store_optab_fn): Handle masked operations.
28557         (expand_mask_store_lanes_optab_fn): New macro.
28558         (direct_mask_load_lanes_optab_supported_p): Likewise.
28559         (direct_mask_store_lanes_optab_supported_p): Likewise.
28560         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
28561         parameter.
28562         (vect_load_lanes_supported): Likewise.
28563         * tree-vect-data-refs.c (strip_conversion): New function.
28564         (can_group_stmts_p): Likewise.
28565         (vect_analyze_data_ref_accesses): Use it instead of checking
28566         for a pair of assignments.
28567         (vect_store_lanes_supported): Take a masked_p parameter.
28568         (vect_load_lanes_supported): Likewise.
28569         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
28570         vect_store_lanes_supported and vect_load_lanes_supported.
28571         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
28572         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
28573         parameter.  Don't allow gaps for masked accesses.
28574         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
28575         and vect_load_lanes_supported.
28576         (get_load_store_type): Take a masked_p parameter and update
28577         call to get_group_load_store_type.
28578         (vectorizable_store): Update call to get_load_store_type.
28579         Handle IFN_MASK_STORE_LANES.
28580         (vectorizable_load): Update call to get_load_store_type.
28581         Handle IFN_MASK_LOAD_LANES.
28583 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28584             Alan Hayward  <alan.hayward@arm.com>
28585             David Sherwood  <david.sherwood@arm.com>
28587         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
28588         modes for SVE.
28589         * config/aarch64/aarch64-protos.h
28590         (aarch64_sve_struct_memory_operand_p): Declare.
28591         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
28592         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
28593         (VPRED, vpred): Handle SVE structure modes.
28594         * config/aarch64/constraints.md (Utx): New constraint.
28595         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
28596         (aarch64_sve_struct_nonimmediate_operand): New predicates.
28597         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
28598         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
28599         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
28600         structure modes.  Split into pieces after RA.
28601         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
28602         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
28603         New patterns.
28604         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
28605         SVE structure modes.
28606         (aarch64_classify_address): Likewise.
28607         (sizetochar): Move earlier in file.
28608         (aarch64_print_operand): Handle SVE register lists.
28609         (aarch64_array_mode): New function.
28610         (aarch64_sve_struct_memory_operand_p): Likewise.
28611         (TARGET_ARRAY_MODE): Redefine.
28613 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28614             Alan Hayward  <alan.hayward@arm.com>
28615             David Sherwood  <david.sherwood@arm.com>
28617         * target.def (array_mode): New target hook.
28618         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
28619         * doc/tm.texi: Regenerate.
28620         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
28621         * hooks.c (hook_optmode_mode_uhwi_none): New function.
28622         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
28623         targetm.array_mode.
28624         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
28625         type sizes.
28627 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28628             Alan Hayward  <alan.hayward@arm.com>
28629             David Sherwood  <david.sherwood@arm.com>
28631         * fold-const.c (fold_binary_loc): Check the argument types
28632         rather than the result type when testing for a vector operation.
28634 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28636         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
28637         * doc/tm.texi: Regenerate.
28639 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28640             Alan Hayward  <alan.hayward@arm.com>
28641             David Sherwood  <david.sherwood@arm.com>
28643         * doc/invoke.texi (-msve-vector-bits=): Document new option.
28644         (sve): Document new AArch64 extension.
28645         * doc/md.texi (w): Extend the description of the AArch64
28646         constraint to include SVE vectors.
28647         (Upl, Upa): Document new AArch64 predicate constraints.
28648         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
28649         enum.
28650         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
28651         (msve-vector-bits=): New option.
28652         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
28653         SVE when these are disabled.
28654         (sve): New extension.
28655         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
28656         modes.  Adjust their number of units based on aarch64_sve_vg.
28657         (MAX_BITSIZE_MODE_ANY_MODE): Define.
28658         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
28659         aarch64_addr_query_type.
28660         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
28661         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
28662         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
28663         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
28664         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
28665         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
28666         (aarch64_simd_imm_zero_p): Delete.
28667         (aarch64_check_zero_based_sve_index_immediate): Declare.
28668         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28669         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28670         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28671         (aarch64_sve_float_mul_immediate_p): Likewise.
28672         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28673         rather than an rtx.
28674         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
28675         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
28676         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
28677         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
28678         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
28679         (aarch64_regmode_natural_size): Likewise.
28680         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
28681         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
28682         left one place.
28683         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
28684         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
28685         for VG and the SVE predicate registers.
28686         (V_ALIASES): Add a "z"-prefixed alias.
28687         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
28688         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
28689         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
28690         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
28691         (REG_CLASS_NAMES): Add entries for them.
28692         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
28693         and the predicate registers.
28694         (aarch64_sve_vg): Declare.
28695         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
28696         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
28697         (REGMODE_NATURAL_SIZE): Define.
28698         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
28699         SVE macros.
28700         * config/aarch64/aarch64.c: Include cfgrtl.h.
28701         (simd_immediate_info): Add a constructor for series vectors,
28702         and an associated step field.
28703         (aarch64_sve_vg): New variable.
28704         (aarch64_dbx_register_number): Handle VG and the predicate registers.
28705         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
28706         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
28707         (VEC_ANY_DATA, VEC_STRUCT): New constants.
28708         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
28709         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
28710         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
28711         (aarch64_get_mask_mode): New functions.
28712         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
28713         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
28714         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
28715         predicate modes and predicate registers.  Explicitly restrict
28716         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
28717         to store a vector mode if it is recognized by
28718         aarch64_classify_vector_mode.
28719         (aarch64_regmode_natural_size): New function.
28720         (aarch64_hard_regno_caller_save_mode): Return the original mode
28721         for predicates.
28722         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
28723         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
28724         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
28725         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
28726         functions.
28727         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
28728         does not overlap dest if the function is frame-related.  Handle
28729         SVE constants.
28730         (aarch64_split_add_offset): New function.
28731         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
28732         them aarch64_add_offset.
28733         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
28734         and update call to aarch64_sub_sp.
28735         (aarch64_add_cfa_expression): New function.
28736         (aarch64_expand_prologue): Pass extra temporary registers to the
28737         functions above.  Handle the case in which we need to emit new
28738         DW_CFA_expressions for registers that were originally saved
28739         relative to the stack pointer, but now have to be expressed
28740         relative to the frame pointer.
28741         (aarch64_output_mi_thunk): Pass extra temporary registers to the
28742         functions above.
28743         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
28744         IP0 and IP1 values for SVE frames.
28745         (aarch64_expand_vec_series): New function.
28746         (aarch64_expand_sve_widened_duplicate): Likewise.
28747         (aarch64_expand_sve_const_vector): Likewise.
28748         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
28749         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
28750         into the register, rather than emitting a SET directly.
28751         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
28752         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
28753         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
28754         (offset_9bit_signed_scaled_p): New functions.
28755         (aarch64_replicate_bitmask_imm): New function.
28756         (aarch64_bitmask_imm): Use it.
28757         (aarch64_cannot_force_const_mem): Reject expressions involving
28758         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
28759         (aarch64_classify_index): Handle SVE indices, by requiring
28760         a plain register index with a scale that matches the element size.
28761         (aarch64_classify_address): Handle SVE addresses.  Assert that
28762         the mode of the address is VOIDmode or an integer mode.
28763         Update call to aarch64_classify_symbol.
28764         (aarch64_classify_symbolic_expression): Update call to
28765         aarch64_classify_symbol.
28766         (aarch64_const_vec_all_in_range_p): New function.
28767         (aarch64_print_vector_float_operand): Likewise.
28768         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
28769         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
28770         and the FP immediates 1.0 and 0.5.
28771         (aarch64_print_address_internal): Handle SVE addresses.
28772         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
28773         (aarch64_regno_regclass): Handle predicate registers.
28774         (aarch64_secondary_reload): Handle big-endian reloads of SVE
28775         data modes.
28776         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
28777         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
28778         (aarch64_convert_sve_vector_bits): New function.
28779         (aarch64_override_options): Use it to handle -msve-vector-bits=.
28780         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28781         rather than an rtx.
28782         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
28783         Handle SVE vector and predicate modes.  Accept VL-based constants
28784         that need only one temporary register, and VL offsets that require
28785         no temporary registers.
28786         (aarch64_conditional_register_usage): Mark the predicate registers
28787         as fixed if SVE isn't available.
28788         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
28789         Return true for SVE vector and predicate modes.
28790         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
28791         rather than an unsigned int.  Handle SVE modes.
28792         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
28793         SVE modes.
28794         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
28795         if SVE is enabled.
28796         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28797         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28798         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28799         (aarch64_sve_float_mul_immediate_p): New functions.
28800         (aarch64_sve_valid_immediate): New function.
28801         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
28802         Explicitly reject structure modes.  Check for INDEX constants.
28803         Handle PTRUE and PFALSE constants.
28804         (aarch64_check_zero_based_sve_index_immediate): New function.
28805         (aarch64_simd_imm_zero_p): Delete.
28806         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
28807         vector modes.  Accept constants in the range of CNT[BHWD].
28808         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
28809         ask for an Advanced SIMD mode.
28810         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
28811         (aarch64_simd_vector_alignment): Handle SVE predicates.
28812         (aarch64_vectorize_preferred_vector_alignment): New function.
28813         (aarch64_simd_vector_alignment_reachable): Use it instead of
28814         the vector size.
28815         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
28816         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
28817         functions.
28818         (MAX_VECT_LEN): Delete.
28819         (expand_vec_perm_d): Add a vec_flags field.
28820         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
28821         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
28822         (aarch64_evpc_ext): Don't apply a big-endian lane correction
28823         for SVE modes.
28824         (aarch64_evpc_rev): Rename to...
28825         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
28826         (aarch64_evpc_rev_global): New function.
28827         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
28828         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
28829         MAX_VECT_LEN.
28830         (aarch64_evpc_sve_tbl): New function.
28831         (aarch64_expand_vec_perm_const_1): Update after rename of
28832         aarch64_evpc_rev.  Handle SVE permutes too, trying
28833         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
28834         than aarch64_evpc_tbl.
28835         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
28836         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
28837         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
28838         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
28839         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
28840         (aarch64_expand_sve_vcond): New functions.
28841         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
28842         of aarch64_vector_mode_p.
28843         (aarch64_dwarf_poly_indeterminate_value): New function.
28844         (aarch64_compute_pressure_classes): Likewise.
28845         (aarch64_can_change_mode_class): Likewise.
28846         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
28847         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
28848         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
28849         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
28850         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
28851         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
28852         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
28853         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
28854         constraints.
28855         (Dn, Dl, Dr): Accept const as well as const_vector.
28856         (Dz): Likewise.  Compare against CONST0_RTX.
28857         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
28858         of "vector" where appropriate.
28859         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
28860         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
28861         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
28862         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
28863         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
28864         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
28865         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
28866         (v_int_equiv): Extend to SVE modes.
28867         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
28868         mode attributes.
28869         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
28870         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
28871         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
28872         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
28873         (SVE_COND_FP_CMP): New int iterators.
28874         (perm_hilo): Handle the new unpack unspecs.
28875         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
28876         attributes.
28877         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
28878         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
28879         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
28880         (aarch64_equality_operator, aarch64_constant_vector_operand)
28881         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
28882         (aarch64_sve_nonimmediate_operand): Likewise.
28883         (aarch64_sve_general_operand): Likewise.
28884         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
28885         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
28886         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
28887         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
28888         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
28889         (aarch64_sve_float_arith_immediate): Likewise.
28890         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
28891         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
28892         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
28893         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
28894         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
28895         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
28896         (aarch64_sve_float_arith_operand): Likewise.
28897         (aarch64_sve_float_arith_with_sub_operand): Likewise.
28898         (aarch64_sve_float_mul_operand): Likewise.
28899         (aarch64_sve_vec_perm_operand): Likewise.
28900         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
28901         (aarch64_mov_operand): Accept const_poly_int and const_vector.
28902         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
28903         as well as const_vector.
28904         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
28905         in file.  Use CONST0_RTX and CONSTM1_RTX.
28906         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
28907         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
28908         Use aarch64_simd_imm_zero.
28909         * config/aarch64/aarch64-sve.md: New file.
28910         * config/aarch64/aarch64.md: Include it.
28911         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
28912         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
28913         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
28914         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
28915         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
28916         (sve): New attribute.
28917         (enabled): Disable instructions with the sve attribute unless
28918         TARGET_SVE.
28919         (movqi, movhi): Pass CONST_POLY_INT operaneds through
28920         aarch64_expand_mov_immediate.
28921         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
28922         CNT[BHSD] immediates.
28923         (movti): Split CONST_POLY_INT moves into two halves.
28924         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
28925         Split additions that need a temporary here if the destination
28926         is the stack pointer.
28927         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
28928         (*add<mode>3_poly_1): New instruction.
28929         (set_clobber_cc): New expander.
28931 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28933         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
28934         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
28935         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
28936         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
28937         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
28938         Change innermode from fixed_mode_size to machine_mode.
28939         (simplify_subreg): Update call accordingly.  Handle a constant-sized
28940         subreg of a variable-length CONST_VECTOR.
28942 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28943             Alan Hayward  <alan.hayward@arm.com>
28944             David Sherwood  <david.sherwood@arm.com>
28946         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
28947         (add_offset_to_base): New function, split out from...
28948         (create_mem_ref): ...here.  When handling a scale other than 1,
28949         check first whether the address is valid without the offset.
28950         Add it into the base if so, leaving the index and scale as-is.
28952 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
28954         PR c++/83778
28955         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
28956         fold_for_warn before checking if arg2 is INTEGER_CST.
28958 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
28960         * config/rs6000/predicates.md (load_multiple_operation): Delete.
28961         (store_multiple_operation): Delete.
28962         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
28963         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
28964         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
28965         guarded by TARGET_STRING.
28966         (rs6000_output_load_multiple): Delete.
28967         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28968         OPTION_MASK_STRING / TARGET_STRING handling.
28969         (print_operand) <'N', 'O'>: Add comment that these are unused now.
28970         (const rs6000_opt_masks) <"string">: Change mask to 0.
28971         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
28972         (MASK_STRING): Delete.
28973         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
28974         parts.  Simplify.
28975         (load_multiple): Delete.
28976         (*ldmsi8): Delete.
28977         (*ldmsi7): Delete.
28978         (*ldmsi6): Delete.
28979         (*ldmsi5): Delete.
28980         (*ldmsi4): Delete.
28981         (*ldmsi3): Delete.
28982         (store_multiple): Delete.
28983         (*stmsi8): Delete.
28984         (*stmsi7): Delete.
28985         (*stmsi6): Delete.
28986         (*stmsi5): Delete.
28987         (*stmsi4): Delete.
28988         (*stmsi3): Delete.
28989         (movmemsi_8reg): Delete.
28990         (corresponding unnamed define_insn): Delete.
28991         (movmemsi_6reg): Delete.
28992         (corresponding unnamed define_insn): Delete.
28993         (movmemsi_4reg): Delete.
28994         (corresponding unnamed define_insn): Delete.
28995         (movmemsi_2reg): Delete.
28996         (corresponding unnamed define_insn): Delete.
28997         (movmemsi_1reg): Delete.
28998         (corresponding unnamed define_insn): Delete.
28999         * config/rs6000/rs6000.opt (mno-string): New.
29000         (mstring): Replace by deprecation warning stub.
29001         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29003 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29005         * regrename.c (regrename_do_replace): If replacing the same
29006         reg multiple times, try to reuse last created gen_raw_REG.
29008         PR debug/81155
29009         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29010         main to workaround a bug in GDB.
29012 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29014         PR target/83737
29015         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29017 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29019         PR rtl-optimization/80481
29020         * ira-color.c (get_cap_member): New function.
29021         (allocnos_conflict_by_live_ranges_p): Use it.
29022         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29023         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29025 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29027         PR target/83628
29028         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29029         (*saddl_se_1): Ditto.
29030         (*ssubsi_1): Ditto.
29031         (*ssubl_se_1): Ditto.
29033 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29035         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29036         rather than wi::to_widest for DR_INITs.
29037         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29038         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29039         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29040         INTEGER_CSTs.
29041         (vect_analyze_group_access_1): Note that here.
29043 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29045         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29046         polynomial type sizes.
29048 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29050         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29051         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29052         (gimple_add_tmp_var): Likewise.
29054 2018-01-12  Martin Liska  <mliska@suse.cz>
29056         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29057         (gimple_alloc_sizes): Likewise.
29058         (dump_gimple_statistics): Use PRIu64 in printf format.
29059         * gimple.h: Change uint64_t to int.
29061 2018-01-12  Martin Liska  <mliska@suse.cz>
29063         * tree-core.h: Use uint64_t instead of int.
29064         * tree.c (tree_node_counts): Likewise.
29065         (tree_node_sizes): Likewise.
29066         (dump_tree_statistics): Use PRIu64 in printf format.
29068 2018-01-12  Martin Liska  <mliska@suse.cz>
29070         * Makefile.in: As qsort_chk is implemented in vec.c, add
29071         vec.o to linkage of gencfn-macros.
29072         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29073         passing the info to record_node_allocation_statistics.
29074         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29075         and pass the info.
29076         * ggc-common.c (struct ggc_usage): Add operator== and use
29077         it in operator< and compare function.
29078         * mem-stats.h (struct mem_usage): Likewise.
29079         * vec.c (struct vec_usage): Remove operator< and compare
29080         function. Can be simply inherited.
29082 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29084         PR target/81616
29085         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29086         * tree-ssa-math-opts.c: Include domwalk.h.
29087         (convert_mult_to_fma_1): New function.
29088         (fma_transformation_info): New type.
29089         (fma_deferring_state): Likewise.
29090         (cancel_fma_deferring): New function.
29091         (result_of_phi): Likewise.
29092         (last_fma_candidate_feeds_initial_phi): Likewise.
29093         (convert_mult_to_fma): Added deferring logic, split actual
29094         transformation to convert_mult_to_fma_1.
29095         (math_opts_dom_walker): New type.
29096         (math_opts_dom_walker::after_dom_children): New method, body moved
29097         here from pass_optimize_widening_mul::execute, added deferring logic
29098         bits.
29099         (pass_optimize_widening_mul::execute): Moved most of code to
29100         math_opts_dom_walker::after_dom_children.
29101         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29102         * config/i386/i386.c (ix86_option_override_internal): Added
29103         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29105 2018-01-12  Richard Biener  <rguenther@suse.de>
29107         PR debug/83157
29108         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29109         inline instance vars.
29111 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29113         PR target/81819
29114         * config/rx/rx.c (rx_is_restricted_memory_address):
29115         Handle SUBREG case.
29117 2018-01-12  Richard Biener  <rguenther@suse.de>
29119         PR tree-optimization/80846
29120         * target.def (split_reduction): New target hook.
29121         * targhooks.c (default_split_reduction): New function.
29122         * targhooks.h (default_split_reduction): Declare.
29123         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29124         target requests first reduce vectors by combining low and high
29125         parts.
29126         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29127         (get_vectype_for_scalar_type_and_size): Export.
29128         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29129         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29130         * doc/tm.texi: Regenerate.
29131         * config/i386/i386.c (ix86_split_reduction): Implement
29132         TARGET_VECTORIZE_SPLIT_REDUCTION.
29134 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29136         PR target/83368
29137         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29138         in PIC mode except for TARGET_VXWORKS_RTP.
29139         * config/sparc/sparc.c: Include cfgrtl.h.
29140         (TARGET_INIT_PIC_REG): Define.
29141         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29142         (sparc_pic_register_p): New predicate.
29143         (sparc_legitimate_address_p): Use it.
29144         (sparc_legitimize_pic_address): Likewise.
29145         (sparc_delegitimize_address): Likewise.
29146         (sparc_mode_dependent_address_p): Likewise.
29147         (gen_load_pcrel_sym): Remove 4th parameter.
29148         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29149         (sparc_expand_prologue): Do not call load_got_register here.
29150         (sparc_flat_expand_prologue): Likewise.
29151         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29152         (sparc_use_pseudo_pic_reg): New function.
29153         (sparc_init_pic_reg): Likewise.
29154         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29155         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29157 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29159         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29160         Add item for branch_cost.
29162 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29164         PR rtl-optimization/83565
29165         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29166         not extend the result to a larger mode for rotate operations.
29167         (num_sign_bit_copies1): Likewise.
29169 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29171         PR target/40411
29172         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29173         -symbolic.
29174         Use values-Xc.o for -pedantic.
29175         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29177 2018-01-12  Martin Liska  <mliska@suse.cz>
29179         PR ipa/83054
29180         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29181         New function.
29182         (possible_polymorphic_call_targets): Use it.
29183         (ipa_devirt): Likewise.
29185 2018-01-12  Martin Liska  <mliska@suse.cz>
29187         * profile-count.h (enum profile_quality): Use 0 as invalid
29188         enum value of profile_quality.
29190 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29192         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29193         -mext-string options.
29195 2018-01-12  Richard Biener  <rguenther@suse.de>
29197         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29198         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29199         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29200         Likewise.
29201         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29203 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29205         * configure.ac (--with-long-double-format): Add support for the
29206         configuration option to change the default long double format on
29207         PowerPC systems.
29208         * config.gcc (powerpc*-linux*-*): Likewise.
29209         * configure: Regenerate.
29210         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29211         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29212         used without modification.
29214 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29216         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29217         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29218         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29219         MISC_BUILTIN_SPEC_BARRIER.
29220         (rs6000_init_builtins): Likewise.
29221         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29222         enum value.
29223         (speculation_barrier): New define_insn.
29224         * doc/extend.texi: Document __builtin_speculation_barrier.
29226 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29228         PR target/83203
29229         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29230         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29231         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29232         iterators.
29233         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29234         integral modes instead of "ss" and "sd".
29235         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29236         vectors with 32-bit and 64-bit elements.
29237         (vecdupssescalarmodesuffix): New mode attribute.
29238         (vec_dup<mode>): Use it.
29240 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29242         PR target/83330
29243         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29244         frame if argument is passed on stack.
29246 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29248         PR target/82682
29249         * ree.c (combine_reaching_defs): Optimize also
29250         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29251         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29253 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29255         PR middle-end/83189
29256         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29258 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29260         PR middle-end/83718
29261         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29262         after they are computed.
29264 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29266         PR tree-optimization/83695
29267         * gimple-loop-linterchange.cc
29268         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29269         reset cached scev information after interchange.
29270         (pass_linterchange::execute): Remove call to scev_reset_htab.
29272 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29274         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29275         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29276         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29277         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29278         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29279         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29280         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29281         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29282         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29283         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29284         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29285         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29286         (V_lane_reg): Likewise.
29287         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29288         New define_expand.
29289         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29290         (vfmal_lane_low<mode>_intrinsic,
29291         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29292         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29293         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29294         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29295         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29296         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29298 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29300         * config/arm/arm-cpus.in (fp16fml): New feature.
29301         (ALL_SIMD): Add fp16fml.
29302         (armv8.2-a): Add fp16fml as an option.
29303         (armv8.3-a): Likewise.
29304         (armv8.4-a): Add fp16fml as part of fp16.
29305         * config/arm/arm.h (TARGET_FP16FML): Define.
29306         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29307         when appropriate.
29308         * config/arm/arm-modes.def (V2HF): Define.
29309         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29310         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29311         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29312         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29313         vfmsl_low, vfmsl_high): New set of builtins.
29314         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29315         (vfml_op): New code attribute.
29316         (VFMLHALVES): New int iterator.
29317         (VFML, VFMLSEL): New mode attributes.
29318         (V_reg): Define mapping for V2HF.
29319         (V_hi, V_lo): New mode attributes.
29320         (VF_constraint): Likewise.
29321         (vfml_half, vfml_half_selector): New int attributes.
29322         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29323         define_expand.
29324         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29325         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29326         New define_insn.
29327         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29328         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29329         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29330         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29331         documentation.
29332         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29333         Document new effective target and option set.
29335 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29337         * config/arm/arm-cpus.in (armv8_4): New feature.
29338         (ARMv8_4a): New fgroup.
29339         (armv8.4-a): New arch.
29340         * config/arm/arm-tables.opt: Regenerate.
29341         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29342         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29343         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29344         Add matching rules for -march=armv8.4-a and extensions.
29345         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29347 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29349         PR target/81821
29350         * config/rx/rx.md (BW): New mode attribute.
29351         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29353 2018-01-11  Richard Biener  <rguenther@suse.de>
29355         PR tree-optimization/83435
29356         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29357         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29358         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29360 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29361             Alan Hayward  <alan.hayward@arm.com>
29362             David Sherwood  <david.sherwood@arm.com>
29364         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29365         field.
29366         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29367         (aarch64_print_address_internal): Use it to check for a zero offset.
29369 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29370             Alan Hayward  <alan.hayward@arm.com>
29371             David Sherwood  <david.sherwood@arm.com>
29373         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29374         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29375         Return a poly_int64 rather than a HOST_WIDE_INT.
29376         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29377         rather than a HOST_WIDE_INT.
29378         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29379         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29380         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29381         final_offset from HOST_WIDE_INT to poly_int64.
29382         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29383         to_constant when getting the number of units in an Advanced SIMD
29384         mode.
29385         (aarch64_builtin_vectorized_function): Check for a constant number
29386         of units.
29387         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29388         GET_MODE_SIZE.
29389         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29390         attribute instead of GET_MODE_NUNITS.
29391         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29392         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29393         GET_MODE_SIZE for fixed-size registers.
29394         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29395         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29396         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29397         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29398         (aarch64_print_operand, aarch64_print_address_internal)
29399         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29400         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29401         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29402         Handle polynomial GET_MODE_SIZE.
29403         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29404         wider than SImode without modification.
29405         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29406         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29407         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29408         passing and returning SVE modes.
29409         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29410         rather than GEN_INT.
29411         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29412         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29413         (aarch64_allocate_and_probe_stack_space): Likewise.
29414         (aarch64_layout_frame): Cope with polynomial offsets.
29415         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29416         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29417         polynomial offsets.
29418         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29419         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29420         poly_int64 rather than a HOST_WIDE_INT.
29421         (aarch64_get_separate_components, aarch64_process_components)
29422         (aarch64_expand_prologue, aarch64_expand_epilogue)
29423         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29424         (aarch64_anchor_offset): New function, split out from...
29425         (aarch64_legitimize_address): ...here.
29426         (aarch64_builtin_vectorization_cost): Handle polynomial
29427         TYPE_VECTOR_SUBPARTS.
29428         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29429         GET_MODE_NUNITS.
29430         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29431         number of elements from the PARALLEL rather than the mode.
29432         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29433         rather than GET_MODE_BITSIZE.
29434         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29435         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29436         (aarch64_expand_vec_perm_const_1): Handle polynomial
29437         d->perm.length () and d->perm elements.
29438         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29439         Apply to_constant to d->perm elements.
29440         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29441         polynomial CONST_VECTOR_NUNITS.
29442         (aarch64_move_pointer): Take amount as a poly_int64 rather
29443         than an int.
29444         (aarch64_progress_pointer): Avoid temporary variable.
29445         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29446         the mode attribute instead of GET_MODE.
29448 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29449             Alan Hayward  <alan.hayward@arm.com>
29450             David Sherwood  <david.sherwood@arm.com>
29452         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29453         x exists before using it.
29454         (aarch64_add_constant_internal): Rename to...
29455         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29456         src and dest rtxes.  Handle the case in which they're different,
29457         including when the offset is zero.  Replace scratchreg with an rtx.
29458         Use 2 additions if there is no spare register into which we can
29459         move a 16-bit constant.
29460         (aarch64_add_constant): Delete.
29461         (aarch64_add_offset): Replace reg with separate src and dest
29462         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29463         Use aarch64_add_offset_1.
29464         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29465         an rtx rather than an int.  Take the delta as a poly_int64
29466         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29467         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29468         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29469         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29470         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29471         and aarch64_add_sp.
29472         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29473         aarch64_add_constant.
29475 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29477         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29478         Use scalar_float_mode.
29480 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29482         * config/aarch64/aarch64-simd.md
29483         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29484         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29485         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29486         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29487         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29488         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29489         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29490         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29491         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29492         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29494 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29496         PR target/83514
29497         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29498         targ_options->x_arm_arch_string is non NULL.
29500 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
29502         * config/aarch64/aarch64.h
29503         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
29505 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
29507         PR target/82096
29508         * expmed.c (emit_store_flag_force): Swap if const op0
29509         and change VOIDmode to mode of op0.
29511 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29513         PR rtl-optimization/83761
29514         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
29515         than bytes to mode_for_size.
29517 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29519         PR middle-end/83189
29520         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
29521         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
29522         profile.
29524 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29526         PR middle-end/83575
29527         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
29528         when in layout mode.
29529         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
29530         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
29531         partition fixup.
29533 2018-01-10  Michael Collison  <michael.collison@arm.com>
29535         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
29536         * config/aarch64/aarch64-option-extension.def: Add
29537         AARCH64_OPT_EXTENSION of 'fp16fml'.
29538         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29539         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
29540         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
29541         * config/aarch64/constraints.md (Ui7): New constraint.
29542         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
29543         (VFMLA_SEL_W): Ditto.
29544         (f16quad): Ditto.
29545         (f16mac1): Ditto.
29546         (VFMLA16_LOW): New int iterator.
29547         (VFMLA16_HIGH): Ditto.
29548         (UNSPEC_FMLAL): New unspec.
29549         (UNSPEC_FMLSL): Ditto.
29550         (UNSPEC_FMLAL2): Ditto.
29551         (UNSPEC_FMLSL2): Ditto.
29552         (f16mac): New code attribute.
29553         * config/aarch64/aarch64-simd-builtins.def
29554         (aarch64_fmlal_lowv2sf): Ditto.
29555         (aarch64_fmlsl_lowv2sf): Ditto.
29556         (aarch64_fmlalq_lowv4sf): Ditto.
29557         (aarch64_fmlslq_lowv4sf): Ditto.
29558         (aarch64_fmlal_highv2sf): Ditto.
29559         (aarch64_fmlsl_highv2sf): Ditto.
29560         (aarch64_fmlalq_highv4sf): Ditto.
29561         (aarch64_fmlslq_highv4sf): Ditto.
29562         (aarch64_fmlal_lane_lowv2sf): Ditto.
29563         (aarch64_fmlsl_lane_lowv2sf): Ditto.
29564         (aarch64_fmlal_laneq_lowv2sf): Ditto.
29565         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
29566         (aarch64_fmlalq_lane_lowv4sf): Ditto.
29567         (aarch64_fmlsl_lane_lowv4sf): Ditto.
29568         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
29569         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
29570         (aarch64_fmlal_lane_highv2sf): Ditto.
29571         (aarch64_fmlsl_lane_highv2sf): Ditto.
29572         (aarch64_fmlal_laneq_highv2sf): Ditto.
29573         (aarch64_fmlsl_laneq_highv2sf): Ditto.
29574         (aarch64_fmlalq_lane_highv4sf): Ditto.
29575         (aarch64_fmlsl_lane_highv4sf): Ditto.
29576         (aarch64_fmlalq_laneq_highv4sf): Ditto.
29577         (aarch64_fmlsl_laneq_highv4sf): Ditto.
29578         * config/aarch64/aarch64-simd.md:
29579         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
29580         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29581         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
29582         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29583         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
29584         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
29585         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
29586         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
29587         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
29588         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
29589         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
29590         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
29591         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
29592         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
29593         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
29594         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
29595         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
29596         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
29597         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
29598         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
29599         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
29600         (vfmlsl_low_u32): Ditto.
29601         (vfmlalq_low_u32): Ditto.
29602         (vfmlslq_low_u32): Ditto.
29603         (vfmlal_high_u32): Ditto.
29604         (vfmlsl_high_u32): Ditto.
29605         (vfmlalq_high_u32): Ditto.
29606         (vfmlslq_high_u32): Ditto.
29607         (vfmlal_lane_low_u32): Ditto.
29608         (vfmlsl_lane_low_u32): Ditto.
29609         (vfmlal_laneq_low_u32): Ditto.
29610         (vfmlsl_laneq_low_u32): Ditto.
29611         (vfmlalq_lane_low_u32): Ditto.
29612         (vfmlslq_lane_low_u32): Ditto.
29613         (vfmlalq_laneq_low_u32): Ditto.
29614         (vfmlslq_laneq_low_u32): Ditto.
29615         (vfmlal_lane_high_u32): Ditto.
29616         (vfmlsl_lane_high_u32): Ditto.
29617         (vfmlal_laneq_high_u32): Ditto.
29618         (vfmlsl_laneq_high_u32): Ditto.
29619         (vfmlalq_lane_high_u32): Ditto.
29620         (vfmlslq_lane_high_u32): Ditto.
29621         (vfmlalq_laneq_high_u32): Ditto.
29622         (vfmlslq_laneq_high_u32): Ditto.
29623         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
29624         (AARCH64_FL_FOR_ARCH8_4): New.
29625         (AARCH64_ISA_F16FML): New ISA flag.
29626         (TARGET_F16FML): New feature flag for fp16fml.
29627         (doc/invoke.texi): Document new fp16fml option.
29629 2018-01-10  Michael Collison  <michael.collison@arm.com>
29631         * config/aarch64/aarch64-builtins.c:
29632         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
29633         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29634         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
29635         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
29636         (AARCH64_ISA_SHA3): New ISA flag.
29637         (TARGET_SHA3): New feature flag for sha3.
29638         * config/aarch64/iterators.md (sha512_op): New int attribute.
29639         (CRYPTO_SHA512): New int iterator.
29640         (UNSPEC_SHA512H): New unspec.
29641         (UNSPEC_SHA512H2): Ditto.
29642         (UNSPEC_SHA512SU0): Ditto.
29643         (UNSPEC_SHA512SU1): Ditto.
29644         * config/aarch64/aarch64-simd-builtins.def
29645         (aarch64_crypto_sha512hqv2di): New builtin.
29646         (aarch64_crypto_sha512h2qv2di): Ditto.
29647         (aarch64_crypto_sha512su0qv2di): Ditto.
29648         (aarch64_crypto_sha512su1qv2di): Ditto.
29649         (aarch64_eor3qv8hi): Ditto.
29650         (aarch64_rax1qv2di): Ditto.
29651         (aarch64_xarqv2di): Ditto.
29652         (aarch64_bcaxqv8hi): Ditto.
29653         * config/aarch64/aarch64-simd.md:
29654         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
29655         (aarch64_crypto_sha512su0qv2di): Ditto.
29656         (aarch64_crypto_sha512su1qv2di): Ditto.
29657         (aarch64_eor3qv8hi): Ditto.
29658         (aarch64_rax1qv2di): Ditto.
29659         (aarch64_xarqv2di): Ditto.
29660         (aarch64_bcaxqv8hi): Ditto.
29661         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
29662         (vsha512h2q_u64): Ditto.
29663         (vsha512su0q_u64): Ditto.
29664         (vsha512su1q_u64): Ditto.
29665         (veor3q_u16): Ditto.
29666         (vrax1q_u64): Ditto.
29667         (vxarq_u64): Ditto.
29668         (vbcaxq_u16): Ditto.
29669         * config/arm/types.md (crypto_sha512): New type attribute.
29670         (crypto_sha3): Ditto.
29671         (doc/invoke.texi): Document new sha3 option.
29673 2018-01-10  Michael Collison  <michael.collison@arm.com>
29675         * config/aarch64/aarch64-builtins.c:
29676         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
29677         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29678         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
29679         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
29680         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
29681         (AARCH64_ISA_SM4): New ISA flag.
29682         (TARGET_SM4): New feature flag for sm4.
29683         * config/aarch64/aarch64-simd-builtins.def
29684         (aarch64_sm3ss1qv4si): Ditto.
29685         (aarch64_sm3tt1aq4si): Ditto.
29686         (aarch64_sm3tt1bq4si): Ditto.
29687         (aarch64_sm3tt2aq4si): Ditto.
29688         (aarch64_sm3tt2bq4si): Ditto.
29689         (aarch64_sm3partw1qv4si): Ditto.
29690         (aarch64_sm3partw2qv4si): Ditto.
29691         (aarch64_sm4eqv4si): Ditto.
29692         (aarch64_sm4ekeyqv4si): Ditto.
29693         * config/aarch64/aarch64-simd.md:
29694         (aarch64_sm3ss1qv4si): Ditto.
29695         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
29696         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
29697         (aarch64_sm4eqv4si): Ditto.
29698         (aarch64_sm4ekeyqv4si): Ditto.
29699         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
29700         (sm3part_op): Ditto.
29701         (CRYPTO_SM3TT): Ditto.
29702         (CRYPTO_SM3PART): Ditto.
29703         (UNSPEC_SM3SS1): New unspec.
29704         (UNSPEC_SM3TT1A): Ditto.
29705         (UNSPEC_SM3TT1B): Ditto.
29706         (UNSPEC_SM3TT2A): Ditto.
29707         (UNSPEC_SM3TT2B): Ditto.
29708         (UNSPEC_SM3PARTW1): Ditto.
29709         (UNSPEC_SM3PARTW2): Ditto.
29710         (UNSPEC_SM4E): Ditto.
29711         (UNSPEC_SM4EKEY): Ditto.
29712         * config/aarch64/constraints.md (Ui2): New constraint.
29713         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
29714         * config/arm/types.md (crypto_sm3): New type attribute.
29715         (crypto_sm4): Ditto.
29716         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
29717         (vsm3tt1aq_u32): Ditto.
29718         (vsm3tt1bq_u32): Ditto.
29719         (vsm3tt2aq_u32): Ditto.
29720         (vsm3tt2bq_u32): Ditto.
29721         (vsm3partw1q_u32): Ditto.
29722         (vsm3partw2q_u32): Ditto.
29723         (vsm4eq_u32): Ditto.
29724         (vsm4ekeyq_u32): Ditto.
29725         (doc/invoke.texi): Document new sm4 option.
29727 2018-01-10  Michael Collison  <michael.collison@arm.com>
29729         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
29730         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
29731         (AARCH64_FL_FOR_ARCH8_4): New.
29732         (AARCH64_FL_V8_4): New flag.
29733         (doc/invoke.texi): Document new armv8.4-a option.
29735 2018-01-10  Michael Collison  <michael.collison@arm.com>
29737         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29738         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
29739         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
29740         * config/aarch64/aarch64-option-extension.def: Add
29741         AARCH64_OPT_EXTENSION of 'sha2'.
29742         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
29743         (crypto): Disable sha2 and aes if crypto disabled.
29744         (crypto): Enable aes and sha2 if enabled.
29745         (simd): Disable sha2 and aes if simd disabled.
29746         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
29747         New flags.
29748         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
29749         (TARGET_SHA2): New feature flag for sha2.
29750         (TARGET_AES): New feature flag for aes.
29751         * config/aarch64/aarch64-simd.md:
29752         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
29753         conditional on TARGET_AES.
29754         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
29755         (aarch64_crypto_sha1hsi): Make pattern conditional
29756         on TARGET_SHA2.
29757         (aarch64_crypto_sha1hv4si): Ditto.
29758         (aarch64_be_crypto_sha1hv4si): Ditto.
29759         (aarch64_crypto_sha1su1v4si): Ditto.
29760         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
29761         (aarch64_crypto_sha1su0v4si): Ditto.
29762         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
29763         (aarch64_crypto_sha256su0v4si): Ditto.
29764         (aarch64_crypto_sha256su1v4si): Ditto.
29765         (doc/invoke.texi): Document new aes and sha2 options.
29767 2018-01-10  Martin Sebor  <msebor@redhat.com>
29769         PR tree-optimization/83781
29770         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
29771         as string arrays.
29773 2018-01-11  Martin Sebor  <msebor@gmail.com>
29774             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29776         PR tree-optimization/83501
29777         PR tree-optimization/81703
29779         * tree-ssa-strlen.c (get_string_cst): Rename...
29780         (get_string_len): ...to this.  Handle global constants.
29781         (handle_char_store): Adjust.
29783 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
29784             Jim Wilson  <jimw@sifive.com>
29786         * config/riscv/riscv-protos.h (riscv_output_return): New.
29787         * config/riscv/riscv.c (struct machine_function): New naked_p field.
29788         (riscv_attribute_table, riscv_output_return),
29789         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
29790         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
29791         (riscv_compute_frame_info): Only compute frame->mask if not a naked
29792         function.
29793         (riscv_expand_prologue): Add early return for naked function.
29794         (riscv_expand_epilogue): Likewise.
29795         (riscv_function_ok_for_sibcall): Return false for naked function.
29796         (riscv_set_current_function): New.
29797         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
29798         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
29799         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
29800         * doc/extend.texi (RISC-V Function Attributes): New.
29802 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
29804         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
29805         check for 128-bit long double before checking TCmode.
29806         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
29807         128-bit long doubles before checking TFmode or TCmode.
29808         (FLOAT128_IBM_P): Likewise.
29810 2018-01-10  Martin Sebor  <msebor@redhat.com>
29812         PR tree-optimization/83671
29813         * builtins.c (c_strlen): Unconditionally return zero for the empty
29814         string.
29815         Use -Warray-bounds for warnings.
29816         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
29817         for non-constant array indices with COMPONENT_REF, arrays of
29818         arrays, and pointers to arrays.
29819         (gimple_fold_builtin_strlen): Determine and set length range for
29820         non-constant character arrays.
29822 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
29824         PR middle-end/81897
29825         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
29826         empty blocks.
29828 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
29830         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
29832 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29834         PR target/83399
29835         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
29836         VECTOR_MEM_ALTIVEC_OR_VSX_P.
29837         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
29838         indexed_or_indirect_operand predicate.
29839         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
29840         (*vsx_le_perm_load_v8hi): Likewise.
29841         (*vsx_le_perm_load_v16qi): Likewise.
29842         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
29843         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
29844         (*vsx_le_perm_store_v8hi): Likewise.
29845         (*vsx_le_perm_store_v16qi): Likewise.
29846         (eight unnamed splitters): Likewise.
29848 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29850         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
29851         * config/rs6000/emmintrin.h: Likewise.
29852         * config/rs6000/mmintrin.h: Likewise.
29853         * config/rs6000/xmmintrin.h: Likewise.
29855 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
29857         PR c++/43486
29858         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
29859         "public_flag".
29860         * tree.c (tree_nop_conversion): Return true for location wrapper
29861         nodes.
29862         (maybe_wrap_with_location): New function.
29863         (selftest::check_strip_nops): New function.
29864         (selftest::test_location_wrappers): New function.
29865         (selftest::tree_c_tests): Call it.
29866         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
29867         (maybe_wrap_with_location): New decl.
29868         (EXPR_LOCATION_WRAPPER_P): New macro.
29869         (location_wrapper_p): New inline function.
29870         (tree_strip_any_location_wrapper): New inline function.
29872 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
29874         PR target/83735
29875         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
29876         stack_realign_offset for the largest alignment of stack slot
29877         actually used.
29878         (ix86_find_max_used_stack_alignment): New function.
29879         (ix86_finalize_stack_frame_flags): Use it.  Set
29880         max_used_stack_alignment if we don't realign stack.
29881         * config/i386/i386.h (machine_function): Add
29882         max_used_stack_alignment.
29884 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
29886         * config/arm/arm.opt (-mbranch-cost): New option.
29887         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
29888         account.
29890 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
29892         PR target/83629
29893         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
29894         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
29896 2018-01-10  Richard Biener  <rguenther@suse.de>
29898         PR debug/83765
29899         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
29900         early out so it also covers the case where we have a non-NULL
29901         origin.
29903 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29905         PR tree-optimization/83753
29906         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
29907         for non-strided grouped accesses if the number of elements is 1.
29909 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29911         PR target/81616
29912         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
29913         * i386.h (TARGET_USE_GATHER): Define.
29914         * x86-tune.def (X86_TUNE_USE_GATHER): New.
29916 2018-01-10  Martin Liska  <mliska@suse.cz>
29918         PR bootstrap/82831
29919         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
29920         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
29921         partitioning.
29922         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
29923         CLEANUP_NO_PARTITIONING is not set.
29925 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29927         * doc/rtl.texi: Remove documentation of (const ...) wrappers
29928         for vectors, as a partial revert of r254296.
29929         * rtl.h (const_vec_p): Delete.
29930         (const_vec_duplicate_p): Don't test for vector CONSTs.
29931         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
29932         * expmed.c (make_tree): Likewise.
29934         Revert:
29935         * common.md (E, F): Use CONSTANT_P instead of checking for
29936         CONST_VECTOR.
29937         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
29938         checking for CONST_VECTOR.
29940 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29942         PR middle-end/83575
29943         * predict.c (force_edge_cold): Handle in more sane way edges
29944         with no prediction.
29946 2018-01-09  Carl Love  <cel@us.ibm.com>
29948         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
29949         V4SI, V4SF types.
29950         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
29951         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
29952         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
29953         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
29954         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
29955         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
29956         * config/rs6000/rs6000-protos.h: Add extern defition for
29957         rs6000_generate_float2_double_code.
29958         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
29959         function.
29960         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
29961         (float2_v2df): Add define_expand.
29963 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
29965         PR target/83628
29966         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
29967         op_mode in the force_to_mode call.
29969 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29971         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
29972         instead of checking each element individually.
29973         (aarch64_evpc_uzp): Likewise.
29974         (aarch64_evpc_zip): Likewise.
29975         (aarch64_evpc_ext): Likewise.
29976         (aarch64_evpc_rev): Likewise.
29977         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
29978         instead of checking each element individually.  Return true without
29979         generating rtl if
29980         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
29981         whether all selected elements come from the same input, instead of
29982         checking each element individually.  Remove calls to gen_rtx_REG,
29983         start_sequence and end_sequence and instead assert that no rtl is
29984         generated.
29986 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29988         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
29989         order of HIGH and CONST checks.
29991 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29993         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
29994         if the destination isn't an SSA_NAME.
29996 2018-01-09  Richard Biener  <rguenther@suse.de>
29998         PR tree-optimization/83668
29999         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30000         move prologue...
30001         (canonicalize_loop_form): ... here, renamed from ...
30002         (canonicalize_loop_closed_ssa_form): ... this and amended to
30003         swap successor edges for loop exit blocks to make us use
30004         the RPO order we need for initial schedule generation.
30006 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30008         PR tree-optimization/64811
30009         * match.pd: When optimizing comparisons with Inf, avoid
30010         introducing or losing exceptions from comparisons with NaN.
30012 2018-01-09  Martin Liska  <mliska@suse.cz>
30014         PR sanitizer/82517
30015         * asan.c (shadow_mem_size): Add gcc_assert.
30017 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30019         Don't save registers in main().
30021         PR target/83738
30022         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30023         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30024         * config/avr/avr.c (avr_set_current_function): Don't error if
30025         naked, OS_task or OS_main are specified at the same time.
30026         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30027         OS_main.
30028         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30029         attribute.
30030         * common/config/avr/avr-common.c (avr_option_optimization_table):
30031         Switch on -mmain-is-OS_task for optimizing compilations.
30033 2018-01-09  Richard Biener  <rguenther@suse.de>
30035         PR tree-optimization/83572
30036         * graphite.c: Include cfganal.h.
30037         (graphite_transform_loops): Connect infinite loops to exit
30038         and remove fake edges at the end.
30040 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30042         * ipa-inline.c (edge_badness): Revert accidental checkin.
30044 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30046         PR ipa/80763
30047         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30048         symbols; not inline clones.
30050 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30052         PR target/83507
30053         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30054         hard registers.  Formatting fixes.
30056         PR preprocessor/83722
30057         * gcc.c (try_generate_repro): Pass
30058         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30059         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30060         do_report_bug.
30062 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30063             Kito Cheng  <kito.cheng@gmail.com>
30065         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30066         (riscv_leaf_function_p): Delete.
30067         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30069 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30071         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30072         function.
30073         (do_ifelse): New function.
30074         (do_isel): New function.
30075         (do_sub3): New function.
30076         (do_add3): New function.
30077         (do_load_mask_compare): New function.
30078         (do_overlap_load_compare): New function.
30079         (expand_compare_loop): New function.
30080         (expand_block_compare): Call expand_compare_loop() when appropriate.
30081         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30082         option description.
30083         (-mblock-compare-inline-loop-limit): New option.
30085 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30087         PR target/83677
30088         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30089         Reverse order of second and third operands in first alternative.
30090         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30091         of first and second elements in UNSPEC_VPERMR vector.
30092         (altivec_expand_vec_perm_le): Likewise.
30094 2018-01-08  Jeff Law  <law@redhat.com>
30096         PR rtl-optimizatin/81308
30097         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30098         (process_switch): If group_case_labels makes a change, then set
30099         cfg_altered.
30100         (pass_convert_switch::execute): If a switch is converted, then
30101         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30103         PR rtl-optimization/81308
30104         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30105         splitting insns.
30107 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30109         PR target/83663 - Revert r255946
30110         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30111         generation for cases where splatting a value is not useful.
30112         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30113         across a vec_duplicate and a paradoxical subreg forming a vector
30114         mode to a vec_concat.
30116 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30118         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30119         -march=armv8.3-a variants.
30120         * config/arm/t-multilib: Likewise.
30121         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30123 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30125         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30126         to generate rtl.
30127         (cceq_ior_compare_complement): Give it a name so I can use it, and
30128         change boolean_or_operator predicate to boolean_operator so it can
30129         be used to generate a crand.
30130         (eqne): New code iterator.
30131         (bd/bd_neg): New code_attrs.
30132         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30133         a single define_insn.
30134         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30135         decrement (bdnzt/bdnzf/bdzt/bdzf).
30136         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30137         with the new names of the branch decrement patterns, and added the
30138         names of the branch decrement conditional patterns.
30140 2018-01-08  Richard Biener  <rguenther@suse.de>
30142         PR tree-optimization/83563
30143         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30144         cache.
30146 2018-01-08  Richard Biener  <rguenther@suse.de>
30148         PR middle-end/83713
30149         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30151 2018-01-08  Richard Biener  <rguenther@suse.de>
30153         PR tree-optimization/83685
30154         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30155         references to abnormals.
30157 2018-01-08  Richard Biener  <rguenther@suse.de>
30159         PR lto/83719
30160         * dwarf2out.c (output_indirect_strings): Handle empty
30161         skeleton_debug_str_hash.
30162         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30164 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30166         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30167         (emit_store_direct): Likewise.
30168         (arc_trampoline_adjust_address): Likewise.
30169         (arc_asm_trampoline_template): New function.
30170         (arc_initialize_trampoline): Use asm_trampoline_template.
30171         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30172         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30173         * config/arc/arc.md (flush_icache): Delete pattern.
30175 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30177         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30178         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30179         munaligned-access.
30181 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30183         PR target/83681
30184         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30185         by not USED_FOR_TARGET.
30186         (make_pass_resolve_sw_modes): Likewise.
30188 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30190         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30191         USED_FOR_TARGET.
30193 2018-01-08  Richard Biener  <rguenther@suse.de>
30195         PR middle-end/83580
30196         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30198 2018-01-08  Richard Biener  <rguenther@suse.de>
30200         PR middle-end/83517
30201         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30203 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30205         PR middle-end/81897
30206         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30207         basic blocks with a small number of successors.
30208         (convert_control_dep_chain_into_preds): Improve handling of
30209         forwarder blocks.
30210         (dump_predicates): Split apart into...
30211         (dump_pred_chain): ...here...
30212         (dump_pred_info): ...and here.
30213         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30214         (can_chain_union_be_invalidated_p): Improve check for invalidation
30215         of paths.
30216         (uninit_uses_cannot_happen): Avoid unnecessary if
30217         convert_control_dep_chain_into_preds yielded nothing.
30219 2018-01-06  Martin Sebor  <msebor@redhat.com>
30221         PR tree-optimization/83640
30222         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30223         subtracting negative offset from size.
30224         (builtin_access::overlap): Adjust offset bounds of the access to fall
30225         within the size of the object if possible.
30227 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30229         PR rtl-optimization/83699
30230         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30231         extract_bit_field_as_subreg to cases in which the extracted
30232         value is also a vector.
30234         * lra-constraints.c (process_alt_operands): Test for the equivalence
30235         substitutions when detecting a possible reload cycle.
30237 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30239         PR debug/83480
30240         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30241         by default if flag_selective_schedling{,2}.  Formatting fixes.
30243         PR rtl-optimization/83682
30244         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30245         if it has non-VECTOR_MODE element mode.
30246         (vec_duplicate_p): Likewise.
30248         PR middle-end/83694
30249         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30250         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30252 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30254         PR target/83604
30255         * config/i386/i386-builtin.def
30256         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30257         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30258         Require also OPTION_MASK_ISA_AVX512F in addition to
30259         OPTION_MASK_ISA_GFNI.
30260         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30261         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30262         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30263         to OPTION_MASK_ISA_GFNI.
30264         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30265         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30266         OPTION_MASK_ISA_AVX512BW.
30267         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30268         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30269         addition to OPTION_MASK_ISA_GFNI.
30270         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30271         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30272         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30273         to OPTION_MASK_ISA_GFNI.
30274         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30275         a requirement for all ISAs rather than any of them with a few
30276         exceptions.
30277         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30278         processing.
30279         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30280         bitmasks to be enabled with 3 exceptions, instead of requiring any
30281         enabled ISA with lots of exceptions.
30282         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30283         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30284         Change avx512bw in isa attribute to avx512f.
30285         * config/i386/sgxintrin.h: Add license boilerplate.
30286         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30287         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30288         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30289         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30290         defined.
30291         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30292         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30293         temporarily sse2 rather than sse if not enabled already.
30295         PR target/83604
30296         * config/i386/sse.md (VI248_VLBW): Rename to ...
30297         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30298         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30299         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30300         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30301         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30302         mode iterator instead of VI248_VLBW.
30304 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30306         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30307         (record_modified): Skip clobbers; add debug output.
30308         (param_change_prob): Use sreal frequencies.
30310 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30312         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30313         punt for user-aligned variables.
30315 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30317         * tree-chrec.c (chrec_contains_symbols): Return true for
30318         POLY_INT_CST.
30320 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30322         PR target/82439
30323         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30324         of (x|y) == x for BICS pattern.
30326 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30328         PR tree-optimization/83605
30329         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30330         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30331         can throw.
30333 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30335         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30336         * config/epiphany/rtems.h: New file.
30338 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30339             Uros Bizjak  <ubizjak@gmail.com>
30341         PR target/83554
30342         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30343         QIreg_operand instead of register_operand predicate.
30344         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30345         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30346         comments instead of -fmitigate[-_]rop.
30348 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30350         PR bootstrap/81926
30351         * cgraphunit.c (symbol_table::compile): Switch to text_section
30352         before calling assembly_start debug hook.
30353         * run-rtl-passes.c (run_rtl_passes): Likewise.
30354         Include output.h.
30356 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30358         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30359         range_int_cst_p rather than !symbolic_range_p before calling
30360         extract_range_from_multiplicative_op_1.
30362 2018-01-04  Jeff Law  <law@redhat.com>
30364         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30365         redundant test in assertion.
30367 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30369         * doc/rtl.texi: Document machine_mode wrapper classes.
30371 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30373         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30374         using tree_to_uhwi.
30376 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30378         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30379         the VEC_PERM_EXPR fold to fail.
30381 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30383         PR debug/83585
30384         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30385         to switched_sections.
30387 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30389         PR target/83680
30390         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30391         test for d.testing.
30393 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30395         PR target/83387
30396         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30397         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30399 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30401         PR debug/83666
30402         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30403         is BLKmode and bitpos not zero or mode change is needed.
30405 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30407         PR target/83675
30408         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30409         TARGET_VIS2.
30411 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30413         PR target/83628
30414         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30415         instead of MULT rtx.  Update all corresponding splitters.
30416         (*saddl_se): Ditto.
30417         (*ssub<modesuffix>): Ditto.
30418         (*ssubl_se): Ditto.
30419         (*cmp_sadd_di): Update split patterns.
30420         (*cmp_sadd_si): Ditto.
30421         (*cmp_sadd_sidi): Ditto.
30422         (*cmp_ssub_di): Ditto.
30423         (*cmp_ssub_si): Ditto.
30424         (*cmp_ssub_sidi): Ditto.
30425         * config/alpha/predicates.md (const23_operand): New predicate.
30426         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30427         Look for ASHIFT, not MULT inner operand.
30428         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30430 2018-01-04  Martin Liska  <mliska@suse.cz>
30432         PR gcov-profile/83669
30433         * gcov.c (output_intermediate_file): Add version to intermediate
30434         gcov file.
30435         * doc/gcov.texi: Document new field 'version' in intermediate
30436         file format. Fix location of '-k' option of gcov command.
30438 2018-01-04  Martin Liska  <mliska@suse.cz>
30440         PR ipa/82352
30441         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30443 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30445         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30447 2018-01-03  Martin Sebor  <msebor@redhat.com>
30449         PR tree-optimization/83655
30450         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30451         checking calls with invalid arguments.
30453 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30455         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30456         (vectorizable_mask_load_store): Delete.
30457         (vectorizable_call): Return false for masked loads and stores.
30458         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30459         instead of gimple_assign_rhs1.
30460         (vectorizable_load): Handle IFN_MASK_LOAD.
30461         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30463 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30465         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30466         split out from..,
30467         (vectorizable_mask_load_store): ...here.
30468         (vectorizable_load): ...and here.
30470 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30472         * tree-vect-stmts.c (vect_build_all_ones_mask)
30473         (vect_build_zero_merge_argument): New functions, split out from...
30474         (vectorizable_load): ...here.
30476 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30478         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30479         split out from...
30480         (vectorizable_mask_load_store): ...here.
30481         (vectorizable_store): ...and here.
30483 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30485         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30486         split out from...
30487         (vectorizable_mask_load_store): ...here.
30489 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30491         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30492         (vect_model_store_cost): Take a vec_load_store_type instead of a
30493         vect_def_type.
30494         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30495         (vect_model_store_cost): Take a vec_load_store_type instead of a
30496         vect_def_type.
30497         (vectorizable_mask_load_store): Update accordingly.
30498         (vectorizable_store): Likewise.
30499         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
30501 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30503         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
30504         IFN_MASK_LOAD calls here rather than...
30505         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
30507 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30508             Alan Hayward  <alan.hayward@arm.com>
30509             David Sherwood  <david.sherwood@arm.com>
30511         * expmed.c (extract_bit_field_1): For vector extracts,
30512         fall back to extract_bit_field_as_subreg if vec_extract
30513         isn't available.
30515 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30516             Alan Hayward  <alan.hayward@arm.com>
30517             David Sherwood  <david.sherwood@arm.com>
30519         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
30520         they are variable or constant sized.
30521         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
30522         slots for constant-sized data.
30524 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30525             Alan Hayward  <alan.hayward@arm.com>
30526             David Sherwood  <david.sherwood@arm.com>
30528         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
30529         handling COND_EXPRs with boolean comparisons, try to find a better
30530         basis for the mask type than the boolean itself.
30532 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30534         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
30535         is calculated and how it can be overridden.
30536         * genmodes.c (max_bitsize_mode_any_mode): New variable.
30537         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
30538         if defined.
30539         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
30540         if nonzero.
30542 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30543             Alan Hayward  <alan.hayward@arm.com>
30544             David Sherwood  <david.sherwood@arm.com>
30546         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
30547         Remove the mode argument.
30548         (aarch64_simd_valid_immediate): Remove the mode and inverse
30549         arguments.
30550         * config/aarch64/iterators.md (bitsize): New iterator.
30551         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
30552         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
30553         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
30554         aarch64_simd_valid_immediate.
30555         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
30556         (aarch64_reg_or_bic_imm): Likewise.
30557         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
30558         with an insn_type enum and msl with a modifier_type enum.
30559         Replace element_width with a scalar_mode.  Change the shift
30560         to unsigned int.  Add constructors for scalar_float_mode and
30561         scalar_int_mode elements.
30562         (aarch64_vect_float_const_representable_p): Delete.
30563         (aarch64_can_const_movi_rtx_p)
30564         (aarch64_simd_scalar_immediate_valid_for_move)
30565         (aarch64_simd_make_constant): Update call to
30566         aarch64_simd_valid_immediate.
30567         (aarch64_advsimd_valid_immediate_hs): New function.
30568         (aarch64_advsimd_valid_immediate): Likewise.
30569         (aarch64_simd_valid_immediate): Remove mode and inverse
30570         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
30571         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
30572         and aarch64_float_const_representable_p on the result.
30573         (aarch64_output_simd_mov_immediate): Remove mode argument.
30574         Update call to aarch64_simd_valid_immediate and use of
30575         simd_immediate_info.
30576         (aarch64_output_scalar_simd_mov_immediate): Update call
30577         accordingly.
30579 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30580             Alan Hayward  <alan.hayward@arm.com>
30581             David Sherwood  <david.sherwood@arm.com>
30583         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
30584         (mode_nunits): Likewise CONST_MODE_NUNITS.
30585         * machmode.def (ADJUST_NUNITS): Document.
30586         * genmodes.c (mode_data::need_nunits_adj): New field.
30587         (blank_mode): Update accordingly.
30588         (adj_nunits): New variable.
30589         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
30590         parameter.
30591         (emit_mode_size_inline): Set need_bytesize_adj for all modes
30592         listed in adj_nunits.
30593         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
30594         listed in adj_nunits.  Don't emit case statements for such modes.
30595         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
30596         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
30597         nothing if adj_nunits is nonnull.
30598         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
30599         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
30600         (emit_mode_fbit): Update use of print_maybe_const_decl.
30601         (emit_move_size): Likewise.  Treat the array as non-const
30602         if adj_nunits.
30603         (emit_mode_adjustments): Handle adj_nunits.
30605 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30607         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
30608         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
30609         (VECTOR_MODES): Use it.
30610         (make_vector_modes): Take the prefix as an argument.
30612 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30613             Alan Hayward  <alan.hayward@arm.com>
30614             David Sherwood  <david.sherwood@arm.com>
30616         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
30617         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
30618         for MODE_VECTOR_BOOL.
30619         * machmode.def (VECTOR_BOOL_MODE): Document.
30620         * genmodes.c (VECTOR_BOOL_MODE): New macro.
30621         (make_vector_bool_mode): New function.
30622         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
30623         MODE_VECTOR_BOOL.
30624         * lto-streamer-in.c (lto_input_mode_table): Likewise.
30625         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
30626         Likewise.
30627         * stor-layout.c (int_mode_for_mode): Likewise.
30628         * tree.c (build_vector_type_for_mode): Likewise.
30629         * varasm.c (output_constant_pool_2): Likewise.
30630         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
30631         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
30632         for MODE_VECTOR_BOOL.
30633         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
30634         of mode class checks.
30635         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
30636         instead of a list of mode class checks.
30637         (expand_vector_scalar_condition): Likewise.
30638         (type_for_widest_vector_mode): Handle BImode as an inner mode.
30640 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30641             Alan Hayward  <alan.hayward@arm.com>
30642             David Sherwood  <david.sherwood@arm.com>
30644         * machmode.h (mode_size): Change from unsigned short to
30645         poly_uint16_pod.
30646         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
30647         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30648         or if measurement_type is not polynomial.
30649         (fixed_size_mode::includes_p): Check for constant-sized modes.
30650         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
30651         return a poly_uint16 rather than an unsigned short.
30652         (emit_mode_size): Change the type of mode_size from unsigned short
30653         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
30654         (emit_mode_adjustments): Cope with polynomial vector sizes.
30655         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30656         for GET_MODE_SIZE.
30657         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30658         for GET_MODE_SIZE.
30659         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
30660         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
30661         * caller-save.c (setup_save_areas): Likewise.
30662         (replace_reg_with_saved_mem): Likewise.
30663         * calls.c (emit_library_call_value_1): Likewise.
30664         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
30665         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
30666         (gen_lowpart_for_combine): Likewise.
30667         * convert.c (convert_to_integer_1): Likewise.
30668         * cse.c (equiv_constant, cse_insn): Likewise.
30669         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
30670         (cselib_subst_to_values): Likewise.
30671         * dce.c (word_dce_process_block): Likewise.
30672         * df-problems.c (df_word_lr_mark_ref): Likewise.
30673         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
30674         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
30675         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
30676         (rtl_for_decl_location): Likewise.
30677         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
30678         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
30679         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
30680         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
30681         (expand_expr_real_1): Likewise.
30682         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
30683         (pad_below): Likewise.
30684         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30685         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
30686         * ira.c (get_subreg_tracking_sizes): Likewise.
30687         * ira-build.c (ira_create_allocno_objects): Likewise.
30688         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
30689         (ira_sort_regnos_for_alter_reg): Likewise.
30690         * ira-costs.c (record_operand_costs): Likewise.
30691         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
30692         (resolve_simple_move): Likewise.
30693         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
30694         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
30695         (lra_constraints): Likewise.
30696         (CONST_POOL_OK_P): Reject variable-sized modes.
30697         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
30698         (add_pseudo_to_slot, lra_spill): Likewise.
30699         * omp-low.c (omp_clause_aligned_alignment): Likewise.
30700         * optabs-query.c (get_best_extraction_insn): Likewise.
30701         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30702         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
30703         (expand_mult_highpart, valid_multiword_target_p): Likewise.
30704         * recog.c (offsettable_address_addr_space_p): Likewise.
30705         * regcprop.c (maybe_mode_change): Likewise.
30706         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
30707         * regrename.c (build_def_use): Likewise.
30708         * regstat.c (dump_reg_info): Likewise.
30709         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
30710         (find_reloads, find_reloads_subreg_address): Likewise.
30711         * reload1.c (eliminate_regs_1): Likewise.
30712         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
30713         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
30714         (simplify_binary_operation_1, simplify_subreg): Likewise.
30715         * targhooks.c (default_function_arg_padding): Likewise.
30716         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
30717         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
30718         (verify_gimple_assign_ternary): Likewise.
30719         * tree-inline.c (estimate_move_cost): Likewise.
30720         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30721         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
30722         (get_address_cost_ainc): Likewise.
30723         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
30724         (vect_supportable_dr_alignment): Likewise.
30725         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30726         (vectorizable_reduction): Likewise.
30727         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
30728         (vectorizable_operation, vectorizable_load): Likewise.
30729         * tree.c (build_same_sized_truth_vector_type): Likewise.
30730         * valtrack.c (cleanup_auto_inc_dec): Likewise.
30731         * var-tracking.c (emit_note_insn_var_location): Likewise.
30732         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
30733         (ADDR_VEC_ALIGN): Likewise.
30735 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30736             Alan Hayward  <alan.hayward@arm.com>
30737             David Sherwood  <david.sherwood@arm.com>
30739         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
30740         unsigned short.
30741         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30742         or if measurement_type is polynomial.
30743         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
30744         * combine.c (make_extraction): Likewise.
30745         * dse.c (find_shift_sequence): Likewise.
30746         * dwarf2out.c (mem_loc_descriptor): Likewise.
30747         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
30748         (extract_bit_field, extract_low_bits): Likewise.
30749         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
30750         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
30751         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
30752         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
30753         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30754         * reload.c (find_reloads): Likewise.
30755         * reload1.c (alter_reg): Likewise.
30756         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
30757         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
30758         * tree-if-conv.c (predicate_mem_writes): Likewise.
30759         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
30760         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
30761         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
30762         * valtrack.c (dead_debug_insert_temp): Likewise.
30763         * varasm.c (mergeable_constant_section): Likewise.
30764         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
30766 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30767             Alan Hayward  <alan.hayward@arm.com>
30768             David Sherwood  <david.sherwood@arm.com>
30770         * expr.c (expand_assignment): Cope with polynomial mode sizes
30771         when assigning to a CONCAT.
30773 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30774             Alan Hayward  <alan.hayward@arm.com>
30775             David Sherwood  <david.sherwood@arm.com>
30777         * machmode.h (mode_precision): Change from unsigned short to
30778         poly_uint16_pod.
30779         (mode_to_precision): Return a poly_uint16 rather than an unsigned
30780         short.
30781         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
30782         or if measurement_type is not polynomial.
30783         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
30784         in which the mode is already known to be a scalar_int_mode.
30785         * genmodes.c (emit_mode_precision): Change the type of mode_precision
30786         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
30787         initializer.
30788         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30789         for GET_MODE_PRECISION.
30790         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30791         for GET_MODE_PRECISION.
30792         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
30793         as polynomial.
30794         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
30795         (expand_field_assignment, make_extraction): Likewise.
30796         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
30797         (get_last_value): Likewise.
30798         * convert.c (convert_to_integer_1): Likewise.
30799         * cse.c (cse_insn): Likewise.
30800         * expr.c (expand_expr_real_1): Likewise.
30801         * lra-constraints.c (simplify_operand_subreg): Likewise.
30802         * optabs-query.c (can_atomic_load_p): Likewise.
30803         * optabs.c (expand_atomic_load): Likewise.
30804         (expand_atomic_store): Likewise.
30805         * ree.c (combine_reaching_defs): Likewise.
30806         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
30807         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
30808         * tree.h (type_has_mode_precision_p): Likewise.
30809         * ubsan.c (instrument_si_overflow): Likewise.
30811 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30812             Alan Hayward  <alan.hayward@arm.com>
30813             David Sherwood  <david.sherwood@arm.com>
30815         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
30816         polynomial numbers of units.
30817         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
30818         (valid_vector_subparts_p): New function.
30819         (build_vector_type): Remove temporary shim and take the number
30820         of units as a poly_uint64 rather than an int.
30821         (build_opaque_vector_type): Take the number of units as a
30822         poly_uint64 rather than an int.
30823         * tree.c (build_vector_from_ctor): Handle polynomial
30824         TYPE_VECTOR_SUBPARTS.
30825         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
30826         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
30827         (build_vector_from_val): If the number of units is variable,
30828         use build_vec_duplicate_cst for constant operands and
30829         VEC_DUPLICATE_EXPR otherwise.
30830         (make_vector_type): Remove temporary is_constant ().
30831         (build_vector_type, build_opaque_vector_type): Take the number of
30832         units as a poly_uint64 rather than an int.
30833         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
30834         VECTOR_CST_NELTS.
30835         * cfgexpand.c (expand_debug_expr): Likewise.
30836         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
30837         (store_constructor, expand_expr_real_1): Likewise.
30838         (const_scalar_mask_from_tree): Likewise.
30839         * fold-const-call.c (fold_const_reduction): Likewise.
30840         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
30841         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
30842         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
30843         (fold_relational_const): Likewise.
30844         (native_interpret_vector): Likewise.  Change the size from an
30845         int to an unsigned int.
30846         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
30847         TYPE_VECTOR_SUBPARTS.
30848         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
30849         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
30850         duplicating a non-constant operand into a variable-length vector.
30851         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
30852         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
30853         * ipa-icf.c (sem_variable::equals): Likewise.
30854         * match.pd: Likewise.
30855         * omp-simd-clone.c (simd_clone_subparts): Likewise.
30856         * print-tree.c (print_node): Likewise.
30857         * stor-layout.c (layout_type): Likewise.
30858         * targhooks.c (default_builtin_vectorization_cost): Likewise.
30859         * tree-cfg.c (verify_gimple_comparison): Likewise.
30860         (verify_gimple_assign_binary): Likewise.
30861         (verify_gimple_assign_ternary): Likewise.
30862         (verify_gimple_assign_single): Likewise.
30863         * tree-pretty-print.c (dump_generic_node): Likewise.
30864         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30865         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
30866         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
30867         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
30868         (vect_shift_permute_load_chain): Likewise.
30869         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
30870         (expand_vector_condition, optimize_vector_constructor): Likewise.
30871         (lower_vec_perm, get_compute_type): Likewise.
30872         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30873         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
30874         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
30875         (vect_recog_mask_conversion_pattern): Likewise.
30876         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
30877         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
30878         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30879         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
30880         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
30881         (vectorizable_shift, vectorizable_operation, vectorizable_store)
30882         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
30883         (supportable_widening_operation): Likewise.
30884         (supportable_narrowing_operation): Likewise.
30885         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
30886         Likewise.
30887         * varasm.c (output_constant): Likewise.
30889 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30890             Alan Hayward  <alan.hayward@arm.com>
30891             David Sherwood  <david.sherwood@arm.com>
30893         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
30894         so that both the length == 3 and length != 3 cases set up their
30895         own permute vectors.  Add comments explaining why we know the
30896         number of elements is constant.
30897         (vect_permute_load_chain): Likewise.
30899 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30900             Alan Hayward  <alan.hayward@arm.com>
30901             David Sherwood  <david.sherwood@arm.com>
30903         * machmode.h (mode_nunits): Change from unsigned char to
30904         poly_uint16_pod.
30905         (ONLY_FIXED_SIZE_MODES): New macro.
30906         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
30907         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
30908         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
30909         New typedefs.
30910         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
30911         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
30912         or if measurement_type is not polynomial.
30913         * genmodes.c (ZERO_COEFFS): New macro.
30914         (emit_mode_nunits_inline): Make mode_nunits_inline return a
30915         poly_uint16.
30916         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
30917         Use ZERO_COEFFS when emitting initializers.
30918         * data-streamer.h (bp_pack_poly_value): New function.
30919         (bp_unpack_poly_value): Likewise.
30920         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30921         for GET_MODE_NUNITS.
30922         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30923         for GET_MODE_NUNITS.
30924         * tree.c (make_vector_type): Remove temporary shim and make
30925         the real function take the number of units as a poly_uint64
30926         rather than an int.
30927         (build_vector_type_for_mode): Handle polynomial nunits.
30928         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
30929         * emit-rtl.c (const_vec_series_p_1): Likewise.
30930         (gen_rtx_CONST_VECTOR): Likewise.
30931         * fold-const.c (test_vec_duplicate_folding): Likewise.
30932         * genrecog.c (validate_pattern): Likewise.
30933         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
30934         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30935         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
30936         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
30937         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
30938         * rtlanal.c (subreg_get_info): Likewise.
30939         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30940         (vect_grouped_load_supported): Likewise.
30941         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
30942         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
30943         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
30944         (simplify_const_unary_operation, simplify_binary_operation_1)
30945         (simplify_const_binary_operation, simplify_ternary_operation)
30946         (test_vector_ops_duplicate, test_vector_ops): Likewise.
30947         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
30948         instead of CONST_VECTOR_NUNITS.
30949         * varasm.c (output_constant_pool_2): Likewise.
30950         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
30951         explicit-encoded elements in the XVEC for variable-length vectors.
30953 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30955         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
30957 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30958             Alan Hayward  <alan.hayward@arm.com>
30959             David Sherwood  <david.sherwood@arm.com>
30961         * coretypes.h (fixed_size_mode): Declare.
30962         (fixed_size_mode_pod): New typedef.
30963         * builtins.h (target_builtins::x_apply_args_mode)
30964         (target_builtins::x_apply_result_mode): Change type to
30965         fixed_size_mode_pod.
30966         * builtins.c (apply_args_size, apply_result_size, result_vector)
30967         (expand_builtin_apply_args_1, expand_builtin_apply)
30968         (expand_builtin_return): Update accordingly.
30970 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30972         * cse.c (hash_rtx_cb): Hash only the encoded elements.
30973         * cselib.c (cselib_hash_rtx): Likewise.
30974         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
30975         CONST_VECTOR encoding.
30977 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30978             Jeff Law  <law@redhat.com>
30980         PR target/83641
30981         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
30982         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
30983         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
30984         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
30986         PR target/83641
30987         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
30988         explicitly probe *sp in a noreturn function if there were any callee
30989         register saves or frame pointer is needed.
30991 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30993         PR debug/83621
30994         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
30995         BLKmode for ternary, binary or unary expressions.
30997         PR debug/83645
30998         * var-tracking.c (delete_vta_debug_insn): New inline function.
30999         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31000         insns from get_insns () to NULL instead of each bb separately.
31001         Use delete_vta_debug_insn.  No longer static.
31002         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31003         delete_vta_debug_insns callers.
31004         * rtl.h (delete_vta_debug_insns): Declare.
31005         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31006         instead of variable_tracking_main.
31008 2018-01-03  Martin Sebor  <msebor@redhat.com>
31010         PR tree-optimization/83603
31011         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31012         arguments past the endof the argument list in functions declared
31013         without a prototype.
31014         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31015         Avoid checking when arguments are null.
31017 2018-01-03  Martin Sebor  <msebor@redhat.com>
31019         PR c/83559
31020         * doc/extend.texi (attribute const): Fix a typo.
31021         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31022         issuing -Wsuggest-attribute for void functions.
31024 2018-01-03  Martin Sebor  <msebor@redhat.com>
31026         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31027         offset_int::from instead of wide_int::to_shwi.
31028         (maybe_diag_overlap): Remove assertion.
31029         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31030         * gimple-ssa-sprintf.c (format_directive): Same.
31031         (parse_directive): Same.
31032         (sprintf_dom_walker::compute_format_length): Same.
31033         (try_substitute_return_value): Same.
31035 2018-01-03  Jeff Law  <law@redhat.com>
31037         PR middle-end/83654
31038         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31039         non-constant residual for zero at runtime and avoid probing in
31040         that case.  Reorganize code for trailing problem to mirror handling
31041         of the residual.
31043 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31045         PR tree-optimization/83501
31046         * tree-ssa-strlen.c (get_string_cst): New.
31047         (handle_char_store): Call get_string_cst.
31049 2018-01-03  Martin Liska  <mliska@suse.cz>
31051         PR tree-optimization/83593
31052         * tree-ssa-strlen.c: Include tree-cfg.h.
31053         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31054         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31055         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31056         to false.
31057         (strlen_dom_walker::before_dom_children): Call
31058         gimple_purge_dead_eh_edges. Dump tranformation with details
31059         dump flags.
31060         (strlen_dom_walker::before_dom_children): Update call by adding
31061         new argument cleanup_eh.
31062         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31064 2018-01-03  Martin Liska  <mliska@suse.cz>
31066         PR ipa/83549
31067         * cif-code.def (VARIADIC_THUNK): New enum value.
31068         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31069         thunks.
31071 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31073         * sse.md (mov<mode>_internal): Tighten condition for when to use
31074         vmovdqu<ssescalarsize> for TI and OI modes.
31076 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31078         Update copyright years.
31080 2018-01-03  Martin Liska  <mliska@suse.cz>
31082         PR ipa/83594
31083         * ipa-visibility.c (function_and_variable_visibility): Skip
31084         functions with noipa attribure.
31086 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31088         * gcc.c (process_command): Update copyright notice dates.
31089         * gcov-dump.c (print_version): Ditto.
31090         * gcov.c (print_version): Ditto.
31091         * gcov-tool.c (print_version): Ditto.
31092         * gengtype.c (create_file): Ditto.
31093         * doc/cpp.texi: Bump @copying's copyright year.
31094         * doc/cppinternals.texi: Ditto.
31095         * doc/gcc.texi: Ditto.
31096         * doc/gccint.texi: Ditto.
31097         * doc/gcov.texi: Ditto.
31098         * doc/install.texi: Ditto.
31099         * doc/invoke.texi: Ditto.
31101 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31103         * vector-builder.h (vector_builder::m_full_nelts): Change from
31104         unsigned int to poly_uint64.
31105         (vector_builder::full_nelts): Update prototype accordingly.
31106         (vector_builder::new_vector): Likewise.
31107         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31108         (vector_builder::operator ==): Likewise.
31109         (vector_builder::finalize): Likewise.
31110         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31111         Take the number of elements as a poly_uint64 rather than an
31112         unsigned int.
31113         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31114         from unsigned int to poly_uint64.
31115         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31116         (vec_perm_indices::new_vector): Likewise.
31117         (vec_perm_indices::length): Likewise.
31118         (vec_perm_indices::nelts_per_input): Likewise.
31119         (vec_perm_indices::input_nelts): Likewise.
31120         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31121         number of elements per input as a poly_uint64 rather than an
31122         unsigned int.  Use the original encoding for variable-length
31123         vectors, rather than clamping each individual element.
31124         For the second and subsequent elements in each pattern,
31125         clamp the step and base before clamping their sum.
31126         (vec_perm_indices::series_p): Handle polynomial element counts.
31127         (vec_perm_indices::all_in_range_p): Likewise.
31128         (vec_perm_indices_to_tree): Likewise.
31129         (vec_perm_indices_to_rtx): Likewise.
31130         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31131         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31132         (tree_vector_builder::new_binary_operation): Handle polynomial
31133         element counts.  Return false if we need to know the number
31134         of elements at compile time.
31135         * fold-const.c (fold_vec_perm): Punt if the number of elements
31136         isn't known at compile time.
31138 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31140         * vec-perm-indices.h (vec_perm_builder): Change element type
31141         from HOST_WIDE_INT to poly_int64.
31142         (vec_perm_indices::element_type): Update accordingly.
31143         (vec_perm_indices::clamp): Handle polynomial element_types.
31144         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31145         (vec_perm_indices::all_in_range_p): Likewise.
31146         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31147         than shwi trees.
31148         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31149         polynomial vec_perm_indices element types.
31150         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31151         * fold-const.c (fold_vec_perm): Likewise.
31152         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31153         * tree-vect-generic.c (lower_vec_perm): Likewise.
31154         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31155         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31156         element type to HOST_WIDE_INT.
31158 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31159             Alan Hayward  <alan.hayward@arm.com>
31160             David Sherwood  <david.sherwood@arm.com>
31162         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31163         rather than an int.  Use plus_constant.
31164         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31165         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31167 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31168             Alan Hayward  <alan.hayward@arm.com>
31169             David Sherwood  <david.sherwood@arm.com>
31171         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31172         a HOST_WIDE_INT to a poly_int64.
31174 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31175             Alan Hayward  <alan.hayward@arm.com>
31176             David Sherwood  <david.sherwood@arm.com>
31178         * calls.c (load_register_parameters): Cope with polynomial
31179         mode sizes.  Require a constant size for BLKmode parameters
31180         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31181         forces a parameter to be padded at the lsb end in order to
31182         fill a complete number of words, require the parameter size
31183         to be ordered wrt UNITS_PER_WORD.
31185 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31186             Alan Hayward  <alan.hayward@arm.com>
31187             David Sherwood  <david.sherwood@arm.com>
31189         * reload1.c (spill_stack_slot_width): Change element type
31190         from unsigned int to poly_uint64_pod.
31191         (alter_reg): Treat mode sizes as polynomial.
31193 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31194             Alan Hayward  <alan.hayward@arm.com>
31195             David Sherwood  <david.sherwood@arm.com>
31197         * reload.c (complex_word_subreg_p): New function.
31198         (reload_inner_reg_of_subreg, push_reload): Use it.
31200 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31201             Alan Hayward  <alan.hayward@arm.com>
31202             David Sherwood  <david.sherwood@arm.com>
31204         * lra-constraints.c (process_alt_operands): Reject matched
31205         operands whose sizes aren't ordered.
31206         (match_reload): Refer to this check here.
31208 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31209             Alan Hayward  <alan.hayward@arm.com>
31210             David Sherwood  <david.sherwood@arm.com>
31212         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31213         that the mode size is in the set {1, 2, 4, 8, 16}.
31215 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31216             Alan Hayward  <alan.hayward@arm.com>
31217             David Sherwood  <david.sherwood@arm.com>
31219         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31220         Use plus_constant instead of gen_rtx_PLUS.
31222 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31223             Alan Hayward  <alan.hayward@arm.com>
31224             David Sherwood  <david.sherwood@arm.com>
31226         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31227         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31228         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31229         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31230         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31231         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31232         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31233         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31234         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31235         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31236         a poly_int64.
31237         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31238         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31239         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31240         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31241         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31242         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31243         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31244         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31245         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31246         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31247         function.
31248         * expr.c (emit_move_resolve_push): Treat the input and result
31249         of PUSH_ROUNDING as a poly_int64.
31250         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31251         (emit_push_insn): Likewise.
31252         * lra-eliminations.c (mark_not_eliminable): Likewise.
31253         * recog.c (push_operand): Likewise.
31254         * reload1.c (elimination_effects): Likewise.
31255         * rtlanal.c (nonzero_bits1): Likewise.
31256         * calls.c (store_one_arg): Likewise.  Require the padding to be
31257         known at compile time.
31259 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31260             Alan Hayward  <alan.hayward@arm.com>
31261             David Sherwood  <david.sherwood@arm.com>
31263         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31264         Use plus_constant instead of gen_rtx_PLUS.
31266 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31267             Alan Hayward  <alan.hayward@arm.com>
31268             David Sherwood  <david.sherwood@arm.com>
31270         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31271         rather than an int.
31273 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31274             Alan Hayward  <alan.hayward@arm.com>
31275             David Sherwood  <david.sherwood@arm.com>
31277         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31278         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31279         via stack temporaries.  Treat the mode size as polynomial too.
31281 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31282             Alan Hayward  <alan.hayward@arm.com>
31283             David Sherwood  <david.sherwood@arm.com>
31285         * expr.c (expand_expr_real_2): When handling conversions involving
31286         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31287         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31288         as a poly_uint64 too.
31290 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31291             Alan Hayward  <alan.hayward@arm.com>
31292             David Sherwood  <david.sherwood@arm.com>
31294         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31296 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31297             Alan Hayward  <alan.hayward@arm.com>
31298             David Sherwood  <david.sherwood@arm.com>
31300         * combine.c (can_change_dest_mode): Handle polynomial
31301         REGMODE_NATURAL_SIZE.
31302         * expmed.c (store_bit_field_1): Likewise.
31303         * expr.c (store_constructor): Likewise.
31304         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31305         and polynomial REGMODE_NATURAL_SIZE.
31306         (gen_lowpart_common): Likewise.
31307         * reginfo.c (record_subregs_of_mode): Likewise.
31308         * rtlanal.c (read_modify_subreg_p): Likewise.
31310 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31311             Alan Hayward  <alan.hayward@arm.com>
31312             David Sherwood  <david.sherwood@arm.com>
31314         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31315         numbers of elements.
31317 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31318             Alan Hayward  <alan.hayward@arm.com>
31319             David Sherwood  <david.sherwood@arm.com>
31321         * match.pd: Cope with polynomial numbers of vector elements.
31323 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31324             Alan Hayward  <alan.hayward@arm.com>
31325             David Sherwood  <david.sherwood@arm.com>
31327         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31328         in a POINTER_PLUS_EXPR.
31330 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31331             Alan Hayward  <alan.hayward@arm.com>
31332             David Sherwood  <david.sherwood@arm.com>
31334         * omp-simd-clone.c (simd_clone_subparts): New function.
31335         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31336         (ipa_simd_modify_function_body): Likewise.
31338 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31339             Alan Hayward  <alan.hayward@arm.com>
31340             David Sherwood  <david.sherwood@arm.com>
31342         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31343         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31344         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31345         (expand_vector_condition, vector_element): Likewise.
31346         (subparts_gt): New function.
31347         (get_compute_type): Use subparts_gt.
31348         (count_type_subparts): Delete.
31349         (expand_vector_operations_1): Use subparts_gt instead of
31350         count_type_subparts.
31352 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31353             Alan Hayward  <alan.hayward@arm.com>
31354             David Sherwood  <david.sherwood@arm.com>
31356         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31357         (vect_compile_time_alias): ...this new function.  Do the calculation
31358         on poly_ints rather than trees.
31359         (vect_prune_runtime_alias_test_list): Update call accordingly.
31361 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31362             Alan Hayward  <alan.hayward@arm.com>
31363             David Sherwood  <david.sherwood@arm.com>
31365         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31366         numbers of units.
31367         (vect_schedule_slp_instance): Likewise.
31369 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31370             Alan Hayward  <alan.hayward@arm.com>
31371             David Sherwood  <david.sherwood@arm.com>
31373         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31374         constant and extern definitions for variable-length vectors.
31375         (vect_get_constant_vectors): Note that the number of units
31376         is known to be constant.
31378 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31379             Alan Hayward  <alan.hayward@arm.com>
31380             David Sherwood  <david.sherwood@arm.com>
31382         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31383         of units as polynomial.  Choose between WIDE and NARROW based
31384         on multiple_p.
31386 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31387             Alan Hayward  <alan.hayward@arm.com>
31388             David Sherwood  <david.sherwood@arm.com>
31390         * tree-vect-stmts.c (simd_clone_subparts): New function.
31391         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31393 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31394             Alan Hayward  <alan.hayward@arm.com>
31395             David Sherwood  <david.sherwood@arm.com>
31397         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31398         vectors as polynomial.  Use build_index_vector for
31399         IFN_GOMP_SIMD_LANE.
31401 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31402             Alan Hayward  <alan.hayward@arm.com>
31403             David Sherwood  <david.sherwood@arm.com>
31405         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31406         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31407         for variable-length vectors.
31408         (vectorizable_mask_load_store): Treat the number of units as
31409         polynomial, asserting that it is constant if the condition has
31410         already been enforced.
31411         (vectorizable_store, vectorizable_load): Likewise.
31413 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31414             Alan Hayward  <alan.hayward@arm.com>
31415             David Sherwood  <david.sherwood@arm.com>
31417         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31418         of units as polynomial.  Punt if we can't tell at compile time
31419         which vector contains the final result.
31421 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31422             Alan Hayward  <alan.hayward@arm.com>
31423             David Sherwood  <david.sherwood@arm.com>
31425         * tree-vect-loop.c (vectorizable_induction): Treat the number
31426         of units as polynomial.  Punt on SLP inductions.  Use an integer
31427         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31428         cast of such a series for variable-length floating-point
31429         reductions.
31431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31432             Alan Hayward  <alan.hayward@arm.com>
31433             David Sherwood  <david.sherwood@arm.com>
31435         * tree.h (build_index_vector): Declare.
31436         * tree.c (build_index_vector): New function.
31437         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31438         of units as polynomial, forcibly converting it to a constant if
31439         vectorizable_reduction has already enforced the condition.
31440         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31441         to create a {1,2,3,...} vector.
31442         (vectorizable_reduction): Treat the number of units as polynomial.
31443         Choose vectype_in based on the largest scalar element size rather
31444         than the smallest number of units.  Enforce the restrictions
31445         relied on above.
31447 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31448             Alan Hayward  <alan.hayward@arm.com>
31449             David Sherwood  <david.sherwood@arm.com>
31451         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31452         number of units as polynomial.
31454 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31455             Alan Hayward  <alan.hayward@arm.com>
31456             David Sherwood  <david.sherwood@arm.com>
31458         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31459         * target.def (autovectorize_vector_sizes): Return the vector sizes
31460         by pointer, using vector_sizes rather than a bitmask.
31461         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31462         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31463         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31464         Likewise.
31465         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31466         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31467         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31468         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31469         * omp-general.c (omp_max_vf): Likewise.
31470         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31471         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31472         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31473         * tree-vect-slp.c (vect_slp_bb): Likewise.
31474         * doc/tm.texi: Regenerate.
31475         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31476         to a poly_uint64.
31477         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31478         the vector size as a poly_uint64 rather than an unsigned int.
31479         (current_vector_size): Change from an unsigned int to a poly_uint64.
31480         (get_vectype_for_scalar_type): Update accordingly.
31481         * tree.h (build_truth_vector_type): Take the size and number of
31482         units as a poly_uint64 rather than an unsigned int.
31483         (build_vector_type): Add a temporary overload that takes
31484         the number of units as a poly_uint64 rather than an unsigned int.
31485         * tree.c (make_vector_type): Likewise.
31486         (build_truth_vector_type): Take the number of units as a poly_uint64
31487         rather than an unsigned int.
31489 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31490             Alan Hayward  <alan.hayward@arm.com>
31491             David Sherwood  <david.sherwood@arm.com>
31493         * target.def (get_mask_mode): Take the number of units and length
31494         as poly_uint64s rather than unsigned ints.
31495         * targhooks.h (default_get_mask_mode): Update accordingly.
31496         * targhooks.c (default_get_mask_mode): Likewise.
31497         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31498         * doc/tm.texi: Regenerate.
31500 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31501             Alan Hayward  <alan.hayward@arm.com>
31502             David Sherwood  <david.sherwood@arm.com>
31504         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
31505         * omp-general.c (omp_max_vf): Likewise.
31506         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
31507         (expand_omp_simd): Handle polynomial safelen.
31508         * omp-low.c (omplow_simd_context): Add a default constructor.
31509         (omplow_simd_context::max_vf): Change from int to poly_uint64.
31510         (lower_rec_simd_input_clauses): Update accordingly.
31511         (lower_rec_input_clauses): Likewise.
31513 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31514             Alan Hayward  <alan.hayward@arm.com>
31515             David Sherwood  <david.sherwood@arm.com>
31517         * tree-vectorizer.h (vect_nunits_for_cost): New function.
31518         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
31519         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
31520         (vect_analyze_slp_cost): Likewise.
31521         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
31522         (vect_model_load_cost): Likewise.
31524 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31525             Alan Hayward  <alan.hayward@arm.com>
31526             David Sherwood  <david.sherwood@arm.com>
31528         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
31529         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
31530         from an unsigned int * to a poly_uint64_pod *.
31531         (calculate_unrolling_factor): New function.
31532         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
31534 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31535             Alan Hayward  <alan.hayward@arm.com>
31536             David Sherwood  <david.sherwood@arm.com>
31538         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
31539         from an unsigned int to a poly_uint64.
31540         (_loop_vec_info::slp_unrolling_factor): Likewise.
31541         (_loop_vec_info::vectorization_factor): Change from an int
31542         to a poly_uint64.
31543         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
31544         (vect_get_num_vectors): New function.
31545         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
31546         (vect_get_num_copies): Use vect_get_num_vectors.
31547         (vect_analyze_data_ref_dependences): Change max_vf from an int *
31548         to an unsigned int *.
31549         (vect_analyze_data_refs): Change min_vf from an int * to a
31550         poly_uint64 *.
31551         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31552         than an unsigned HOST_WIDE_INT.
31553         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
31554         (vect_analyze_data_ref_dependence): Change max_vf from an int *
31555         to an unsigned int *.
31556         (vect_analyze_data_ref_dependences): Likewise.
31557         (vect_compute_data_ref_alignment): Handle polynomial vf.
31558         (vect_enhance_data_refs_alignment): Likewise.
31559         (vect_prune_runtime_alias_test_list): Likewise.
31560         (vect_shift_permute_load_chain): Likewise.
31561         (vect_supportable_dr_alignment): Likewise.
31562         (dependence_distance_ge_vf): Take the vectorization factor as a
31563         poly_uint64 rather than an unsigned HOST_WIDE_INT.
31564         (vect_analyze_data_refs): Change min_vf from an int * to a
31565         poly_uint64 *.
31566         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
31567         vfm1 as a poly_uint64 rather than an int.  Make the same change
31568         for the returned bound_scalar.
31569         (vect_gen_vector_loop_niters): Handle polynomial vf.
31570         (vect_do_peeling): Likewise.  Update call to
31571         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
31572         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
31573         be constant.
31574         * tree-vect-loop.c (vect_determine_vectorization_factor)
31575         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
31576         (vect_get_known_peeling_cost): Likewise.
31577         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
31578         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
31579         (vect_transform_loop): Likewise.  Use the lowest possible VF when
31580         updating the upper bounds of the loop.
31581         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
31582         rather than an int.
31583         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
31584         polynomial unroll factors.
31585         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
31586         (vect_make_slp_decision): Likewise.
31587         (vect_supported_load_permutation_p): Likewise, and polynomial
31588         vf too.
31589         (vect_analyze_slp_cost): Handle polynomial vf.
31590         (vect_slp_analyze_node_operations): Likewise.
31591         (vect_slp_analyze_bb_1): Likewise.
31592         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31593         than an unsigned HOST_WIDE_INT.
31594         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
31595         (vectorizable_load): Handle polynomial vf.
31596         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
31597         a poly_uint64.
31598         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
31600 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31601             Alan Hayward  <alan.hayward@arm.com>
31602             David Sherwood  <david.sherwood@arm.com>
31604         * match.pd: Handle bit operations involving three constants
31605         and try to fold one pair.
31607 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31609         * tree-vect-loop-manip.c: Include gimple-fold.h.
31610         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
31611         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
31612         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
31613         Add a path that uses a step of VF instead of 1, but disable it
31614         for now.
31615         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
31616         and niters_no_overflow parameters.  Update calls to
31617         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
31618         Create a new SSA name if the latter choses to use a ste other
31619         than zero, and return it via niters_vector_mult_vf_var.
31620         * tree-vect-loop.c (vect_transform_loop): Update calls to
31621         vect_do_peeling, vect_gen_vector_loop_niters and
31622         slpeel_make_loop_iterate_ntimes.
31623         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
31624         (vect_gen_vector_loop_niters): Update declarations after above changes.
31626 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
31628         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
31629         128-bit round to integer instructions.
31630         (ceil<mode>2): Likewise.
31631         (btrunc<mode>2): Likewise.
31632         (round<mode>2): Likewise.
31634 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31636         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
31637         unaligned VSX load/store on P8/P9.
31638         (expand_block_clear): Allow the use of unaligned VSX
31639         load/store on P8/P9.
31641 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31643         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
31644         New function.
31645         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
31646         swap associated with both a load and a store.
31648 2018-01-02  Andrew Waterman  <andrew@sifive.com>
31650         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
31651         * config/riscv/riscv.md (clear_cache): Use it.
31653 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
31655         * web.c: Remove out-of-date comment.
31657 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31659         * expr.c (fixup_args_size_notes): Check that any existing
31660         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
31661         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
31662         (emit_single_push_insn): ...here.
31664 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31666         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
31667         (const_vector_encoded_nelts): New function.
31668         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
31669         (const_vector_int_elt, const_vector_elt): Declare.
31670         * emit-rtl.c (const_vector_int_elt_1): New function.
31671         (const_vector_elt): Likewise.
31672         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
31673         of CONST_VECTOR_ELT.
31675 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31677         * expr.c: Include rtx-vector-builder.h.
31678         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
31679         directly on the tree encoding.
31680         (const_vector_from_tree): Likewise.
31681         * optabs.c: Include rtx-vector-builder.h.
31682         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
31683         sequence of "u" values.
31684         * vec-perm-indices.c: Include rtx-vector-builder.h.
31685         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
31686         directly on the vec_perm_indices encoding.
31688 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31690         * doc/rtl.texi (const_vector): Describe new encoding scheme.
31691         * Makefile.in (OBJS): Add rtx-vector-builder.o.
31692         * rtx-vector-builder.h: New file.
31693         * rtx-vector-builder.c: Likewise.
31694         * rtl.h (rtx_def::u2): Add a const_vector field.
31695         (CONST_VECTOR_NPATTERNS): New macro.
31696         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
31697         (CONST_VECTOR_DUPLICATE_P): Likewise.
31698         (CONST_VECTOR_STEPPED_P): Likewise.
31699         (CONST_VECTOR_ENCODED_ELT): Likewise.
31700         (const_vec_duplicate_p): Check for a duplicated vector encoding.
31701         (unwrap_const_vec_duplicate): Likewise.
31702         (const_vec_series_p): Check for a non-duplicated vector encoding.
31703         Say that the function only returns true for integer vectors.
31704         * emit-rtl.c: Include rtx-vector-builder.h.
31705         (gen_const_vec_duplicate_1): Delete.
31706         (gen_const_vector): Call gen_const_vec_duplicate instead of
31707         gen_const_vec_duplicate_1.
31708         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
31709         (gen_const_vec_duplicate): Use rtx_vector_builder.
31710         (gen_const_vec_series): Likewise.
31711         (gen_rtx_CONST_VECTOR): Likewise.
31712         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
31713         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31714         Build a new vector rather than modifying a CONST_VECTOR in-place.
31715         (handle_special_swappables): Update call accordingly.
31716         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
31717         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31718         Build a new vector rather than modifying a CONST_VECTOR in-place.
31719         (handle_special_swappables): Update call accordingly.
31721 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31723         * simplify-rtx.c (simplify_const_binary_operation): Use
31724         CONST_VECTOR_ELT instead of XVECEXP.
31726 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31728         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
31729         the selector elements to be different from the data elements
31730         if the selector is a VECTOR_CST.
31731         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
31732         ssizetype for the selector.
31734 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31736         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
31737         before testing each element individually.
31738         * tree-vect-generic.c (lower_vec_perm): Likewise.
31740 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31742         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
31743         * selftest-run-tests.c (selftest::run_tests): Call it.
31744         * vector-builder.h (vector_builder::operator ==): New function.
31745         (vector_builder::operator !=): Likewise.
31746         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
31747         (vec_perm_indices::all_from_input_p): New function.
31748         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31749         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
31750         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
31751         instead of reading the VECTOR_CST directly.  Detect whether both
31752         vector inputs are the same before constructing the vec_perm_indices,
31753         and update the number of inputs argument accordingly.  Use the
31754         utility functions added above.  Only construct sel2 if we need to.
31756 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31758         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
31759         the broadcast of the low byte.
31760         (expand_mult_highpart): Use an explicit encoding for the permutes.
31761         * optabs-query.c (can_mult_highpart_p): Likewise.
31762         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
31763         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31764         (vectorizable_bswap): Likewise.
31765         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
31766         explicit encoding for the power-of-2 permutes.
31767         (vect_permute_store_chain): Likewise.
31768         (vect_grouped_load_supported): Likewise.
31769         (vect_permute_load_chain): Likewise.
31771 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31773         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
31774         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
31775         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
31776         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31777         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
31778         (vect_gen_perm_mask_any): Likewise.
31780 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31782         * int-vector-builder.h: New file.
31783         * vec-perm-indices.h: Include int-vector-builder.h.
31784         (vec_perm_indices): Redefine as an int_vector_builder.
31785         (auto_vec_perm_indices): Delete.
31786         (vec_perm_builder): Redefine as a stand-alone class.
31787         (vec_perm_indices::vec_perm_indices): New function.
31788         (vec_perm_indices::clamp): Likewise.
31789         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
31790         (vec_perm_indices::new_vector): New function.
31791         (vec_perm_indices::new_expanded_vector): Update for new
31792         vec_perm_indices class.
31793         (vec_perm_indices::rotate_inputs): New function.
31794         (vec_perm_indices::all_in_range_p): Operate directly on the
31795         encoded form, without computing elided elements.
31796         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
31797         encoding.  Update for new vec_perm_indices class.
31798         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
31799         the given vec_perm_builder.
31800         (expand_vec_perm_var): Update vec_perm_builder constructor.
31801         (expand_mult_highpart): Use vec_perm_builder instead of
31802         auto_vec_perm_indices.
31803         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
31804         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
31805         or double series encoding as appropriate.
31806         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
31807         vec_perm_indices instead of auto_vec_perm_indices.
31808         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31809         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31810         (vect_permute_store_chain): Likewise.
31811         (vect_grouped_load_supported): Likewise.
31812         (vect_permute_load_chain): Likewise.
31813         (vect_shift_permute_load_chain): Likewise.
31814         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31815         (vect_transform_slp_perm_load): Likewise.
31816         (vect_schedule_slp_instance): Likewise.
31817         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31818         (vectorizable_mask_load_store): Likewise.
31819         (vectorizable_bswap): Likewise.
31820         (vectorizable_store): Likewise.
31821         (vectorizable_load): Likewise.
31822         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
31823         vec_perm_indices instead of auto_vec_perm_indices.  Use
31824         tree_to_vec_perm_builder to read the vector from a tree.
31825         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
31826         vec_perm_builder instead of a vec_perm_indices.
31827         (have_whole_vector_shift): Use vec_perm_builder and
31828         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
31829         truncation to calc_vec_perm_mask_for_shift.
31830         (vect_create_epilog_for_reduction): Likewise.
31831         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
31832         from auto_vec_perm_indices to vec_perm_indices.
31833         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31834         instead of changing individual elements.
31835         (aarch64_vectorize_vec_perm_const): Use new_vector to install
31836         the vector in d.perm.
31837         * config/arm/arm.c (expand_vec_perm_d::perm): Change
31838         from auto_vec_perm_indices to vec_perm_indices.
31839         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31840         instead of changing individual elements.
31841         (arm_vectorize_vec_perm_const): Use new_vector to install
31842         the vector in d.perm.
31843         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
31844         Update vec_perm_builder constructor.
31845         (rs6000_expand_interleave): Likewise.
31846         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
31847         (rs6000_expand_interleave): Likewise.
31849 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31851         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
31852         to qimode could truncate the indices.
31853         * optabs.c (expand_vec_perm_var): Likewise.
31855 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31857         * Makefile.in (OBJS): Add vec-perm-indices.o.
31858         * vec-perm-indices.h: New file.
31859         * vec-perm-indices.c: Likewise.
31860         * target.h (vec_perm_indices): Replace with a forward class
31861         declaration.
31862         (auto_vec_perm_indices): Move to vec-perm-indices.h.
31863         * optabs.h: Include vec-perm-indices.h.
31864         (expand_vec_perm): Delete.
31865         (selector_fits_mode_p, expand_vec_perm_var): Declare.
31866         (expand_vec_perm_const): Declare.
31867         * target.def (vec_perm_const_ok): Replace with...
31868         (vec_perm_const): ...this new hook.
31869         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
31870         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
31871         * doc/tm.texi: Regenerate.
31872         * optabs.def (vec_perm_const): Delete.
31873         * doc/md.texi (vec_perm_const): Likewise.
31874         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
31875         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
31876         expand_vec_perm for constant permutation vectors.  Assert that
31877         the mode of variable permutation vectors is the integer equivalent
31878         of the mode that is being permuted.
31879         * optabs-query.h (selector_fits_mode_p): Declare.
31880         * optabs-query.c: Include vec-perm-indices.h.
31881         (selector_fits_mode_p): New function.
31882         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
31883         is defined, instead of checking whether the vec_perm_const_optab
31884         exists.  Use targetm.vectorize.vec_perm_const instead of
31885         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
31886         fit in the vector mode before using a variable permute.
31887         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
31888         vec_perm_indices instead of an rtx.
31889         (expand_vec_perm): Replace with...
31890         (expand_vec_perm_const): ...this new function.  Take the selector
31891         as a vec_perm_indices rather than an rtx.  Also take the mode of
31892         the selector.  Update call to shift_amt_for_vec_perm_mask.
31893         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
31894         Use vec_perm_indices::new_expanded_vector to expand the original
31895         selector into bytes.  Check whether the indices fit in the vector
31896         mode before using a variable permute.
31897         (expand_vec_perm_var): Make global.
31898         (expand_mult_highpart): Use expand_vec_perm_const.
31899         * fold-const.c: Includes vec-perm-indices.h.
31900         * tree-ssa-forwprop.c: Likewise.
31901         * tree-vect-data-refs.c: Likewise.
31902         * tree-vect-generic.c: Likewise.
31903         * tree-vect-loop.c: Likewise.
31904         * tree-vect-slp.c: Likewise.
31905         * tree-vect-stmts.c: Likewise.
31906         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
31907         Delete.
31908         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
31909         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
31910         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
31911         (aarch64_vectorize_vec_perm_const): ...this new function.
31912         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31913         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31914         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
31915         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
31916         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31917         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31918         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
31919         into...
31920         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
31921         check for NEON modes.
31922         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
31923         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
31924         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
31925         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
31926         into...
31927         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
31928         the old VEC_PERM_CONST conditions.
31929         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
31930         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
31931         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
31932         (ia64_vectorize_vec_perm_const_ok): Merge into...
31933         (ia64_vectorize_vec_perm_const): ...this new function.
31934         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
31935         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
31936         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
31937         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
31938         * config/mips/mips.c (mips_expand_vec_perm_const)
31939         (mips_vectorize_vec_perm_const_ok): Merge into...
31940         (mips_vectorize_vec_perm_const): ...this new function.
31941         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
31942         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
31943         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
31944         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
31945         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
31946         (rs6000_expand_vec_perm_const): Delete.
31947         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
31948         Delete.
31949         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31950         (altivec_expand_vec_perm_const_le): Take each operand individually.
31951         Operate on constant selectors rather than rtxes.
31952         (altivec_expand_vec_perm_const): Likewise.  Update call to
31953         altivec_expand_vec_perm_const_le.
31954         (rs6000_expand_vec_perm_const): Delete.
31955         (rs6000_vectorize_vec_perm_const_ok): Delete.
31956         (rs6000_vectorize_vec_perm_const): New function.
31957         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31958         an element count and rtx array.
31959         (rs6000_expand_extract_even): Update call accordingly.
31960         (rs6000_expand_interleave): Likewise.
31961         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
31962         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
31963         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
31964         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
31965         (rs6000_expand_vec_perm_const): Delete.
31966         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31967         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31968         (altivec_expand_vec_perm_const_le): Take each operand individually.
31969         Operate on constant selectors rather than rtxes.
31970         (altivec_expand_vec_perm_const): Likewise.  Update call to
31971         altivec_expand_vec_perm_const_le.
31972         (rs6000_expand_vec_perm_const): Delete.
31973         (rs6000_vectorize_vec_perm_const_ok): Delete.
31974         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
31975         reference to the SPE evmerge intructions.
31976         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31977         an element count and rtx array.
31978         (rs6000_expand_extract_even): Update call accordingly.
31979         (rs6000_expand_interleave): Likewise.
31980         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
31981         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
31982         new function.
31983         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31985 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31987         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
31988         vector mode and that that mode matches the mode of the data
31989         being permuted.
31990         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
31991         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
31992         directly using expand_vec_perm_1 when forcing selectors into
31993         registers.
31994         (expand_vec_perm_var): New function, split out from expand_vec_perm.
31996 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31998         * optabs-query.h (can_vec_perm_p): Delete.
31999         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32000         * optabs-query.c (can_vec_perm_p): Split into...
32001         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32002         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32003         particular selector is valid.
32004         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32005         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32006         (vect_grouped_load_supported): Likewise.
32007         (vect_shift_permute_load_chain): Likewise.
32008         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32009         (vect_transform_slp_perm_load): Likewise.
32010         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32011         (vectorizable_bswap): Likewise.
32012         (vect_gen_perm_mask_checked): Likewise.
32013         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32014         implementations of variable permutation vectors into account
32015         when deciding which selector to use.
32016         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32017         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32018         with a false third argument.
32019         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32020         to test whether the constant selector is valid and can_vec_perm_var_p
32021         to test whether a variable selector is valid.
32023 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32025         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32026         * optabs-query.c (can_vec_perm_p): Likewise.
32027         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32028         instead of vec_perm_indices.
32029         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32030         (vect_gen_perm_mask_checked): Likewise,
32031         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32032         (vect_gen_perm_mask_checked): Likewise,
32034 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32036         * optabs-query.h (qimode_for_vec_perm): Declare.
32037         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32038         (qimode_for_vec_perm): ...this new function.
32039         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32041 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32043         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32044         does not have a conditional at the top.
32046 2018-01-02  Richard Biener  <rguenther@suse.de>
32048         * ipa-inline.c (big_speedup_p): Fix expression.
32050 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32052         PR target/81616
32053         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32054         for generic 4->6.
32056 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32058         PR target/81616
32059         Generic tuning.
32060         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32061         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32062         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32063         cond_taken_branch_cost 3->4.
32065 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32067         PR tree-optimization/83581
32068         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32069         TODO_cleanup_cfg if any changes have been made.
32071         PR middle-end/83608
32072         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32073         convert_modes if target mode has the right side, but different mode
32074         class.
32076         PR middle-end/83609
32077         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32078         last argument when extracting from CONCAT.  If either from_real or
32079         from_imag is NULL, use expansion through memory.  If result is not
32080         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32081         the parts directly to inner mode, if even that fails, use expansion
32082         through memory.
32084         PR middle-end/83623
32085         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32086         check for bswap in mode rather than HImode and use that in expand_unop
32087         too.
32089 Copyright (C) 2018 Free Software Foundation, Inc.
32091 Copying and distribution of this file, with or without modification,
32092 are permitted in any medium without royalty provided the copyright
32093 notice and this notice are preserved.