2018-11-14 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
bloba3b86595f2047bd9ea5d35cf6de3db66e7922b13
1 2018-11-14  Richard Biener  <rguenther@suse.de>
3         PR tree-optimization/88019
4         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
5         COND_EXPR arguments to non-trapping overflow.
7 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
9         * config/arc/arc.h (reg_class): Reorder registers classes, remove
10         unused register classes.
11         (REG_CLASS_NAMES): Likewise.
12         (REG_CLASS_CONTENTS): Likewise.
13         (FIXED_REGISTERS): Make lp_count fixed.
14         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
15         (PROGRAM_COUNTER_REGNO): Remove.
16         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
17         register classes, use constants for register numbers, remove
18         reg_class_contents references.
19         (arc_process_double_reg_moves): Add asserts.
20         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
21         lra_in_progress predicate.
22         (arc_init_reg_tables): Remove unused register classes.
23         (arc_register_move_cost): Likewise.
24         (arc_preferred_reload_class): Likewise.
25         (hwloop_optimize): Update rtx patterns involving lp_count
26         register.
27         (arc_return_address_register): Rename ILINK1, INLINK2 regnums
28         macros.
29         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
30         CHEAP_CORE_REGS.  Former one will be used for LRA.
31         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
32         one will be used for LRA.
33         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
34         one will be used for LRA.
35         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
36         Former one will be used for LRA.
37         ("f"): Delete constraint.
38         ("k"): Likewise.
39         ("e"): Likewise.
40         ("l"): Change it from register constraint to constraint.
41         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
42         (movhi_insn): Likewise.
43         (movsi_insn): Update pattern.
44         (arc_lp): Likewise.
45         (dbnz): Likewise.
46         (stack_tie): Remove 'b' constraint letter.
47         (R4_REG): Define.
48         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
49         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
50         (R57_REG, R59_REG, PCL_REG): Likewise.
51         (ILINK1_REGNUM): Renamed to ILINK1_REG.
52         (ILINK2_REGNUM): Renamed to ILINK2_REG.
53         (Rgp): Remove.
54         (SP_REGS): Likewise.
55         (Rcw): Remove unused reg classes.
56         * config/arc/predicates.md (dest_reg_operand): Just default on
57         register_operand predicate.
58         (mpy_dest_reg_operand): Likewise.
59         (move_dest_operand): Use macros instead of constants.
60         * config/arc/arc.opt (mlra): Switch to lra as default.
62 2018-11-14  Richard Biener  <rguenther@suse.de>
64         PR tree-optimization/87974
65         * tree-vect-loop.c (vectorizable_reduction): When computing
66         the vectorized reduction PHI vector type ignore constant
67         and external defs.
69 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
71         PR tree-optimization/87977
72         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
73         stmt, build a new one and replace the old one with it.  Formatting fix.
74         Call release_ssa_name (x) if !has_other_use and !delete_div.
75         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
76         verify lhs of stmt is still def.
78 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
80         PR rtl-optimization/87507
81         * lower-subreg.c (operand_for_swap_move_operator): New function.
82         (simple_move): Strip simple operators.
83         (find_pseudo_copy): Likewise.
84         (resolve_operand_for_swap_move_operator): New function.
85         (resolve_simple_move): Strip simple operators and swap operands.
87 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
89         PR tree-optimization/87898
90         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
91         (ipa_simd_modify_function_body): Remove debug stmts where the first
92         argument was changed into a non-decl.
94 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
96         PR rtl-optimization/87899
97         * lra-lives.c (start_living): Update white space in comment.
98         (enum point_type): New.
99         (sparseset_contains_pseudos_p): New function.
100         (update_pseudo_point): Likewise.
101         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
102         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
103         handling.  Move early exit after adding conflicts.
104         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
105         if regno is already live.  Remove all handling of program points.
106         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
107         after adding conflicts.  Remove all handling of program points and
108         ignore_reg_for_conflicts.
109         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
110         and do not guard call to mark_pseudo_live.
111         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
112         and do not guard call to mark_pseudo_dead.
113         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
114         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
115         Use new function update_pseudo_point.  Handle register copies by
116         removing the source register from the live set.  Handle INOUT operands.
117         Update to the next program point using the unused_set, dead_set and
118         start_dying sets.
119         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
121 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
123         * builtins.c: Replace "source_location" with "location_t".
124         * diagnostic-show-locus.c: Likewise.
125         * diagnostic.c: Likewise.
126         * dumpfile.c: Likewise.
127         * gcc-rich-location.h: Likewise.
128         * genmatch.c: Likewise.
129         * gimple.h: Likewise.
130         * gimplify.c: Likewise.
131         * input.c: Likewise.
132         * input.h: Likewise.  Eliminate the typedef.
133         * omp-expand.c: Likewise.
134         * selftest.h: Likewise.
135         * substring-locations.h (get_source_location_for_substring):
136         Rename to..
137         (get_location_within_string): ...this.
138         * tree-cfg.c: Replace "source_location" with "location_t".
139         * tree-cfgcleanup.c: Likewise.
140         * tree-diagnostic.c: Likewise.
141         * tree-into-ssa.c: Likewise.
142         * tree-outof-ssa.c: Likewise.
143         * tree-parloops.c: Likewise.
144         * tree-phinodes.c: Likewise.
145         * tree-phinodes.h: Likewise.
146         * tree-ssa-loop-ivopts.c: Likewise.
147         * tree-ssa-loop-manip.c: Likewise.
148         * tree-ssa-phiopt.c: Likewise.
149         * tree-ssa-phiprop.c: Likewise.
150         * tree-ssa-threadupdate.c: Likewise.
151         * tree-ssa.c: Likewise.
152         * tree-ssa.h: Likewise.
153         * tree-vect-loop-manip.c: Likewise.
155 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
157         * doc/extend.texi: Fix typo in the weakref description.
159 2018-11-13  Richard Biener  <rguenther@suse.de>
161         PR tree-optimization/86991
162         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
163         group building until we have successfully detected the SLP
164         reduction.
165         (vect_is_simple_reduction): Remove fixup code here.
167 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
169         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
170         (dump_gimple_stmt): Use it.
171         (dump_gimple_stmt_loc): Likewise.
172         (dump_gimple_expr): Likewise.
173         (dump_gimple_expr_loc): Likewise.
174         (dump_generic_expr): Likewise.
175         (dump_generic_expr_loc): Likewise.
176         (dump_printf): Likewise.
177         (dump_printf_loc): Likewise.
178         (dump_dec): Likewise.
179         (dump_dec): Likewise.
180         (dump_hex): Likewise.
181         (dump_symtab_node): Likewise.
182         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
183         Guard dump call with dump_enabled_p.
184         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
185         * graphite-optimize-isl.c (optimize_isl): Likewise.
186         * graphite.c (graphite_transform_loops): Likewise.
187         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
188         * tree-parloops.c (parallelize_loops): Likewise.
189         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
190         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
191         (vect_prune_runtime_alias_test_list): Likewise.
192         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
193         (vect_estimate_min_profitable_iters): Likewise.
194         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
195         (vect_build_slp_tree_2): Likewise.
196         (vect_supported_load_permutation_p): Likewise.
197         (vect_slp_analyze_operations): Likewise.
198         (vect_slp_analyze_bb_1): Likewise.
199         (vect_slp_bb): Likewise.
200         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
201         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
202         (pass_slp_vectorize::execute): Likewise.
203         (increase_alignment): Likewise.
205 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
207         PR ipa/87955
208         * ipa-inline.c (report_inline_failed_reason): Guard calls to
209         cl_target_option_print_diff and cl_optimization_print_diff with
210         if (dump_file).
212 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
214         * doc/invoke.texi (-fsave-optimization-record): Note that the
215         output is compressed.
216         * optinfo-emit-json.cc: Include <zlib.h>.
217         (optrecord_json_writer::write): Compress the output.
219 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
221         * tree-vrp.c (value_range_base::dump): Dump type.
222         Do not use INF nomenclature for 1-bit types.
223         (dump_value_range): Group all variants to common dumping code.
224         (debug): New overloaded functions for value_ranges.
225         (value_range_base::dump): Remove no argument version.
226         (value_range::dump): Same.
228 2018-11-13  Richard Biener  <rguenther@suse.de>
230         PR tree-optimization/87931
231         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
232         nested cycles we support to latch computations vectorizable_reduction
233         handles.
235 2018-11-13  Martin Liska  <mliska@suse.cz>
237         PR tree-optimization/87885
238         * cfghooks.c (account_profile_record): Rename
239         to ...
240         (profile_record_check_consistency): ... this.
241         Calculate missing num_mismatched_freq_in.
242         (profile_record_account_profile): New function
243         that calculates time and size of a function.
244         * cfghooks.h (struct profile_record): Remove
245         all tuples.
246         (struct cfg_hooks): Remove after_pass flag.
247         (account_profile_record): Rename to ...
248         (profile_record_check_consistency): ... this.
249         (profile_record_account_profile): New.
250         * cfgrtl.c (rtl_account_profile_record): Remove
251         after_pass flag.
252         * passes.c (check_profile_consistency): Do only
253         checking.
254         (account_profile): Calculate size and time of
255         function only.
256         (pass_manager::dump_profile_report): Reformat
257         output.
258         (execute_one_ipa_transform_pass): Call
259         consistency check before clean upand call account_profile
260         after a clean up is done.
261         (execute_one_pass): Call check_profile_consistency and
262         account_profile instead of using after_pass flag..
263         * tree-cfg.c (gimple_account_profile_record): Likewise.
265 2018-11-13  Martin Liska  <mliska@suse.cz>
267         PR sanitizer/87930
268         * config/i386/i386.c (ix86_option_override_internal): Error
269         about usage -mabi=ms and -fsanitize={,kernel-}address.
271 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
273         * config/aarch64/aarch64.c
274         (aarch64_vectorize_preferred_vector_alignment): Change return type to
275         poly_uint64.
276         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
277         alignment being a poly int.
278         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
279         return type to poly_uint64.
280         * target.def (default_preferred_vector_alignment): Likewise.
281         * targhooks.c (default_preferred_vector_alignment): Likewise.
282         * targhooks.h (default_preferred_vector_alignment): Likewise.
283         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
284         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
285         poly int.
286         (vect_update_misalignment_for_peel): Likewise.
287         (vect_enhance_data_refs_alignment): Likewise.
288         (vect_find_same_alignment_drs): Likewise.
289         (vect_duplicate_ssa_name_ptr_info): Likewise.
290         (vect_setup_realignment): Likewise.
291         (vect_can_force_dr_alignment_p): Change alignment parameter type to
292         poly_uint64.
293         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
294         mask with a compile time variable vector alignment.
295         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
296         int.
297         (vect_do_peeling): Exit early if vector alignment is not constant.
298         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
299         a poly int.
300         (vectorizable_store): Likewise.
301         (vectorizable_load): Likweise.
302         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
303         poly_uint64.
304         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
305         poly int.
306         (vect_can_force_dr_alignment_p): Change alignment parameter type to
307         poly_uint64.
309 2018-11-13  Richard Biener  <rguenther@suse.de>
311         PR tree-optimization/87962
312         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
313         detect outer reduction for disqualifying in-loop uses.
315 2018-11-13  Richard Biener  <rguenther@suse.de>
317         PR tree-optimization/87967
318         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
319         for constants for the scalar loop.
321 2018-11-13  Alan Modra  <amodra@gmail.com>
323         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
324         integer constants when -mcmodel=medium.
326 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
328         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
329         * config/arc/elf.h (PROFILE_HOOK): Define.
330         * config/arc/linux.h (PROFILE_HOOK): Likewise.
332 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
334         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
335         jump table data in the text section.
337 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
339         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
340         to fit the eh_return pattern.
341         * config/arc/arc.md (eh_return): Define.
342         (VUNSPEC_ARC_EH_RETURN): Likewise.
343         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
344         new implementation.
345         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
348 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
350         * common/config/arc/arc-common.c (arc_option_optimization_table):
351         Millicode optimization is default on for size optimizations.
352         * config/arc/arc-protos.h (arc_check_multi): New function.
353         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
354         (ENTER_LEAVE_START_REG): Define.
355         (ENTER_LEAVE_END_REG): Likewise.
356         (arc_override_options): Disable millicode when long calls option
357         is on.
358         (arc_frame_info): Change it from int to bool.
359         (arc_compute_frame_size): Clean up.
360         (arc_save_restore): Remove.
361         (frame_save_reg): New function.
362         (frame_restore_reg): Likewise.
363         (arc_enter_leave_p): Likewise.
364         (arc_save_callee_saves): Likewise.
365         (arc_restore_callee_saves): Likewise.
366         (arc_save_callee_enter): Likewise.
367         (arc_restore_callee_leave): Likewise.
368         (arc_save_callee_milli): Likewise.
369         (arc_restore_callee_milli): Likewise.
370         (arc_expand_prologue): Reimplement to emit enter/leave
371         instructions.
372         (arc_expand_epilogue): Likewise.
373         (arc_check_multi): New function.
374         * config/arc/arc.md (push_multi_fp): New pattern.
375         (push_multi_fp_blink): Likewise.
376         (pop_multi_fp): Likewise.
377         (pop_multi_fp_blink): Likewise.
378         (pop_multi_fp_ret): Likewise.
379         (pop_multi_fp_blink_ret): Likewise.
380         * config/arc/arc.opt (mmillicode): Update option.
381         (mcode-density-frame): New option.
382         * config/arc/predicates.md (push_multi_operand): New predicate.
383         (pop_multi_operand): Likewise.
384         * doc/invoke.texi (ARC): Update ARC options information.
387 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
389         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
390         * config/arc/arc.c (operands_ok_ldd_std): New function.
391         (mem_ok_for_ldd_std): Likewise.
392         (gen_operands_ldd_std): Likewise.
393         * config/arc/arc.md: Add peephole2 rules for std/ldd.
395 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
397         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
398         test on stack_usage_file.
399         (lang_dependent_init): Do not open the .su file if generating LTO.
401 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
403         PR rtl-optimization/87918
404         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
405         simplify_gen_relational rather than simplify_gen_binary.
407 2018-11-13  Richard Biener  <rguenther@suse.de>
409         * tree-ssanames.h (set_range_info): Use value_range_base.
410         (get_range_info): Likewise.
411         * tree-ssanames.c (set_range_info): Likewise.
412         (get_range_info): Likewise.
413         * tree-vrp.c (value_range_base::union_helper): Split
414         out common parts of value_range[_base]::union_.
415         (value_range_base::union_): Update.
416         (value_range::union_): Likewise.
417         (determine_value_range_1): Use value_range_base.
418         (determine_value_range): Likewise.
419         * tree-vrp.h (value_range_base::union_helper): Move ...
420         (value_range::union_helper): ... from here.
422 2018-11-13  Alan Modra  <amodra@gmail.com>
424         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
425         offset for PRE_DEC.
426         (rs6000_secondary_reload_gpr): Don't call find_replacement.
428 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
430         PR middle-end/59634
431         * doc/invoke.texi (Optimize Options): Clarify that the
432         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
433         apply to data cache size.
435 2018-11-13  Alan Modra  <amodra@gmail.com>
437         * config/rs6000/predicates.md (logical_const_operand),
438         (logical_operand): Correct comment.
439         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
441 2018-11-13  Alan Modra  <amodra@gmail.com>
443         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
445 2018-11-13  Alan Modra  <amodra@gmail.com>
447         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
448         description.
449         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
450         covered by alternative.
451         (movcc_internal1): Ignore h for register preference.
452         (mov<mode>_hardfloat64): Likewise.
453         (mov<mode>_softfloat): Ignore c, l, h for register preference.
455 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
457         PR preprocessor/47823
458         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
459         documentation to...
460         (Pragmas): ...here.
461         * doc/extend.texi (Pragmas): Note additional pragmas documented
462         in the CPP manual.
464 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
466         PR middle-end/86677
467         PR middle-end/87528
468         * tree-scalar-evolution.c (expression_expensive_p): Make BUILTIN POPCOUNT
469         as expensive when backend does not define it.
471 2018-11-12  Fredrik Noring  <noring@nocrew.org>
473         * config.gcc: Update with-llsc defaults for MIPS r5900.
475 2018-11-12  Martin Liska  <mliska@suse.cz>
477         PR gcov-profile/87442
478         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
479         options.
480         * doc/invoke.texi: Document them.
481         * tree-profile.c (parse_profile_filter): New.
482         (parse_profile_file_filtering): Likewise.
483         (release_profile_file_filtering): Likewise.
484         (include_source_file_for_profile): Likewise.
485         (tree_profiling): Filter source files based on the
486         newly added options.
488 2018-11-12  Martin Liska  <mliska@suse.cz>
490         PR target/87903
491         * doc/extend.texi: Add missing values for __builtin_cpu_is and
492         __builtin_cpu_supports for x86 target.
494 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
496         PR middle-end/21110
497         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
498         have to be Pmode.
499         (Arithmetic): Likewise for "lo_sum".
501 2018-11-12  Renlin Li  <renlin.li@arm.com>
503         PR target/87815
504         * dse.c (get_stored_val): Add check for compile-time constantness
505         of gap.
507 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
509         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
510         (ARMv8_5a): New fgroup.
511         (armv8.5-a): New arch.
512         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
513         sb and predres.
514         * config/arm/arm-tables.opt: Regenerate.
515         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
516         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
517         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
518         Add matching rules for -march=armv8.5-a and extensions.
519         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
520         Add sb and predres to all armv8-a except armv8.5-a.
522 2018-11-12  Richard Biener  <rguenther@suse.de>
524         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
525         overload for single value.
526         (value_range[_base]::set_nonnull): New.
527         (value_range[_base]::set_null): Likewise.
528         (value_range): Document bitmap copying behavior, mark
529         copy constructor and assignment operator deleted.
530         (value_range::move): New.
531         (value_range::set_and_canonicalize): Default bitmap to zero.
532         (set_value_range_to_nonnull): Remove.
533         (set_value_range_to_null): Likewise.
534         (set_value_range): Likewise.
535         (set_value_range_to_value): Likewise.
536         (extract_range_from_unary_expr): Work on value_range_base.
537         (extract_range_from_binary_expr_1): Likewise.  Rename to...
538         (extract_range_from_binary_expr): ... this.
539         * tree-vrp.c (value_range::update): Clear equiv bitmap
540         if required.
541         (value_range::move): New, move equiv bitmap.
542         (value_range_base::set_undefined): Avoid assignment.
543         (value_range::set_undefined): Likewise.
544         (value_range_base::set_varying): Likewise.
545         (value_range::set_varying): Likewise.
546         (set_value_range): Remove.
547         (value_range_base::set): New overload for value.
548         (value_range::set): Likewise.
549         (set_value_range_to_nonnull): Remove.
550         (value_range_base::set_nonnull): New.
551         (value_range::set_nonnull): Likewise.
552         (set_value_range_to_null): Remove.
553         (value_range_base::set_null): New.
554         (value_range::set_null): Likewise.
555         (range_is_null): Work on value_range_base.
556         (range_is_nonnull): Likewise.
557         (ranges_from_anti_range): Likewise.
558         (extract_range_into_wide_ints): Likewise.
559         (extract_range_from_multiplicative_op): Likewise.
560         (extract_range_from_binary_expr): Likewise.  Update for API changes.
561         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
562         handling.
563         (value_range::intersect_helper): Avoid copy and assignment.
564         (value_range::union_helper): Likewise.
565         (determine_value_range_1): Adjust.
566         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
567         Avoid assignment by using move.
568         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
569         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
570         Likewise.
571         * tree-ssanames.c (get_range_info): Likewise.
572         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
573         * vr-values.c (vr_values::get_value_range): Adjust.
574         (vr_values::update_value_range): Likewise.
575         (symbolic_range_based_on_p): Work on value_range_base.
576         (vr_values::extract_range_from_binary_expr): Use value_range_base.
577         (vr_values::extract_range_from_unary_expr): Likewise.
578         (vr_values::extract_range_from_cond_expr): Avoid assignment.
579         (vr_values::extract_range_from_comparison): Adjust.
580         (vr_values::check_for_binary_op_overflow): Use value_range_base.
581         (vr_values::extract_range_basic): Adjust.
582         (vr_values::adjust_range_with_scev): Likewise.
583         (vr_values::vrp_visit_assignment_or_call): Likewise.
584         (vr_values::get_vr_for_comparison): Change API to avoid
585         assignment and copy construction.
586         (vr_values::compare_name_with_value): Adjust accordingly.
587         (vr_values::compare_names): Likewise.
588         (vr_values::extract_range_from_phi_node): Avoid assignment and
589         bogus in-place modify of equiv bitmap.
590         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
591         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
592         for extract_range_from_unary_expr API change.
593         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
595 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
597         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
599 2018-11-12  Richard Biener  <rguenther@suse.de>
601         * tree-vrp.h (value_range_base::symbolic_p,
602         value_range_base::constant_p, value_range_base::zero_p,
603         value_range_base::singleton_p): Move from value_range.
604         (value_range::dump): Add.
605         * gimple-ssa-evrp-analyze.c
606         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
607         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
608         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
609         Use set_varying.
610         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
611         (value_range::constant_p): Likewise.
612         (value_range::singleton_p): Likewise.
613         (value_range_base::dump): Add.
614         (set_value_range_to_undefined): Remove.
615         (set_value_range_to_varying): Likewise.
616         (range_int_cst_p): Take value_range_base argument.
617         (range_int_cst_singleton_p): Likewise.
618         (value_range_constant_singleton): Likewise.
619         (vrp_set_zero_nonzero_bits): Likewise.
620         (extract_range_from_multiplicative_op): Use set_varying.
621         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
622         (extract_range_from_unary_expr): Likewise.
623         (dump_value_range_base): Change to overload of dump_value_range.
624         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
625         (vrp_prop::visit_stmt): Likewise.
626         (value_range::intersect_helper): Likewise.
627         (value_range::union_helper): Likewise.
628         (determine_value_range_1): Likewise.
630 2018-11-12  Richard Biener  <rguenther@suse.de>
632         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
633         (set_value_range_to_null): Likewise.
634         * vr-values.c (vr_values::extract_range_from_comparison):
635         Clear equiv for constant singleton ranges.
637 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
639         * config/i386/sse.md: Combine VFIXUPIMM* patterns
640         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
641         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
642         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
643         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
644         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
645         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
647 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
649         PR c/69502
650         * doc/extend.texi (Common Type Attributes): For the align type
651         attribute, copy language about decreasing alignment from the
652         corresponding variable attribute.
654 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
656         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
657         -frounding-math.
659 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
661         PR c++/43105
662         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
663         -frtti and -fno-rtti code.
665 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
667         PR c/26366
668         * doc/extend.texi (Other Builtins): Document probability associated
669         with __builtin_expect.
671 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
673         PR target/87928
674         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
675         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
676         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
677         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
679 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
681         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
683 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
685         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
687 2018-11-11  Richard Biener  <rguenther@suse.de>
689         * tree-vrp.h (class value_range_base): New base class for
690         value_range containing all but the m_equiv member.
691         (dump_value_range_base): Add.
692         (range_includes_zero_p): Work on value_range_base.
693         * tree-vrp.c (value_range_base::set): Split out base handling
694         from...
695         (value_range::set): this.
696         (value_range::set_equiv): New.
697         (value_range_base::value_range_base): New constructors.
698         (value_range_base::check): Split out base handling from...
699         (value_range::check): this.
700         (value_range::equal_p): Refactor in terms of
701         ignore_equivs_equal_p which is now member of the base.
702         (value_range_base::set_undefined): New.
703         (value_range_base::set_varying): Likewise.
704         (value_range_base::dump):Split out base handling from...
705         (value_range::dump): this.
706         (value_range_base::set_and_canonicalize): Split out base handling
707         from...
708         (value_range::set_and_canonicalize): this.
709         (value_range_base::union_): New.
710         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
711         for m_vr.
712         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
713         instead of value_range everywhere.
714         (ipcp_vr_lattice::print): Use dump_value_range_base.
715         (ipcp_vr_lattice::meet_with): Adjust.
716         (ipcp_vr_lattice::meet_with_1): Likewise.
717         (ipa_vr_operation_and_type_effects): Likewise.
718         (propagate_vr_across_jump_function): Likewise.
719         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
720         (ipa_get_value_range): Likewise.
721         (ipa_set_jfunc_vr): Likewise.
722         (ipa_compute_jump_functions_for_edge): Likewise.
724 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
726         PR middle-end/65703
727         * doc/invoke.texi (Optimize Options): Add @opindex entries
728         for the positive forms of -fno-xxx and -mno-xxx options
729         that were lacking them.
731 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
733         * combine.c (make_more_copies): Only make an intermediate copy if the
734         dest of a move is a pseudo.
736 2018-11-09  Maya Rashish  <coypu@sdf.org>
738         PR target/87221
739         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
740         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
742 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
744         PR driver/41179
745         PR middle-end/65703
746         * doc/invoke.texi (Optimize Options): Clarify default behavior
747         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
749 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
751         PR tree-optimization/87940
752         * expr.c (string_constant): Don't strip NOPS in subexpressions.
753         Fold PLUS_EXPR correctly.
755 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
757         PR target/87762
758         * config/s390/s390.c (s390_safe_relative_long_p): New function.
759         (annotate_constant_pool_refs): Skip insns which support
760         relative addressing.
761         (annotate_constant_pool_refs_1): New helper function.
762         (find_constant_pool_ref): Skip insns which support relative
763         addression.
764         (find_constant_pool_ref_1): New helper function.
765         (replace_constant_pool_ref): Skip insns which support
766         relative addressing.
767         (replace_constant_pool_ref_1): New helper function.
768         (s390_mainpool_start): Adapt to the new signature.
769         (s390_mainpool_finish): Likewise.
770         (s390_chunkify_start): Likewise.
771         (s390_chunkify_finish): Likewise.
772         (pass_s390_early_mach::execute): Likewise.
773         (s390_prologue_plus_offset): Likewise.
774         (s390_emit_prologue): Likewise.
775         (s390_emit_epilogue): Likewise.
777 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
779         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
780         but unsupported lastprivate with conditional modifier.
782 2018-11-09  Jeff Law  <law@redhat.com>
784         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
785         unused argument better.  Add gcc_unreachable to silence warning.
787 2018-11-09  Martin Sebor  <msebor@redhat.com>
789         PR middle-end/81824
790         * attribs.c (has_attribute): New helper function.
791         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
792         * attribs.h (decls_mismatched_attributes): Declare.
793         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
794         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
795         * common.opt (-Wattribute-alias): Take an argument.
796         (-Wno-attribute-alias): New option.
797         * doc/extend.texi (Common Function Attributes): Document copy.
798         (Common Variable Attributes): Same.
799         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
800         (-Wattribute-alias): Document new option argument.
802 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
804         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
805         matching criteria.  Remove unused array initializer.
807 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
808             Jinsong Ji  <jji@us.ibm.com>
810         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
811         constraints by introducing a new temporary.
812         (_mm_cvtss_si64): Likewise.
814 2018-11-09  Martin Liska  <mliska@suse.cz>
816         * common.opt: Add -fipa-stack-alignment flag.
817         * doc/invoke.texi: Document it.
818         * final.c (rest_of_clean_state): Guard stack
819         shrinking with flag.
821 2018-11-09  Martin Liska  <mliska@suse.cz>
823         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
824         to ...
825         (ipa_discover_variable_flags): ... this.
826         * common.opt: Come up with new flag -fipa-reference-addressable.
827         * doc/invoke.texi: Document it.
828         * ipa-reference.c (propagate): Call the renamed fn.
829         * ipa-visibility.c (whole_program_function_and_variable_visibility):
830         Likewise.
831         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
832         ...
833         (ipa_discover_variable_flags): ... this.  Discover
834         non-addressable variables only with the newly added flag.
835         * opts.c: Enable the newly added flag with -O1 and higher
836         optimization level.
838 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
840         * json.cc (selftest::test_writing_literals): Fix comment.
842 2018-11-09  Martin Liska  <mliska@suse.cz>
844         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
845         string to a stack buffer.
846         (aarch64_parse_cpu): Likewise.
847         (aarch64_parse_tune): Likewise.
849 2018-11-09  Richard Biener  <rguenther@suse.de>
851         PR tree-optimization/87953
852         * tree-vect-loop.c (vectorizable_reduction): For analysis
853         always pass ops[0] to vectorizable_condition.
855 2018-11-09  Stafford Horne  <shorne@gmail.com>
856             Richard Henderson  <rth@twiddle.net>
857             Joel Sherrill  <joel@rtems.org>
859         * common/config/or1k/or1k-common.c: New file.
860         * config/or1k/*: New.
861         * config.gcc (or1k*-*-*): New.
862         * configure.ac (or1k*-*-*): New test for openrisc tls.
863         * configure: Regenerated.
864         * doc/install.texi: Document OpenRISC triplets.
865         * doc/invoke.texi: Document OpenRISC arguments.
866         * doc/md.texi: Document OpenRISC.
868 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
870         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
871         (arm7tdmi-s): Delete CPU.
872         (arm710t): Add aliases for arm720t and arm740t.
873         (arm720t, arm740t): Delete CPUs.
874         (arm920t): Add aliases for arm920, arm922t and arm940t.
875         (arm920, arm922t, arm940t): Delete CPUs.
876         (arm10tdmi): Add alias for arm1020t.
877         (arm1020t): Delete CPU.
878         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
879         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
880         (arm10e): Add aliases for arm1020e and arm1022e.
881         (arm1020e, arm1022e): Delete CPU.
882         * config/arm/arm.md (generic_sched): Remove entries that are now
883         handled by aliases.
884         (generic_vfp): Likewise.
885         * config/arm/arm1020e.md: Simplify tuning selection based on alias
886         changes.
887         * config/arm/arm-tune.md: Regenerated.
888         * config/arm/arm-tables.opt: Regenerated.
890 2018-11-09  Richard Biener  <rguenther@suse.de>
892         PR tree-optimization/87621
893         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
894         op with only phi inputs.
895         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
896         (ch_base::copy_headers): Run CSE on copied loop headers.
897         (pass_ch_vect::process_loop_p): Simplify.
899 2018-11-09  Alexandre Oliva <oliva@adacore.com>
901         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
902         for non-w64 x86_64 biarch.
904 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
906         PR rtl-optimization/86438
907         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
908         of in_b for the compare if in_b is SET_DEST.
910         PR target/87793
911         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
912         non-toplevel UNSPEC.
914 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
916         * tree-vrp.c (value_range::check): Do not access internals
917         directly.
918         (value_range::singleton_p): Same.
919         (value_range::type): Same.
920         (vrp_finalize): Use value_range API.
922 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
924         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
926 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
928         * vr-values.c (vr_values::get_value_range): Use value_range API
929         instead of piecing together ranges.
930         (vr_values::update_value_range): Same.
932 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
934         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
935         of performing ad-hoc calculations.
936         * tree-ssanames.c (set_range_info): New overloaded function
937         accepting value_range &.
938         (get_range_info): Same.
939         * tree-ssanames.h (set_range_info_raw): Remove.
940         (set_range_info): New prototype.
941         (get_range_info): Same.
942         * tree-vrp.h (value_range::null_p): Rename to zero_p.
943         * tree-vrp.c (value_range::null_p): Same.
945 2018-11-09  Jan Hubicka  <jh@suse.cz>
947         * tree.c (fld_type_variant_equal_p): Test user align flag.
948         (flt_type_variant): Copy user align flag.
949         (fld_incomplete_type_of): Clear it.
951 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
953         * config/arm/neon.md (div<mode>3): New pattern.
955 2018-11-08  Andi Kleen  <ak@linux.intel.com>
957         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
958         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
959         (ix86_handle_option): Handle OPT_mptwrite.
960         * config/i386/cpuid.h (bit_PTWRITE): Add.
961         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
962         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
963         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
964         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
965         (ix86_option_override_internal): Handle PTA_PTWRITE.
966         (ix86_valid_target_attribute_inner_p): Define ptwrite.
967         (def_builtin2): Force UINT64 to be 64bit only.
968         * config/i386/i386.h (TARGET_PTWRITE): Add.
969         (TARGET_PTWRITE_P): Add.
970         (PTA_PTWRITE): Add.
971         * config/i386/i386.md: Define ptwrite.
972         * config/i386/i386.opt: Add -mptwrite.
973         * config/i386/immintrin.h (_ptwrite64): Add.
974         (_ptwrite32): Add
975         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
976         * doc/invoke.texi: Document -mptwrite.
978 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
980         PR rtl-optimization/87600
981         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
982         * lra-constraints.c (process_alt_operands): Skip illegal hard
983         register usage.  Prefer reloading non hard register operands.
985 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
987         PR other/36572
988         * doc/invoke.texi (Optimize Options): Clarify default behavior
989         for -fno-sched-interblock and -fno-sched-spec.
991 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
993         * collect2.c (linker_select):  Add USE_LLD_LD.
994         (ld_suffixes): Add ld.lld.
995         (main): Handle -fuse-ld=lld.
996         * common.opt (-fuse-ld=lld): New option.
997         * doc/invoke.texi (-fuse-ld=lld): Document.
998         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1000 2018-11-08  Paul Koning  <ni1d@arrl.net>
1002         * config/pdp11/constraints.md: Add "Z" series constraints for use
1003         with pre-dec and post-inc addressing.
1004         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1005         (pdp11_expand_operands): Add int argument (word count).
1006         (pdp11_sp_frame_offset): Delete.
1007         (pdp11_cmp_length): New function.
1008         (pushpop_regeq): New function.
1009         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1010         Add hook.
1011         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1012         frame layout.
1013         (pdp11_initial_elimination_offset): Ditto.
1014         (pdp11_expand_operands): Add word count argument.  Bugfixes.
1015         (output_move_multiple): Change how pointer adjustment is done.
1016         (pdp11_gen_int_label): Correct format.
1017         (output_ascii): Ditto.
1018         (pdp11_asm_output_var): Add code for DEC assembler case.
1019         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1020         value.
1021         (legitimate_const_double_p): Ditto.
1022         (pdp11_register_move_cost): Adjust for new register classes.
1023         (pdp11_regno_reg_class): Ditto.
1024         (expand_block_move): Delete.
1025         (pushpop_regeq): New function.
1026         (pdp11_legitimate_address_p): Bugfix in check for constant
1027         offset.
1028         (pdp11_sp_frame_offset): Delete.
1029         (pdp11_reg_save_size): New helper function for new frame layout.
1030         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1031         (pdp11_expand_shift): Bugfix in check for constant shift count.
1032         (pdp11_shift_length): Ditto.
1033         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1034         (pdp11_cmp_length): New function.
1035         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1036         some compile options.
1037         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1038         (CALL_USED_REGISTERS): Ditto.
1039         (ELIMINABLE_REGS): Ditto.
1040         (REGISTER_NAMES): Ditto.
1041         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1042         constraints.
1043         (REG_CLASS_NAMES): Ditto.
1044         (REG_CLASS_CONTENTS): Ditto.  Also remove
1045         HARD_FRAME_POINTER_REGNUM.
1046         (CPU_REG_CLASS): New macro.
1047         (CLASS_MAX_NREGS): Adjust for new register classes.
1048         (FUNCTION_PROFILER): Make no-op.
1049         (may_call_alloca): Remove unused declaration.
1050         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1051         (ASM_OUTPUT_SKIP): Fix format.
1052         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1053         (HARD_FRAME_POINTER_REGNUM): Remove.
1054         (return): Delete.
1055         (*rts): Rename.  Remove epilogue related checks.
1056         (cmpsi, cmpdi): New insn.
1057         (cbranch<mode>4): Change to apply to SI and DI modes as well.
1058         (mov<mode>): Change constraints to enforce that push/pop
1059         destination cannot use the same register as source.
1060         (*mov<mode><cc_cc>): Ditto.
1061         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1062         at assembly output rather than as RTL expander.
1063         (zero_extendqihi2): Bugfix in check for same registers.
1064         (adddi3_nocc): Bugfix in check for constant operand.
1065         (addsi3_nocc): Ditto.
1066         (subdi3_nocc): Ditto.
1067         (subsi3_nocc): Ditto.
1068         (negdi2_nocc): Copy input to pdp11_expand_operands.
1069         (negsi2_nocc): Ditto.
1070         (bswap2_nocc): Ditto.
1071         * config/pdp11/pdp11.opt (mlra): Fix documentation.
1072         * config/pdp11/t-pdp11: Use -Os.
1074 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
1076         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1077         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1078         (gen_comm_data): Print out CPU alias tables.
1079         (check_cpu): Match aliases when checking the CPU name.
1080         * config/arm/arm-protos.h (cpu_alias): New structure.
1081         (cpu_option): Add entry for aliases.
1082         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1083         strongarm1100 and strongarm1110.
1084         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1085         (config/arm/arm-generic.md): Remove redundant references to
1086         strongarm110, strongarm1100 and strongarm1110.
1087         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1088         Scan aliases for additional hints.
1089         (arm_parse_cpu_option_name): Also match a cpu name against the list
1090         of aliases.
1091         * config/arm/arm-tables.opt: Regenerated.
1092         * config/arm/arm-tune.md: Regenerated.
1094 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1096         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1097         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1098         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1099         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1100         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1101         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1102         * gengtype.c (open_base_files): Add omp-general.h.
1103         * gimple.c (gimple_build_omp_critical):
1104         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1105         gimple_omp_taskgroup_set_clauses.
1106         (gimple_build_omp_atomic_load): Add mo argument, call
1107         gimple_omp_atomic_set_memory_order.
1108         (gimple_build_omp_atomic_store): Likewise.
1109         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1110         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1111         instead of GSS_OMP.
1112         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1113         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1114         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1115         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1116         different value for GF_OMP_ATOMIC_NEED_VALUE.
1117         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1118         comments.
1119         (struct gimple_statement_omp_single_layout): And remove here.
1120         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1121         than gimple_statement_omp_single_layout.
1122         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1123         GIMPLE_OMP_TEAMS.
1124         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1125         (gimple_omp_subcode): Formatting fix.
1126         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1127         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1128         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1129         gimple_omp_teams_host, gimple_omp_teams_set_host,
1130         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1131         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1132         gimple_omp_taskgroup_set_clauses): New inline functions.
1133         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1134         (gimple_build_omp_atomic_store): Likewise.
1135         (gimple_omp_atomic_seq_cst_p): Remove.
1136         (gimple_omp_atomic_memory_order): New function.
1137         (gimple_omp_atomic_set_seq_cst): Remove.
1138         (gimple_omp_atomic_set_memory_order): New function.
1139         (gimple_build_omp_taskgroup): Add clauses argument.
1140         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1141         (dump_gimple_omp_task): Print taskwait with depend clauses.
1142         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1143         dump_omp_atomic_memory_order.
1144         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1145         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1146         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1147         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1148         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1149         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1150         ORT_UNTIED_TASKLOOP enumerators.
1151         (enum gimplify_defaultmap_kind): New.
1152         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1153         target_map_pointers_as_0len_arrays members, add defaultmap.
1154         (new_omp_context): Initialize defaultmap member.
1155         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1156         (maybe_fold_stmt): Don't fold even in host teams regions.
1157         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1158         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1159         ctx->omp_firstprivatize_variable.
1160         (omp_add_variable): Don't add private/firstprivate for VLAs in
1161         ORT_TASKGROUP.
1162         (omp_default_clause): Print "taskloop" rather than "task" if
1163         ORT_*TASKLOOP.
1164         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1165         Handle new defaultmap clause kinds.
1166         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1167         iterator to be lastprivate or private.  Fix up diagnostics if linear
1168         is used on collapse>1 simd iterator.
1169         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1170         (gimplify_omp_depend): New function.
1171         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1172         combined parallel master taskloop{, simd} if taskloop has
1173         firstprivate, lastprivate or reduction clause.  Handle
1174         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1175         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1176         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1177         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1178         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1179         lastprivate.
1180         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1181         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
1182         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1183         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1184         (gimplify_omp_task): Handle taskwait with depend clauses.
1185         (gimplify_omp_for): Add shared clause on parallel for combined
1186         parallel master taskloop{, simd} if taskloop has firstprivate,
1187         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1188         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1189         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1190         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1191         __for_end and __for_range temporaries on OMP_PARALLEL for
1192         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1193         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1194         sandwiched in between two taskloops.
1195         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1196         instead of ctx->omp_firstprivatize_variable.
1197         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1198         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1199         host teams, use gimplify_and_return_first etc. for body like
1200         for target or target data constructs, and at the end call
1201         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1202         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1203         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1204         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1205         gimple_omp_atomic_set_seq_cst calls.
1206         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1207         case, handle taskgroup clauses.
1208         * lto-streamer-out.c (hash_tree): Handle
1209         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1210         * Makefile.in (GTFILES): Add omp-general.h.
1211         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1212         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1213         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1214         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1215         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1216         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1217         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1218         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1219         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1220         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1221         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1222         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1223         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1224         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1225         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1226         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1227         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1228         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1229         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1230         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1231         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1232         non-worksharing loops.
1233         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1234         (determine_parallel_type): Don't combine parallel with worksharing
1235         which has _reductemp_ clause.
1236         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1237         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1238         if there is nonmonotonic modifier or if there is no modifier and no
1239         ordered clause.  For dynamic and guided schedule without monotonic
1240         and nonmonotonic modifier, default to nonmonotonic.
1241         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1242         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1243         task reductions.
1244         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1245         there are any reduction clauses.
1246         (expand_taskwait_call): New function.
1247         (expand_teams_call): New function.
1248         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1249         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1250         depend clauses.
1251         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1252         of worksharing loops with task reductions.
1253         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1254         expansion of worksharing loops with task reductions.
1255         (expand_omp_sections): Handle expansion of sections with task
1256         reductions.
1257         (expand_omp_synch): For host teams call expand_omp_taskreg.
1258         (omp_memory_order_to_memmodel): New function.
1259         (expand_omp_atomic_load, expand_omp_atomic_store,
1260         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1261         instead of gimple_omp_atomic_seq_cst_p.
1262         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1263         depend clauses as a standalone directive.
1264         * omp-general.c (enum omp_requires): New variable.
1265         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1266         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1267         GT_EXPR loops depending on incr sign.  Formatting fixes.
1268         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1269         (enum omp_requires): New enum.
1270         (omp_requires_mask): Declare.
1271         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1272         Fix comment typos.
1273         * omp-low.c (struct omp_context): Add task_reductions and
1274         task_reduction_map fields.
1275         (is_host_teams_ctx): New function.
1276         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1277         (use_pointer_for_field): Use is_global_var instead of
1278         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1279         in outer contexts.
1280         (build_outer_var_ref): Ignore taskgroup outer contexts.
1281         (delete_omp_context): Release task_reductions and task_reduction_map.
1282         (scan_sharing_clauses): Don't add any fields for reduction clause on
1283         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1284         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1285         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1286         Handle OMP_CLAUSE_NONTEMPORAL.
1287         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1288         needed.
1289         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1290         clauses with task modifier.
1291         (scan_omp_task): Handle taskwait with depend clauses.
1292         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1293         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1294         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1295         (scan_omp_for): Fix comment formatting.
1296         (scan_omp_teams): Handle host teams constructs.
1297         (check_omp_nesting_restrictions): Allow teams with no outer
1298         OpenMP context.  Adjust diagnostics for teams strictly nested into
1299         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1300         inside of simd regions.
1301         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1302         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1303         taskreg_nesting_level while scanning host teams construct.
1304         (task_reduction_read): New function.
1305         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1306         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1307         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1308         second argument create_tmp_var if it is NULL.  Don't ignore shared
1309         clauses in is_host_teams_ctx contexts.  Handle
1310         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1311         clauses.
1312         (lower_reduction_clauses): Ignore reduction clauses with task
1313         modifier.  Remove second argument create_tmp_var if it is NULL.
1314         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1315         (lower_send_clauses): Ignore reduction clauses with task modifier.
1316         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1317         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1318         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1319         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1320         contexts.
1321         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1322         lower_omp_task_reductions): New functions.
1323         (lower_omp_sections): Handle reduction clauses with taskgroup
1324         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1325         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1326         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1327         modifiers.
1328         (lower_omp_taskgroup): Handle taskgroup reductions.
1329         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1330         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1331         (lower_depend_clauses): If there are any
1332         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1333         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1334         seen, assume lowering is done already and return early.  Set kind
1335         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1336         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1337         parallel construct.  Handle reduction clause on taskloop construct.
1338         Handle taskwait with depend clauses.
1339         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1340         for host teams constructs.
1341         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1342         nontemporal and _reductemp_ clause entries.
1343         (omp_clause_code_name): Likewise.
1344         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1345         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1346         * tree-core.h (enum omp_clause_code): Add
1347         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1348         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1349         (struct tree_base): Add omp_atomic_memory_order field into union.
1350         Remove OMP_ATOMIC_SEQ_CST comment.
1351         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1352         and OMP_CLAUSE_DEPEND_DEPOBJ.
1353         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1354         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1355         OpenMP constructs with body and clauses operands.
1356         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1357         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1358         (OMP_TASKGROUP_CLAUSES): Define.
1359         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1360         OMP_CLAUSE__LOOPTEMP_.
1361         (OMP_ATOMIC_SEQ_CST): Remove.
1362         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1363         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1364         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1365         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1366         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1367         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1368         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1369         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1370         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1371         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1372         Define.
1373         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1374         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1375         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1376         (convert_local_omp_clauses): Likewise.  Remove useless test.
1377         * tree-parloops.c (create_call_for_reduction_1): Pass
1378         OMP_MEMORY_ORDER_RELAXED as new argument to
1379         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1380         * tree-pretty-print.c (dump_omp_iterators): New function.
1381         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1382         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1383         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1384         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1385         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1386         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1387         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1388         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1389         (dump_omp_atomic_memory_order): New function.
1390         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1391         taskwait with depend clauses.
1392         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1393         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1394         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1395         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1396         write_ts_omp_clause_tree_pointers): Likewise.
1398 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1400         PR ipa/86395
1401         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1402         "-missed", "-note", and "-all" sub-options.
1403         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1404         API.
1405         (can_early_inline_edge_p): Likewise.
1406         (want_early_inline_function_p): Likewise.
1407         (want_inline_self_recursive_call_p): Likewise.
1408         (recursive_inlining): Likewise.
1409         (inline_small_functions): Likewise.
1410         (flatten_function): Likewise.
1411         (ipa_inline): Likewise.
1412         (inline_always_inline_functions): Likewise.
1413         (early_inline_small_functions): Likewise.
1414         (early_inliner): Likewise.
1415         * tree-inline.c (expand_call_inline): Likewise.
1417 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1419         * pretty-print.c (pp_format): Handle %f.
1420         (selftest::test_pp_format): Add test of %f.
1421         * pretty-print.h (pp_double): New macro.
1423 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1425         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1426         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1427         (dump_pretty_printer::decode_format): Implement "%C" for
1428         cgraph_node *.
1429         (selftest::test_capture_of_dump_calls): Rename "where" to
1430         "stmt_loc".  Convert test_decl to a function decl and set its
1431         location.  Add a symbol_table_test RAII instance and a
1432         cgraph_node, using it to test "%C" and dump_symtab_node.
1434 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1436         PR middle-end/87916
1437         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1439 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1441         * cgraph.c: Include "selftest.h".
1442         (saved_symtab): New variable.
1443         (selftest::symbol_table_test::symbol_table_test): New ctor.
1444         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1445         (selftest::test_symbol_table_test): New test.
1446         (selftest::cgraph_c_tests): New.
1447         * cgraph.h (saved_symtab): New decl.
1448         (selftest::symbol_table_test): New class.
1449         * selftest-run-tests.c (selftest::run_tests): Call
1450         selftest::cgraph_c_tests.
1451         * selftest.h (selftest::cgraph_c_tests): New decl.
1453 2018-11-08  Richard Biener  <rguenther@suse.de>
1455         * tree-data-ref.h (lambda_int): New typedef.
1456         (lambda_vector_gcd): Adjust.
1457         (lambda_vector_new): Likewise.
1458         (lambda_matrix_new): Likewise.
1459         * tree-data-ref.c  (print_lambda_vector): Adjust.
1461 2018-11-08  Richard Biener  <rguenther@suse.de>
1463         PR tree-optimization/87929
1464         * tree-complex.c (expand_complex_comparison): Clean EH.
1466 2018-11-08  Martin Liska  <mliska@suse.cz>
1468         * doc/extend.texi: Reword.
1469         * predict.c (expr_expected_value_1): Likewise.
1471 2018-11-08  Richard Biener  <rguenther@suse.de>
1473         PR tree-optimization/87913
1474         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1475         of extreme values to ordered comparisons.
1477 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1479         PR middle-end/42726
1480         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1482 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1484         * doc/invoke.texi: Remove leading dash from @opindex entries
1485         throughout the file.
1487 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1489         PR driver/80828
1490         * doc/invoke.texi (Option Summary): Add -e and --entry.
1491         (Link Options): Likewise.
1493 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1495         PR 87926
1496         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1497         --disable-checking bootstrap.
1499 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1501         * configure: Regenerated.
1503 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1505         PR c/87691
1506         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1507         to the mode of the widest field iff the widest field has mode class
1508         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1509         reference.
1511 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1513         * common.opt: Add -Wattribute-warning.
1514         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1515         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1516         call to allow user configure behavior of "warning" attribute.
1518 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1520         * target.def: Put @: after every vs., e.g., and i.e. where it is
1521         followed by whitespace.
1522         * doc/extend.texi: Ditto.
1523         * doc/fragments.texi: Ditto.
1524         * doc/gimple.texi: Ditto.
1525         * doc/implement-c.texi: Ditto.
1526         * doc/install.texi: Ditto.
1527         * doc/invoke.texi: Ditto.
1528         * doc/md.texi: Ditto.
1529         * doc/plugins.texi: Ditto.
1530         * doc/rtl.texi: Ditto.
1531         * doc/sourcebuild.texi: Ditto.
1532         * doc/tm.texi.in: Ditto.
1533         * doc/ux.texi: Ditto.
1534         * doc/tm.texi: Regenerate.
1536 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1538         * config/arm/arm-cpus.in (ares): New entry.
1539         * config/arm/arm-tables.opt: Regenerate.
1540         * config/arm/arm-tune.md: Likewise.
1541         * doc/invoke.texi (ARM Options): Document ares.
1543 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1545         * config/aarch64/aarch64-cores.def (ares): Define.
1546         * config/aarch64/aarch64-tune.md: Regenerate.
1547         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1549 2018-11-07  Jan Hubicka  <jh@suse.cz>
1551         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1552         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1553         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1554         TYPE_DECL.
1556 2018-11-07  Richard Biener  <rguenther@suse.de>
1558         PR tree-optimization/87914
1559         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1560         of nested cycles.
1561         (vectorizable_reduction): Handle shifts and rotates by dispatching
1562         to vectorizable_shift.
1563         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1564         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1565         def cases.
1566         (vectorizable_shift): Export and handle being called as
1567         vect_nested_cycle.
1568         (vect_analyze_stmt): Call vectorizable_shift after
1569         vectorizable_reduction.
1570         * tree-vectorizer.h (vectorizable_shift): Declare.
1572 2018-11-07  Jan Hubicka  <jh@suse.cz>
1574         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1575         than const decls in TREE_VALUE of enum.
1576         (dump_type_inheritance_graph): Improve duplicate dumping.
1577         (free_enum_values): New.
1578         (build_type_inheritance_graph): Use it.
1579         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
1580         which are not main variants or not ODR types.
1581         (verify_type_variant): Expect variants to have no TYPE_VALUES.
1583 2018-11-07  Richard Biener  <rguenther@suse.de>
1585         * ipa-inline.c (want_inline_small_function_p): Compute
1586         big_speedup_p lazily and last.
1588 2018-11-07  Jan Hubicka  <jh@suse.cz>
1590         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
1591         building incomplete variant of complete type.
1592         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
1593         variant of complete type.
1595 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1597         * config/mips/mips.c: Fix typo in documentation of
1598         mips_loongson_ext2_prefetch_cookie.
1599         (mips_option_override): fix brain twister logical.
1600         * config/mips/mips.h: Fix typo in documentation of
1601         ISA_HAS_CTZ_CTO and define pattern.
1602         * config/mips/mips.md (prefetch): Hoist EXT2 above
1603         the 2EF/EXT block.
1604         (prefetch_indexed): Hoist EXT2 above the EXT block.
1606 2018-11-07  Jan Hubicka  <jh@suse.cz>
1608         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
1609         return and parameter types of function and method types.
1610         (free_lang_data_in_cgraph): Update.
1612 2018-11-07  Martin Liska  <mliska@suse.cz>
1614         PR rtl-optimization/87868
1615         * postreload-gcse.c (eliminate_partially_redundant_load): Set
1616         threshold to max_count if we would overflow.
1617         * profile-count.h: Make max_count a public constant.
1619 2018-11-07  Martin Liska  <mliska@suse.cz>
1621         * mem-stats.h: Fix GNU coding style.
1623 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1625         * config/mips/gs264e.md: New.
1626         * config/mips/mips-cpus.def: Define gs264e.
1627         * config/mips/mips-tables.opt: Regenerate.
1628         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1629         gs264e.
1630         (mips_issue_rate): Add support for gs264e.
1631         (mips_multipass_dfa_lookahead): Likewise.
1632         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
1633         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
1634         (MIPS_ASE_MSA_SPEC): New.
1635         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
1636         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
1637         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
1638         * config/mips/mips.md: Include gs264e.md.
1639         (processor): Add gs264e.
1640         * config/mips/mips.opt (MSA): Use Mask instead of Var.
1641         * doc/invoke.texi: Add gs264e to supported architectures.
1643 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1645         * config/mips/gs464e.md: New.
1646         * config/mips/mips-cpus.def: Define gs464e.
1647         * config/mips/mips-tables.opt: Regenerate.
1648         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1649         gs464e.
1650         (mips_issue_rate): Add support for gs464e.
1651         (mips_multipass_dfa_lookahead): Likewise.
1652         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
1653         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
1654         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
1655         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
1656         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
1657         * config/mips/mips.md: Include gs464e.md.
1658         (processor): Add gs464e.
1659         * doc/invoke.texi: Add gs464e to supported architectures.
1661 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1663         * config/mips/loongson3a.md: Rename to ...
1664         * config/mips/gs464.md: ... here.
1665         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
1666         as an alias of gs464 processor.
1667         * config/mips/mips-tables.opt: Regenerate.
1668         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
1669         instead of PROCESSOR_LOONGSON_3A.
1670         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
1671         TUNE_LOONGSON_3A.
1672         (mips_option_override): Enable MMI and EXT for gs464.
1673         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
1674         Rename TUNE_LOONGSON_3A to TUNE_GS464.
1675         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
1676         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
1677         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
1678         TARGET_LOONGSON_3A.
1679         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
1680         (processor): Add gs464;
1681         * doc/invoke.texi: Add gs464 to supported architectures.
1683 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1685         * config/mips/mips-protos.h
1686         (mips_loongson_ext2_prefetch_cookie): New prototype.
1687         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
1688         (mips_option_override): Enable TARGET_LOONGSON_EXT when
1689         TARGET_LOONGSON_EXT2 is true.
1690         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
1691         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
1692         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
1693         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
1694         TARGET_LOONGSON_EXT2.
1695         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
1696         (define_insn "ctz<mode>2"): New insn pattern.
1697         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
1698         (define_insn "prefetch_indexed_<mode>"): Include
1699         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
1700         * config/mips/mips.opt (-mloongson-ext2): Add option.
1701         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
1703 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1705         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
1706         __mips_loongson_ext.
1707         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
1708         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
1709         -mloongson-ext.
1710         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
1711         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
1712         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
1713         instead of TARGET_LOONGSON_3A.
1714         * config/mips/mips.opt (-mloongson-ext): Add option.
1715         * gcc/doc/invoke.texi (-mloongson-ext): Document.
1717 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1719         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
1720         * config/mips/loongson.md: Move to ...
1721         * config/mips/loongson-mmi.md: here; Adjustment.
1722         * config/mips/loongson.h: Move to ...
1723         State as deprecated. Include loongson-mmiintrin.h for back
1724         compatibility and warning.
1725         * config/mips/loongson-mmiintrin.h: ... here.
1726         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
1727         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
1728         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1729         (mips_option_override): Make sure MMI use hard float;
1730         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
1731         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
1732         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
1733         TARGET_LOONGSON_VECTORS.
1734         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
1735         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
1736         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
1737         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
1738         -mloongson-mmi.
1739         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
1740         TARGET_LOONGSON_VECTORS.
1741         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
1742         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1743         (Loongson MMI patterns): Include loongson-mmi.md instead of
1744         loongson.md.
1745         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
1746         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
1748 2018-11-07  Richard Biener  <rguenther@suse.de>
1750         PR lto/87906
1751         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
1752         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
1754 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
1756         PR rtl-optimization/87874
1757         * lra.c (lra_substitute_pseudo): Do not create a subreg for
1758         const wide ints.
1760 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
1762         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
1763         if not in indexed or indirect form.
1764         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
1765         (bswap<mode>2_store): Ditto.
1767 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
1769         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
1770         the UNSPEC.
1772 2018-11-06  Richard Biener  <rguenther@suse.de>
1774         PR tree-optimization/86850
1775         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
1776         instead of src.m_vec.
1778 2018-11-06  Jan Hubicka  <jh@suse.cz>
1780         * tree.c (fld_simplified_type_name): Break out form ...
1781         (free_lang_data_in_type): ... here.
1782         (fld_type_variant_equal_p): Use it.
1784 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1786         * config/default-d.c: Include memmodel.h.
1788         * config/sol2-d.c: New file.
1789         * config/t-sol2 (sol2-d.o): New rule.
1790         * config.gcc <*-*-solaris2*>: Set d_target_objs,
1791         target_has_targetdm.
1793 2018-11-06  Jan Hubicka  <jh@suse.cz>
1795         * tree.c (fld_type_variant): Also copy alignment; be sure that
1796         new variant is equal.
1798 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1800         PR target/87762
1801         * config/s390/s390.md: Add relative_long attribute.
1803 2018-11-06  Jan Hubicka  <jh@suse.cz>
1805         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
1806         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
1807         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
1808         stream TYPE_NEEDS_CONSTRUCTING.
1809         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
1810         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
1812 2018-11-06  Richard Biener  <rguenther@suse.de>
1814         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
1815         dump-scope ...
1816         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
1818 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1820         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
1821         REG_CLASS_CONTENTS[GEN_REGS].
1822         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
1824 2018-11-06  Jan Hubicka  <jh@suse.cz>
1826          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
1828 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1830         PR sanitizer/80953
1831         * config/sol2.h (ASAN_CC1_SPEC): Define.
1832         (LD_WHOLE_ARCHIVE_OPTION): Define.
1833         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
1834         (ASAN_REJECT_SPEC): Provide default.
1835         (LIBASAN_EARLY_SPEC): Define.
1836         (LIBTSAN_EARLY_SPEC): Define.
1837         (LIBLSAN_EARLY_SPEC): Define.
1838         * config/i386/sol2.h (CC1_SPEC): Redefine.
1839         (ASAN_REJECT_SPEC): Define.
1841         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
1842         (TARGET_ASAN_SHADOW_OFFSET): Define.
1843         (sparc_asan_shadow_offset): New function.
1844         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
1845         (ASAN_REJECT_SPEC): Define.
1847 2018-11-06  Jan Hubicka  <jh@suse.cz>
1849         * tree.c (fld_type_variant): Copy canonical type.
1850         (fld_incomplete_type_of): Check that canonical types looks sane;
1851         copy canonical type.
1852         (verify_type): Accept when incomplete type has complete canonical type.
1854 2018-11-06  Jan Hubicka  <jh@suse.cz>
1856         * tree.c (free_lang_data): Reset overwite_assembler_name,
1857         print_xnode, print_decl, print_type and print_identifier of
1858         langhooks.
1860 2018-11-06  Richard Biener  <rguenther@suse.de>
1862         PR tree-optimization/87889
1863         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1864         Do nothing if old and new arg are the same
1866 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
1868         PR target/87723
1869         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
1870         attributes for operands 3 and 4.
1872 2018-11-06  Richard Biener  <rguenther@suse.de>
1874         PR middle-end/18041
1875         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
1876         matching bitfield insertion.
1878 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
1880         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
1881         (reg_next_debug_use): New.
1882         (attempt_change): Propagate adjusted expression into affected
1883         debug insns.
1884         (merge_in_block): Track uses in debug insns.
1885         (pass_inc_dec::execute): Allocate and release
1886         reg_next_debug_use.
1888 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
1890         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
1891         (_mm512_fixupimm_round_pd): Update parameters and builtin.
1892         (_mm512_maskz_fixupimm_round_pd): Ditto.
1893         (_mm512_fixupimm_round_ps): Ditto.
1894         (_mm512_maskz_fixupimm_round_ps): Ditto.
1895         (_mm_fixupimm_round_sd): Ditto.
1896         (_mm_maskz_fixupimm_round_sd): Ditto.
1897         (_mm_fixupimm_round_ss): Ditto.
1898         (_mm_maskz_fixupimm_round_ss): Ditto.
1899         (_mm512_fixupimm_pd): Ditto.
1900         (_mm512_maskz_fixupimm_pd): Ditto.
1901         (_mm512_fixupimm_ps): Ditto.
1902         (_mm512_maskz_fixupimm_ps): Ditto.
1903         (_mm_fixupimm_sd): Ditto.
1904         (_mm_maskz_fixupimm_sd): Ditto.
1905         (_mm_fixupimm_ss): Ditto.
1906         (_mm_maskz_fixupimm_ss): Ditto.
1907         (_mm512_mask_fixupimm_round_pd): Update builtin.
1908         (_mm512_mask_fixupimm_round_ps): Ditto.
1909         (_mm_mask_fixupimm_round_sd): Ditto.
1910         (_mm_mask_fixupimm_round_ss): Ditto.
1911         (_mm512_mask_fixupimm_pd): Ditto.
1912         (_mm512_mask_fixupimm_ps): Ditto.
1913         (_mm_mask_fixupimm_sd): Ditto.
1914         (_mm_mask_fixupimm_ss): Ditto.
1915         * config/i386/avx512vlintrin.h:
1916         (_mm256_fixupimm_pd): Update parameters and builtin.
1917         (_mm256_maskz_fixupimm_pd): Ditto.
1918         (_mm256_fixupimm_ps): Ditto.
1919         (_mm256_maskz_fixupimm_ps): Ditto.
1920         (_mm_fixupimm_pd): Ditto.
1921         (_mm_maskz_fixupimm_pd): Ditto.
1922         (_mm_fixupimm_ps): Ditto.
1923         (_mm_maskz_fixupimm_ps): Ditto.
1924         (_mm256_mask_fixupimm_pd): Update builtin.
1925         (_mm256_mask_fixupimm_ps): Ditto.
1926         (_mm_mask_fixupimm_pd): Ditto.
1927         (_mm_mask_fixupimm_ps): Ditto.
1928         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
1929         * config/i386/i386-builtin.def: Update builtin definitions.
1930         * config/i386/i386.c: Handle new builtin types and remove useless ones.
1931         * config/i386/sse.md: Update VFIXUPIMM* patterns.
1932         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1933         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1934         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
1935         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1936         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1937         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
1938         * config/i386/subst.md:
1939         (round_saeonly_sd_mask_operand4): Add new subst_attr.
1940         (round_saeonly_sd_mask_op4): Ditto.
1941         (round_saeonly_expand_operand5): Ditto.
1942         (round_saeonly_expand): Update.
1944 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
1946         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1948 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
1950         PR rtl-optimization/87871
1951         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
1953 2018-11-05  Paul Koning  <ni1d@arrl.net>
1955         * doc/sourcebuild.texi (target attributes): Document new "inf"
1956         effective target keyword.
1958 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1960         * config/s390/s390.c (s390_register_move_cost): Increase costs for
1961         moves involving the CC reg.
1963 2018-11-05  Richard Biener  <rguenther@suse.de>
1965         PR tree-optimization/87873
1966         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
1967         argument.
1968         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
1969         * tree-vect-loop.c (vect_transform_loop): When splitting the
1970         loop exit also create forwarder PHIs for constants.
1971         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1972         Handle constant to_arg, add extra checking we match up the correct
1973         PHIs.
1975 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1977         * config/s390/s390.md: QImode and HImode for load on condition.
1979 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1981         * config/s390/predicates.md: Fix typo.
1982         * config/s390/s390.md: Allow immediates for load on condition.
1984 2018-11-05  Martin Liska  <mliska@suse.cz>
1986         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
1987         * bitmap.h (struct bitmap_usage): Likewise.
1988         * ggc-common.c (SCALE): Remove.
1989         (LABEL): Likewise.
1990         (struct ggc_usage): Use SIZE_AMOUNT. And update
1991         compare method.
1992         * ggc-page.c (SCALE): Remove.
1993         (STAT_LABEL): Remove.
1994         (ggc_print_statistics): Use SIZE_AMOUNT.
1995         * gimple.h (SCALE): Remove.
1996         (LABEL): Likewise.
1997         * input.c (ONE_K): Remove.
1998         (ONE_M): Likewise.
1999         (SCALE): Likewise.
2000         (STAT_LABEL): Likewise.
2001         (FORMAT_AMOUNT): Likewise.
2002         (dump_line_table_statistics): Use SIZE_AMOUNT.
2003         * mem-stats.h (struct mem_usage): Likewise.
2004         * rtl.c (dump_rtx_statistics): Likewise.
2005         (rtx_alloc_counts): Change type to size_t.
2006         (rtx_alloc_sizes): Likewise.
2007         (rtx_count_cmp): New.
2008         (dump_rtx_statistics): Sort first based on counts.
2009         * tree.c (tree_nodes_cmp): New.
2010         (tree_codes_cmp): New.
2011         (dump_tree_statistics): Sort first based on counts.
2012         * system.h (ONE_K): New.
2013         (ONE_M): Likewise.
2014         (SIZE_SCALE): Likewise.
2015         (SIZE_LABEL): Likewise.
2016         (SIZE_AMOUNT): Likewise.
2017         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2018         * tree-dfa.c (dump_dfa_stats): Likewise.
2019         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2020         * tree-ssanames.c (ssanames_print_statistics): Likewise.
2021         * tree.c (dump_tree_statistics): Likewise.
2022         * vec.c (struct vec_usage): Likewise.
2023         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2024         trigger a -Werror=format-overflow with
2025         --enable-gather-detailed-stats.
2027 2018-11-05  Martin Liska  <mliska@suse.cz>
2029         * mem-stats.h (mem_alloc_description::release_instance_overhead):
2030         Return T *.
2031         * vec.c (struct vec_usage): Register m_element_size.
2032         (vec_prefix::register_overhead): New arguments: elements and
2033         element_size.
2034         (vec_prefix::release_overhead): Subtract elements.
2035         * vec.h (struct vec_prefix): Change signature.
2036         (va_heap::reserve): Pass proper arguments.
2037         (va_heap::release): Likewise.
2039 2018-11-05  Martin Liska  <mliska@suse.cz>
2041         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2042         style.
2043         * vec.c: Likewise.
2045 2018-11-05  Richard Biener  <rguenther@suse.de>
2047         * tree-scalar-evolution.h (final_value_replacement_loop): Update
2048         prototype.
2049         * tree-scalar-evolution.c (final_value_replacement_loop): Return
2050         whether anything was done.
2051         (scev_const_prop): Remove constant propagation part, fold
2052         remains into ...
2053         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2054         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2055         conditionally.
2057 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
2059         PR tree-optimization/87859
2060         * gimple-ssa-store-merging.c (struct merged_store_group): Add
2061         only_constants and first_nonmergeable_order members.
2062         (merged_store_group::merged_store_group): Initialize them.
2063         (merged_store_group::do_merge): Clear only_constants member if
2064         adding something other than INTEGER_CST store.
2065         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2066         stores with order >= first_nonmergeable_order.  Use
2067         merged_store->only_constants instead of always recomputing it.
2068         Set merged_store->first_nonmergeable_order if we've skipped any
2069         stores.  Attempt to merge overlapping INTEGER_CST stores that
2070         we would otherwise skip.
2072         PR sanitizer/87837
2073         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2075 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2077         PR target/87853
2078         * config/i386/emmintrin.h (__v16qs): New to cope with option
2079         -funsigned-char.
2080         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2081         (_mm_cmplt_epi8): Likewise.
2082         (_mm_cmpgt_epi8): Likewise.
2084 2018-11-05  Richard Biener  <rguenther@suse.de>
2086         PR rtl-optimization/87852
2087         * fwprop.c (use_killed_between): Only consider single-defs of the
2088         use whose definition statement dominates the use.
2090 2018-11-05  Martin Liska  <mliska@suse.cz>
2092         PR web/87829
2093         * doc/invoke.texi: Remove options that are
2094         not disabled with -Os.
2096 2018-11-05  Martin Liska  <mliska@suse.cz>
2098         PR c/87811
2099         * doc/extend.texi: Update constrain about the last argument
2100         of __builtin_expect_with_probability.
2102 2018-11-05  Martin Liska  <mliska@suse.cz>
2104         PR c/87811
2105         * predict.c (expr_expected_value_1): Verify
2106         that last argument is a real constants and emit
2107         error.
2109 2018-11-05  Martin Liska  <mliska@suse.cz>
2111         PR gcov-profile/77698
2112         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2113         only in LTO mode.
2115 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2117         PR tree-optimization/86572
2118         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2120 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2122         PR tree-optimization/87672
2123         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2124         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2126 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2128         PR middle-end/58372
2129         * cfgexpand.c (pass_expand::execute): Move the call to
2130         finish_eh_generation in front of the call to expand_stack_alignment.
2132 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2134         * common/config/i386/i386-common.c (processor_alias_table): Add
2135         znver2 entry.
2136         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2137         (case ${target}): Add znver2.
2138         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2139         -march=native recognize znver2 processors.
2140         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2141         * config/i386/i386.c (m_znver2): New definition.
2142         (m_ZNVER): New definition.
2143         (m_AMD_MULTIPLE): Includes m_znver2.
2144         (processor_cost_table): Add znver2 entry.
2145         (processor_target_table): Add znver2 entry.
2146         (get_builtin_code_for_version): Set priority for
2147         PROCESSOR_ZNVER2.
2148         (processor_model): Add M_AMDFAM17H_ZNVER2.
2149         (arch_names_table): Ditto.
2150         (ix86_reassociation_width): Include znver2.
2151         * config/i386/i386.h (TARGET_znver2): New definition.
2152         (struct ix86_size_cost): Add TARGET_ZNVER2.
2153         (enum processor_type): Add PROCESSOR_ZNVER2.
2154         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2155         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2156         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2157         (ix86_adjust_cost): Add znver2.
2158         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2159         * gcc/doc/extend.texi: Add details about znver2.
2160         * gcc/doc/invoke.texi: Add details about znver2.
2162 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2164         PR target/87079
2166         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2167         pattern.
2169 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2171         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2172         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2174 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2176         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2177         shrink-wrapping when -mtrack-speculation.
2179 2018-11-02  Richard Biener  <rguenther@suse.de>
2181         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2182         (pop_cost_one_pair): Do not free pair.
2183         (pop_best_coalesce): Likewise.
2184         (create_coalesce_list): Initialize obstack.
2185         (delete_coalesce_list): Free obstack.
2186         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2187         (add_cost_one_coalesce): Likewise.
2188         (struct live_track): Remove bitmap pointer indirections.
2189         (new_live_track): Adjust.
2190         (delete_live_track): Likewise.
2191         (live_track_remove_partition): Likewise.
2192         (live_track_add_partition): Likewise.
2193         (live_track_live_p): Likewise.
2194         (live_track_process_def): Likewise.
2195         (live_track_clear_base_vars): Likewise.
2197 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2199         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2200         * configure: Regenerate.
2201         * config.in: Regenerate.
2202         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2203         HAVE_AS_REGISTER_PSEUDO_OP guard.
2204         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2205         !HAVE_AS_REGISTER_PSEUDO_OP support.
2206         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2207         guard.
2209 2018-11-02  Richard Biener  <rguenther@suse.de>
2211         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2212         with is_type_die.
2214 2018-11-02  Richard Biener  <rguenther@suse.de>
2216         PR tree-optimization/87776
2217         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2218         executable when iterating but running into rpo-vn-max-loop-depth
2219         and not eliding the iteration.
2221 2018-11-30  Jan Hubicka  <jh@suse.cz>
2223         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2224         TYPE_DECL.
2226 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2228         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2229         prototype.
2230         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2231         Combine with rs6000_address_for_fpconvert.
2232         (rs6000_address_for_fpconvert): Combine with
2233         rs6000_force_indexed_or_indirect_mem.
2234         (rs6000_expand_vector_init): Change function call from
2235         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2236         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2237         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2238         (floatsi<mode>2_lfiwax_mem): Ditto.
2239         (floatunssi<mode>2_lfiwzx): Ditto.
2240         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2241         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2242         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2243         (fix_trunc<mode>si2_stfiwx): Ditto.
2244         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2245         (float_<mode>si2_hw): Ditto.
2246         (floatuns_<mode>si2_hw): Ditto.
2247         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2248         (vsx_splat_<mode>): Ditto.
2250 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2252         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2253         at end of assembler input text.
2254         * configure: Regenerate.
2256 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2258         PR tree-optimization/87826
2259         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2260         negative or larger or equal to type's precision.
2262 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2264         * opts.c (default_options_table): Do not enable
2265         OPT_fdelayed_branch at -Og.
2266         * doc/invoke.texi (-fdelayed-branch): Document it.
2268 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2270         * optabs-libfuncs.c (build_libfunc_function_visibility):
2271         New, split out from...
2272         (build_libfunc_function): ... here.
2273         (init_one_libfunc_visibility): New, split out from ...
2274         (init_one_libfunc): ... here.
2276         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2277         scratch register need not be early-clobber.  Document the reason
2278         why we cannot use ST<OP>.
2280 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2282         PR bootstrap/82856
2283         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2284         line for second argument of AC_DEFINE_UNQUOTED.
2285         * doc/install.texi (Tools/packages necessary for modifying GCC):
2286         Update to autoconf 2.69 and automake 1.15.1.
2287         * aclocal.m4, config.in, configure: Regenerate.
2289 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2291         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2292         initialization.
2294 2018-10-31  Martin Liska  <mliska@suse.cz>
2296         PR driver/83193
2297         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2298         Add new argument invalid_extension.
2299         (aarch64_get_all_extension_candidates): New function.
2300         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2301         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2302         new argument.
2303         (aarch64_get_all_extension_candidates): New function.
2304         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2305         argument invalid_extension.
2306         (aarch64_parse_cpu): Likewise.
2307         (aarch64_print_hint_for_extensions): New function.
2308         (aarch64_validate_mcpu): Provide hint about invalid extension.
2309         (aarch64_validate_march): Likewise.
2310         (aarch64_handle_attr_arch): Pass new argument.
2311         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2312         (aarch64_handle_attr_isa_flags): Likewise.
2314 2018-10-31  Richard Biener  <rguenther@suse.de>
2316         PR middle-end/70359
2317         PR middle-end/86270
2318         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2319         copy generation to useful cases.  Place the copy before
2320         the definition of the backedge value when possible.
2322 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2324         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2325         * config/arc/arc.c (arc_active_insn): New function.
2326         (check_store_cacheline_hazard): Likewise.
2327         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2328         (arc_override_options): Disable delay slot scheduler for older
2329         A7.
2330         (arc_store_addr_hazard_p): New implementation, old one renamed to
2331         ...
2332         (arc_store_addr_hazard_internal_p): Renamed.
2333         (arc_reorg): Don't combine into brcc instructions which are part
2334         of hardware hazard solution.
2335         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2336         (tune_arc700): Likewise.
2337         * config/arc/arc.opt (arc7xx): New tune value.
2338         * config/arc/arc700.md: Improve A7 scheduler.
2340 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2342         * config/arc/arc.c (arc_override_options): Remove
2343         TARGET_COMPACT_CASESI.
2344         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2345         (CASE_VECTOR_MODE): Likewise.
2346         (CASE_VECTOR_PC_RELATIVE): Likewise.
2347         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2348         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2349         (ADDR_VEC_ALIGN): Update.
2350         (ASM_OUTPUT_CASE_LABEL): Undefine.
2351         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2352         (TARGET_BI_BIH): Define.
2353         (DEFAULT_BRANCH_INDEX): Likewise.
2354         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2355         instructions, remove compact_casesi use case.
2356         (casesi_compact_jump): Remove.
2357         (casesi_dispatch): New pattern.
2358         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2359         compact_casesi option.
2360         * doc/invoke.texi: Document mbranch-index option.
2362 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2364         * config/arc/arc.c (arc_get_tp): Remove function.
2365         (arc_emit_call_tls_get_addr): Likewise.
2366         (arc_call_tls_get_addr): New function.
2367         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2368         * config/arc/arc.md (tls_load_tp_soft): Remove.
2369         (tls_gd_get_addr): Likewise.
2371 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2373         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2374         (mulsi3_600_lib): Remove pattern.
2375         (umulsi3_highpart_600_lib_le): Likewise.
2376         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2377         (umulsidi3): Remove call to umulsidi3_600_lib.
2378         (umulsidi3_600_lib): Remove pattern.
2379         (peephole2): Remove peephole using the above deprecated patterns.
2381 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2383         PR target/87374
2384         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2385         use of -mslow-flash-data and -mword-relocations.
2386         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2387         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2388         flag_pic.
2389         * doc/invoke.texi (-mword-relocations): Mention conflict with
2390         -mslow-flash-data.
2391         (-mslow-flash-data): Reciprocally.
2393 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2395         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2396         16-byte modes held in GP registers to use an even regno.
2398         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2399         (aarch64_atomic_ldop_supported_p): Remove.
2400         (aarch64_gen_atomic_ldop): Remove.
2401         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2402         Fully expand LSE operations here.
2403         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2404         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2405         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2406         and use ATOMIC_LDOP instead; use register_operand for the input;
2407         drop the split and emit insns directly.
2408         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2409         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2410         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2412         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2413         (aarch64_gen_atomic_ldop): Don't call it.
2414         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2415         Use aarch64_reg_or_zero.
2416         (aarch64_atomic_exchange<ALLI>): Likewise.
2417         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2418         operand 0; use aarch64_reg_or_zero for input; merge ...
2419         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2421         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2422         (aarch64_split_compare_and_swap): Use it.
2423         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2424         test oldval against the proper predicate.
2425         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2426         Use nonmemory_operand for expected.
2427         (cas_short_expected_pred): New.
2428         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2429         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2430         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2431         (aarch64_plushi_operand): New.
2433         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2434         Force oldval into the rval register for TARGET_LSE; emit the compare
2435         during initial expansion so that it may be deleted if unused.
2436         (aarch64_gen_atomic_cas): Remove.
2437         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2438         Change =&r to +r for operand 0; use match_dup for operand 2;
2439         remove is_weak and mod_f operands as unused.  Drop the split
2440         and merge with...
2441         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2442         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2443         (@aarch64_atomic_cas<GPI>): Similarly.
2445 2018-10-31  Richard Biener  <rguenther@suse.de>
2447         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2448         using ABSU_EXPR.
2450 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2452         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2453         * config/aarch64/aarch64.md: Include saphira.md
2454         * config/aarch64/saphira.md: New file for pipeline description.
2456 2018-10-30  Martin Sebor  <msebor@redhat.com>
2458         PR middle-end/87041
2459         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2460         inlining context.
2461         (sprintf_dom_walker::compute_format_length):
2462         Avoid setting POSUNDER4K here.
2463         (get_destination_size): Handle null argument values.
2464         (get_user_idx_format): New function.
2465         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2466         functions, including user-defined with attribute format printf.
2467         Use %G to include inlining context.
2468         Set POSUNDER4K here.
2470 2018-10-30  Jan Hubicka  <jh@suse.cz>
2472         * params.def (lto-partitions): Bump from 32 to 128.
2474 2018-10-30  Jan Hubicka  <jh@suse.cz>
2476         * tree.c
2477         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2478         head in file.
2479         (free_lang_data_in_type): Forward declare.
2480         (fld_type_variant_equal_p): New function.
2481         (fld_type_variant): New function
2482         (fld_incomplete_types): New hash.
2483         (fld_incomplete_type_of): New function
2484         (fld_simplfied-type): New function.
2485         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2486         (free_lang_data): Allocate and free fld_incomplete_type; update call
2487         of free_lang_data_in_decl.
2489 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2491         * gcov.c (output_lines): Remove duplicate line.
2493 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2495         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2496         if not in indexed or indirect form.
2497         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2498         (bswapdi2_store): Ditto.
2499         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2500         helper function.
2501         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2502         Prototype for helper function.
2504 2018-10-30  Martin Sebor  <msebor@redhat.com>
2506         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2507         (target, pragma GCC optimize, pragma GCC target): Ditto.
2509 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2511         * doc/extend.texi: Fix prototype and description of
2512         __builtin_expect_with_probability.
2514 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2516         * cgraph.h (clone_function_name_1): Replaced by new
2517           clone_function_name_numbered that takes name as string; for
2518           privatize_symbol_name_1 use only.
2519           (clone_function_name): Renamed to
2520           clone_function_name_numbered to be explicit about numbering.
2521           (clone_function_name): New two-argument function that does
2522           not number its output.
2523           (clone_function_name): New three-argument function that
2524           takes a number to append to its output.
2525         * cgraphclones.c (duplicate_thunk_for_node):
2526           (clone_function_name_1): Renamed.
2527           (clone_function_name_numbered): Two new functions.
2528           (clone_function_name): Improved documentation.
2529           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2530         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2531         * final.c (final_scan_insn_1): Use the new clone_function_name
2532           without numbering.
2533         * multiple_target.c (create_dispatcher_calls): Ditto.
2534           (create_target_clone): Ditto.
2535         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2536         * omp-low.c (create_omp_child_function_name): Ditto.
2537         * omp-simd-clone.c (simd_clone_create): Ditto.
2538         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2539           new clone_function_name without numbering.
2541 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2543         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2544         Assert that the allocation size is not zero.
2546 2018-10-30  Richard Biener  <rguenther@suse.de>
2548         PR tree-optimization/87800
2549         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2550         non-induction or reduction PHIs.
2552 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2554         * config/aarch64/falkor-tag-collision-avoidance.c
2555         (execute_tag_collision_avoidance): Call df_note_add_problem.
2557 2018-10-30  Martin Liska  <mliska@suse.cz>
2559         * doc/extend.texi: Fix typo in documentation
2560         of __builtin_expect_with_probability.
2562 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2564         PR c++/87721
2565         * input.c (get_substring_ranges_for_loc): Detect if
2566         linemap_resolve_location gives us a NULL map, and reject
2567         this case.
2569 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2571         * config.gcc (xstormy16-*-elf): Set tm_d_file.
2573 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2574             Martin Sebor  <msebor@redhat.com>
2575             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2577         * doc/ux.texi (Quoting): New subsection, adapted from material at
2578         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
2579         MartinSebor and ManuelLopezIbanez.
2580         (Fix-it hints): Note that fix-it hints shouldn't be marked for
2581         translation.
2583 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
2585         PR middle-end/87469
2586         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
2587         max value.
2589 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2591         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
2593 2018-10-29  Paul Koning  <ni1d@arrl.net>
2595         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
2597 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2599         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
2600         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
2601         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
2602         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
2603         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
2604         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
2605         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
2606         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
2607         Change 'vector' to '__vector'.
2608         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
2609         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
2610         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
2611         _mm_avg_pu16): Likewise.  And, whitespace corrections.
2613 2018-10-29  Richard Biener  <rguenther@suse.de>
2615         PR tree-optimization/87785
2616         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
2617         internal defs.
2619 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2621         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
2623 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2625         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
2626         ports configuration macro, defaults to "gnu".
2627         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
2628         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
2630 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2632         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
2634 2018-10-29  Richard Biener  <rguenther@suse.de>
2636         PR tree-optimization/87790
2637         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
2638         (vect_make_slp_decision): Adjust.
2639         (vect_slp_analyze_bb_1): Likewise.
2640         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
2641         edges.
2643 2018-10-29  Richard Biener  <rguenther@suse.de>
2645         PR tree-optimization/87785
2646         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
2647         and processing.
2648         (vect_build_slp_tree): Likewise.
2649         (vect_gather_slp_loads): New function.
2650         (vect_analyze_slp_instance): Gather loads separately from the
2651         SLP tree build.
2653 2018-10-29  Martin Liska  <mliska@suse.cz>
2655         * Makefile.in: Make dependency to json.o.
2656         * doc/gcov.texi: Document new JSON format, remove
2657         old intermediate format documentation.
2658         * gcov.c (struct function_info): Come up with m_name and
2659         m_demangled_name.
2660         (function_info::function_info): Initialize it.
2661         (function_info::~function_info): Release it.
2662         (main): Rename flag_intermediate_format to flag_json_format.
2663         (print_usage): Describe --json-format.
2664         (process_args): Set flag_json_format.
2665         (output_intermediate_line): Remove.
2666         (output_intermediate_json_line): Likewise.
2667         (get_gcov_intermediate_filename): Return new extension
2668         ".gcov.json.gz".
2669         (output_intermediate_file): Implement JSON emission.
2670         (output_json_intermediate_file): Implement JSON emission.
2671         (generate_results): Use ::get_name for function name.
2672         Handle JSON output file.
2673         (read_graph_file): Use ::get_name instead of cplus_demangle.
2674         (read_count_file): Likewise.
2675         (solve_flow_graph): Likewise.
2676         (add_line_counts): Likewise.
2677         (accumulate_line_counts): Use new flag_json_format.
2678         (output_function_details): Use ::get_name instead of cplus_demangle.
2679         (output_lines): Likewise.
2680         * json.cc (test_writing_literals): Add new tests.
2681         * json.h (class literal): Add new boolean constructor.
2683 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
2685         PR rtl-optimization/87701
2686         PR rtl-optimization/87780
2687         * combine.c (make_more_copies): Rewrite.
2689 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
2691         * doc/generic.texi (ABSU_EXPR): Document.
2692         * match.pd (absu(x)*absu(x) -> x*x): Handle.
2693         (absu(absu(X)) -> absu(X)): Likewise.
2694         (absu(-X) -> absu(X)): Likewise.
2695         (absu(X)  where X is nonnegative -> X): Likewise.
2697 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
2699         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
2700         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
2701         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
2702         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
2703         (s-tm-texi): Also check timestamp on d-target.def.
2704         (generated_files): Add TM_D_H and d-target-hooks-def.h.
2705         (build/genhooks.o): Also depend on D_TARGET_DEF.
2706         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
2707         variables.
2708         * config/aarch64/aarch64-d.c: New file.
2709         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
2710         Define.
2711         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
2712         prototype.
2713         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
2714         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
2715         * config/arm/arm-d.c: New file.
2716         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
2717         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
2718         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2719         * config/arm/t-arm (arm-d.o): New rule.
2720         * config/default-d.c: New file.
2721         * config/glibc-d.c: New file.
2722         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2723         * config/i386/i386-d.c: New file.
2724         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
2725         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
2726         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2727         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
2728         * config/i386/t-i386 (i386-d.o): New rule.
2729         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2730         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2731         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
2732         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2733         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2734         * config/mips/mips-d.c: New file.
2735         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
2736         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
2737         * config/mips/t-mips (mips-d.o): New rule.
2738         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2739         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2740         * config/powerpcspe/powerpcspe-d.c: New file.
2741         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
2742         New prototype.
2743         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2744         Support GNU D by using 0 as the language type.
2745         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
2746         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
2747         * config/riscv/riscv-d.c: New file.
2748         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
2749         prototype.
2750         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
2751         * config/riscv/t-riscv (riscv-d.o): New rule.
2752         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2753         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2754         * config/rs6000/rs6000-d.c: New file.
2755         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
2756         prototype.
2757         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2758         Support GNU D by using 0 as the language type.
2759         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
2760         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
2761         * config/s390/s390-d.c: New file.
2762         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
2763         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
2764         * config/s390/t-s390 (s390-d.o): New rule.
2765         * config/sparc/sparc-d.c: New file.
2766         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
2767         prototype.
2768         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
2769         * config/sparc/t-sparc (sparc-d.o): New rule.
2770         * config/t-glibc (glibc-d.o): New rule.
2771         * configure: Regenerated.
2772         * configure.ac (tm_d_file): New variable.
2773         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
2774         * doc/contrib.texi (Contributors): Add self for the D frontend.
2775         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
2776         * doc/install.texi (Configuration): Mention libphobos as an option for
2777         --enable-shared.  Mention d as an option for --enable-languages.
2778         (Testing): Mention check-d as a target.
2779         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
2780         name suffixes.  Mention d as a -x option.
2781         * doc/sourcebuild.texi (Top Level): Mention libphobos.
2782         * doc/standards.texi (Standards): Add section on D language.
2783         * doc/tm.texi: Regenerated.
2784         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
2785         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
2786         * dwarf2out.c (is_dlang): New function.
2787         (gen_compile_unit_die): Use DW_LANG_D for D.
2788         (declare_in_namespace): Return module die for D, instead of adding
2789         extra declarations into the namespace.
2790         (gen_namespace_die): Generate DW_TAG_module for D.
2791         (gen_decl_die): Handle CONST_DECLSs for D.
2792         (dwarf2out_decl): Likewise.
2793         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
2794         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
2795         kinds of aggregates.
2796         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
2797         * genhooks.c: Include d/d-target.def.
2799 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
2801         PR target/85669
2802         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2803         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2804         (STACK_DYNAMIC_OFFSET): Likewise.
2806 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
2808         PR target/80024
2809         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
2810         error message.
2812 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
2814         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
2815         return type and other typos.
2817 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
2819         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2820         a shorter sequence with fewer branches.
2821         (emit_final_str_compare_gpr): Ditto.
2823 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2825         * config/rs6000/tmmintrin.h: New file.
2826         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
2828 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2830         * config/rs6000/mmintrin.h: Enable 32bit compilation.
2831         * config/rs6000/xmmintrin.h: Likewise.
2833 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2835         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2837 2018-10-26  Richard Biener  <rguenther@suse.de>
2839         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
2840         and wrapper.
2841         (vect_mark_slp_stmts_relevant): Likewise.
2842         (vect_detect_hybrid_slp_stmts): Likewise.
2843         (vect_bb_slp_scalar_cost): Likewise.
2844         (vect_remove_slp_scalar_calls): Likewise.
2846 2018-10-26  Jan Hubicka  <jh@suse.cz>
2848         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
2849         (warn_types_mismatch): Fix walk of DECL_NAME.
2850         (odr_types_equivalent_p): Fix overactive assert.
2852 2018-10-26  Richard Biener  <rguenther@suse.de>
2854         PR tree-optimization/87105
2855         * tree-vectorizer.h (_slp_tree::refcnt): New member.
2856         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
2857         refcnt.
2858         (vect_create_new_slp_node): Initialize refcnt to one.
2859         (bst_traits): Move.
2860         (scalar_stmts_set_t, bst_fail): Remove.
2861         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
2862         (vect_build_slp_tree): Add bst_map argument and lookup
2863         already created SLP nodes.
2864         (vect_print_slp_tree): Handle a SLP graph, print SLP node
2865         addresses.
2866         (vect_slp_rearrange_stmts): Handle a SLP graph.
2867         (vect_analyze_slp_instance): Adjust and free SLP nodes from
2868         the CSE map.  Fix indenting.
2869         (vect_schedule_slp_instance): Add short-cut.
2871 2018-10-26  Martin Liska  <mliska@suse.cz>
2873         PR testsuite/86158
2874         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
2875         addr_expr and not with pointers.
2877 2018-10-26  Jan Hubicka  <jh@suse.cz>
2879         * tree.c (free_lang_data_in_type): Only check main variants.
2880         * ipa-devirt.c (warn_odr): Make static.
2881         (types_same_for_odr): Drop strict variant.
2882         (types_odr_comparable): Likewise.
2883         (odr_or_derived_type_p): Look for main variants.
2884         (odr_name_hasher::equal): Cleanup comment.
2885         (odr_subtypes_equivalent): Add warn and warned arguments; check main
2886         variants.
2887         (type_variants_equivalent_p): break out from ...
2888         (odr_types_equivalent): ... here; go for main variants where needed.
2889         (warn_odr): ... here; turn static.
2890         (warn_types_mismatch): Compare mangled names of main variants.
2891         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
2892         (type_with_linkage_p): Sanity check that we look at main variant.
2893         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
2894         * tree.h (types_same_for_odr): Drop strict argument.
2896 2018-10-26  Richard Biener  <rguenther@suse.de>
2898         PR tree-optimization/87746
2899         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
2900         Simplify and fix WRT strided store groups with size not
2901         equal to step in element count.
2902         (vect_analyze_group_access_1): Dump the whole group.
2904 2018-10-25  Carl Love  <cel@us.ibm.com>
2906         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
2907         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
2908         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
2909         precicion entry for each overloaded builtin.
2910         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
2911         VSCEDPUO): Rename overloaded name.
2912         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
2913         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
2914         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
2915         define_expand for xscmpexqp instruction.
2916         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
2918 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2919             Jinsong Ji <jji@us.ibm.com>
2921         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
2922         function with vec_sl.
2923         (_mm_slli_epi32): Likewise.
2924         (_mm_slli_epi64): Likewise.
2925         (_mm_srai_epi16): Replace deprecated function with vec_sra.
2926         (_mm_srai_epi32): Likewise.
2927         (_mm_srli_epi16): Replace deprecated function with vec_sr.
2928         (_mm_srli_epi32): Likewise.
2929         (_mm_srli_epi64): Likewise.
2930         (_mm_sll_epi16): Replace deprecated function with vec_sl.
2931         (_mm_sll_epi32): Likewise.
2932         (_mm_sll_epi64): Likewise.
2933         (_mm_sra_epi16): Replace deprecated function with vec_sra.
2934         (_mm_sra_epi32): Likewise.
2935         (_mm_srl_epi16): Replace deprecated function with vec_sr.
2936         (_mm_srl_epi32): Likewise.
2937         (_mm_srl_epi64): Likewise.
2939 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2940             Jinsong Ji <jji@us.ibm.com>
2942         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
2943         comparison operators with vec_cmp* for compatibility due to
2944         unfortunate history; clean up formatting and use types more
2945         appropriately.
2946         (_mm_sll_epi32): Likewise.
2947         (_mm_sll_epi64): Likewise.
2948         (_mm_srl_epi16): Likewise.
2949         (_mm_srl_epi32): Likewise.
2950         (_mm_srl_epi64): Likewise.
2952 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2953             Jinsong Ji <jji@us.ibm.com>
2955         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
2956         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
2957         __vector __bool int.  Use vec_cmpgt in preference to deprecated
2958         function vec_vcmpgtfp.
2959         (_mm_max_ps): Likewise.
2961 2018-10-25  Jeff Law  <law@redhat.com>
2963         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
2964         if returning false.
2966 2018-10-25  Martin Sebor  <msebor@redhat.com>
2968         * doc/extend.texi (aligned): Expand attribute description.
2969         (Alignment): Rename section.  Discuss function arguments.
2971 2018-10-25  Jan Hubicka  <jh@suse.cz>
2973         * ipa-devirt.c (main_odr_variant): Remove.
2974         (hash_odr_name, types_same_for_odr, types_odr_comparable,
2975         odr_name_hasher::equal, odr_subtypes_equivalent_p):
2976         Drop use of main_odr_variant.
2977         (add_type_duplicate): Silence confused warnings on integer types.
2978         (get_odr_type): Always look for main variant.
2979         (register_odr_type): Simplify.
2981 2018-10-25  Richard Biener  <rguenther@suse.de>
2983         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
2984         Initialize ng to silence error with release checking bootstrap.
2986 2018-10-25  Richard Biener  <rguenther@suse.de>
2988         * tree-if-conv.c: Include tree-ssa-sccvn.h.
2989         (tree_if_conversion): Run CSE on the if-converted loop body.
2991 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2993         * config/s390/constraints.md (ZL): New constraint.
2994         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
2995         operands.
2996         * config/s390/s390.md (movdi_larl): Remove.
2997         (movdi_64): Add the LARL alternative.
2999 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3001         PR bootstrap/87747
3002         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3003         (RTX_CODE_HWINT_P): New macro.
3004         (rtx_code_size): Use RTX_CODE_HWINT_P ().
3006 2018-10-25  Jan Hubicka  <jh@suse.cz>
3008         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3009         is anonymous.
3011 2018-10-25  Richard Biener  <rguenther@suse.de>
3013         PR tree-optimization/87665
3014         PR tree-optimization/87745
3015         * tree-vectorizer.h (get_earlier_stmt): Remove.
3016         (get_later_stmt): Pick up UID from the original non-pattern stmt.
3018 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
3020         * options.texi (Deprecated): Move list to Var section.
3022 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
3023             Jinsong Ji <jji@us.ibm.com>
3025         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3026         __vector long to __vector long long.
3027         (_mm_cvtpd_ps): Likewise.
3028         (_mm_cvttpd_epi32): Likewise.
3029         (_mm_cvtpi32_pd): Likewise.
3030         (_mm_unpackhi_epi64): Likewise.
3031         (_mm_unpacklo_epi64): Likewise.
3033 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
3035         PR rtl-optimization/87720
3036         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3038 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
3040         * gimple-ssa-isolate-paths.c
3041         (find_implicit_erroneous_behavior): Do not change code if the
3042         pass is running for warnings only.
3043         (find_explicit_erroneous_behavior): Likewise.
3045 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
3047         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3048         Define as rs6000_mangle_decl_assembler_name.
3049         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3050         long double to IEEE long double, switch the names of the long
3051         double built-in functions to be <func>f128 instead of <func>l.
3053 2018-10-24  Martin Sebor  <msebor@redhat.com>
3055         * doc/extend.texi (nonnull): List no-argument form.  Reference
3056         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3058 2018-10-24  Richard Biener  <rguenther@suse.de>
3060         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3062 2018-10-24  Martin Liska  <mliska@suse.cz>
3064         PR tree-optimization/84436
3065         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3066         Remove.
3067         (switch_conversion::contains_linear_function_p): New.
3068         (switch_conversion::build_one_array): Support linear
3069         transformation on input.
3070         * tree-switch-conversion.h (struct switch_conversion): Add
3071         contains_linear_function_p declaration.
3073 2018-10-24  Richard Biener  <rguenther@suse.de>
3075         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3076         if its argument is CONSTANT_CLASS_P.
3078 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3080         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3081         it is wrong for forward declarations.
3083 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3085         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3086         SYMBOL_REF_P () check.
3088 2018-10-24  Richard Biener  <rguenther@suse.de>
3090         PR tree-optimization/87105
3091         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3092         dump classification.
3093         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3094         stores by splitting the affected group after the fact.
3095         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3096         fail the SLP build because of size constraints.
3098 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3100         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3101         * configure: Regenerate.
3102         * config.in: Regenerate.
3103         * varasm.c (mergeable_string_section): Use readonly_data_section
3104         if linker doesn't support SHF_MERGE with alignment > 8.
3105         (mergeable_constant_section): Likewise.
3107 2018-10-24  Richard Biener  <rguenther@suse.de>
3109         PR tree-optimization/84013
3110         * tree-ssa-structalias.c (struct msdi_data): New struct for
3111         marshalling data to walk_stmt_load_store_ops.
3112         (maybe_set_dependence_info): Refactor as callback for
3113         walk_stmt_load_store_ops.
3114         (compute_dependence_clique): Set restrict info on all stmt kinds.
3116 2018-10-24  Martin Liska  <mliska@suse.cz>
3118         * cgraph.c (cgraph_node::dump):
3119         Remove reduntant dumps and make tp_first_run dump more compact.
3121 2018-10-24  Richard Biener  <rguenther@suse.de>
3123         PR tree-optimization/87665
3124         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3125         to reflect reality.
3127 2018-10-12  Jeff Law  <law@redhat.com>
3129         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3130         for H8/S.
3132 2018-10-23  Richard Biener  <rguenther@suse.de>
3134         * tree-vrp.c (add_assert_info): Guard dump_printf with
3135         dump_enabled_p.
3136         * gimple-ssa-evrp-analyze.c
3137         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3138         Use value_range::ignore_equivs_equal_p.
3140 2018-10-23  Richard Biener  <rguenther@suse.de>
3142         PR tree-optimization/87105
3143         PR tree-optimization/87608
3144         * passes.def (pass_all_early_optimizations): Add early phi-opt
3145         after dce.
3146         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3147         addition to debug stmts.
3148         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3149         and abs replacement early.
3150         * tree-cfg.c (gimple_empty_block_p): Likewise.
3152 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3154         PR target/86383
3155         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3156         specified to configure.
3157         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3159 2018-10-23  Richard Biener  <rguenther@suse.de>
3161         PR tree-optimization/87700
3162         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3164 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3166         PR target/87674
3167         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3168         second argument from __mmask16 to __mmask8.
3169         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3170         _mm_mask_packs_epi32): Likewise.
3171         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3172         Likewise.
3173         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3175 2018-10-23  Richard Biener  <rguenther@suse.de>
3177         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3179 2018-10-23  Richard Biener  <rguenther@suse.de>
3181         PR tree-optimization/86144
3182         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3183         over simd attribute.
3185 2018-10-23  Richard Biener  <rguenther@suse.de>
3187         PR tree-optimization/87693
3188         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3189         the case we do not find the taken edge.
3191 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3192             Jinsong Ji  <jji@us.ibm.com>
3194         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3195         (_mm_store_pd): Use unaligned vector type for pointer cast.
3196         (_mm_maskmoveu_si128): Likewise.
3197         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3198         (_mm_store_ps): Use unaligned vector type for pointer cast.
3200 2018-10-22  Paul Koning  <ni1d@arrl.net>
3202         * symtab.c (symtab_node::increase_alignment): Correct max
3203         alignment check.
3205 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3207         PR tree-optimization/87633
3208         * match.pd: Do not generate unordered integer comparisons.
3210 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3212         PR rtl-optimization/87600
3213         * combine.c: Add include of expr.h.
3214         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3215         register to a pseudo.
3216         (make_more_copies): New function, add a copy to a new pseudo after
3217         the moves from hard registers into pseudos.
3218         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3219         later.  Call make_more_copies.
3221 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3223         * lra-constraints.c (process_alt_operands): New local array,
3224         matching_early_clobber.  Check matching_early_clobber before
3225         decrementing reject, and set matching_early_clobber after.
3227 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3229         PR target/87598
3230         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3231         call output_addr_const and hope for the best.
3233 2018-10-22  Richard Biener  <rguenther@suse.de>
3235         * gimple-ssa-evrp-analyze.c
3236         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3237         smarter about what ranges to use.
3238         * tree-vrp.c (add_assert_info): Dump here.
3239         (register_edge_assert_for_2): Instead of here at multiple but
3240         not all places.
3242         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3243         * gcc.dg/predict-6.c: Adjust.
3244         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3245         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3246         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3248 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3249         Richard Biener  <rguenther@suse.de>
3251         * bitmap.h: Update data structure documentation, including a
3252         description of bitmap views as either linked-lists or splay trees.
3253         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3254         (struct bitmap_head_def): Likewise.
3255         (bitmap_list_view, bitmap_tree_view): New prototypes.
3256         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3257         tree_form fields.
3258         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3259         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3260         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3261         released bitmap element here.
3262         (bitmap_element_free): Remove.
3263         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3264         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3265         this function similar ones such that linked-list bitmap implementation
3266         functions are grouped.
3267         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3268         and moved for grouping.
3269         (bitmap_list_insert_element_after): Renamed from
3270         bitmap_elt_insert_after, and moved for grouping.
3271         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3272         (bitmap_tree_link_left, bitmap_tree_link_right,
3273         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3274         bitmap_tree_link_element, bitmap_tree_unlink_element,
3275         bitmap_tree_find_element): New functions for splay-tree bitmap
3276         implementation.
3277         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3278         Renamed and moved, see above entries.
3279         (bitmap_tree_listify_from): New function to convert part of a splay
3280         tree bitmap to a linked-list bitmap.
3281         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3282         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3283         (bitmap_find_bit): Remove.
3284         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3285         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3286         Handle splay tree bitmaps.
3287         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3288         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3289         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3290         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3291         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3292         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3293         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3294         corresponding changes to use linked-list specific bitmap_element
3295         manipulation functions as applicable for efficiency.
3296         (bitmap_tree_to_vec): New function.
3297         (debug_bitmap_elt_file): New function split out from ...
3298         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3299         (bitmap_print): Likewise.
3301         PR tree-optimization/63155
3302         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3303         SSA edge worklists.
3304         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3305         in tree-view.
3307 2018-10-22  Martin Liska  <mliska@suse.cz>
3309         PR tree-optimization/87686
3310         Revert
3311         2018-08-29  Martin Liska  <mliska@suse.cz>
3313         * tree-switch-conversion.c (switch_conversion::expand):
3314         Strenghten assumption about gswitch statements.
3316 2018-10-22  Martin Liska  <mliska@suse.cz>
3318         * ipa-icf.c (sem_item::compare_attributes): Remove.
3319         (sem_item::compare_referenced_symbol_properties): Use
3320         attribute_list_equal instead.
3321         (sem_function::equals_wpa): Likewise.
3322         * ipa-icf.h: Remove compare_attributes.
3324 2018-10-22  Richard Biener  <rguenther@suse.de>
3326         PR middle-end/87682
3327         * mem-stats.h (mem_usage::operator==): Fix pasto.
3329 2018-10-22  Richard Biener  <rguenther@suse.de>
3331         PR tree-optimization/87640
3332         * tree-vrp.c (set_value_range_with_overflow): Decompose
3333         incomplete result.
3334         (extract_range_from_binary_expr_1): Adjust.
3336 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3338         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3339         (stmt_can_throw_external): Likewise.
3340         (stmt_can_throw_internal): Likewise.
3341         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3342         (lower_eh_constructs_2): Likewise.
3343         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3344         (stmt_can_throw_external): Likewise.
3345         (stmt_can_throw_internal): Likewise.
3346         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3347         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3348         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3349         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3350         (pass_lower_eh_dispatch::execute): Pass cfun to
3351         stmt_can_throw_external.
3352         (cleanup_empty_eh): Likewise.
3353         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3354         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3355         stmt_can_throw_external instead of pushing it to cfun.
3356         (symbol_table::create_edge): Likewise.
3357         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3358         stmt_can_throw_internal.
3359         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3360         to stmt_could_throw_p.
3361         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3362         stmt_can_throw_internal.
3363         (pass_store_merging::execute): Likewise.
3364         * gimple-ssa-strength-reduction.c
3365         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3366         stmt_could_throw_p.
3367         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3368         stmt_can_throw_internal.
3369         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3370         to stmt_can_throw_external.
3371         (check_stmt): Pass cfun to stmt_could_throw_p.
3372         (check_stmt): Pass cfun to stmt_can_throw_external.
3373         (pass_nothrow::execute): Likewise.
3374         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3375         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3376         stmt_can_throw_internal.
3377         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3378         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3379         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3380         * tree-complex.c (expand_complex_libcall): Pass cfun to
3381         stmt_could_throw_p and to stmt_can_throw_internal.
3382         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3383         * tree-inline.c (copy_edges_for_bb): Likewise.
3384         (maybe_move_debug_stmts_to_successors): Likewise.
3385         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3386         stmt_could_throw_p.
3387         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3388         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3389         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3390         stmt_can_throw_internal.
3391         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3392         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3393         stmt_could_throw_p.
3394         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3395         stmt_can_throw_internal.
3396         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3397         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3398         stmt_could_throw_p.
3399         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3400         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3401         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3402         (convert_mult_to_fma_1): Likewise.
3403         (convert_to_divmod): Likewise.
3404         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3405         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3406         * tree-ssa-propagate.c
3407         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3408         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3409         (maybe_optimize_range_tests): Likewise.
3410         (linearize_expr_tree): Likewise.
3411         (reassociate_bb): Likewise.
3412         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3413         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3414         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3415         (handle_char_store): Likewise.
3416         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3417         stmt_can_throw_internal.
3418         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3419         stmt_could_throw_p.
3420         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3421         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3422         (vectorizable_simd_clone_call): Likewise.
3423         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3424         (gimple_stringop_fixed_value): Likewise.
3426 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3428         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3429         literal pool references.
3430         (s390_check_qrst_address): Adapt to the new behavior of
3431         s390_loadrelative_operand_p ().
3433 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3435         PR target/72782
3436         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3438 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3440         PR target/72782
3441         * config/i386/sse.md (*<code><mode>3_bcst): New.
3443 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3445         PR target/72782
3446         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3447         V4DI, V16SI and V8DI.
3448         (*sub<mode>3<mask_name>_bcst): New.
3449         (*add<mode>3<mask_name>_bcst): Likewise.
3451 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3452             Jinsong Ji  <jji@us.ibm.com>
3454         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3455         __m64 with __vector unsigned long long for compatibility.
3456         (_mm_movemask_epi8): Likewise.
3457         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3458         (_mm_cvttps_pi32): Likewise.
3459         (_mm_cvtpi32_ps): Likewise.
3460         (_mm_cvtps_pi16): Likewise.
3461         (_mm_loadh_pi): Likewise.
3462         (_mm_storeh_pi): Likewise.
3463         (_mm_movehl_ps): Likewise.
3464         (_mm_movelh_ps): Likewise.
3465         (_mm_loadl_pi): Likewise.
3466         (_mm_storel_pi): Likewise.
3467         (_mm_movemask_ps): Likewise.
3468         (_mm_shuffle_pi16): Likewise.
3470 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3472         PR target/72782
3473         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3474         __builtin_ia32_vfnmsubpd512_mask.
3475         (_mm512_mask_fnmsub_round_pd): Likewise.
3476         (_mm512_fnmsub_pd): Likewise.
3477         (_mm512_mask_fnmsub_pd): Likewise.
3478         (_mm512_maskz_fnmsub_round_pd): Use
3479         __builtin_ia32_vfnmsubpd512_maskz.
3480         (_mm512_maskz_fnmsub_pd): Likewise.
3481         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3482         (_mm512_mask_fnmsub_round_ps): Likewise.
3483         (_mm512_fnmsub_ps): Likewise.
3484         (_mm512_mask_fnmsub_ps): Likewise.
3485         (_mm512_maskz_fnmsub_round_ps): Use
3486         __builtin_ia32_vfnmsubps512_maskz.
3487         (_mm512_maskz_fnmsub_ps): Likewise.
3488         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3489         __builtin_ia32_vfnmsubpd256_mask.
3490         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3491         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3492         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3493         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3494         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3495         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3496         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3497         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3498         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3499         __builtin_ia32_vfnmsubpd.
3500         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3501         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3502         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3503         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3504         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3505         * config/i386/i386-builtin.def: Add
3506         __builtin_ia32_vfnmsubpd256_mask,
3507         __builtin_ia32_vfnmsubpd256_maskz,
3508         __builtin_ia32_vfnmsubpd128_mask,
3509         __builtin_ia32_vfnmsubpd128_maskz,
3510         __builtin_ia32_vfnmsubps256_mask,
3511         __builtin_ia32_vfnmsubps256_maskz,
3512         __builtin_ia32_vfnmsubps128_mask,
3513         __builtin_ia32_vfnmsubps128_maskz,
3514         __builtin_ia32_vfnmsubpd512_mask,
3515         __builtin_ia32_vfnmsubpd512_maskz,
3516         __builtin_ia32_vfnmsubps512_mask,
3517         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3518         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3519         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3520         __builtin_ia32_vfnmsubpd256.
3521         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3522         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3523         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3524         Likewise.
3525         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3526         Likewise.
3527         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3528         Likewise.
3529         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3531 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3533         PR target/72782
3534         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3535         __builtin_ia32_vfnmaddpd512_mask.
3536         (_mm512_mask_fnmadd_round_pd): Likewise.
3537         (_mm512_fnmadd_pd): Likewise.
3538         (_mm512_mask_fnmadd_pd): Likewise.
3539         (_mm512_maskz_fnmadd_round_pd): Use
3540         __builtin_ia32_vfnmaddpd512_maskz.
3541         (_mm512_maskz_fnmadd_pd): Likewise.
3542         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3543         (_mm512_mask_fnmadd_round_ps): Likewise.
3544         (_mm512_fnmadd_ps): Likewise.
3545         (_mm512_mask_fnmadd_ps): Likewise.
3546         (_mm512_maskz_fnmadd_round_ps): Use
3547         __builtin_ia32_vfnmaddps512_maskz.
3548         (_mm512_maskz_fnmadd_ps): Likewise.
3549         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3550         __builtin_ia32_vfnmaddpd256_mask.
3551         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3552         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3553         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3554         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3555         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3556         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3557         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3558         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3559         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3560         __builtin_ia32_vfnmaddpd.
3561         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3562         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3563         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3564         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3565         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3566         * config/i386/i386-builtin.def: Add
3567         __builtin_ia32_vfnmaddpd256_mask,
3568         __builtin_ia32_vfnmaddpd256_maskz,
3569         __builtin_ia32_vfnmaddpd128_mask,
3570         __builtin_ia32_vfnmaddpd128_maskz,
3571         __builtin_ia32_vfnmaddps256_mask,
3572         __builtin_ia32_vfnmaddps256_maskz,
3573         __builtin_ia32_vfnmaddps128_mask,
3574         __builtin_ia32_vfnmaddps128_maskz,
3575         __builtin_ia32_vfnmaddpd512_mask,
3576         __builtin_ia32_vfnmaddpd512_maskz,
3577         __builtin_ia32_vfnmaddps512_mask,
3578         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
3579         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
3580         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
3581         __builtin_ia32_vfnmaddpd256.
3582         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
3583         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
3584         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
3585         Likewise.
3586         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
3587         Likewise.
3588         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
3589         Likewise.
3590         (fmai_vmfnmadd_<mode><round_name>): Likewise.
3592 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3594         PR target/72782
3595         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
3596         __builtin_ia32_vfmsubpd512_mask.
3597         (_mm512_mask_fmsub_round_pd): Likewise.
3598         (_mm512_fmsub_pd): Likewise.
3599         (_mm512_mask_fmsub_pd): Likewise.
3600         (_mm512_maskz_fmsub_round_pd): Use
3601         __builtin_ia32_vfmsubpd512_maskz.
3602         (_mm512_maskz_fmsub_pd): Likewise.
3603         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
3604         (_mm512_mask_fmsub_round_ps): Likewise.
3605         (_mm512_fmsub_ps): Likewise.
3606         (_mm512_mask_fmsub_ps): Likewise.
3607         (_mm512_maskz_fmsub_round_ps): Use
3608         __builtin_ia32_vfmsubps512_maskz.
3609         (_mm512_maskz_fmsub_ps): Likewise.
3610         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
3611         __builtin_ia32_vfmsubpd256_mask.
3612         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
3613         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3614         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
3615         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3616         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3617         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
3618         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
3619         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
3620         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
3621         __builtin_ia32_vfmsubpd.
3622         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
3623         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
3624         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
3625         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
3626         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
3627         * config/i386/i386-builtin.def: Add
3628         __builtin_ia32_vfmsubpd256_mask,
3629         __builtin_ia32_vfmsubpd256_maskz,
3630         __builtin_ia32_vfmsubpd128_mask,
3631         __builtin_ia32_vfmsubpd128_maskz,
3632         __builtin_ia32_vfmsubps256_mask,
3633         __builtin_ia32_vfmsubps256_maskz,
3634         __builtin_ia32_vfmsubps128_mask,
3635         __builtin_ia32_vfmsubps128_maskz,
3636         __builtin_ia32_vfmsubpd512_mask,
3637         __builtin_ia32_vfmsubpd512_maskz,
3638         __builtin_ia32_vfmsubps512_mask,
3639         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
3640         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
3641         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
3642         __builtin_ia32_vfmsubpd256.
3643         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
3644         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
3645         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
3646         Likewise.
3647         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
3648         Likewise.
3649         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
3650         Likewise.
3651         (fmai_vmfmsub_<mode><round_name>): Likewise.
3653 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3655         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
3656         Remove plus.  Renamed to ...
3657         (*sub<mode>3<mask_name>_bcst): This.
3658         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
3659         (*add<mode>3<mask_name>_bcst): This.
3661 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3663         PR target/72782
3664         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
3666 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3668         PR target/87662
3669         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
3670         (_mm_or_epi32): Likewise.
3671         (_mm256_xor_epi32): Likewise.
3672         (_mm_xor_epi32): Likewise.
3673         (_mm256_or_epi64): Likewise.
3674         (_mm_or_epi64): Likewise.
3675         (_mm256_xor_epi64): Likewise.
3676         (_mm_xor_epi64): Likewise.
3678 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
3680         PR target/72782
3681         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
3683 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
3685         PR middle-end/87647
3686         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
3688 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
3690         * doc/ux.texi: Move @section directly after @node.
3692 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
3694         PR middle-end/85488
3695         PR middle-end/87649
3696         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
3697         depend closely nested inside of loop with ordered clause with
3698         a parameter.
3700 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
3702         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
3703         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
3704         * doc/ux.texi: New file.
3706 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
3708         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
3709         be the first CR field allocated.
3711 2018-10-19  Richard Biener  <rguenther@suse.de>
3713         PR target/87657
3714         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
3715         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
3717 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3719         PR target/72782
3720         * config/i386/sse.md
3721         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
3722         (*add<mode>3<mask_name>_bcst_2): Likewise.
3724 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3726         * config/i386/sse.md
3727         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3728         Replace nonimmediate_operand with register_operand.
3729         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3730         Likewise.
3731         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3732         Likewise.
3734 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3736         PR rtl-optimization/87596
3737         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
3738         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
3739         for instructions in FROM..TO range.
3741 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
3743         * cfgexpand.c (expand_one_var): Use specific wording in error message
3744         for non-local frame variables.
3745         * stor-layout.c (layout_decl): Do not issue a warning for them.
3747 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
3749         * haifa-sched.c (priority): Add force_recompute parameter.
3750         (apply_replacement): Call priority () with force_recompute = true.
3751         (restore_pattern): Likewise.
3753 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3755         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
3756         HOST_BITS_PER_WIDE_INT.
3757         (test_vector_ops_duplicate): Likewise.
3759 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3761         PR target/72782
3762         * config/i386/sse.md (VF_AVX512): New.
3763         (avx512bcst): Likewise.
3764         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3765         Likewise.
3766         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3767         Likewise.
3768         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3769         Likewise.
3771 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
3773         * doc/invoke.texi (-dumpversion): Improve grammar.
3774         (-dumpfullversion): Make more consistent with -dumpversion.
3776 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
3778         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
3779         Set JUMP_LABEL to the jump insn.
3780         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
3781         Predict emitted jump and add label to jump insn.
3783 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
3785         PR tree-optimization/87562
3786         * input.c (get_substring_ranges_for_loc): Use
3787         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
3788         getting the linemap for the endpoint.  Verify that it's either
3789         in the same linemap as the start point's spelling location, or
3790         at least in the same file.
3792 2018-10-18  Richard Biener  <rguenther@suse.de>
3794         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
3795         feed width-specific load/store costs through ix86_vec_cost.
3796         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
3797         (k8_cost): Likewise.
3798         (bdver_cost): Likewise.
3799         (znver1_cost): Likewise.
3800         (btver1_cost): Likewise.
3801         (btver2_cost): Likewise.
3803 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3805         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
3806         to simplify subreg of vec_merge.
3808 2018-10-18  Richard Biener  <rguenther@suse.de>
3810         * config/i386/i386.c: Fix costing of vector FMA.
3812 2018-10-18  Richard Biener  <rguenther@suse.de>
3814         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
3815         and argument.
3816         (ix86_builtin_vectorization_cost): For vec_construct properly
3817         cost insertion into SSE regs.
3818         (...): Adjust calls to ix86_vec_cost.
3820 2018-10-18  Richard Biener  <rguenther@suse.de>
3822         PR middle-end/87087
3823         Revert
3824         2018-02-07  Richard Biener  <rguenther@suse.de>
3826         PR tree-optimization/84204
3827         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3828         this place.
3830 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3832         PR target/87537
3833         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
3834         of vec_duplicate.
3835         (test_vector_ops_duplicate): Add test for a scalar subreg of a
3836         VEC_MERGE of a VEC_DUPLICATE.
3838 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3840         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
3841         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
3842         * doc/standards.texi (C Language): Document C2X.
3843         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
3844         (rl78_option_override): Handle "GNU C2X" language name.
3846 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3848         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
3849         Document C17 as published in 2018.
3851 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
3853         PR middle-end/87623
3854         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
3855         bail out if both sides do not have the same storage order.
3857 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3859         * bitmap.c (bitmap_head::dump): New.
3860         * bitmap.h (bitmap_head): Add dump().
3861         * gimple-ssa-evrp-analyze.c
3862         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
3863         (evrp_range_analyzer::set_ssa_range_info): Same.
3864         (evrp_range_analyzer::record_ranges_from_phis): Same.
3865         (evrp_range_analyzer::record_ranges_from_stmt): Same.
3866         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
3867         * gimple-ssa-sprintf.c (get_int_range): Same.
3868         (format_integer): Same.
3869         (sprintf_dom_walker::handle_gimple_call): Same.
3870         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
3871         (ipcp_vr_lattice::top_p): Same.
3872         (ipcp_vr_lattice::bottom_p): Same.
3873         (ipcp_vr_lattice::set_to_bottom): Same.
3874         (ipa_vr_operation_and_type_effects): Same.
3875         (propagate_vr_across_jump_function): Same.
3876         (ipcp_store_vr_results): Same.
3877         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
3878         (ipa_print_node_jump_functions_for_edge): Same.
3879         (ipa_get_value_range): Same.
3880         (ipa_compute_jump_functions_for_edge): Same.
3881         (ipa_write_jump_function): Same.
3882         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
3883         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3884         Same.
3885         * vr-values.c (set_value_range_to_nonnegative): Same.
3886         (set_value_range_to_truthvalue): Same.
3887         (vr_values::get_value_range): Same.
3888         (vr_values::set_defs_to_varying): Same.
3889         (vr_values::update_value_range): Same.
3890         (symbolic_range_based_on_p): Same.
3891         (vr_values::op_with_boolean_value_range_p): Same.
3892         (vr_values::extract_range_for_var_from_comparison_expr): Same.
3893         (vr_values::extract_range_from_ssa_name): Same.
3894         (vr_values::extract_range_from_binary_expr): Same.
3895         (vr_values::extract_range_from_unary_expr): Same.
3896         (vr_values::extract_range_from_cond_expr): Same.
3897         (vr_values::extract_range_from_comparison): Same.
3898         (vr_values::check_for_binary_op_overflow): Same.
3899         (vr_values::extract_range_basic): Same.
3900         (vr_values::extract_range_from_assignment): Same.
3901         (compare_ranges): Same.
3902         (compare_range_with_value): Same.
3903         (vr_values::adjust_range_with_scev): Same.
3904         (vrp_valueize): Same.
3905         (vrp_valueize_1): Same.
3906         (vr_values::get_vr_for_comparison): Same.
3907         (vr_values::compare_name_with_value): Same.
3908         (vr_values::compare_names): Same.
3909         (vr_values::vrp_evaluate_conditional): Same.
3910         (find_case_label_ranges): Same.
3911         (vr_values::vrp_visit_switch_stmt): Same.
3912         (vr_values::extract_range_from_phi_node): Same.
3913         (vr_values::simplify_div_or_mod_using_ranges): Same.
3914         (vr_values::simplify_bit_ops_using_ranges): Same.
3915         (test_for_singularity): Same.
3916         (range_fits_type_p): Same.
3917         (vr_values::simplify_cond_using_ranges_1): Same.
3918         (vr_values::simplify_switch_using_ranges): Same.
3919         (vr_values::simplify_float_conversion_using_ranges): Same.
3920         (vr_values::two_valued_val_range_p): Same.
3921         (vr_values::add_equivalence): Move to value_range::equiv_add.
3922         * vr-values.h (vr_values::add_equivalence): Remove.
3923         (VR_INITIALIZER): Remove.
3924         * tree-vrp.c (value_range::set): New.
3925         (value_range::equiv_add): New.
3926         (value_range::value_range): New.
3927         (value_range::deep_copy): New.
3928         (value_range::check): New.
3929         (value_range::equal_p): New.
3930         (value_range::ignore_equivs_equal_p): New.
3931         (value_range::operator==): New.
3932         (value_range::operator!=): New.
3933         (value_range::symbolic_p): New.
3934         (value_range::numeric_p): New.
3935         (value_range::set_undefined): New.
3936         (value_range::set_varying): New.
3937         (value_range::may_contain_p): New.
3938         (value_range::equiv_clear): New.
3939         (value_range::singleton_p): New.
3940         (value_range::intersect): New.
3941         (value_range::dump): New.
3942         (value_range::set_and_canonicalize): New.
3943         (set_value_range): Adjust for value_range API.
3944         (set_value_range_to_undefined): Same.
3945         (set_value_range_to_varying): Same.
3946         (set_and_canonicalize_value_range): Same.
3947         (set_value_range_to_nonnull): Same.
3948         (set_value_range_to_null): Same.
3949         (range_is_null): Same.
3950         (range_is_nonnull): Same.
3951         (range_int_cst_p): Same.
3952         (range_int_cst_singleton_p): Same.
3953         (symbolic_range_p): Same.
3954         (range_includes_zero_p): Same.
3955         (value_range_constant_singleton): Same.
3956         (vrp_set_zero_nonzero_bits): Same.
3957         (ranges_from_anti_range): Same.
3958         (extract_range_into_wide_ints): Same.
3959         (extract_range_from_multiplicative_op): Same.
3960         (set_value_range_with_overflow): Same.
3961         (extract_range_from_binary_expr_1): Same.
3962         (extract_range_from_unary_expr): Same.
3963         (dump_value_range): Same.
3964         (debug_value_range): Same.
3965         (vrp_prop::check_array_ref): Same.
3966         (vrp_prop::check_mem_ref): Same.
3967         (vrp_prop::vrp_initialize): Same.
3968         (vrp_prop::visit_stmt): Same.
3969         (intersect_ranges): Same.
3970         (vrp_prop::visit_phi): Same.
3971         (vrp_prop::vrp_finalize): Same.
3972         (determine_value_range_1): Same.
3973         (determine_value_range): Same.
3974         (vrp_intersect_ranges_1): Rename to...
3975         (vrp_intersect_1): this.
3976         (vrp_intersect_ranges): Rename to...
3977         (value_range::intersect_helper): ...this.
3978         (vrp_meet_1): Rename to...
3979         (value_range::union_helper): ...this.
3980         (vrp_meet): Rename to...
3981         (value_range::union_): ...this.
3982         (copy_value_range): Remove.
3983         * tree-vrp.h (struct value_range): Rewrite into a proper class.
3984         (value_range::vrtype): New.
3985         (value_range::type): New.
3986         (value_range::equiv): New.
3987         (value_range::min): New.
3988         (value_range::max): New.
3989         (value_range::varying_p): New.
3990         (value_range::undefined_p): New.
3991         (value_range::null_p): New.
3992         (value_range::equiv_add): New.
3993         (copy_value_range): Remove.
3995 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
3997         * Makefile.in (SELFTEST_TARGETS): New.
3998         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
3999         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4000         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4001         c/Make-lang.in.
4002         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4003         (selftest-c++-gdb, selftest-c++-valgrind): Move to
4004         cp/Make-lang.in.
4005         * configure: Regenerate.
4006         * configure.ac (selftest_languages): New.
4008 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4010         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4011         overflow wraps argument.
4012         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4013         wide_int_range_multiplicative_op.
4014         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4015         overflow wraps argument.
4016         (wide_int_range_multiplicative_op): Same.
4017         (wide_int_range_lshift): Same.
4018         (wide_int_range_div): Same.
4019         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4020         (wide_int_range_lshift): Same.
4021         (wide_int_range_div): Same.
4023 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4025         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4026         use sign as argument.
4027         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4028         wide_int_range_shift_undefined_p.
4030 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4032         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4033         Rename to...
4034         (@despeculate_copy<ALLI_TI:mode>): ... This.
4035         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4036         switch statement.
4038 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4040         * config.gcc: Obsolete *-*-solaris2.10*.
4041         * doc/install.texi (Specific, *-*-solaris2*): Document it.
4043 2018-10-12  Jeff Law  <law@redhat.com>
4045         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4046         reg + sym +- const_int addressing modes.
4048 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4050         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4051         * diagnostic-show-locus.c (layout::layout): Apply the minimum
4052         margin width.
4053         (layout::start_annotation_line): Only print up to 3 of the
4054         margin character, to avoid touching the left-hand side.
4055         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4056         minimum margin width, as set by test_diagnostic_context's ctor.
4057         (selftest::test_fixit_insert_containing_newline): Likewise.
4058         (selftest::test_fixit_insert_containing_newline_2): Likewise.
4059         (selftest::test_line_numbers_multiline_range): Clear
4060         dc.min_margin_width.
4061         * diagnostic.c (diagnostic_initialize): Initialize
4062         min_margin_width.
4063         * diagnostic.h (struct diagnostic_context): Add field
4064         "min_margin_width".
4065         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4066         * opts.c (common_handle_option): Handle
4067         OPT_fdiagnostics_minimum_margin_width_.
4068         * selftest-diagnostic.c
4069         (selftest::test_diagnostic_context::test_diagnostic_context):
4070         Initialize min_margin_width to 6.
4071         * toplev.c (general_init): Initialize global_dc->min_margin_width.
4073 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4075         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4076         Fix usage of "error_at_rich_loc" in the comment.
4078 2018-10-15  Renlin Li  <renlin.li@arm.com>
4080         PR target/87563
4081         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4082         if-conversioned loop when it contains ifn with types not
4083         supported by backend.
4084         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4085         (direct_internal_fn_supported_p): New helper function.
4086         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4088 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4090         PR target/87572
4091         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4092         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4093         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4095 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4097         PR tree-optimization/87022
4098         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4099         bits in dist vector rather than the first one.
4101 2018-10-15  Richard Biener  <rguenther@suse.de>
4103         PR middle-end/87610
4104         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4105         (visit_loadstore): When a used restrict tag escaped verify that
4106         the points-to solution of "other" pointers do not include
4107         escaped.
4108         (compute_dependence_clique): If a used restrict tag escaped
4109         communicated that down to visit_loadstore.
4111 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4113         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4114         into reg if it isn't a general operand.
4116 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4118         PR target/87599
4119         * config/i386/sse.md (*vec_dupv2di): Add register source to
4120         movddup.
4122 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4124         PR target/87572
4125         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4126         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4127         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4128         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4130 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4132         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4133         (notice_args_size): Set it in the current trace if no insn that can
4134         throw internally has been seen yet.
4135         (connect_traces): When connecting args_size between traces, allow the
4136         incoming values not to match if there is an insn setting it before the
4137         first insn that can throw internally; in that case, force the creation
4138         of a CFI note on this latter insn.
4140 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4142         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4143         the base class, because using the injected-class-name was not clearly
4144         specified until DR 176.
4146 2018-10-12  Paul Koning  <ni1d@arrl.net>
4148         * config/pdp11/pdp11.md (doloop_end): New expander.
4149         (doloop_end_insn): renamed from "doloop_end".
4150         (addqi3): New pattern.
4151         (subqi3): New pattern.
4152         * config/pdp11/predicates.md (incdec_operand): New predicate.
4154 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4156         PR middle-end/81376
4157         * real.c (format_helper::can_represent_integral_type_p): New function
4158         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4159         * match.pd: New pattern.
4161 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4163         * configure.ac: Introduce --enable-large-address-aware
4164         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4165         * doc/install.texi: Document it.
4166         * configure, config.in: Rebuilt.
4167         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4168         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4169         (LINK_SPEC): Insert it.
4170         * config/i386/mingw-w64.h: Likewise.
4172         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4174 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4176         PR rtl-optimization/87600
4177         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4179 2018-10-12  Paul Koning  <ni1d@arrl.net>
4181         * doc/md.texi (doloop_end): Document that the pattern code may
4182         need to check operand mode.
4184 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4186         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4187         to zero-extend between int and floating-point registers.
4188         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4189         ldp into floating-point registers.  Add type and arch attributes.
4190         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4191         Use f_loads for type attribute.
4193 2018-10-11  Martin Sebor  <msebor@redhat.com>
4195         * doc/extend.texi (attribute packed): Correct typos.
4197 2018-10-11  Martin Sebor  <msebor@redhat.com>
4199         * doc/extend.texi (attribute flatten): Mention interaction with
4200         noinline.
4202 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4204         PR target/87156
4205         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4206         Set new_decl virtual flag to zero.
4208 2018-10-11  Martin Sebor  <msebor@redhat.com>
4210         PR middle-end/87593
4211         * doc/extend.texi (attribute format_arg): Discuss using multiple
4212         attributes on a single function.
4214 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4216         PR tree-optimization/86829
4217         * match.pd (sin (atan (x))): New simplification rules.
4218         (cos (atan (x))): Likewise.
4219         * real.c (build_sinatan_real): New function.
4220         * real.h (build_sinatan_real): Prototype.
4222 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4224         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4225         function.
4226         (fold_mergeeo_helper): New helper function.
4227         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4228         intrinsics.  Correct some whitespace indentation issues.
4230 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4232         PR target/87511
4233         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4234         Use HOST_WIDE_INT_1U for shift.
4236 2018-10-11  Doug Rupp  <rupp@adacore.com>
4237             Olivier Hainque  <hainque@adacore.com>
4239         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4240         Pass --relax to the linker for RTPs.
4241         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4243 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4244             Jan Hubicka  <jh@suse.cz>
4245             Martin Jambor  <mjambor@suse.cz>
4247         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4248         the same elements are repeated rather than printing all of them.
4249         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4250         "repeated" elements.
4251         * read-rtl-function.c (test_loading_repeat): New function.
4252         (read_rtl_function_c_tests): Call test_loading_repeat.
4253         * rtl-tests.c (test_dumping_repeat): New function.
4254         (rtl_tests_c_tests): Call test_dumping_repeat.
4256 2018-10-11  Richard Biener  <rguenther@suse.de>
4258         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4259         bdver?_cost): Unify to ...
4260         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4261         * config/i386/i386.c (processor_cost_table): Adjust.
4263 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4265         PR middle-end/87574
4266         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4267         the thunk when expanding to GIMPLE.
4269 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4271         * varasm.c (mergeable_string_section): Don't try to move zero-length
4272         strings to the merge section.
4274 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4276         PR target/87573
4277         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4279 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4281         PR target/87550
4282         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4283         to special_args set.
4285 2018-10-10  Richard Biener  <rguenther@suse.de>
4287         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4288         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4289         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4290         and recursing and pattern terminating the recursion on SSE
4291         vector width using ix86_expand_reduc.
4292         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4293         width and recursing and SSE2 vector variant doing the final
4294         reduction with ix86_expand_reduc.
4295         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4296         with terminating the recursion at AVX level, splitting that
4297         to SSE there.
4299 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4301         * genmatch.c (error_cb): Rename to...
4302         (diagnostic_cb): ...this, converting int params to enums.
4303         (fatal_at): Update for renaming.
4304         (warning_at): Likewise.
4305         (main): Likewise.
4306         * input.c (selftest::ebcdic_execution_charset::apply):
4307         Update for renaming of...
4308         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4309         to...
4310         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4311         converting level and reason to enums.
4312         (class selftest::lexer_error_sink): Rename to...
4313         (class selftest::lexer_test_options): ...this, renaming field
4314         "m_errors" to "m_diagnostics".
4315         (selftest::lexer_test_options::apply): Update for renaming of...
4316         (selftest::lexer_test_options::on_error): ...this, renaming to...
4317         (selftest::lexer_test_options::on_diagnostic): ...this
4318         converting level and reason to enums.
4319         (selftest::test_lexer_string_locations_raw_string_unterminated):
4320         Update for renamings.
4321         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4322         "reason".
4324 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4326         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4327         * config/rs6000/pmmintrin.h: New file.
4329 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4331         PR tree-optimization/86659
4332         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4334 2018-10-09  Richard Biener  <rguenther@suse.de>
4336         PR tree-optimization/63155
4337         * tree-ssa-structalias.c: Include tree-ssa.h.
4338         (get_constraint_for_ssa_var): For undefs return nothing_id.
4339         (find_func_aliases): Cleanup PHI handling.
4341 2018-10-09  Richard Biener  <rguenther@suse.de>
4343         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4344         replacements.
4346 2018-10-09  Martin Liska  <mliska@suse.cz>
4348         * asan.c (asan_emit_stack_protection): If a stack variable
4349         is located in a same file as current function, then emit
4350         line info into variable definition string.
4352 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4354         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4355         information.
4357 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4359         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4360         on the thunk.
4362 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4364         PR tree-optimization/86659
4365         * gimple-match.h (struct gimple_match_op): Add reverse field.
4366         (gimple_match_op::set_op): New overloaded method.
4367         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4368         the REF_REVERSE_STORAGE_ORDER flag on the value.
4369         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4370         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4371         
4372 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4374         PR middle-end/63155
4375         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4376         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4378 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4380         PR target/87517
4381         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4382         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4384 2018-10-08  Richard Biener  <rguenther@suse.de>
4386         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4387         cost the same as AVX128 ones.
4389 2018-10-08  Paul Koning  <ni1d@arrl.net>
4391         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4392         (expand_block_move): New function.
4393         * config/pdp11/pdp11.c (output_block_move): Remove.
4394         (expand_block_move): New function.
4395         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4396         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4397         (*movmemhi1): Remove.
4399 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4401         * config/s390/2827.md: Increase latencies for some FP instructions.
4403 2018-10-08  Richard Biener  <rguenther@suse.de>
4405         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4406         Open a dump scope.
4407         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4408         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4409         (add_stmt_cost): Dump return value of the hook.
4411 2018-10-08  Richard Biener  <rguenther@suse.de>
4413         PR tree-optimization/63155
4414         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4415         (ssa_propagation_engine::ssa_propagate): Remove redundant
4416         bitmap bit clearing.
4418 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4420         PR rtl-optimization/86939
4421         PR rtl-optimization/87479
4422         * ira.h (non_conflicting_reg_copy_p): New prototype.
4423         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4424         (make_hard_regno_dead): Don't add conflicts for register
4425         ignore_reg_for_conflicts.
4426         (make_object_dead): Likewise.
4427         (non_conflicting_reg_copy_p): New function.
4428         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4429         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4430         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4431         (make_hard_regno_dead): Don't add conflicts for register
4432         ignore_reg_for_conflicts.  Remove special conflict handling of
4433         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4434         check_pic_pseudo_p and update callers.
4435         (mark_pseudo_dead): Don't add conflicts for register
4436         ignore_reg_for_conflicts.
4437         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4439 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4440             Jim Wilson  <jimw@sifive.com>
4442         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4443         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4444         new pattern using HONOR_SNANS that emits one extra instruction.
4446 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4448         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4449         patterns): Merge SI and DI patterns to a GPR pattern.
4450         (unnamed define_insn and define_split for record form of that): Merge
4451         to a single define_insn_and_split pattern.
4453 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4455         PR c++/56856
4456         * input.c (expand_location_to_spelling_point): Add param "aspect"
4457         and use rather than hardcoding LOCATION_ASPECT_CARET.
4458         (get_substring_ranges_for_loc): Handle the case of a single token
4459         within a macro expansion.
4460         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4461         defaulting to LOCATION_ASPECT_CARET.
4463 2018-10-05  Paul Koning  <ni1d@arrl.net>
4465         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4466         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4467         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4468         (pdp11_guard_type): New function.
4470 2018-10-05  Paul Koning  <ni1d@arrl.net>
4472         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4473         * config/pdp11/pdp11.opt (mfloat32): Remove.
4474         (mfloat64): Remove.
4475         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4476         (pdp11 -mfloat64): Remove.
4478 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4480         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4481         (*cmp<mode>_cc_i387): Ditto.
4482         (*cmpu<mode>_cc_i387): Ditto.
4483         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4484         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4485         "scratch" argument.
4486         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4487         Emit x86_sahf_1 pattern.
4488         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4489         (ix86_expand_carry_flag_compare): Ditto.
4491 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4493         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4494         to reg_or_0_operand.  Add "C" constraint.
4495         (*cmpxf_cc_i387): Ditto.
4496         (*cmp<mode>_i387): Change operand 2 predicate
4497         to nonimm_or_0_operand.  Add "C" constraint.
4498         (*cmp<mode>_cc_i387): Ditto.
4499         (*cmp<mode>_0_i387): Remove insn pattern.
4500         (*cmp<mode>_0_cc_i387): Ditto.
4502 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4504         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4505         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4506         from vector_move_operand.  Update all uses.
4508 2018-10-05  Martin Sebor  <msebor@redhat.com>
4510         PR tree-optimization/87490
4511         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4512         consistently.
4514 2018-10-05  Richard Biener  <rguenther@suse.de>
4516         PR tree-optimization/63155
4517         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4518         vertical space in dumpfiles.
4519         * tree-ssa-propagate.h
4520         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4521         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4522         (ssa_edge_worklist_back): Likewise.
4523         (curr_order): Likewise.
4524         (cfg_blocks_get): Remove abstraction.
4525         (cfg_blocks_add): Likewise.
4526         (cfg_blocks_empty_p): Likewise.
4527         (add_ssa_edge): Add to current or next worklist based on
4528         RPO index.
4529         (add_control_edge): Likewise.
4530         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4531         into ...
4532         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4533         iteration from CFG and SSA edge worklist so we process
4534         everything in RPO order, prioritizing forward progress
4535         over iteration.
4536         (ssa_prop_init): Allocate new worklists, do not dump
4537         immediate uses.
4538         (ssa_prop_fini): Free new worklists.
4540 2018-10-05  Richard Biener  <rguenther@suse.de>
4542         * tree-core.h (tree_block::abstract_flag): Remove.
4543         (tree_block::block_num): Make full 32bits.
4544         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4545         * tree.h (BLOCK_ABSTRACT): Remove.
4546         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4547         resulting from BLOCK_ABSTRACT being always false.
4548         (gen_inlined_subroutine_die): Likewise.
4549         (gen_block_die): Likewise.
4550         * tree.c (block_ultimate_origin): Likewise.
4551         * tree-pretty-print.c (dump_block_node): Remove code dealing
4552         with BLOCK_ABSTRACT.
4553         * tree-ssa-live.c (dump_scope_block): Likewise.
4554         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4555         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4557 2018-10-05   Richard Biener  <rguenther@suse.de>
4559         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4560         is asked for initialize mode to the component mode of the
4561         vector type.
4563 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4565         PR target/87522
4566         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4567         assembler for -mavx.
4568         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4570 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4572         PR target/87509
4573         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4574         RS6000_BTM_DFP.
4575         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4576         to be DImode.  When using mffscrn, force the operand to a register.
4578 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
4580         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
4581         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
4582         X87MODEF mode iterator.
4583         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
4584         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
4585         X87MODEF mode iterator.
4587 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
4589         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
4590         -Wno-prio-ctor-dtor.
4592 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4594         * Makefile.in (OBJS): Add opt-problem.o.
4595         * dump-context.h: Include "selftest.h.
4596         (selftest::temp_dump_context): New forward decl.
4597         (class dump_context): Make friend of class
4598         selftest::temp_dump_context.
4599         (dump_context::dump_loc_immediate): New decl.
4600         (class dump_pretty_printer): Move here from dumpfile.c.
4601         (class temp_dump_context): Move to namespace selftest.
4602         (temp_dump_context::temp_dump_context): Add param
4603         "forcibly_enable_dumping".
4604         (selftest::verify_dumped_text):
4605         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
4606         (selftest::verify_item):
4607         (ASSERT_IS_TEXT): Move here from dumpfile.c.
4608         (ASSERT_IS_TREE): Likewise.
4609         (ASSERT_IS_GIMPLE): Likewise.
4610         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
4611         to...
4612         (dump_context::dump_loc_immediate): ...this new function.
4613         (class dump_pretty_printer): Move to dump-context.h.
4614         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
4615         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
4616         (temp_dump_context::temp_dump_context): Move to "selftest"
4617         namespace.  Add param "forcibly_enable_dumping", and use it to
4618         conditionalize the use of m_pp;
4619         (selftest::verify_dumped_text): Make non-static.
4620         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
4621         (selftest::verify_item): Make non-static.
4622         (ASSERT_IS_TEXT): Move to dump-context.h.
4623         (ASSERT_IS_TREE): Likewise.
4624         (ASSERT_IS_GIMPLE): Likewise.
4625         (selftest::test_capture_of_dump_calls): Pass "true" for new
4626         param of temp_dump_context.
4627         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
4628         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
4629         TDF_COMPARE_DEBUG.
4630         * opt-problem.cc: New file.
4631         * opt-problem.h: New file.
4632         * optinfo-emit-json.cc
4633         (selftest::test_building_json_from_dump_calls): Pass "true" for
4634         new param of temp_dump_context.
4635         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
4636         (optinfo::emit_for_opt_problem): New function.
4637         (optinfo::emit): Clarity which emit_item is used.
4638         * optinfo.h (optinfo::get_dump_location): New accessor.
4639         (optinfo::emit_for_opt_problem): New decl.
4640         (optinfo::emit): Make const.
4641         * selftest-run-tests.c (selftest::run_tests): Call
4642         selftest::opt_problem_cc_tests.
4643         * selftest.h (selftest::opt_problem_cc_tests): New decl.
4644         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
4645         bool to opt_result, converting fprintf messages to
4646         opt_result::failure_at calls.  Add "stmt" param for use by the
4647         failure_at calls.
4648         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
4649         (runtime_alias_check_p): Convert return type from bool to
4650         opt_result, converting dump_printf calls to
4651         opt_result::failure_at, using the statement DDR_A for their
4652         location.
4653         (find_data_references_in_stmt): Convert return type from bool to
4654         opt_result, converting "return false" to opt_result::failure_at
4655         with a new message.
4656         * tree-data-ref.h: Include "opt-problem.h".
4657         (dr_analyze_innermost): Convert return type from bool to opt_result,
4658         and add a const gimple * param.
4659         (find_data_references_in_stmt): Convert return type from bool to
4660         opt_result.
4661         (runtime_alias_check_p): Likewise.
4662         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
4663         dr_analyze_innermost.
4664         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
4665         Convert return type from bool to opt_result, adding a message for
4666         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
4667         (vect_analyze_data_ref_dependence): Convert return type from bool
4668         to opt_result.  Change sense of return type from "false"
4669         effectively meaning "no problems" to "false" meaning a problem,
4670         so that "return false" becomes "return opt_result::success".
4671         Convert "return true" calls to opt_result::failure_at, using
4672         the location of statement A rather than vect_location.
4673         (vect_analyze_data_ref_dependences): Convert return type from bool
4674         to opt_result.
4675         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
4676         calls to opt_result::failure_at, using the stmt location rather
4677         than vect_location.
4678         (vect_verify_datarefs_alignment): Convert return type from bool
4679         to opt_result.
4680         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
4681         into multiple more-tightly-scoped copies.
4682         (vect_analyze_data_refs_alignment): Convert return type from bool
4683         to opt_result.
4684         (vect_analyze_data_ref_accesses): Likewise, converting a
4685         "return false" to a "return opt_result::failure_at", adding a
4686         new message.
4687         (vect_prune_runtime_alias_test_list): Convert return type from
4688         bool to opt_result, converting dump_printf_loc to
4689         opt_result::failure_at.  Add a %G to show the pertinent statement,
4690         and use the stmt's location rather than vect_location.
4691         (vect_find_stmt_data_reference): Convert return type from
4692         bool to opt_result, converting dump_printf_loc to
4693         opt_result::failure_at, using stmt's location.
4694         (vect_analyze_data_refs):  Convert return type from bool to
4695         opt_result.  Convert "return false" to "return
4696         opt_result::failure_at", adding messages as needed.
4697         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
4698         type from bool to opt_result.
4699         (vect_determine_vf_for_stmt): Likewise.
4700         (vect_determine_vectorization_factor): Likewise, converting
4701         dump_printf_loc to opt_result::failure_at, using location of phi
4702         rather than vect_location.
4703         (vect_analyze_loop_form_1): Convert return type from bool to
4704         opt_result, converting dump_printf_loc calls, retaining the use of
4705         vect_location.
4706         (vect_analyze_loop_form): Convert return type from loop_vec_info
4707         to opt_loop_vec_info.
4708         (vect_analyze_loop_operations): Convert return type from bool to
4709         opt_result, converting dump_printf_loc calls, using the location
4710         of phi/stmt rather than vect_location where available.  Convert
4711         various "return false" to "return opt_result::failure_at" with
4712         "unsupported phi" messages.
4713         (vect_get_datarefs_in_loop): Convert return type from bool to
4714         opt_result.  Add a message for the
4715         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
4716         (vect_analyze_loop_2): Convert return type from bool to
4717         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
4718         each "goto again;", adding new messages where needed.
4719         Add "unsupported grouped {store|load}" messages.
4720         (vect_analyze_loop): Convert return type from loop_vec_info to
4721         opt_loop_vec_info.
4722         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
4723         bool to opt_result.
4724         * tree-vect-stmts.c (process_use): Likewise, converting
4725         dump_printf_loc call and using stmt location, rather than
4726         vect_location.
4727         (vect_mark_stmts_to_be_vectorized): Likeise.
4728         (vect_analyze_stmt): Likewise, adding a %G.
4729         (vect_get_vector_types_for_stmt): Convert return type from bool to
4730         opt_result, converting dump_printf_loc calls and using stmt
4731         location, rather than vect_location.
4732         (vect_get_mask_type_for_stmt): Convert return type from tree to
4733         opt_tree, converting dump_printf_loc calls and using stmt location.
4734         * tree-vectorizer.c: Include "opt-problem.h.
4735         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
4736         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
4737         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
4738         enabled, use it to report at the top level "couldn't vectorize
4739         loop" followed by the problem.
4740         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
4741         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
4742         to opt_result.
4743         (vect_analyze_stmt): Likewise.
4744         (vect_get_vector_types_for_stmt): Likewise.
4745         (tree vect_get_mask_type_for_stmt): Likewise.
4746         (vect_analyze_data_ref_dependences): Likewise.
4747         (vect_enhance_data_refs_alignment): Likewise.
4748         (vect_analyze_data_refs_alignment): Likewise.
4749         (vect_verify_datarefs_alignment): Likewise.
4750         (vect_analyze_data_ref_accesses): Likewise.
4751         (vect_prune_runtime_alias_test_list): Likewise.
4752         (vect_find_stmt_data_reference): Likewise.
4753         (vect_analyze_data_refs): Likewise.
4754         (vect_analyze_loop): Convert return type from loop_vec_info to
4755         opt_loop_vec_info.
4756         (vect_analyze_loop_form): Likewise.
4757         (vect_analyze_slp): Convert return type from bool to opt_result.
4759 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4761         * doc/invoke.texi (-fopt-info): Document new "internals"
4762         sub-option.
4763         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
4764         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
4765         MSG_ALL_KINDS.
4766         (optinfo_verbosity_options): Add "internals".
4767         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
4768         (dump_context::apply_dump_filter_p): New member function.
4769         (dump_context::dump_loc): Use apply_dump_filter_p rather than
4770         explicitly masking the dump_kind.
4771         (dump_context::begin_scope): Increment the scope depth first.  Use
4772         apply_dump_filter_p rather than explicitly masking the dump_kind.
4773         (dump_context::emit_item): Use apply_dump_filter_p rather than
4774         explicitly masking the dump_kind.
4775         (dump_dec): Likewise.
4776         (dump_hex): Likewise.
4777         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
4778         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
4779         (opt_info_switch_p): Update handling of default
4780         MSG_OPTIMIZED_LOCATIONS to cope with default of
4781         MSG_PRIORITY_USER_FACING.
4782         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
4783         masking the dump_kind.
4784         (selftest::test_capture_of_dump_calls): Update test_dump_context
4785         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
4786         than MSG_ALL.  Generalize scope test to be run at all four
4787         combinations of with/without MSG_PRIORITY_USER_FACING and
4788         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
4789         for each of the two values.
4790         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
4791         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
4792         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
4793         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
4794         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
4795         with MSG_PRIORITY_*.
4796         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
4797         dump messages as MSG_PRIORITY_USER_FACING.
4798         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
4799         about the interaction with MSG_PRIORITY_*.
4801 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4803         * varasm.c (output_constant): Add new parameter merge_strings.
4804         Make strings properly zero terminated in merge string sections.
4805         (mergeable_string_section): Don't fail if the last char is non-zero.
4806         (assemble_variable_contents): Handle merge string sections.
4807         (assemble_variable): Likewise.
4808         (assemble_constant_contents): Likewise.
4809         (output_constant_def_contents): Likewise.
4810         (output_constructor_array_range,
4811         output_constructor_regular_field): Adjust call to output_constant.
4812         (output_object_block): Adjust call to assemble_constant_contents
4813         and assemble_variable_contents.
4815 2018-10-04  Martin Liska  <mliska@suse.cz>
4817         PR c/87483
4818         * cgraphunit.c (process_function_and_variable_attributes):
4819         Warn about a function with alias attribute and a body.
4821 2018-10-04  Martin Liska  <mliska@suse.cz>
4823         PR ipa/82625
4824         * multiple_target.c (redirect_to_specific_clone): New function.
4825         (ipa_target_clone): Use it.
4826         * tree-inline.c: Fix comment.
4828 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4830         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
4831         fields.
4832         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
4833         (gcc::dump_manager::register_pass): New member function, adapted
4834         from loop body in gcc::pass_manager::register_pass, adding a
4835         call to update_dfi_for_opt_info.
4836         (gcc::dump_manager::opt_info_enable_passes): Store the
4837         -fopt-info options into the new fields.  Move the loop
4838         bodies into...
4839         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
4840         function.
4841         * dumpfile.h (struct opt_pass): New forward decl.
4842         (gcc::dump_manager::register_pass): New decl.
4843         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
4844         (class gcc::dump_manager): Add fields "m_optgroup_flags",
4845         "m_optinfo_flags", and "m_optinfo_filename".
4846         * passes.c (gcc::pass_manager::register_pass): Move all of the
4847         dump-handling code to gcc::dump_manager::register_pass.
4849 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
4851         PR rtl-optimization/87466
4852         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
4853         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
4854         * doc/tm.texi: Regenerate.
4855         * ira-lives.c (process_bb_node_lives): Use the new target hook.
4856         * lra-lives.c (process_bb_lives): Likewise.
4857         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
4858         Define.
4860 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
4862         * params.c (add_params): Fix initialization.
4864 2018-10-04  Martin Liska  <mliska@suse.cz>
4866         PR gcov-profile/84107
4867         * tree-profile.c (init_ic_make_global_vars):
4868         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
4869         Come up with new ic_tuple* variables.  Emit
4870         __gcov_indirect_call{,_topn} variables.
4871         (gimple_gen_ic_profiler): Access the variable
4872         and emit gimple.
4873         (gimple_gen_ic_func_profiler): Access
4874         __gcov_indirect_call.callee field.
4875         (gimple_init_gcov_profiler): Use ptr_type_node.
4876         * value-prof.c (gimple_ic): Use ptr_type_node.
4878 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4880         PR tree-optimization/85787
4881         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
4882         into this function and add support for detecting multiple phis.
4883         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
4885 2018-10-04  Martin Liska  <mliska@suse.cz>
4887         PR ipa/87491
4888         * ipa-inline.c (inline_to_all_callers_1):
4889         Call ultimate_alias_target for node being inlined.
4891 2018-10-03  Jeff Law  <law@redhat.com>
4893         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
4894         target's wchar_t.
4895         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
4896         * tree.h (get_typenode_from_name): Prototype.
4898 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4900         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
4901         Change operand 2 predicate to nonimmediate_operand.
4902         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4904 2018-10-03  Martin Sebor  <msebor@redhat.com>
4905             Jeff Law  <law@redhat.com>
4907         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
4908         initialize it.
4909         (get_string_length): Detect unterminated arrays.
4910         (format_string): Same.
4911         (format_directive): Warn about unterminated arrays.
4912         (handle_gimple_call): Mark statements with no_warning as needed.
4914 2018-10-03  Jim Wilson  <jimw@sifive.com>
4916         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
4917         also define __riscv_abi_rve.  Delete trailing white space.
4919 2018-10-03  Paul Koning  <ni1d@arrl.net>
4921         Enable LRA register allocator for PDP11.
4922         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
4923         (R): Likewise.
4924         (D): Likewise.
4925         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
4926         * config/pdp11/pdp11.opt (-mlra): New option.
4927         * doc/invoke.texi (PDP-11 Options): Document -mlra.
4929 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4931         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
4932         (*<absneg:code>extend<mode>xf2): Ditto.
4934 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4936         PR tree-optimization/87415
4937         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
4938         precision fields.
4940 2018-10-02  Jeff Law  <law@redhat.com>
4942         * gimple-fold.c (get_range_strlen): Only set *nonstr when
4943         an unterminated string is discovered.  Bubble up range
4944         even for unterminated strings.
4945         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
4946         indicates the string was not terminated via NONSTR.
4948 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4950         * tree-vrp.c (extract_range_from_unary_expr): Special case all
4951         pointer conversions.
4952         Do not do anything special for anti-ranges.
4954 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
4956         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
4957         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
4959 2018-10-03  Martin Liska  <mliska@suse.cz>
4961         PR gcov-profile/86109
4962         * coverage.c (coverage_begin_function): Do not
4963         mark lambdas as artificial.
4964         * tree-core.h (struct GTY): Remove tm_clone_flag
4965         and introduce new lambda_function.
4966         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
4968 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4970         PR target/87474
4971         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
4972         P8_VECTOR and VSX are enabled.
4974 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4976         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
4977         0x3907 as CPU model number.
4979 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4981         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
4982         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
4983         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
4984         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
4985         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
4986         * config/s390/s390.md: Likewise. Rename also the cpu attribute
4987         value from arch12 to z14.
4989 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4991         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
4992         (isinfxf2): Ditto.
4993         (isinf<mode>2): Ditto.
4995 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4997         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
4998         before emitting fxam.  Perform calculations in XFmode.
5000 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5002         * match.pd (((X /[ex] A) +- B) * A): New transformation.
5004 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5006         PR middle-end/87319
5007         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5008         * tree.c (signed_or_unsigned_type_for): Handle complex.
5010 2018-10-02  Jeff Law  <law@redhat.com>
5012         * gimple-fold.c (get_range_strlen): Remove dead code.
5014 2018-10-02  Martin Sebor  <msebor@redhat.com>
5015             Jeff Law  <law@redhat.com>
5017         * builtins.c (unterminated_array): Add new arguments.
5018         If argument is not terminated, bubble up size and exact
5019         state to callers.
5020         (expand_builtin_strnlen): Detect, avoid expanding
5021         and diagnose unterminated arrays.
5022         (c_strlen): Fill in offset of start of unterminated strings.
5023         * builtins.h (unterminated_array): Update prototype.
5025 2018-10-02  Richard Biener  <rguenther@suse.de>
5027         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5028         of haddv4df, first reduce to SSE width and exploit the fact
5029         that we only need element zero with the reduction result.
5030         (reduc_plus_scal_v2df): Likewise.
5032 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5034         * dojump.h (do_jump): Delete.
5035         (do_jump_1): Likewise.
5036         (split_comparison): Move around.
5037         * dojump.c (do_jump): Make static.
5038         (do_jump_1): Likewise.
5039         (jumpifnot): Move around.
5040         (jumpifnot_1): Likewise.
5041         (jumpif): Likewise.
5042         (jumpif_1): Likewise.
5043         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5045 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5047         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5048         insns in the delay slot and add_insn_after for the jump insn.
5050 2018-10-02  Richard Biener  <rguenther@suse.de>
5052         * tree-inline.c (expand_call_inline): Use the location of
5053         the callee declaration for the inline-entry marker.
5054         * final.c (notice_source_line): Remove special-casing of
5055         NOTE_INSN_INLINE_ENTRY.
5057 2018-10-01  Carl Love  <cel@us.ibm.com>
5059         PR 69431
5060         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5061         (__builtin_mtfsb0): New.
5062         (__builtin_mtfsb1): New.
5063         ( __builtin_set_fpscr_rn): New.
5064         (__builtin_set_fpscr_drn): New.
5065         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5066         (rs6000_expand_set_fpscr_rn_builtin): Add.
5067         (rs6000_expand_set_fpscr_drn_builtin): Add.
5068         (rs6000_expand_builtin): Add case statement entries for
5069         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5070         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5071         RS6000_BUILTIN_MFFSL.
5072         (rs6000_init_builtins): Add ftype initialization and def_builtin
5073         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5074         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5075         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5076         rs6000_mffscdrn): Add define_insn.
5077         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5078         * doc/extend.texi: Add documentation for the builtins.
5080 2018-10-01  Richard Biener  <rguenther@suse.de>
5082         PR tree-optimization/87465
5083         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5084         causing branch miscounts.
5086 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5088         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5089         aarch64_option_default_param):  New.
5090         (params.h): Include.
5091         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5092         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5093         stack-clash protection validation code.
5095 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5097         * params.c (validate_param): New.
5098         (add_params): Use it.
5099         (set_param_value): Refactor param validation into validate_param.
5100         (diagnostic.h): Include.
5101         * diagnostic.h (diagnostic_ready_p): New.
5103 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5105         * params.c (set_param_value):
5106         Add index of parameter being validated.
5107         * common/common-target.def (option_validate_param): New.
5108         * common/common-targhooks.h (default_option_validate_param): New.
5109         * common/common-targhooks.c (default_option_validate_param): New.
5110         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5111         * doc/tm.texi: Regenerate.
5113 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5115         PR target/86486
5116         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5117         Add validation for stack-clash parameters and set defaults.
5119 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5121         PR target/86486
5122         * configure.ac: Add stack-clash-protection-guard-size.
5123         * doc/install.texi: Document it.
5124         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5125         * params.def: Update comment for guard-size.
5126         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5127         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5128         * configure: Regenerate.
5130 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5132         PR target/86486
5133         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5134         STACK_DYNAMIC_OFFSET): New.
5135         * config/aarch64/aarch64.c (aarch64_layout_frame):
5136         Update outgoing args size.
5137         (aarch64_stack_clash_protection_alloca_probe_range,
5138         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5140 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5142         PR target/86486
5143         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5144         probe ranges.
5145         * target.def (stack_clash_protection_alloca_probe_range): New.
5146         (stack_clash_protection_final_dynamic_probe): Remove.
5147         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5148         (default_stack_clash_protection_final_dynamic_probe): Remove.
5149         * targhooks.c: Likewise.
5150         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5151         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5152         * doc/tm.texi: Regenerate.
5154 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5156         PR target/86486
5157         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5158         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5159         aarch64_clamp_to_uimm12_shift): New.
5160         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5161         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5163 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5165         PR target/86486
5166         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5168 2018-10-01  Jeff Law  <law@redhat.com>
5169             Richard Sandiford <richard.sandiford@linaro.org>
5170             Tamar Christina  <tamar.christina@arm.com>
5172         PR target/86486
5173         * config/aarch64/aarch64.md
5174         (probe_stack_range): Add k (SP) constraint.
5175         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5176         STACK_CLASH_MAX_UNROLL_PAGES): New.
5177         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5178         stack probes for stack clash.
5179         (aarch64_allocate_and_probe_stack_space): New.
5180         (aarch64_expand_prologue): Use it.
5181         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5182         (aarch64_sub_sp): Add emit_move_imm optional param.
5184 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5186         PR tree-optimization/87261
5187         * match.pd: Remove trailing whitespace.
5188         Add (x & y) | ~(x | y) -> ~(x ^ y),
5189         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5191 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5193         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5194         constraints.
5195         (ashlsi3_insn): Update instruction constraints.
5196         (ashrsi3_insn): Likewise.
5197         (rotrsi3): Likewise.
5198         (add_shift): Likewise.
5199         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5200         avoids placing in the limm field small constants which, otherwise,
5201         could end into a small instruction.
5203 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5205         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5206         destination register is not odd-even.
5207         (umaddsidi4_split): Likewise.
5209 2018-10-01  Richard Biener  <rguenther@suse.de>
5211         * tree-inline.c (expand_call_inline): Store origin of fn
5212         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5213         * tree.c (block_ultimate_origin): Simplify and do some
5214         checking.
5216 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5218         * config/i386/mmx.md (EMMS): New int iterator.
5219         (emms): New int attribute.
5220         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5221         EMMS int iterator.  Explicitly declare clobbers.
5222         (mmx_emms): Remove expander.
5223         (mmx_femms): Ditto.
5224         * config/i386/predicates.md (emms_operation): Remove predicate.
5225         (vzeroall_pattern): New predicate.
5226         (vzeroupper_pattern): Rename from vzeroupper_operation.
5227         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5228         vzeroupper_pattern and vzeroall_pattern predicates.
5230 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5232         PR rtl-optimization/86939
5233         * ira-lives.c (make_hard_regno_born): Rename from this...
5234         (make_hard_regno_live): ... to this.  Remove update to conflict
5235         information.  Update function comment.
5236         (make_hard_regno_dead): Add conflict information update.  Update
5237         function comment.
5238         (make_object_born): Rename from this...
5239         (make_object_live): ... to this.  Remove update to conflict information.
5240         Update function comment.
5241         (make_object_dead):  Add conflict information update.  Update function
5242         comment.
5243         (mark_pseudo_regno_live): Call make_object_live.
5244         (mark_pseudo_regno_subword_live): Likewise.
5245         (mark_hard_reg_dead): Update function comment.
5246         (mark_hard_reg_live): Call make_hard_regno_live.
5247         (process_bb_node_lives): Likewise.
5248         * lra-lives.c (make_hard_regno_born): Rename from this...
5249         (make_hard_regno_live): ... to this.  Remove update to conflict
5250         information.  Remove now uneeded check_pic_pseudo_p argument.
5251         Update function comment.
5252         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5253         to conflict information.  Update function comment.
5254         (mark_pseudo_live): Remove update to conflict information.  Update
5255         function comment.
5256         (mark_pseudo_dead): Add conflict information update.
5257         (mark_regno_live): Call make_hard_regno_live.
5258         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5259         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5261 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5263         PR target/87370
5264         * config/i386/i386.c (construct_container): Use TImode for
5265         BLKmode values in 2 integer registers.
5267 2018-09-29  Jeff Law  <law@redhat.com>
5269         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5270         c_strlen rather than just a tree *.
5271         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5272         Update recursive calls appropriately.  If caller did not provide a
5273         suitable data pointer, create a local one.  When a non-terminated
5274         string is discovered, bubble up information about the string via the
5275         c_strlen_data object.
5276         * builtins.h (c_strlen): Update prototype.
5277         (c_strlen_data): New structure.
5278         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5279         For a type 2 call, if c_strlen indicates a non-terminated string
5280         use the length of the non-terminated string.
5281         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5283 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5285         PR target/87467
5286         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5287         __m512d type for __A argument rather than __m512.
5289 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5291         * match.pd (simple_comparison): Don't optimize if either operand is
5292         a function pointer when target needs function pointer canonicalization.
5294 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5296         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5297         power5 .. power9 to remove indirection.
5298         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5299         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5300         ASM_CPU_476_SPEC): Delete.
5301         (ASM_CPU_SPEC): Adjust.
5302         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5303         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5305 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5307         * config.in: Delete HAVE_AS_DCI.
5308         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5309         * config/rs6000/rs6000.h: Ditto.
5310         * configure.ac: Delete HAVE_AS_DCI.
5311         * configure: Regenerate.
5313 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5315         * config.in (HAVE_AS_LWSYNC): Delete.
5316         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5317         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5318         do it as a .long .
5319         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5320         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5321         as a .long .
5322         * configure.ac: Delete HAVE_AS_LWSYNC.
5323         * configure: Regenerate.
5325 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5326             Pierre-Marie de Rodat  <derodat@adacore.com>
5328         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5329         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5330         (cgraph_node::create_thunk): Add indirect_offset parameter.
5331         (thunk_adjust): Likewise.
5332         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5333         and initialize the corresponding field with it.
5334         (cgraph_node::dump): Dump indirect_offset field.
5335         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5336         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5337         (thunk_adjust): Add indirect_offset parameter and deal with it.
5338         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5339         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5340         or if the thunk is external or local.  Fix formatting.  Do not chain
5341         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5342         if any, in the GIMPLE representation.
5343         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5344         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5345         (input_node): Read indirect_offset field.
5346         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5347         call to thunk_adjust.
5348         * tree-nested.c (struct nesting_info): Add thunk_p field.
5349         (create_nesting_tree): Set it.
5350         (convert_all_function_calls): Copy static chain from targets to thunks.
5351         (finalize_nesting_tree_1): Return early for thunks.
5352         (unnest_nesting_tree_1): Do not finalize thunks.
5353         (gimplify_all_functions): Do not gimplify thunks.
5355 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5357         * opt-suggestions.c (option_proposer::build_option_suggestions):
5358         Release "option_values".
5360 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5362         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5363         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5364         * dumpfile.c (kind_as_string): New function.
5365         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5366         kind_as_string to vary the prefix based on dump_kind.
5367         (selftest::test_capture_of_dump_calls): Update for above.
5369 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5371         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5372         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5374 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5376         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5377         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5378         INVALID_REGNUM instead of FPSR_REG.
5379         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5380         * config/i386/i386.md: Update comment of FP compares.
5381         (fldenv): Do not clobber FPSR_REG.
5383 2018-09-28  Richard Biener  <rguenther@suse.de>
5385         * tree.h (BLOCK_ORIGIN): New.
5386         * omp-expand.c (grid_expand_target_grid_body): Assign
5387         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5388         * tree-inline.c (remap_block): Likewise.
5389         * auto-profile.c (get_function_decl_from_block): Simplify
5390         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5391         * langhooks.c (lhd_print_error_function): Likewise.
5392         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5393         Likewise.
5394         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5395         * tree.c (block_nonartificial_location): Likewise.
5396         (block_ultimate_origin): Likewise.
5397         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5398         no longer needed LTO case.
5400 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5401             Jan Hubicka  <jh@suse.cz>
5402             Martin Jambor  <mjambor@suse.cz>
5404         * simplify-rtx.c (simplify_merge_mask): New function.
5405         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5406         same masks are used in op1 or op2.
5407         (test_vec_merge): New function.
5408         (test_vector_ops): Call test_vec_merge.
5410 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5412         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5413         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5414         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5415         (cypress_costs): Set it.
5416         (supersparc_costs): Likewise.
5417         (hypersparc_costs): Likewise.
5418         (leon_cost): Likewise.
5419         (leon3_costs): Likewise.
5420         (sparclet_costs): Likewise.
5421         (ultrasparc_costs): Likewise.
5422         (ultrasparc_costs): Likewise.
5423         (niagara_costs): Likewise.
5424         (niagara2_costs): Likewise.
5425         (niagara3_costs): Likewise.
5426         (niagara4_costs): Likewise.
5427         (niagara7_costs): Likewise.
5428         (m8_costs): Likewise.
5429         (TARGET_CAN_FOLLOW_JUMP): Define.
5430         (pass_work_around_errata::gate): Minor tweak.
5431         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5432         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5433         Automaitcally clear MASK_FSMULD mask for V7 processors.
5434         (sparc_can_follow_jump): New static function.
5435         (output_ubranch): Deal with CROSSING_JUMP_P.
5436         (sparc_use_sched_lookahead): Rewrite using switch statement.
5437         (sparc_issue_rate): Reorder.
5438         (sparc_branch_cost): New function.
5440 2018-09-27  Martin Sebor  <msebor@redhat.com>
5442         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5443         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5444         (int_fits_type_p): Same.
5446 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5448         * config/i386/i386.md (FPCR_REG): Remove.
5449         (UNSPEC_FLDCW): Remove.
5450         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5451         (x86_fldcw_1): Remove insn pattern.
5452         (fnstenv): Do not clobber FPCR_REG.
5453         (fldenv): Ditto.
5454         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5455         (CALL_USED_REGISTERS): Ditto.
5456         (REG_ALLOC_ORDER): Ditto.
5457         (REG_CLASS_CONTENTS): Ditto.
5458         (HI_REGISTER_NAMES): Ditto.
5459         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5460         of numerical constants.
5461         * config/i386/i386.c (regclass_map): Remove fpsr register.
5462         (dbx_register_map): Ditto.
5463         (dbx64_register_map): Ditto.
5464         (svr4_dbx_register_map): Ditto.
5465         (print_reg): Do not handle FPCR_REG.
5467 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5469         PR target/87149
5470         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5471         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5472         Delete, always treat as true.
5473         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5474         Ditto.  Simplify remaining code.
5475         * config/powerpcspe/powerpcspe.h: Ditto.
5476         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5477         Simplify remaining code.
5478         (rs6000_expand_builtin): Ditto.
5479         * config/rs6000/rs6000.h: Ditto.
5480         * configure.ac: Ditto.
5481         * configure: Regenerate.
5483 2018-09-27  Martin Liska  <mliska@suse.cz>
5485         * coverage.c (get_coverage_counts): Revert the formatting
5486         of missing profile opt info.
5488 2018-09-27  Richard Biener  <rguenther@suse.de>
5490         PR debug/37801
5491         PR debug/87440
5492         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5493         block as we do not output that.
5494         (gen_inlined_subroutine_die): Elide the originally outermost
5495         block, matching what we do for concrete instances.
5496         (decls_for_scope): Add parameter specifying whether to recurse
5497         to subblocks.
5499 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5500             Tom de Vries  <tom@codesourcery.com>
5502         PR 82089
5504         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5505         STORE_FLAG_VALUE == 1.
5507 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5509         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5510         constant definitions.
5511         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5512         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5513         ("speculation_barrier"): New expander definition.
5515 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5517         PR gcov-profile/86957
5518         * common.opt: New warning option -Wmissing-profile.
5519         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5520         * doc/invoke.texi: Document -Wmissing-profile.
5522 2018-09-26  Jim Wilson  <jimw@sifive.com>
5524         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5525         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5527 2018-09-26  Martin Sebor  <msebor@redhat.com>
5529         * tree.c (zerop): Change return type to bool.
5530         (integer_zerop, integer_onep, integer_each_onep): Same.
5531         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5532         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5533         (real_onep, real_minus_onep, chain_index): Same.
5534         (print_type_hash_statistics, type_list_equal): Same.
5535         * tree.h (zerop): Same.
5536         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5537         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5538         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5539         (real_onep, real_minus_onep, chain_index): Same.
5540         (print_type_hash_statistics, type_list_equal): Same.
5542 2018-09-26  Jim Wilson  <jimw@sifive.com>
5544         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5546 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5548         PR target/87414
5549         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5550         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5551         call.
5553 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5555         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5557 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5559         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5560         and FP_SECOND_SSE_REGS.
5561         (REG_CLASS_NAMES): Ditto.
5562         (REG_CLASS_CONTENTS): Ditto.
5563         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5564         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5565         (ix86_preferred_output_reload_class): Ditto.
5566         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5567         clobber constraint to "=&f".
5568         (fix_truncdi_i387): Ditto.
5569         (lrintxfdi2): Ditto.
5570         (fistdi2_<rounding>): Ditto.
5571         (fpremxf4_i387): Change "=u" constraint to "=f".
5572         (fprem1xf4_i387): Ditto.
5573         (sincosxf3): Ditto.
5574         (fptanxf4_i387): Ditto.
5575         (fxtractxf3_i387): Ditto.
5576         (fscalexf4_i387): Ditto.
5577         (atan2xf3): Change "u" constraint to "f".
5578         (fyl2xxf3_i387): Ditto.
5579         (fyl2xp1xf3_i387): Ditto.
5581 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5583         PR target/87439
5584         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5585         for removed I387_MASK_PM entity.
5588 2018-09-26  Jeff Law  <law@redhat.com>
5589         Revert
5590         2018-09-26  Alexey Neyman  <stilor@att.net>
5592         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5593         headers are no longer pulled in by <isl/val.h>.
5595 2018-09-26  Richard Biener  <rguenther@suse.de>
5597         PR debug/87443
5598         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
5599         or concrete instance DIE to the tree.  Create abstract origin
5600         attributes also for concrete instances.
5602 2018-09-26  Alexey Neyman  <stilor@att.net>
5604         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5605         headers are no longer pulled in by <isl/val.h>.
5607 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
5609         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
5610         Use new helper functions.
5611         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
5612         Use new helper functions.
5613         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
5614         aarch_mm_needs_release): New declarations.
5615         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
5616         aarch_mm_needs_release): New.
5618 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
5620         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
5621         (arm32_output_mi_thunk): Deal with long calls.
5623 2018-09-26  Richard Biener  <rguenther@suse.de>
5625         PR debug/87428
5626         PR debug/87362
5627         * tree-inline.c (expand_call_inline): When the location
5628         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
5629         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
5630         the inserted BLOCK to make inlined_function_outer_scope_p
5631         recognize it.
5632         * dwarf2out.c (add_call_src_coords_attributes): Do not add
5633         coords for reserved locations.
5635 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5637         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
5638         (*call_indirect_nonlocal_sysv<mode>): Ditto.
5639         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
5640         (*sibcall_nonlocal_sysv<mode>): Ditto.
5641         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
5642         (<bd>_<mode>): Ditto.
5643         (<bd>tf_<mode>): Ditto.
5645 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5647         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
5648         control string as a list of templates instead of as C code.
5649         (*altivec_movti): Ditto.
5650         * config/rs6000/darwin.md (movdf_low_di): Ditto.
5652 2018-09-25  Jim Wilson  <jimw@sifive.com>
5654         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
5655         when target symbol is weak.
5657 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5659         PR c/87387
5660         * builtins.c (unterminated_array): Simplify.
5661         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
5662         where pointer arithmetic is safe.
5664 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5666         PR target/86987
5667         * config/rs6000/altivec.md (altivec_vspltb): Use
5668         const_0_to_15_operand instead of u5bit_cint_operand.
5669         (*altivec_vspltb_internal): Ditto.
5670         (altivec_vspltb_direct): Ditto.
5671         (altivec_vsplth): Use const_0_to_7_operand instead of
5672         u5bit_cint_operand.
5673         (*altivec_vsplth_internal): Ditto.
5674         (altivec_vsplth_direct): Ditto.
5675         (altivec_vspltw): Use const_0_to_3_operand instead of
5676         u5bit_cint_operand.
5677         (*altivec_vspltw_internal): Ditto.
5678         (altivec_vspltw_direct): Ditto.
5679         (altivec_vspltsf): Ditto.
5680         (*altivec_vspltsf_internal): Ditto.
5681         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
5682         various splats with the proper size immediate.  Reorder the various
5683         cases by ascending size of immediate, and put all such together.
5685 2018-09-25  Richard Biener  <rguenther@suse.de>
5687         PR debug/83941
5688         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
5689         GC-ification.
5690         (maybe_create_die_with_external_ref): Do not create
5691         DW_TAG_imported_unit here.
5692         (add_abstract_origin_attribute): Handle external BLOCK refs.
5693         (dwarf2out_abstract_function): Simplify LTO case.
5694         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
5695         rather than using maybe_create_die_with_external_ref.
5697 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
5699         PR target/71278
5700         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
5701         (frndintxf2_mask_pm_i387): Ditto.
5702         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
5703         Enable for !flag_trapping_math.
5704         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
5705         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
5706         Change operand 1 predicate to nonimmediate_operand.
5707         (attr "i387_cw"): Remove mask_pm.
5708         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
5709         (enum ix86_entity): Remove I387_MASK_PM.
5710         * config/i386/i386.c (ix86_i387_mode_needed): Do not
5711         handle I387_MASK_PM.
5712         (ix86_mode_needed): Ditto.
5713         (ix86_mode_after): Ditto.
5714         (ix86_mode_entry): Ditto.
5715         (ix86_mode_exit): Ditto.
5716         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
5718 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
5720         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
5721         to_update_switch_stmts to vNULL instead of calling create on them
5722         immediately.
5724 2018-09-25  Richard Biener  <rguenther@suse.de>
5726         PR tree-optimization/87402
5727         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
5728         (visit_phi): Re-instantiate handling of supposed to be VARYING
5729         but non-VARYING backedge value.
5731 2018-09-25  Richard Biener  <rguenther@suse.de>
5733         PR debug/83941
5734         * dwarf2out.c (struct sym_off_pair): New.
5735         (external_die_map): New global.
5736         (lookup_decl_die): When in LTO create DIEs lazily from the
5737         external_die_map.
5738         (lookup_block_die): New function, create DIEs lazily in LTO.
5739         (equate_block_to_die): New function.
5740         (dwarf2out_die_ref_for_decl): During WPA get the association
5741         from the external DIE map.
5742         (dwarf2out_register_external_die): Record mapping into the
5743         external DIE map.
5744         (maybe_create_die_with_external_ref): New function split out from
5745         DIE generation part of old dwarf2out_register_external_die.
5746         (add_abstract_origin_attribute): Do not return the DIE.  When
5747         in LTO reference externals directly.
5748         (dwarf2out_abstract_function): When in LTO ignore calls for
5749         decls with external DIEs (already present abstract instances).
5750         (gen_call_site_die): Adjust.
5751         (add_high_low_attributes): Likewise.
5752         (gen_lexical_block_die): Likewise.
5753         (gen_inlined_subroutine_die): Likewie.
5754         (gen_block_die): Likewise.
5755         (dwarf2out_inline_entry): Likewise.
5756         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
5757         DIEs.
5759 2018-09-25  Martin Liska  <mliska@suse.cz>
5761         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
5762         integers and not by a float value.
5764 2018-09-25  Martin Liska  <mliska@suse.cz>
5766         PR fortran/87394
5767         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
5768         instead of NULL.
5769         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
5770         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
5771         * godump.c (go_format_type): Remove extra parenthesis.
5773 2018-09-25  Martin Liska  <mliska@suse.cz>
5775         * alias.c (set_dest_equal_p): Remove unused function.
5776         * config/i386/i386.c (def_builtin_pure2): Likewise.
5777         * diagnostic-show-locus.c (class layout): Remove
5778         unused field.
5779         (layout::layout): Likewise here.
5780         * dump-context.h (class temp_dump_context): Likewise.
5781         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
5782         (add_AT_loclistsptr): Likewise.
5783         (add_AT_offset): Likewise.
5784         (get_AT_hi_pc): Likewise.
5785         (is_comdat_die): Likewise.
5786         (type_is_enum): Likewise.
5787         (ceiling): Likewise.
5788         (add_AT_vms_delta): Likewise.
5789         (is_class_die): Likewise.
5790         * edit-context.c (class line_event): Remove unused field.
5791         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
5792         unused function.
5793         * ipa-cp.c (ipa_get_vr_lat): Likewise.
5794         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
5795         (ok_for_base_p_nonstrict): Likewise.
5796         * tree-chrec.c (is_not_constant_evolution): Likewise.
5797         (chrec_fold_poly_cst): Likewise.
5798         * tree-if-conv.c (has_pred_critical_p): Likewise.
5799         * tree-ssa-coalesce.c (print_exprs): Likewise.
5800         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
5801         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
5802         * tree-vrp.c (value_ranges_intersect_p): Likewise.
5803         (value_range_nonnegative_p): Likewise.
5805 2018-09-25  Martin Liska  <mliska@suse.cz>
5807         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5808         Do not handle "GNU Pascal".
5809         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5810         Likewise.
5811         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
5812         from documentation. Likewise.
5813         * dbxout.c (dbxout_range_type): Likewise.
5814         * doc/cpp.texi: Likewise.
5815         * doc/extend.texi: Likewise.
5816         * doc/frontends.texi: Likewise.
5817         * doc/invoke.texi: Remove Pascal entry.
5818         * tree.def (CLEANUP_POINT_EXPR): Likewise.
5819         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
5821 2018-09-25  Martin Liska  <mliska@suse.cz>
5823         PR middle-end/86078
5824         * doc/invoke.texi: Document all parameters and remove default
5825         of the parameters.
5827 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5829         PR bootstrap/87417
5830         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
5831         contains HOST_WIDE_INTs when computing its size.
5833 2018-09-24  Jim Wilson  <jimw@sifive.com>
5835         PR target/87391
5836         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
5837         not TARGET_RVE.
5838         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
5840 2018-09-24  Andrew Pinski  <apinski@marvell.com>
5842         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
5843         access prev before checking it for NULLness in the
5844         AARCH64_FUSE_CMP_BRANCH case.
5846 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
5848         PR target/82699
5849         * config/i386/i386.c (rest_of_insert_endbranch): Set
5850         endbr_queued_at_entrance to true and don't insert ENDBR if
5851         x86_function_profiler will be called.
5852         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
5853         is true.
5854         * config/i386/i386.h (machine_function): Add
5855         endbr_queued_at_entrance.
5857 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5859         * genattrtab.c (mk_attr_alt): Use alternative_mask.
5860         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
5861         types.
5862         (check_attr_test): Use alternative_mask.
5863         (get_attr_value): Likewise.
5864         (compute_alternative_mask): Use alternative_mask and XWINT.
5865         (make_alternative_compare): Use alternative_mask.
5866         (attr_alt_subset_p): Use XWINT.
5867         (attr_alt_subset_of_compl_p): Likewise.
5868         (attr_alt_intersection): Use alternative_mask and XWINT.
5869         (attr_alt_union): Likewise.
5870         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
5871         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
5872         (simplify_test_exp): Use alternative_mask and XWINT.
5873         (write_test_expr): Use alternative_mask and XWINT, adjust bit
5874         number calculation to support 64 bits.  Generate code that
5875         checks 64-bit masks.
5876         (main): Use alternative_mask.
5877         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
5879 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5881         PR target/80080
5882         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5883         RETURN+USE when returning via %r14.
5885 2018-09-24  Martin Liska  <mliska@suse.cz>
5887         * gcov.c (output_lines): Print colorization legend
5888         for both flag_use_colors and flag_use_hotness_colors.
5889         Reword the help.
5891 2018-09-24  Martin Liska  <mliska@suse.cz>
5893         * coverage.c (get_coverage_counts): Use warning_at
5894         with current_function_decl location. Use %qD in warning
5895         message.
5897 2018-09-24  Martin Liska  <mliska@suse.cz>
5899         * memory-block.h (memory_block_pool::release): Annotate with
5900         valgrind that the memory is not accessible.
5902 2018-09-24  Martin Liska  <mliska@suse.cz>
5904         PR sanitizer/85774
5905         * asan.c: Make asan_handled_variables extern.
5906         * asan.h: Likewise.
5907         * cfgexpand.c (expand_stack_vars): Make sure
5908         a representative is unpoison if another
5909         variable in the partition is handled by
5910         use-after-scope sanitization.
5912 2018-09-24  Richard Biener  <rguenther@suse.de>
5914         PR tree-optimization/63155
5915         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
5916         the worklist when the edge of the respective argument isn't
5917         executable.
5919 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5921         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
5922         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
5923         (MASK_CLASS_P): Update for rename.
5924         (MAYBE_MASK_CLASS_P): Ditto.
5925         (REG_CLASS_NAMES): Update.
5926         (REG_CLASS_CONTENT): Update.
5927         * config/i386/i386.c (regclass_map): Update for MASK_REG
5928         and ALL_MASK_REGS rename.
5929         * config/i386/constraints.md (Yk): Update for rename.
5930         (k): Ditto.
5932 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5934         * config/i386/i386.h (enum reg_class): Remove
5935         EVEX_SSE_REGS and MOD4_SSE_REGS.
5936         (REG_CLASS_NAMES): Update.
5937         (REG_CLASS_CONTENT): Update.
5938         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
5939         registers as ALL_SSE_REGS.
5940         (ix86_additional_allocno_class_p): Remove.
5941         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
5942         (ix86_register_priority): Lower priority of EVEX SSE registers.
5943         Use IN_RANGE macro where appropriate.
5944         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
5945         AVX-5124VNNIW checks.
5946         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
5947         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
5948         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
5949         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
5950         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
5951         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
5952         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
5953         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
5954         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
5955         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
5956         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
5957         * config/i386/constraints.md (Yh): Remove.
5959 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5961         * config/i386/i386.c (regclass_map): Declare integer REX registers
5962         as GENERAL_REGS.
5964 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
5966         * doc/service.texi (Service): Switch the fsf.org link to https.
5968 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5970         PR target/86798
5971         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5972         Define to speculation_safe_value_not_needed.
5974 2018-09-21  Florian Weimer  <fweimer@redhat.com>
5976         PR middle-end/81035
5977         * doc/extend.texi (Common Function Attributes): Mention that
5978         noreturn suppresses tail call optimization.
5980 2018-09-21  Jeff Law  <law@redhat.com>
5982         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
5983         vr_values::cleanup_edges_and_switches.
5984         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
5985         vr_values class.
5986         (identify_jump_threads): Remove EDGE_IGNORE handling.
5987         (execute_vrp): Move handling of to_remove_edges and
5988         to_update_switch_stmts into vr_values class member functions.
5989         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
5990         (to_update_switch_stmts): Likewise.
5991         * vr-values.c: Include cfghooks.h.
5992         (vr_values::vr_values): Initialize to_remove_edges and
5993         to_update_switch_stmts.
5994         (vr_values::~vr_values): Verify to_remove_edges and
5995         to_update_switch_stmts are empty.
5996         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
5997         (vr_values::cleanup_edges_and_switches): New member function.
5998         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
5999         function.  Add new data members.
6001 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
6003         PR tree-optimization/87309
6004         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6005         calls with pflags and alt_flags.
6006         (selftest::test_capture_of_dump_calls): Add test of interaction of
6007         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6009 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6011         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6013 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6015         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6017 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6019         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6020         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6021         (VXWORKS_LIBS_RTP): Minor reordering.
6023 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6025         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6026         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6028 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6030         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6031         (PTRDIFF_TYPE): Likewise.
6033 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6035         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6036         triplet, similar to support for VxWorks7.
6037         * config/vxworks-dummy.h: Provide a default definition
6038         of TARGET_VXWORKS64 to 0.
6040 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6042         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6043         * config/vxworks-dummy.h: here.
6045 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6047         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6049 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
6050             Bo Zhou  <zbo.zhou@hisilicon.com>
6052         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6053         * config/aarch64/aarch64-tune.md: Regenerated.
6054         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6055         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6056         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6058 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
6059             Julian Brown  <julian@codesourcery.com>
6061         * builtins.c (get_builtin_sync_mem): Handle address spaces.
6063 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6065         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6066         if the call takes a static chain.
6068 2018-09-21  Martin Liska  <mliska@suse.cz>
6070         * auto-profile.c (autofdo_source_profile::read): Do not
6071         set sum_all.
6072         (read_profile): Do not add working sets.
6073         (read_autofdo_file): Remove sum_all.
6074         (afdo_callsite_hot_enough_for_early_inline): Remove const
6075         qualifier.
6076         * coverage.c (struct counts_entry): Remove gcov_summary.
6077         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6078         do not support GCOV_TAG_PROGRAM_SUMMARY.
6079         (get_coverage_counts): Remove summary and expected
6080         arguments.
6081         * coverage.h (get_coverage_counts): Likewise.
6082         * doc/gcov-dump.texi: Remove -w option.
6083         * gcov-dump.c (dump_working_sets): Remove.
6084         (main): Do not support '-w' option.
6085         (print_usage): Likewise.
6086         (tag_summary): Likewise.
6087         * gcov-io.c (gcov_write_summary): Do not dump
6088         histogram.
6089         (gcov_read_summary): Likewise.
6090         (gcov_histo_index): Remove.
6091         (gcov_histogram_merge): Likewise.
6092         (compute_working_sets): Likewise.
6093         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6094         it not obsolete.
6095         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6096         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6097         (GCOV_HISTOGRAM_SIZE): Remove.
6098         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6099         (struct gcov_summary): Simplify rapidly just
6100         to runs and sum_max fields.
6101         (gcov_histo_index): Remove.
6102         (NUM_GCOV_WORKING_SETS): Likewise.
6103         (compute_working_sets): Likewise.
6104         * gcov-tool.c (print_overlap_usage_message): Remove
6105         trailing empty line.
6106         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6107         (output_lines): Remove program related line.
6108         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6109         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6110         histogram.
6111         (input_profile_summary): Do not read it.
6112         (merge_profile_summaries): And do not merge it.
6113         (input_symtab): Do not call removed function.
6114         * modulo-sched.c (sms_schedule): Do not print sum_max.
6115         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6116         removed when histogram method was invented.
6117         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6118         mode.
6119         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6120         GCOV coding style.
6121         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6122         and dump selected value.
6123         * profile.c (add_working_set): Remove.
6124         (get_working_sets): Likewise.
6125         (find_working_set): Likewise.
6126         (get_exec_counts): Do not work with working sets.
6127         (read_profile_edge_counts): Do not inform as sum_max is removed.
6128         (compute_branch_probabilities): Likewise.
6129         (compute_value_histograms): Remove argument for call of
6130         get_coverage_counts.
6131         * profile.h: Do not make gcov_summary const.
6133 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6135         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6137 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6139         PR tree-optimization/86990
6140         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6141         Check that the entire merged store group is made of constants only for
6142         overlapping stores.
6144 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6146         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6147         (VTABLE_VERIFICATION_SPEC): Likewise.
6148         (SANITIZER_EARLY_SPEC): Likewise.
6149         (SANITIZER_SPEC): Likewise.
6150         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6151         * doc/invoke.texi (Link Options): Document -r.
6153 2018-09-20  Richard Biener <rguenther@suse.de>
6155         PR middle-end/87054
6156         * gimplify.c (gimplify_expr): Retain alignment of
6157         addressable lvalue in dereference.
6159 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6161         PR bootstrap/87013
6162         * configure.ac: Check for .loc is_stmt support.
6163         * configure, config.in: Rebuilt.
6164         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6165         if not supported.
6167 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6169         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6170         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6171         -misel=no.
6173 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6175         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6176         VECTOR_OTHER.
6177         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6178         case VECTOR_OTHER.
6180 2018-09-20  Marek Polacek  <polacek@redhat.com>
6182         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6184 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6186         PR tree-optimization/87288
6187         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6188         into account when determining PEELING_FOR_NITERS.
6190 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6192         PR tree-optimization/86877
6193         * tree-vect-loop.c (vect_analyze_loop_2): Call
6194         vect_verify_datarefs_alignment.
6196 2018-09-19  Marek Polacek  <polacek@redhat.com>
6198         * doc/invoke.texi: Document -Wclass-conversion.
6200 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6202         * config/pa/pa.c (pa_adjust_priority): Delete.
6203         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6205         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6206         (atomic_storehi): Likewise.
6207         (atomic_storesi): Likewise.
6208         (atomic_loaddi): Restore compare and swap exchange loop code.
6210 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6212         PR rtl-optimization/86902
6213         * combine.c (try_combine): When changing the CC mode used, don't change
6214         an unrelated mode in other_insn to that new CC mode.
6216 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6218         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6219         with %T in place of calls to dump_generic_expr.
6220         (prune_runtime_alias_test_list): Likewise.
6221         (create_runtime_alias_checks): Likewise.
6222         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6223         (vect_analyze_data_ref_dependence): Likewise.
6224         (vect_slp_analyze_data_ref_dependence): Likewise.
6225         (vect_record_base_alignment): Likewise.  Use %G in place of call
6226         to dump_gimple_stmt.
6227         (vect_compute_data_ref_alignment): Likewise.
6228         (verify_data_ref_alignment): Likewise.
6229         (vect_find_same_alignment_drs): Likewise.
6230         (vect_analyze_group_access_1): Likewise.
6231         (vect_analyze_data_ref_accesses): Likewise.
6232         (dependence_distance_ge_vf): Likewise.
6233         (dump_lower_bound): Likewise.
6234         (vect_prune_runtime_alias_test_list): Likewise.
6235         (vect_find_stmt_data_reference): Likewise.
6236         (vect_analyze_data_refs): Likewise.
6237         (vect_create_addr_base_for_vector_ref): Likewise.
6238         (vect_create_data_ref_ptr): Likewise.
6239         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6240         (vect_can_advance_ivs_p): Likewise.
6241         (vect_update_ivs_after_vectorizer): Likewise.
6242         (vect_gen_prolog_loop_niters): Likewise.
6243         (vect_prepare_for_masked_peels): Likewise.
6244         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6245         (vect_determine_vectorization_factor): Likewise.
6246         (vect_is_simple_iv_evolution): Likewise.
6247         (vect_analyze_scalar_cycles_1): Likewise.
6248         (vect_analyze_loop_operations): Likewise.
6249         (report_vect_op): Likewise.
6250         (vect_is_slp_reduction): Likewise.
6251         (check_reduction_path): Likewise.
6252         (vect_is_simple_reduction): Likewise.
6253         (vect_create_epilog_for_reduction): Likewise.
6254         (vect_finalize_reduction:): Likewise.
6255         (vectorizable_induction): Likewise.
6256         (vect_transform_loop_stmt): Likewise.
6257         (vect_transform_loop): Likewise.
6258         (optimize_mask_stores): Likewise.
6259         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6260         (vect_split_statement): Likewise.
6261         (vect_recog_over_widening_pattern): Likewise.
6262         (vect_recog_average_pattern): Likewise.
6263         (vect_determine_min_output_precision_1): Likewise.
6264         (vect_determine_precisions_from_range): Likewise.
6265         (vect_determine_precisions_from_users): Likewise.
6266         (vect_mark_pattern_stmts): Likewise.
6267         (vect_pattern_recog_1): Likewise.
6268         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6269         (vect_record_max_nunits): Likewise.
6270         (vect_build_slp_tree_1): Likewise.
6271         (vect_build_slp_tree_2): Likewise.
6272         (vect_print_slp_tree): Likewise.
6273         (vect_analyze_slp_instance): Likewise.
6274         (vect_detect_hybrid_slp_stmts): Likewise.
6275         (vect_detect_hybrid_slp_1): Likewise.
6276         (vect_slp_analyze_operations): Likewise.
6277         (vect_slp_analyze_bb_1): Likewise.
6278         (vect_transform_slp_perm_load): Likewise.
6279         (vect_schedule_slp_instance): Likewise.
6280         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6281         (vect_mark_stmts_to_be_vectorized): Likewise.
6282         (vect_init_vector_1): Likewise.
6283         (vect_get_vec_def_for_operand): Likewise.
6284         (vect_finish_stmt_generation_1): Likewise.
6285         (vect_check_load_store_mask): Likewise.
6286         (vectorizable_call): Likewise.
6287         (vectorizable_conversion): Likewise.
6288         (vectorizable_operation): Likewise.
6289         (vectorizable_load): Likewise.
6290         (vect_analyze_stmt): Likewise.
6291         (vect_is_simple_use): Likewise.
6292         (vect_get_vector_types_for_stmt): Likewise.
6293         (vect_get_mask_type_for_stmt): Likewise.
6294         * tree-vectorizer.c (increase_alignment): Likewise.
6296 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6298         * doc/rtl.texi: Adjust vec_select description.
6299         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6300         non-constant selectors.
6302 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6304         * config/aarch64/aarch64-protos.h
6305         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6306         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6307         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6308         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6309         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6310         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6311         (AARCH64_ISA_RCPC8_4): New flag.
6312         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6313         to aarch64_offset_9bit_signed_unscaled_p.
6314         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6315         and use stlur.
6316         * config/aarch64/constraints.md (Ust): New constraint.
6317         * config/aarch64/predicates.md.
6318         (aarch64_9bit_offset_memory_operand): New predicate.
6319         (aarch64_rcpc_memory_operand): New predicate.
6321 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6323         PR rtl-optimization/87361
6324         * rtlanal.c (nonzero_bits1): Revert accidental change.
6326 2018-09-19  Richard Biener  <rguenther@suse.de>
6328         PR tree-optimization/87349
6329         PR tree-optimization/87342
6330         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6332 2018-09-18  Marek Polacek  <polacek@redhat.com>
6334         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6335         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6337 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6339         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6341 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6343         PR rtl-optimization/86882
6344         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6346 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6348         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6349         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6351 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6353         PR other/87353
6354         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6356 2018-09-18  Richard Biener  <rguenther@suse.de>
6358         PR middle-end/63155
6359         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6360         (compute_samebase_partition_bases): Likewise.
6361         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6362         (gimple_can_coalesce_p): Simplify.
6364 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6366         Handle a library implementation of ffs calling __builtin_ffs.
6367         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6368         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6369         to __ffsdi2.
6371 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6373         * diagnostic-show-locus.c (class layout_range): Add field
6374         "m_original_idx".
6375         (layout_range::layout_range): Add "original_idx" param and use it
6376         to initialize new field.
6377         (make_range): Use 0 for original_idx.
6378         (layout::layout): Pass in index to calls to
6379         maybe_add_location_range.
6380         (layout::maybe_add_location_range): Add param "original_idx" and
6381         pass it on to layout_range.
6382         (layout::print_any_labels): Pass on range->m_original_idx to
6383         get_text call.
6384         (gcc_rich_location::add_location_if_nearby): Use 0 for
6385         original_idx.
6386         * gcc-rich-location.h (text_range_label::get_text): Update for new
6387         param.
6388         (range_label_for_type_mismatch::get_text): Likewise.
6390 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6392         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6394 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6396         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6397         format_string_diagnostic_t.
6398         (fmtwarn_n): Likewise.
6399         * substring-locations.c
6400         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6401         (format_warning_n_va): Convert to...
6402         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6403         (format_warning_va): Convert to...
6404         (format_string_diagnostic_t::emit_warning_va): ...this.
6405         (format_warning_at_substring): Convert to...
6406         (format_string_diagnostic_t::emit_warning): ...this.
6407         (format_warning_at_substring_n): Convert to...
6408         (format_string_diagnostic_t::emit_warning_n): ...this.
6409         * substring-locations.h (class format_string_diagnostic_t): New
6410         class.
6411         (format_warning_va): Convert to
6412         format_string_diagnostic_t::emit_warning_va.
6413         (format_warning_n_va): Convert to
6414         format_string_diagnostic_t::emit_warning_n_va.
6415         (format_warning_at_substring): Convert to
6416         format_string_diagnostic_t::emit_warning.
6417         (format_warning_at_substring_n): Convert to
6418         format_string_diagnostic_t::emit_warning_n.
6420 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6421             Bernd Schmidt <bernds_cb1@t-online.de>
6423         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6424         SImode args.
6426 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6428         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6429         operand 0 predicate to nonimmediate operand.
6430         (rint<mode>2_frndint): Remove insn pattern.
6431         (rint<mode>2): Change operand 1 predicate to general_operand.
6432         Extend operand 1 to XFmode and generate rintxf2 insn.
6433         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6434         Do not use X87MODEF mode macro.
6435         (frndintxf2_<rounding>_i387): Rename from
6436         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6437         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6438         to XFmode and generate significandxf3 insn.
6440 2018-09-17  Richard Biener  <rguenther@suse.de>
6442         PR tree-optimization/87328
6443         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6444         visiting unexecutable backedges when not iterating.
6445         (do_rpo_vn): Mark all edges not executable even when not
6446         iterating.
6448 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6450         PR c/63886
6451         * doc/invoke.texi (Warning Options): Likewise.
6453 2018-09-17  Richard Biener  <rguenther@suse.de>
6455         PR tree-optimization/87301
6456         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6457         clean EH info from leftover copy assignments.
6459 2018-09-17  Martin Liska  <mliska@suse.cz>
6461         PR gcov-profile/85871
6462         * gcov.c (output_intermediate_file): Fix out of bounds
6463         access.
6465 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6467         * config/arc/arc.c: Object attributes for core4 not reflected
6468         correctly.
6469         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6470         core3).
6472 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6474         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6476 2018-09-17  Martin Liska  <mliska@suse.cz>
6478         * doc/gcov.texi: Document new option --use-hotness-colors.
6479         * gcov.c (struct source_info): Declare new field.
6480         (source_info::source_info): Set default for maximum_count.
6481         (print_usage): Add new -q option.
6482         (process_args): Process it.
6483         (accumulate_line_info): Save src->maximum_count.
6484         (output_line_beginning): Make color line number if
6485         flag_use_hotness_colors is set.
6486         (output_line_details): Pass default argument value.
6487         (output_lines): Pass src->maximum_count.
6489 2018-09-17  Martin Liska  <mliska@suse.cz>
6491         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6492         Use processor_names table.
6493         * config/i386/i386.c (ix86_default_align): Use
6494         processor_cost_table for alignment values.
6495         (ix86_option_override_internal): Use processor_names.
6496         (ix86_function_specific_print): Likewise.
6497         * config/i386/i386.h (struct processor_costs):
6498         Add alignment values.
6499         (struct ptt): Remove and replace with const char *.
6500         * config/i386/x86-tune-costs.h (struct processor_costs):
6501         Declare default alignments for all costs.
6503 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6505         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6506         symbolics or VR_VARYING ranges for ABS_EXPR.
6507         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6508         when range will wrap.
6510 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6512         PR middle-end/86864
6513         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6514         before and after a JUMP_TABLE_DATA.
6516 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6518         PR middle-end/87188
6519         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6520         when one operand is a function pointer.  Use POINTER_TYPE_P and
6521         FUNC_OR_METHOD_TYPE_P.
6522         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6523         * fold-const.c (build_range_check): Likewise.
6524         * match.pd (simple_comparison): Likewise.
6526 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6528         PR c/82967
6529         * spellcheck.c (get_edit_distance_cutoff): New function.
6530         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6531         (selftest::test_get_edit_distance_one_way): ...this.
6532         (selftest::test_get_edit_distance_unit): Rename to...
6533         (selftest::test_get_edit_distance_both_ways): ...this.
6534         (selftest::test_edit_distances): Move tests to this new function,
6535         and test some more pairs of strings.  Update for above renaming.
6536         (selftest::get_old_cutoff): New function.
6537         (selftest::test_get_edit_distance_cutoff): New function.
6538         (selftest::assert_suggested_for): New function.
6539         (ASSERT_SUGGESTED_FOR): New macro.
6540         (selftest::assert_not_suggested_for): New function.
6541         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6542         (selftest::test_suggestions): New function.
6543         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6544         tests to selftest::test_edit_distances and call it.  Add calls to
6545         selftest::test_get_edit_distance_cutoff and
6546         selftest::test_suggestions.
6547         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6548         (best_match::consider): Replace hard-coded cutoff calculation with
6549         a call to...
6550         (best_match::get_cutoff): New declaration.
6551         (best_match::get_best_meaningful_candidate): Likewise.
6553 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6555         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6557 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6559         revert:
6560         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6562         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6563         terminated string literal.
6565 2018-09-14  Martin Sebor  <msebor@redhat.com>
6567         * builtins.c (unterminated_array): Handle ARRAY_REF.
6568         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6569         * builtins.h (unterminated_array): Declare extern.
6570         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6571         arrays.
6572         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6573         calls.
6575 2018-09-14  Martin Sebor  <msebor@redhat.com>
6576             Jeff Law  <law@redhat.com>
6578         * builtins.c (unterminated_array): New.
6579         (expand_builtin_strcpy): Adjust.
6580         (expand_builtin_strcpy_args): Detect unterminated arrays.
6581         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
6582         unterminated arrays.
6583         * gimple-fold.h (get_maxval_strlen): Add argument.
6584         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
6586         * gimple-fold.c (get_range_strlen): Add argument.
6587         (get_maxval_strlen): Adjust.
6588         * gimple-fold.h (get_range_strlen): Add argument.
6590 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
6592         * config/i386/movdirintrin.h: Fix copyright year.
6594 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
6596         * reg-stack.c: Include regs.h.
6597         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
6598         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6599         FIRST_STACK_REG, not DFmode.
6600         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
6601         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
6602         (change stack): Default register mode to the reg_raw_mode of
6603         FIRST_STACK_REG, not DFmode.
6604         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
6605         (*swapxf): Rename from swapxf.
6607 2018-09-14  Carl Love  <cel@us.ibm.com>
6609         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
6610         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
6612 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
6614         PR target/87224
6615         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
6616         alternatives.
6618 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
6620         PR target/85628
6621         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
6623 2018-09-14  Jason Merrill  <jason@redhat.com>
6625         Fix --enable-gather-detailed-mem-stats.
6626         * hash-table.c (hash_table_usage): Change from variable to function.
6627         * hash-table.h: Adjust.
6628         * Makefile.in: Add missing dependencies on hash-table.h.
6630 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6632         PR tree-optimization/87259
6633         PR lto/87283
6634         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
6635         execute_cse_reciprocals_1 has tried transforming.
6637 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
6639         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
6640         VR_VARYING for PLUS/MINUS_EXPR.
6642 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6644         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
6645         formatting.
6647 2018-09-14  Richard Biener  <rguenther@suse.de>
6649         PR middle-end/63155
6650         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
6651         bits for the merged partition.
6653 2018-09-13  Martin Sebor  <msebor@redhat.com>
6654             Bernd Edlinger  <bernd.edlinger@hotmail.de>
6656         * builtins.h (c_srlen): Add argument.
6657         * builtins.c (warn_string_no_nul): New function.
6658         (c_strlen): Add argument and use it.  Update recursive calls.
6659         Pass DECL argument to string_constant to get info on non
6660         terminated strings.  Update *NONSTR as needed.
6661         (fold_builtin_strlen): Add argument to calls to c_strlen.
6662         Warn for unterminated arrays.
6663         (warn_string_no_null): Add prototype.
6664         * expr.c (string_constant): Update arguments.  Update recursive
6665         calls appropriately.  Detect missing NUL terminator and outermost
6666         declaration its missing in.
6667         Improve checks for arrays with nonzero lower bound or elements
6668         that are not a single byte.  Simplify offset computation.
6669         Simplify checks for non-NUL terminated strings.
6670         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
6671         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
6673 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6675         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
6676         correctly.
6677         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
6678         argument.  Fix range checks.
6679         * fold-const.h (c_getstr): Adjust protoype.
6680         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
6681         string is constant but contains no NUL byte.
6683         * expr.c (string_constant): Adjust function comment.
6684         Remove bogus check for zero termination.
6686         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
6688         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
6689         (get_constant_size): Don't make STRING_CSTs larger than they are.
6690         (check_string_literal): New check function for STRING_CSTs.
6691         (output_constant): Use it.
6693 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
6695         PR target/86812
6696         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
6698 2018-09-13  Richard Biener  <rguenther@suse.de>
6700         PR tree-optimization/87263
6701         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
6702         (struct unwind_state): Add max_rpo field.
6703         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
6704         Compute max_rpo, the max RPO number a block can be backwards reached
6705         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
6706         separating it from the iterating mode.
6708 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6710         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
6711         (rfs_decision): New scheduling decision.
6713 2018-09-13  Richard Biener  <rguenther@suse.de>
6715         PR bootstrap/87134
6716         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
6717         (vn_nary_op_insert_pieces_predicated): Do not write useless
6718         valid_dominated_by_p entry outside of the allocated storage.
6720 2018-09-13  Omar Sandoval  <osandov@osandov.com>
6721             Tom de Vries  <tdevries@suse.de>
6723         PR debug/86985
6724         * dwarf2out.c (is_c): New function.
6725         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
6727 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
6729         PR target/85628
6730         * config/aarch64/aarch64.md (*aarch64_bfxil):
6731         Define.
6732         * config/aarch64/constraints.md (Ulc): Define.
6733         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
6734         Define.
6735         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
6736         New function.
6738 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6740         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
6741         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
6742         aarch64_layout_frame call.
6743         (aarch64_expand_epilogue): Likewise.
6744         (aarch64_initial_elimination_offset): Likewise.
6745         (aarch64_get_separate_components): Likewise.
6746         (aarch64_use_return_insn_p): Likewise.
6747         (aarch64_layout_frame): Remove unneeded check.
6749 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6751         * configure.ac: Only append
6752         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
6753         gcc_config_arguments if it was never reconfigured or last reconfigure
6754         was with different arguments.
6755         * configure: Regenerated.
6757 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6758             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6760         PR middle-end/87290
6761         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
6762         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
6764 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6766         PR tree-optimization/87287
6767         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
6768         X % C == 0 to X % (unsigned) C == 0 optimization to ...
6769         * match.pd (X % C == 0): ... here.  New optimization.
6771 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6773         PR middle-end/82853
6774         * expr.h (maybe_optimize_mod_cmp): Declare.
6775         * expr.c (mod_inv): New function.
6776         (maybe_optimize_mod_cmp): New function.
6777         (do_store_flag): Use it.
6778         * cfgexpand.c (expand_gimple_cond): Likewise.
6780 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
6781             Julian Brown  <julian@codesourcery.com>
6783         PR middle-end/86336
6784         * gimplify.c (gimplify_scan_omp_clauses): Set
6785         target_firstprivatize_array_bases in OpenACC parallel and kernels
6786         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
6787         OpenACC data regions.
6789 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
6791         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
6792         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
6794 2018-09-12  Richard Biener  <rguenther@suse.de>
6796         PR tree-optimization/87280
6797         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
6798         edge but unreachable target.
6799         (do_rpo_vn): For conservatively handling a PHI only mark
6800         the backedge executable but not the block reachable.
6802 2018-09-12  Richard Biener  <rguenther@suse.de>
6804         PR tree-optimization/87266
6805         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
6806         visited blocks.
6808 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
6810         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
6811         constants.
6812         ("trunc<BFP:mode><DFP_ALL:mode>2")
6813         ("trunc<DFP_ALL:mode><BFP:mode>2")
6814         ("extend<BFP:mode><DFP_ALL:mode>2")
6815         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
6816         according to the target operand type.
6818 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6819             Andreas Krebbel  <krebbel@linux.ibm.com>
6821         PR tree-optimization/86844
6822         * gimple-ssa-store-merging.c
6823         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
6824         there are any overlapping stores in between them, make sure they are
6825         also coalesced or we give up completely.
6827 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6829         PR middle-end/87248
6830         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
6831         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
6833 2018-09-12  Tom de Vries  <tdevries@suse.de>
6835         * common.opt (gdescribe-dies): Add option.
6836         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
6837         attribute for artifical and nameless decls.
6838         (dwarf2out_register_external_die): Add description attribute to
6839         external reference die.
6840         (add_desc_attribute): New functions.
6841         (gen_subprogram_die): Add description attribute to
6842         DW_TAG_call_site_parameter.
6843         * tree-pretty-print.c (print_generic_expr_to_str): New function.
6844         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
6845         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
6846         -gno-describe-dies.
6847         (@item -gdescribe-dies): Add.
6849 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6851         * tree-vrp.c (vrp_shift_undefined_p): Remove.
6852         (extract_range_from_binary_expr_1: Call
6853         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
6854         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
6855         depend on sign.
6857 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6859         * gimple-ssa-warn-alloca.c
6860         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
6861         field for ALLOCA_BOUND_*_LARGE.
6863 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6865         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
6867 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6869         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
6870         for clobbers.  Remove obsolete comment.
6872 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6874         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
6875         mpxchk, mpxld and mpxst types.
6876         (define_attr length_immediate): Remove all processing of mpx types.
6877         (define_attr prefix_0f): Ditto.
6878         (define_attr memory): Ditto.
6880 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6882         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
6883         (log<mode>2): Change operand 1 predicate to general_operand.
6884         Extend operand 1 to XFmode and generate logxf3 insn.
6885         (log10<mode>2): Change operand 1 predicate to general_operand.
6886         Extend operand 1 to XFmode and generate log10xf3 insn.
6887         (log2<mode>2): Change operand 1 predicate to general_operand.
6888         Extend operand 1 to XFmode and generate log2xf3 insn.
6889         (fyl2xp1_extend<mode>xf3_i387): Remove.
6890         (log1p<mode>2): Change operand 1 predicate to general_operand.
6891         Extend operand 1 to XFmode and generate log1pxf3 insn.
6892         (fxtract_extend<mode>xf3_i387): Remove.
6893         (logb<mode>2): Change operand 1 predicate to general_operand.
6894         Extend operand 1 to XFmode and generate logbxf3 insn.
6895         (ilogb<mode>2): Change operand 1 predicate to general_operand.
6896         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
6897         (significand<mode>2): Change operand 1 predicate to general_operand.
6898         Extend operand 1 to XFmode and generate significandxf3 insn.
6900 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6902         * gcc.c (perror_with_name, pfatal_with_name): Delete.
6903         (load_specs): Use fatal_error.
6904         (DELETE_IF_ORDINARY, process_command): Use error.
6905         (execute, run_attempt): Use fatal_error.
6907 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
6909         * diagnostic-core.h (sorry_at): New prototype.
6910         * diagnostic.c (sorry_at): New function.
6912 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
6914         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
6915         by zero as VR_UNDEFINED.
6917 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6919         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
6920         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
6921         (<sincos>mode2): New expander.
6922         (sincos_extend<mode>xf3_i387): Remove insn pattern.
6923         (sincos -> sin, cos splitters): Remove splitter patterns.
6924         (sincos<mode>3): Change operand 2 predicate to general_operand.
6925         Extend operand 2 to XFmode and generate sincosxf3 insn.
6926         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
6927         Change operand 3 predicate to const1_operand.
6928         (fptan_extend<mode>xf4_i387): Remove insn pattern.
6929         (tanxf2): Update operands in the call to fptanxf4_i387.
6930         (tan<mode>2): Change operand 1 predicate to general_operand.
6931         Extend operand 1 to XFmode and generate tanxf3 insn.
6932         (atan2xf3): Rename from *fpatanxf3_i387.
6933         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
6934         (atan2xf3): Remove expander.
6935         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
6936         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
6937         (atan<mode>2): Change operand 1 predicate to general_operand.
6938         Extend operand 1 to XFmode and generate atanxf3 insn.
6940 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6942         * config/i386/i386.md (x87/SSE constant load splitter): Use
6943         memory_operand instead of nonimmediate_operand for input operand
6944         predicate.
6946 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
6948         * config/i386/i386.md (float partial SSE register stall splitter): Move
6949         splitter near its instruction pattern.
6950         (float_extend partial SSE register stall splitter): Ditto.
6951         (float_truncate partial SSE register stall splitter): Ditto.
6953 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
6955         PR target/86794
6956         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6957         to speculation_safe_value_not_needed.
6959         PR target/85666
6960         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
6961         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
6962         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
6963         leaf_function_p, instead use has_hard_reg_initial_val.
6965 2018-09-09  Nathan Sidwell  <nathan@acm.org>
6967         * gcc.h (pfatal_with_name): Don't declare here.
6968         * gcc.c (pfatal_with_name): Make static.
6970 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
6972         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
6973         earlyclobber.
6975 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
6977         PR rtl-optimization/85458
6978         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
6979         priority hook to reduce the priority of EXPR.
6981 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
6983         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
6984         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
6985         with X87_ENABLE_FLOAT.  Remove preparation code.
6986         (*float<SWI48:mode><MODEF:mode>2): Rename from
6987         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
6988         math using "enabled" attribute.
6989         (*floatdi<MODEF:mode>2_i387): Rename from
6990         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
6991         enable for 32bit targets only.
6992         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
6993         splitter.
6994         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
6995         as operand 1 predicate.  Rewrite as define_insn_and_split.
6996         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
6998 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7000         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7001         to fallthru to FLOAT case.
7003 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7005         PR target/86731
7006         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7007         around folding of vec_sl to handle out of range shift values.
7009 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7011         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7012         Update callers to gen_fix_trunc<mode>_i387_fisttp
7013         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7014         nonimmediate_operand.
7015         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7016         and corresponding splitters.
7017         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7018         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7019         (fix_truncdi_i387_with_temp): Remove insn pattern
7020         and corresponding splitters.
7021         (fix_trunc<mode>_i387): Change operand 0 predicate to
7022         nonimmediate_operand.
7023         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7024         and corresponding splitters.
7025         (*fistdi2_1): Remove.
7026         (fistdi2): Ditto.
7027         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7028         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
7029         (*fist<mode>2_1): Remove.
7030         (fist<mode>2): Ditto.
7031         (fist<mode>2_with_temp): Remove insn pattern and corresponding
7032         splitters.
7033         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
7034         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7035         (fistdi2_<rounding>): Change operand 0 predicate to
7036         nonimmediate_operand.
7037         (fistdi2_<rounding>_with_temp): Remove insn pattern
7038         and corresponding splitters.
7039         (fist<mode>2_<rounding>): Change operand 0 predicate to
7040         nonimmediate_operand.
7041         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7042         and corresponding splitters.
7044         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7046 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7048         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7049         the init value.
7051 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7053         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7054         early gimple folding of vec_splat().
7055         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7056         * gimple-fold.h: Add an extern define for tree_vec_extract().
7058 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7060         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7061         wrappers around TREE_TYPE comparisons.
7063 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7065         PR target/80080
7066         * config/s390/predicates.md: Add nonsym_memory_operand.
7067         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7068         contains a SYMBOL_REF, load it into an intermediate pseudo.
7069         (s390_emit_compare_and_swap): Legitimize operand.
7070         * config/s390/s390.md: Use the new nonsym_memory_operand
7071         with UNSPECV_CAS patterns.
7073 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7075         PR target/80080
7076         * config/s390/s390-passes.def: New file.
7077         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7078         declaration.
7079         (make_pass_s390_early_mach): Add declaration.
7080         * config/s390/s390.c (make_pass_s390_early_mach):
7081         (s390_option_override): Remove dynamic registration.
7082         * config/s390/t-s390: Add s390-passes.def.
7084 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7086         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7087         Remove UNSPEC_LTREL_BASE check.
7088         (annotate_constant_pool_refs): Likewise.
7089         (find_constant_pool_ref): Likewise.
7090         (find_ltrel_base): Removed.
7091         (replace_ltrel_base): Removed.
7092         (s390_mainpool_finish): Remove replace_ltrel_base call.
7093         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7094         (s390_chunkify_finish): Remove replace_ltrel_base call.
7095         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7097 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7099         PR target/86779
7100         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7101         to speculation_safe_value_not_needed.
7103 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7104             Bernd Schmidt  <bernds_cb1@t-online.de>
7106         * config/nvptx/nvptx-opts.h: New file.
7107         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7108         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7109         (ASM_SPEC): Define.
7110         (TARGET_SM35): New macro.
7111         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7112         correct predicate.
7113         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7114         values.
7115         (misa=): New option.
7116         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7118 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7120         * config/i386/i386.md (truncdfsf2): Remove expander.
7121         (truncdfsf2_with_temp): Ditto.
7122         (truncxf<mode>2): Ditto.
7123         (*truncdfsf_fast_mixed): Remove insn pattern.
7124         (*truncdfsf_fast_i387): Ditto.
7125         (*truncdfsf_mixed): Ditto.
7126         (*truncdfsf_i387): Ditto.
7127         (*truncdfsf2_i387_1): Ditto.
7128         (*truncxfsf2_mixed): Ditto.
7129         (*truncxfdf2_mixed): Ditto.
7130         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7131         to call gen_truncxf<mode>2 instead.
7132         (*truncxf<mode>2_i387): Remove.
7133         (reg->reg splitters): Remove splitter pattern.
7134         (reg->mem splitters): Ditto.
7136         (truncdfsf2): New insn pattern.
7137         (truncxf<mode>2): Ditto.
7139 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7141         * tree-ssa-math-opts.c (is_mult_by): New function.
7142         (is_square_of): Use the above.
7143         (optimize_recip_sqrt): New function.
7144         (pass_cse_reciprocals::execute): Use the above.
7146 2018-09-05  Richard Biener  <rguenther@suse.de>
7148         PR bootstrap/87134
7149         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7150         to zero-init the emplaced vec.
7152 2018-09-05  Martin Liska  <mliska@suse.cz>
7154         PR tree-optimization/87205
7155         * tree-switch-conversion.c (pass_lower_switch::execute):
7156         Group cases for switch statements.
7158 2018-09-05  Richard Biener  <rguenther@suse.de>
7160         PR tree-optimization/87217
7161         * tree-ssa-sccvn.c (vuse_valueize): New.
7162         (vn_reference_lookup_pieces): Use it.
7163         (vn_reference_lookup): Likewise.
7165 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7167         PR c++/87137
7168         * stor-layout.c (place_field): Scan forwards to check last
7169         bitfield when ms_bitfield_placement is in effect.
7171 2018-09-05  Richard Biener  <rguenther@suse.de>
7173         PR bootstrap/87225
7174         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7175         return.
7177 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7178             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7180         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7181         * config.gcc (extra_objs): Build it.
7182         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7183         Likewise.
7184         * config/aarch64/aarch64-passes.def
7185         (pass_tag_collision_avoidance): New pass.
7186         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7187         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7188         (aarch64_classify_address): Remove static qualifier.
7189         (aarch64_address_info, aarch64_address_type): Move to...
7190         * config/aarch64/aarch64-protos.h: ... here.
7191         (make_pass_tag_collision_avoidance): New function.
7192         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7193         New tuning flag.
7195 2018-09-05  Martin Liska  <mliska@suse.cz>
7197         * doc/gcov.texi: Update documentation of humar
7198         readable mode.
7199         * gcov.c (format_count): Print one decimal place, it provides
7200         more fine number of situations like '1G' vs. '1.4G'.
7202 2018-09-05  Martin Liska  <mliska@suse.cz>
7204         PR target/87164
7205         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7206         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7207         to generate a MASK value.
7209 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7211         PR debug/86593
7212         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7213         if frame pointer isn't used.
7214         (compute_frame_pointer_to_fb_displacement): Likewise.
7216 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7218         PR target/87198
7219         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7220         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7221         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7222         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7223         and OPTION_MASK_ISA_XSAVEC_UNSET.
7225 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7227         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7228         XOR operations in NAND case.
7230 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7232         * wide-int-range.cc (wide_int_range_convert): New.
7233         * wide-int-range.h (wide_int_range_convert): New.
7234         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7235         code into wide_int_range_convert.
7236         (extract_range_into_wide_ints): Do not munge anti range constants
7237         into the entire domain.  Just return the range back.
7239 2018-09-04  Martin Liska  <mliska@suse.cz>
7241         * genmatch.c (output_line_directive): Add new argument
7242         fnargs.
7243         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7245 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7247         * doc/invoke.texi (Option Summary): Add whitespace.
7249         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7251 2018-09-04  Richard Biener  <rguenther@suse.de>
7253         PR tree-optimization/87211
7254         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7255         backedge value we're supposed to treat as VARYING also number
7256         the PHI to VARYING in case it got a different value-number already.
7258 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7260         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7261         (extract_range_from_binary_expr_1): Do not call
7262         vrp_can_optimize_bit_op.
7263         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7264         static.
7265         (wide_int_range_get_mask_and_bounds): New.
7266         (wide_int_range_optimize_bit_op): New.
7267         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7268         (wide_int_range_bit_and): Same.
7269         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7270         (wide_int_range_optimize_bit_op): New.
7271         (wide_int_range_get_mask_and_bounds): New.
7273 2018-09-04  Richard Biener  <rguenther@suse.de>
7275         PR tree-optimization/87176
7276         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7277         variable.  When value-numbering a virtual PHI node make sure
7278         to not value-number to the backedge value.
7280 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7282         * doc/extend.texi (Long Long, Hex Floats): Document support for
7283         long long and hex floats in more recent versions of ISO C++.
7285 2018-09-03  Richard Biener  <rguenther@suse.de>
7287         PR tree-optimization/87177
7288         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7289         cleanup.
7291 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7293         * bb-reorder.c (edge_order): Convert to C-qsort-style
7294         tri-state comparator.
7295         (reorder_basic_blocks_simple): Change std::stable_sort to
7296         gcc_stablesort.
7298 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7300         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7301         tri-state comparator.
7302         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7304 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7306         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7307         (mergesort): ... here as maximum count for using netsort.
7308         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7309         (gcc_stablesort): New.
7310         * system.h (gcc_stablesort): Declare.
7312 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7314         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7315         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7316         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7318 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7320         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7321         lxsdx and stxsdx alternatives.
7322         (*mov<mode>_hardfloat64): Ditto.
7323         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7325 2018-09-03  Richard Biener  <rguenther@suse.de>
7327         PR tree-optimization/87200
7328         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7329         simplify result.
7331 2018-09-03  Martin Liska  <mliska@suse.cz>
7333         PR tree-optimization/87201
7334         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7335         Fix parenthesis in an expression.
7337 2018-09-03  Richard Biener  <rguenther@suse.de>
7339         PR tree-optimization/87197
7340         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7341         visited.  CSE the VN_INFO hashtable lookup.
7343         PR tree-optimization/87169
7344         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7345         iterating make sure there's no extra backedges from irreducible
7346         regions feeding the header.  Mark the destination block
7347         executable.
7349 2018-09-03  Martin Liska  <mliska@suse.cz>
7351         PR driver/83193
7352         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7353         * common/common-targhooks.c (default_get_valid_option_values):
7354         New function.
7355         * common/common-targhooks.h (default_get_valid_option_values):
7356         Likewise.
7357         * common/config/i386/i386-common.c: Move processor_target_table
7358         from i386.c.
7359         (ix86_get_valid_option_values): New function.
7360         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7361         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7362         (PTA_*): Move all defined masks into i386-common.c.
7363         (ix86_function_specific_restore): Use new processor_cost_table.
7364         * config/i386/i386.h (struct ptt): Moved from i386.c.
7365         (struct pta): Likewise.
7366         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7367         * doc/tm.texi.in: Likewise.
7368         * opt-suggestions.c (option_proposer::suggest_option):
7369         Pass prefix to build_option_suggestions.
7370         (option_proposer::get_completions): Likewise.
7371         (option_proposer::build_option_suggestions): Use the new target
7372         hook.
7373         * opts.c (struct option_help_tuple): New struct.
7374         (print_filtered_help): Use the new target hook.
7376 2018-09-03  Martin Liska  <mliska@suse.cz>
7378         PR middle-end/59521
7379         * predict.c (set_even_probabilities): Add likely_edges
7380         argument and handle cases where we have precisely one
7381         likely edge.
7382         (combine_predictions_for_bb): Catch also likely_edges.
7383         (tree_predict_by_opcode): Handle gswitch statements.
7384         * tree-cfg.h (find_case_label_for_value): New declaration.
7385         (find_taken_edge_switch_expr): Likewise.
7386         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7387         Find pivot in decision tree based on probabily, not by number of
7388         nodes.
7390 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7392         * doc/standards.texi (Standards): Update Objective-C reference.
7394 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7396         * doc/install.texi (Prerequisites): Update link for MPC.
7398 2018-09-01  Michael Matz  <matz@suse.de>
7400         PR tree-optimization/87074
7401         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7402         PHIs for outer-loop uses.
7404 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7406         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7407         * doc/invoke.texi (C Dialect Options): Ditto.
7409 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7411         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7413 2018-08-31  Richard Biener  <rguenther@suse.de>
7415         PR tree-optimization/87168
7416         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7417         (rpo_elim::eliminate_avail): When OP was not visited it must
7418         be available.
7420 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7422         * tree-vrp.c (copy_value_range): Convert param "from" from
7423         "value_range *" to "const value_range *".
7424         (range_is_null): Likewise for param "vr".
7425         (range_int_cst_p): Likewise.
7426         (range_int_cst_singleton_p): Likewise.
7427         (symbolic_range_p): Likewise.
7428         (value_ranges_intersect_p): Likewise for both params.
7429         (value_range_nonnegative_p): Likewise for param "vr".
7430         (value_range_constant_singleton): Likewise.
7431         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7432         (extract_range_into_wide_ints): Likewise for param "vr".
7433         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7434         and "vr1".
7435         (vrp_can_optimize_bit_op): Likewise.
7436         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7437         "vr1_".
7438         (extract_range_from_unary_expr): Likewise.
7439         (debug_value_range): Likewise for param "vr".
7440         (value_range::dump): Add "const" qualifier.
7441         (vrp_prop::check_array_ref): Convert local "vr" from
7442         "value_range *" to "const value_range *".
7443         (vrp_prop::check_mem_ref): Likewise.
7444         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7445         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7446         (vrp_intersect_ranges): Likewise.
7447         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7448         (vrp_prop::vrp_finalize): Likewise.
7449         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7450         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7451         (extract_range_from_unary_expr): Likewise.
7452         (value_range_constant_singleton): Likewise.
7453         (symbolic_range_p): Likewise.
7454         (copy_value_range): Likewise.
7455         (extract_range_from_binary_expr_1): Likewise.
7456         (range_int_cst_p): Likewise.
7457         (vrp_set_zero_nonzero_bits): Likewise.
7458         (range_int_cst_singleton_p): Likewise.
7460 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7462         * config/aarch64/arm_neon.h (vabsd_s64): New.
7463         (vnegd_s64): Likewise.
7465 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7467         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7469 2018-08-31  Martin Liska  <mliska@suse.cz>
7471         * ipa-icf.c (sem_item::add_type): Use
7472         sem_item::m_type_hash_cache.
7473         * ipa-icf.h: Move the cache from sem_item_optimizer
7474         to sem_item.
7476 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7478         * doc/extend.texi (Backwards Compatibility): Remove implicit
7479         extern C leeway of () being (...).
7481 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7483         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7485 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7487         PR target/86684
7488         PR target/87149
7489         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7491 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7493         PR middle-end/87138
7494         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7495         gen_int_mode.  Formatting fixes.
7497 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7499         * target.def (custom_function_descriptors): Improve documentation.
7500         * doc/tm.texi.in (Trampolines): Expand discussion of function
7501         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7502         beginning of the section.
7503         * doc/tm.texi: Regenerated.
7505 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7507         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7508         base class in the initializer list.  This is a workaround for
7509         building with older GCC.
7510         (class auto_bb_flag): Likewise.
7512 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7514         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7515         (altivec_vcmpequ<VI_char>_p): Remove star.
7516         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7517         vector load modes.
7518         (expand_strncmp_vec_sequence): New function.
7519         (emit_final_str_compare_vec): New function.
7520         (expand_strn_compare): Add support for vector strncmp.
7521         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7522         length specification to bytes.
7523         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7524         (vcmpnezb_p): New pattern.
7525         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7526         for option -mstring-compare-inline-limit.
7528 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7530         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7531         (PTA_SKYLAKE): Add PTA_AES.
7532         (PTA_GOLDMONT): Likewise.
7534 2018-08-29  Jan Hubicka  <jh@suse.cz>
7536         PR lto/86517
7537         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7538         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7540 2018-08-29  Jan Hubicka  <jh@suse.cz>
7542         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7543         TYPE_STUB_DECL.
7544         (hash_tree): Do not visit TYPE_STUB_DECL.
7545         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7546         stream TYPE_STUB_DECL.
7547         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7548         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7549         after free_lang_data.
7550         (type_in_anonymous_namespace_p): Likewise.
7552 2018-08-29  Jan Hubicka  <jh@suse.cz>
7554         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7555         comment that it has to be even number.
7556         (class sreal): Change m_sig type to int32_t.
7557         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7558         int64_t for temporary calculations.
7559         (sreal_verify_basics): Drop one bit from minimum and maximum.
7561 2018-08-30  Richard Biener  <rguenther@suse.de>
7563         PR tree-optimization/87147
7564         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7565         (visit_phi): When the degenerate result is from the backedge and
7566         we didn't visit its definition yet drop to VARYING.
7567         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7569 2018-08-29  Jan Hubicka  <jh@suse.cz>
7571         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7572         DECL_VINDEX.
7573         (hash_tree): Likewise.
7575 2018-08-29  Jan Hubicka  <jh@suse.cz>
7577         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
7578         and TYPE_NEXT_REF_TO.
7580 2018-08-29  Jan Hubicka  <jh@suse.cz>
7582         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7583         comment that it has to be even number.
7584         (class sreal): Change m_sig type to int32_t.
7585         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7586         int64_t for temporary calculations.
7587         (sreal_verify_basics): Drop one bit from minimum and maximum.
7589 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
7591         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
7593 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
7595         PR middle-end/86995
7596         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
7597         if to_add is negative.
7599 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7601         PR middle-end/87053
7602         * builtins.c (c_strlen): Improve range checks.
7604 2018-08-29  Martin Sebor  <msebor@redhat.com>
7605             Jeff Law  <law@redhat.com>
7607         PR tree-optimization/86714
7608         PR tree-optimization/86711
7609         * builtins.c (c_strlen): Add arguments to call to string_constant.
7610         * expr.c (string_constant): Add argument.  Detect missing nul
7611         terminator and outermost declaration it's missing in.
7612         * expr.h (string_constant): Add argument.
7613         * fold-const.c (read_from_constant_string): Add arguments to call to
7614         string_constant.
7615         (c_getstr): Likewise.
7616         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7617         to string_constant.
7618         * tree-ssa-strlen.c (get_stridx): Likewise.
7620 2018-08-29  Jan Hubicka  <jh@suse.cz>
7622         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
7623         Do not stream DECL_VINDEX.
7624         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
7625         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
7626         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
7628 2018-08-29  Richard Biener  <rguenther@suse.de>
7630         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
7631         virtual operands that are not default defs to honor region
7632         boundaries.
7633         (rpo_vn_valueize): Remove ineffective code here.
7635 2018-08-29  Richard Biener  <rguenther@suse.de>
7637         PR tree-optimization/87132
7638         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
7639         when skipping defs reachable over backedges.
7641 2018-08-29  Richard Biener  <rguenther@suse.de>
7643         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
7644         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
7645         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
7646         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
7647         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
7648         (vn_reference_lookup_3): Remove use of const_parms.
7649         (free_rpo_vn): Do not free const_parms.
7650         (do_rpo_vn): Do not call init_const_parms.
7651         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
7652         SSA_NAME_POINTS_TO_READONLY_MEMORY.
7653         (call_may_clobber_ref_p_1): Likewise.
7655 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
7657         PR other/86726
7658         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
7659         (-O): Ditto.
7660         (-ftree-scev-cprop): Document.
7662 2018-08-29  Jan Hubicka  <jh@suse.cz>
7664         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
7665         parameters.
7666         (sreal constructor): Update.
7667         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
7668         sreal:operator/): Update.
7670 2018-08-29  Martin Liska  <mliska@suse.cz>
7672         * tree-switch-conversion.c (switch_conversion::expand):
7673         Strenghten assumption about gswitch statements.
7675 2018-08-29  Richard Biener  <rguenther@suse.de>
7677         PR tree-optimization/87117
7678         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
7679         re-value-number released SSA VDEFs.
7681 2018-08-29  Richard Biener  <rguenther@suse.de>
7683         PR tree-optimization/87126
7684         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
7686 2018-08-28  Jim Wilson  <jimw@sifive.com>
7688         * config/riscv/pic.md: Rewrite.
7689         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
7690         invalid address.
7691         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
7692         (SOFTF, default_load, softload, softstore): New.
7694 2018-08-28  Jeff Law  <law@redhat.com>
7696         * fold-const.c (fold_binary_loc): Remove recently added assert.
7698 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
7700         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
7701         to OP parmeter of generated function.
7703 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
7705         PR tree-optimization/87009
7706         * match.pd: Add boolean optimizations.
7708 2018-08-28  Martin Sebor  <msebor@redhat.com>
7710         PR middle-end/86631
7711         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
7712         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
7713         (pass_walloca::gate): Use it.
7714         (alloca_call_type): Same.
7715         (pass_walloca::execute): Same.
7716         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
7718 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
7720         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
7721         GCC_VERSION for usage of "__gcc_dump_printf__" format from
7722         >= 3005 to >= 9000.
7724 2018-08-28  Richard Biener  <rguenther@suse.de>
7726         PR tree-optimization/87124
7727         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
7728         constants before looking up avail.
7730 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
7732         PR middle-end/87099
7733         * calls.c (maybe_warn_nonstring_arg): Punt early if
7734         warn_stringop_overflow is zero.  Don't call get_range_strlen
7735         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
7736         Swap comparison operands to have constants on rhs.  Only use
7737         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
7738         increment lenrng[0].
7740 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
7742         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
7743         use of tree_to_shwi.  Remove duplicated test for the size being
7744         a whole number of bytes.
7746 2018-08-28  Richard Biener  <rguenther@suse.de>
7748         PR tree-optimization/87117
7749         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
7750         Handle removed stmt without LHS (GIMPLE_NOP).
7752 2018-08-28  Richard Biener  <rguenther@suse.de>
7754         PR tree-optimization/87117
7755         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
7756         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
7758 2018-08-28  Richard Biener  <rguenther@suse.de>
7760         PR tree-optimization/87117
7761         * tree-ssa-pre.c (compute_avail): Do not make expressions
7762         with predicated values available.
7763         (get_expr_value_id): Assert we do not run into predicated value
7764         expressions.
7766 2018-08-28  Richard Biener  <rguenther@suse.de>
7768         PR tree-optimization/87117
7769         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
7770         get virtual operands.
7771         (get_expr_operands): Handle STRING_CST like other decls.
7773 2018-08-28  Martin Liska  <mliska@suse.cz>
7775         * tree.h: Update documentation of fndecl_built_in_p
7776         functions.
7779 2018-08-27  Jeff Law  <law@redhat.com>
7780         PR tree-optimization/87110
7781         * tree-ssa-dse.c (compute_trims): Handle non-constant
7782         TYPE_SIZE_UNIT.
7784 2018-08-27  Martin Sebor  <msebor@redhat.com>
7786         PR tree-optimization/86914
7787         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
7789 2018-08-27  Martin Sebor  <msebor@redhat.com>
7791         PR tree-optimization/87112
7792         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
7793         the type of the bound argument.
7795 2018-08-27  Jeff Law  <law@redhat.com>
7797         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
7798         type does not have a TYPE_SIZE_UNIT.
7800 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
7802         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
7803         with include of backend.h.
7805 2018-08-27  Richard Biener  <rguenther@suse.de>
7807         PR tree-optimization/86927
7808         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
7809         use const cond reduction code.
7811 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
7813         PR tree-optimization/85758
7814         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
7816 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7818         PR c++/87091
7819         * diagnostic-show-locus.c (class layout_range): Update for
7820         conversion of show_caret_p to a tri-state.
7821         (layout_range::layout_range): Likewise.
7822         (make_range): Likewise.
7823         (layout::maybe_add_location_range): Likewise.
7824         (layout::should_print_annotation_line_p): Don't show annotation
7825         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
7826         (layout::get_state_at_point): Update for conversion of
7827         show_caret_p to a tri-state.  Bail out early for
7828         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
7829         underlining or source colorization.
7830         (gcc_rich_location::add_location_if_nearby): Update for conversion
7831         of show_caret_p to a tri-state.
7832         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
7833         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
7834         Likewise.
7835         (selftest::test_one_liner_labels): Likewise.
7836         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
7837         conversion of show_caret_p to a tri-state.
7838         * pretty-print.c (text_info::set_location): Likewise.
7839         * pretty-print.h (text_info::set_location): Likewise.
7840         * substring-locations.c (format_warning_n_va): Likewise.
7841         * tree-diagnostic.c (default_tree_printer): Likewise.
7842         * tree-pretty-print.c (newline_and_indent): Likewise.
7844 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7846         PR c++/87091
7847         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
7848         line above for line-insertion fix-it hints.
7849         (selftest::test_fixit_insert_containing_newline): Update the
7850         expected results, and add a test with line-numbering enabled.
7852 2018-08-27  Martin Liska  <mliska@suse.cz>
7854         PR sanitizer/86962
7855         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
7856         params with DECL_HAS_VALUE_EXPR_P.
7858 2018-08-27  Martin Liska  <mliska@suse.cz>
7860         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
7861         selected expansion strategy.
7863 2018-08-27  Martin Liska  <mliska@suse.cz>
7865         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
7866         * builtins.c (is_builtin_fn): Likewise.
7867         * attribs.c (diag_attr_exclusions): Use new function
7868         fndecl_built_in_p and remove check for FUNCTION_DECL if
7869         possible.
7870         (builtin_mathfn_code): Likewise.
7871         (fold_builtin_expect): Likewise.
7872         (fold_call_expr): Likewise.
7873         (fold_builtin_call_array): Likewise.
7874         (fold_call_stmt): Likewise.
7875         (set_builtin_user_assembler_name): Likewise.
7876         (is_simple_builtin): Likewise.
7877         * calls.c (gimple_alloca_call_p): Likewise.
7878         (maybe_warn_nonstring_arg): Likewise.
7879         * cfgexpand.c (expand_call_stmt): Likewise.
7880         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
7881         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
7882         (cgraph_node::verify_node): Likewise.
7883         * cgraphclones.c (build_function_decl_skip_args): Likewise.
7884         (cgraph_node::create_clone): Likewise.
7885         * config/arm/arm.c (arm_insert_attributes): Likewise.
7886         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
7887         * dse.c (scan_insn): Likewise.
7888         * expr.c (expand_expr_real_1): Likewise.
7889         * fold-const.c (operand_equal_p): Likewise.
7890         (fold_binary_loc): Likewise.
7891         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
7892         * gimple-low.c (lower_stmt): Likewise.
7893         * gimple-pretty-print.c (dump_gimple_call): Likewise.
7894         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7895         Likewise.
7896         * gimple.c (gimple_build_call_from_tree): Likewise.
7897         (gimple_call_builtin_p): Likewise.
7898         (gimple_call_combined_fn): Likewise.
7899         * gimplify.c (gimplify_call_expr): Likewise.
7900         (gimple_boolify): Likewise.
7901         (gimplify_modify_expr): Likewise.
7902         (gimplify_addr_expr): Likewise.
7903         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
7904         * ipa-cp.c (determine_versionability): Likewise.
7905         * ipa-fnsummary.c (compute_fn_summary): Likewise.
7906         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
7907         * ipa-split.c (visit_bb): Likewise.
7908         (split_function): Likewise.
7909         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7910         * lto-cgraph.c (input_node): Likewise.
7911         * lto-streamer-out.c (write_symbol): Likewise.
7912         * omp-low.c (setjmp_or_longjmp_p): Likewise.
7913         (lower_omp_1): Likewise.
7914         * predict.c (strip_predict_hints): Likewise.
7915         * print-tree.c (print_node): Likewise.
7916         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
7917         * trans-mem.c (is_tm_irrevocable): Likewise.
7918         (is_tm_load): Likewise.
7919         (is_tm_simple_load): Likewise.
7920         (is_tm_store): Likewise.
7921         (is_tm_simple_store): Likewise.
7922         (is_tm_abort): Likewise.
7923         (tm_region_init_1): Likewise.
7924         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
7925         * tree-cfg.c (verify_gimple_call): Likewise.
7926         (move_stmt_r): Likewise.
7927         (stmt_can_terminate_bb_p): Likewise.
7928         * tree-eh.c (lower_eh_constructs_2): Likewise.
7929         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
7930         * tree-inline.c (remap_gimple_stmt): Likewise.
7931         (copy_bb): Likewise.
7932         (estimate_num_insns): Likewise.
7933         (fold_marked_statements): Likewise.
7934         * tree-sra.c (scan_function): Likewise.
7935         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
7936         (optimize_stack_restore): Likewise.
7937         (pass_fold_builtins::execute): Likewise.
7938         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
7939         (mark_all_reaching_defs_necessary_1): Likewise.
7940         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
7941         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7942         (pass_forwprop::execute): Likewise.
7943         * tree-ssa-loop-im.c (stmt_cost): Likewise.
7944         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7945         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
7946         * tree-ssa-strlen.c (get_string_length): Likewise.
7947         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
7948         (find_func_aliases_for_call): Likewise.
7949         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
7950         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
7951         * tree-tailcall.c (find_tail_calls): Likewise.
7952         * tree.c (need_assembler_name_p): Likewise.
7953         (free_lang_data_in_decl): Likewise.
7954         (get_call_combined_fn): Likewise.
7955         * ubsan.c (is_ubsan_builtin_p): Likewise.
7956         * varasm.c (incorporeal_function_p): Likewise.
7957         * tree.h (DECL_BUILT_IN): Remove and replace with
7958         fndecl_built_in_p.
7959         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
7960         (fndecl_built_in_p): New.
7962 2018-08-27  Martin Liska  <mliska@suse.cz>
7964         PR tree-optimization/86847
7965         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
7966         Dump also subtree probability.
7967         (switch_decision_tree::do_jump_if_equal): New function.
7968         (switch_decision_tree::emit_case_nodes): Handle special
7969         situations in balanced tree that can be emitted much simpler.
7970         Fix calculation of probabilities that happen in tree expansion.
7971         * tree-switch-conversion.h (struct cluster): Add
7972         is_single_value_p.
7973         (struct simple_cluster): Likewise.
7974         (struct case_tree_node): Add new function has_child.
7975         (do_jump_if_equal): New.
7977 2018-08-27  Martin Liska  <mliska@suse.cz>
7979         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
7980         Add new argument to bit_test_cluster constructor.
7981         (bit_test_cluster::emit): Set bits really number of values
7982         handlel by a test.
7983         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
7984         probability argument.
7985         * tree-switch-conversion.h (struct bit_test_cluster):
7986         Add m_handles_entire_switch.
7988 2018-08-27  Martin Liska  <mliska@suse.cz>
7990         PR tree-optimization/86702
7991         * tree-switch-conversion.c (jump_table_cluster::emit):
7992         Make probabilities even for values in jump table
7993         according to number of cases handled.
7994         (switch_decision_tree::compute_cases_per_edge): Pass
7995         argument to reset_out_edges_aux function.
7996         (switch_decision_tree::analyze_switch_statement): Likewise.
7997         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
7998         Make it static.
8000 2018-08-27  Martin Liska  <mliska@suse.cz>
8002         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8003         cfun argument explicitly.
8004         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8005         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8006         function gimple_switch_default_bb.
8007         (convert_switch_statements):
8008         (expand_builtins):
8009         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8010         * stmt.c (label_to_block_fn): Use label_to_block and pass
8011         cfun argument explicitly and use gimple_switch_label_bb.
8012         (expand_case): Likewise.
8013         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8014         cfun argument explicitly. Likewise.
8015         (make_edges_bb): Likewise.
8016         (make_cond_expr_edges): Likewise.
8017         (get_cases_for_edge): Likewise.
8018         (make_gimple_switch_edges): Likewise.
8019         (label_to_block_fn): Likewise.
8020         (label_to_block): Likewise.
8021         (make_goto_expr_edges): Likewise.
8022         (make_gimple_asm_edges): Likewise.
8023         (main_block_label): Likewise.
8024         (group_case_labels_stmt): Likewise.
8025         (find_taken_edge_computed_goto): Likewise.
8026         (find_taken_edge_switch_expr): Likewise.
8027         (gimple_verify_flow_info): Likewise.
8028         (gimple_redirect_edge_and_branch): Likewise.
8029         (gimple_switch_label_bb): New function.
8030         (gimple_switch_default_bb): Likewise.
8031         (gimple_switch_edge): Likewise.
8032         (gimple_switch_default_edge): Likewise.
8033         * tree-cfg.h (label_to_block_fn): Remove and replace ...
8034         (label_to_block): ... with this.
8035         (gimple_switch_label_bb): New.
8036         (gimple_switch_default_bb): Likewise.
8037         (gimple_switch_edge): Likewise.
8038         (gimple_switch_default_edge): Likewise.
8039         * tree-cfgcleanup.c (convert_single_case_switch): Use
8040         new gimple functions and pass new argument to label_to_block.
8041         (cleanup_control_flow_bb):
8042         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8043         cfun argument explicitly.
8044         (make_eh_edges): Likewise.
8045         (redirect_eh_dispatch_edge): Likewise.
8046         (lower_resx): Likewise.
8047         (lower_eh_dispatch): Likewise.
8048         (maybe_remove_unreachable_handlers): Likewise.
8049         (unsplit_eh): Likewise.
8050         (cleanup_empty_eh): Likewise.
8051         (verify_eh_edges): Likewise.
8052         (verify_eh_dispatch_edge): Likewise.
8053         * tree-ssa-dom.c (record_edge_info): Likewise.
8054         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8055         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8056         (thread_through_normal_block): Likewise.
8057         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8058         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8059         * tree-switch-conversion.c (switch_conversion::collect): Use new
8060         gimple functions.
8061         (switch_conversion::check_final_bb): Likewise.
8062         (switch_conversion::gather_default_values): Pass new argument
8063         to label_to_block.
8064         (switch_conversion::build_constructors): Likewise.
8065         (switch_decision_tree::compute_cases_per_edge): Use new
8066         gimple_switch_edge function.
8067         (switch_decision_tree::analyze_switch_statement): Pass new argument
8068         to label_to_block.
8069         (switch_decision_tree::try_switch_expansion): Use
8070         gimple_switch_default_edge.
8071         * tree-vrp.c (find_switch_asserts): Pass new argument
8072         to label_to_block.
8073         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8074         (vr_values::simplify_switch_using_ranges): Likewise.
8076 2018-08-27  Richard Biener  <rguenther@suse.de>
8078         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8079         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8081         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8082         (struct vn_nary_op_s): Add unwind_to member.  Add
8083         predicated_values flag and put result into a union together
8084         with a linked list of vn_pval.
8085         (struct vn_ssa_aux): Add name member to make maintaining
8086         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8087         needed info, dfsnum, low, visited, on_sccstack, use_processed
8088         and range_info_anti_range_p members.
8089         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8090         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8091         New functions.
8092         (vn_valueize): New global.
8093         (vn_context_bb): Likewise.
8094         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8095         VN_INFO_PTR_INFO): Remove.
8096         * tree-ssa-sccvn.c: ... (rewrite)
8097         (pass_fre::execute): For -O2+ initialize loops and run
8098         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8099         run RPO VN in non-optimistic mode.
8100         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8101         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8102         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8103         (rpo-vn-max-loop-depth): Document.
8104         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8105         when valuezing the VUSE signals we walked out of the region.
8106         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8107         (phi_translate): Set VN context block to use for availability
8108         lookup.
8109         (compute_avail): Likewise.
8110         (pre_valueize): New function.
8111         (pass_pre::execute): Adjust to the RPO VN API.
8113         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8114         (propagate_constants_for_unrolling): Remove.
8115         (tree_unroll_loops_completely): Perform value-numbering
8116         on the unrolled bodies loop parent.
8118 2018-08-27  Richard Biener  <rguenther@suse.de>
8120         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8121         for partial antic compute.
8123 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8125         PR rtl-optimization/87065
8126         * combine.c (simplify_if_then_else): Formatting fix.
8127         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8128         check.
8129         (known_cond): Don't return const_true_rtx for vector modes.  Use
8130         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8132 2018-08-27  Martin Liska  <mliska@suse.cz>
8134         PR gcov-profile/87069
8135         * gcov.c (process_file): Record files already processed
8136         and warn about a file being processed multiple times.
8138 2018-08-27  Martin Liska  <mliska@suse.cz>
8140         PR driver/83193
8141         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8142         Set default values for x_aarch64_*_string strings.
8143         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8144         prefix.  For -mabi do not print '=ABI' in help and use
8145         <option_value> format for -msve-vector-bits and -moverride
8146         options.
8148 2018-08-26  Jeff Law  <law@redhat.com>
8150         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8151         "cfg.h"
8153 2018-08-26  Marek Polacek  <polacek@redhat.com>
8155         PR c++/87029, Implement -Wredundant-move.
8156         * doc/invoke.texi: Document -Wredundant-move.
8158 2018-08-25  Martin Sebor  <msebor@redhat.com>
8160         PR tree-optimization/87059
8161         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8162         to the same type as the other.
8163         * fold-const.c (fold_binary_loc): Assert expectation.
8165 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8167         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8168         extraneous parentheses, dead code section and formatting.
8170 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8172         PR c++/87091
8173         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8174         wide enough for jumps in the line-numbering to be visible.
8175         (layout::print_gap_in_line_numbering): New member function.
8176         (layout::calculate_line_spans): When using line numbering, merge
8177         line spans that are only 1 line apart.
8178         (diagnostic_show_locus): When printing line numbers, show gaps in
8179         line numbering directly, rather than printing headers.
8180         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8181         line-numbering with multiple line spans.
8182         (selftest::test_fixit_insert_containing_newline_2): Add test of
8183         line-numbering, in which the spans are close enough to be merged.
8185 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8187         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8188         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8189         * tree-vrp.c (range_is_nonnull): Remove.
8190         (range_includes_zero_p): Accept value_range instead of min/max.
8191         (extract_range_from_binary_expr_1): Do not early bail on
8192         POINTER_PLUS_EXPR.
8193         Use range_includes_zero_p instead of range_is_nonnull.
8194         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8195         of range_is_nonnull.
8196         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8197         special case VR_ANTI_RANGE.
8198         (vrp_finalize): Same.
8199         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8200         instead of min/max.
8201         (range_is_nonnull): Remove.
8202         * vr-values.c (vrp_stmt_computes_nonzero): Use
8203         range_includes_zero_p instead of range_is_nonnull.
8204         (extract_range_basic): Pass value_range to range_includes_zero_p
8205         instead of range_is_nonnull.
8207 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8209         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8210         * emit-rtl.h (rtl_data): Remove return_bnd.
8211         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8212         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8213         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8214         (POINTER_BOUNDS_MODE): Remove definition.
8215         (make_pointer_bounds_mode): Remove.
8216         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8217         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8218         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8219         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8220         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8221         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8222         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8224         * config/i386/i386-modes.def (BND32, BND64): Remove.
8225         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8226         (dbx64_register_map): Ditto.
8227         (svr4_dbx_register_map): Ditto.
8228         (indirect_thunk_bnd_needed): Remove.
8229         (indirect_thunks_bnd_used): Ditto.
8230         (indirect_return_bnd_needed): Ditto.
8231         (indirect_return_via_cx_bnd): Ditto.
8232         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8233         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8234         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8235         (output_indirect_thunk_function): Remove handling of
8236         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8237         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8238         (ix86_save_reg): Remove handling of crtl->return_bnd.
8239         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8240         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8241         and UNSPEC_BNDLX_ADDR.
8242         (ix86_output_indirect_branch_via_reg): Remove handling of
8243         indirect_thunk_prefix_bnd.
8244         (ix86_output_indirect_branch_via_push): Ditto.
8245         (ix86_output_function_return): Ditto.
8246         (ix86_output_indirect_function_return): Ditto.
8247         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8248         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8249         (CALL_USED_REGISTERS): Ditto.
8250         (REG_ALLOC_ORDER): Update for removal of bound registers.
8251         (HI_REGISTER_NAMES): Ditto.
8252         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8253         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8254         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8255         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8256         (FIRST_PSEUDO_REG): Update.
8257         (BND): Remove mode iterator.
8258         * config/i386/predicates.md (bnd_mem_operator): Remove.
8260 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8262         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8263         vectors.
8265 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8267         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8268         the case in which the permute needs only a single element and
8269         repeats for every vector of the result.  Extend that case to
8270         handle variable-length vectors.
8271         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8273 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8275         PR debug/79342
8276         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8277         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8279 2018-08-24  Richard Biener  <rguenther@suse.de>
8281         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8282         bb_flags_allocated members.
8283         (auto_flag): New RAII class for allocating flags.
8284         (auto_edge_flag): New RAII class for allocating edge flags.
8285         (auto_bb_flag): New RAII class for allocating bb flags.
8286         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8287         flag dynamically.
8288         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8289         in favor of temporarily allocated BB flag.
8290         * hsa-brig.c: Re-order includes.
8291         * hsa-dump.c: Likewise.
8292         * hsa-regalloc.c: Likewise.
8293         * print-rtl.c: Likewise.
8294         * profile-count.c: Likewise.
8296 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8298         PR target/86989
8299         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8300         the TOC register.
8302 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8304         PR 87073/bootstrap
8305         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8306         from wide_int_range_multiplicative_op.
8308 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8310         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8311         "permutaion".
8313 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8315         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8316         to 'expanded'.
8318 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8320         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8321         full GENERIC expression used for replacement.
8323 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8325         * tree-vrp.c (abs_extent_range): Remove.
8326         (extract_range_into_wide_ints): Pass wide ints by reference.
8327         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8328         Pass wide ints by reference in all calls to
8329         extract_range_into_wide_ints.
8330         * wide-int-range.cc (wide_int_range_div): New.
8331         * wide-int-range.h (wide_int_range_div): New.
8332         (wide_int_range_includes_zero_p): New.
8333         (wide_int_range_zero_p): New.
8335 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8337         * config/aarch64/aarch64.md (arches): New enum.
8338         (arch): New enum attr.
8339         (arch_enabled): New attr.
8340         (enabled): Now uses arch_enabled only.
8341         (simd, sve, fp16): Removed attribute.
8342         (fp): Attr now defined in terms of 'arch'.
8343         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8344         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8345         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8346         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8347         attributes into 'arch'.
8348         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8349         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8350         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8351         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8352         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8353         'simd' attribute into 'arch'.
8354         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8355         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8356         Convert use of 'fp' attribute to 'arch'.
8357         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8358         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8359         into 'arch'.
8360         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8361         (different modes) Merge 'fp' and 'simd' into 'arch'.
8362         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8363         'simd' into 'arch'.
8365 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8367         PR rtl-optimization/87026
8368         * expmed.c (canonicalize_comparison): If we can no longer create
8369         pseudoregisters, don't.
8371 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8373         PR target/86951
8374         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8375         prototype.
8376         * config/arm/arm.c (speculation_barrier_libfunc): New static
8377         variable.
8378         (arm_init_libfuncs): Initialize it.
8379         (arm_emit_speculation_barrier): New function.
8380         * config/arm/arm.md (speculation_barrier): Call
8381         arm_emit_speculation_barrier for architectures that do not have
8382         DSB or ISB.
8383         (speculation_barrier_insn): Only match on Armv7 or later.
8385 2018-08-23  Richard Biener  <rguenther@suse.de>
8387         PR middle-end/87024
8388         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8389         calls.
8391 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8393         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8394         of single-vector TBLs.
8395         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8396         one input is given.
8398 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8400         PR target/85910
8401         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8402         aarch64_evpc_tbl guard.
8404 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8406         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8407         behaviour.
8409 2018-08-22  Martin Sebor  <msebor@redhat.com>
8411         PR middle-end/87052
8412         * tree-pretty-print.c (pretty_print_string): Add argument.
8413         (dump_generic_node): Call to pretty_print_string with string size.
8415 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8417         PR rtl-optimization/86771
8418         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8419         of two SETs into those two SETs, one to be placed at i2, if that SETs
8420         destination is modified between i2 and i3.
8422 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8424         PR tree-optimization/86725
8425         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8426         function.
8427         (vect_analyze_scalar_cycles_1): Check it.
8429 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8431         PR tree-optimization/86725
8432         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8433         an outer loop phi as a double reduction, make sure that the
8434         single user of the phi result is an inner loop phi.
8436 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8438         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8439         grouped stores with gaps to a strided group.
8441 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8443         * tree-vect-stmts.c (get_group_load_store_type)
8444         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8445         first statement in a group.
8447 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8449         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8450         the sequence used in gcc/gcc.c.
8452 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8454         PR other/704
8455         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8456         building it.
8458 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8460         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8461         Darwin10-specific unwinder-shim.
8462         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8463         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8464         New to cater for Darwin10 Rosetta.
8466 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8468         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8469         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8471 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8473         PR bootstrap/81033
8474         PR target/81733
8475         PR target/52795
8476         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8477         (dwarf2out_switch_text_section): Generate a local label for the
8478         second function sub-section and apply it as the second FDE start
8479         label.
8480         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8481         second sub-section start.
8483 2018-08-22  Richard Biener  <rguenther@suse.de>
8485         PR tree-optimization/86988
8486         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8488 2018-08-22  Richard Biener  <rguenther@suse.de>
8490         PR tree-optimization/86945
8491         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8493 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8495         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8496         a comment about how uses of r2 for .sdata2 come about.
8498 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8500         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8502 2018-08-21  Marek Polacek  <polacek@redhat.com>
8504         PR c++/86981, Implement -Wpessimizing-move.
8505         * doc/invoke.texi: Document -Wpessimizing-move.
8507 2018-08-21  Jan Hubicka  <jh@suse.cz>
8509         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8510         * tree.h (is_redundant_typedef): Remove.
8511         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8513 2018-08-21  Jan Hubicka  <jh@suse.cz>
8515         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8516         when possible.
8518 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8520         * expmed.c (extract_low_bits): Reject invalid subregs early.
8522 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8524         PR middle-end/86121
8525         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8526         behaviour.
8528 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8530         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8531         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8532         * config/vxworks.c: Likewise.
8534 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8536         * config/vxworks.c: Set targetm.have_ctors_dtors
8537         if HAVE_INITFINI_ARRAY_SUPPORT.
8538         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8539         if HAVE_INITFINI_ARRAY_SUPPORT.
8541 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8543         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8544         default search path for VxWorks < 7.
8546 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8548         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8549         (get_format_string): Refer to c_getstr.
8551 2018-08-21  Tom de Vries  <tdevries@suse.de>
8553         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8554         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8555         (debuginfo_early_stop): Declare.
8556         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8557         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8558         (debuginfo_early_stop): New function.
8559         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8560         and debuginfo_early_stop.
8561         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8562         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8563         (general_init): Call debuginfo_early_init.
8564         (finalize): Call debuginfo_fini.
8565         (do_compile): Call debuginfo_init.
8566         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8567         -fdump-early-debug.
8568         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8570 2018-08-21  Tom de Vries  <tdevries@suse.de>
8572         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8573         flag_dump_noaddr and flag_dump_unnumbered.
8575 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
8577         * wide-int-range.cc (wide_int_range_abs): New.
8578         (wide_int_range_order_set): Rename from wide_int_range_min_max.
8579         * wide-int-range.h (wide_int_range_abs): New.
8580         (wide_int_range_min_max): New.
8581         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
8582         case to call wide_int_range_abs.
8583         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
8584         (extract_range_from_abs_expr): Delete.
8586 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
8588         PR target/87033
8589         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
8590         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
8591         for indexed loads.
8593 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8594             Jeff Law <law@redhat.com>
8596         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
8597         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
8599 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
8601         PR other/84889
8602         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
8603         (decl_attributes): Likewise.
8604         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
8605         instance.
8606         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8607         * diagnostic-core.h (class auto_diagnostic_group): New class.
8608         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
8609         (diagnostic_report_diagnostic): Handle the first diagnostics within
8610         a group.
8611         (emit_diagnostic): Add auto_diagnostic_group instance.
8612         (inform): Likewise.
8613         (inform_n): Likewise.
8614         (warning): Likewise.
8615         (warning_at): Likewise.
8616         (warning_n): Likewise.
8617         (pedwarn): Likewise.
8618         (permerror): Likewise.
8619         (error): Likewise.
8620         (error_n): Likewise.
8621         (error_at): Likewise.
8622         (sorry): Likewise.
8623         (fatal_error): Likewise.
8624         (internal_error): Likewise.
8625         (internal_error_no_backtrace): Likewise.
8626         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
8627         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
8628         * diagnostic.h (struct diagnostic_context): Add fields
8629         "diagnostic_group_nesting_depth",
8630         "diagnostic_group_emission_count", "begin_group_cb",
8631         "end_group_cb".
8632         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
8633         Add auto_diagnostic_group instance(s).
8634         (find_explicit_erroneous_behavior): Likewise.
8635         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
8636         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
8637         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
8638         (gimplify_va_arg_expr): Likewise.
8639         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
8640         (HSA_SORRY_AT): Likewise.
8641         * ipa-devirt.c (compare_virtual_tables): Likewise.
8642         (warn_odr): Likewise.
8643         * multiple_target.c (expand_target_clones): Likewise.
8644         * opts-common.c (cmdline_handle_error): Likewise.
8645         * reginfo.c (globalize_reg): Likewise.
8646         * substring-locations.c (format_warning_n_va): Likewise.
8647         * tree-inline.c (expand_call_inline): Likewise.
8648         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
8649         * tree-ssa-loop-niter.c
8650         (do_warn_aggressive_loop_optimizations): Likewise.
8651         * tree-ssa-uninit.c (warn_uninit): Likewise.
8652         * tree.c (warn_deprecated_use): Likewise.
8654 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
8656         PR target/87014
8657         * config/i386/i386.md (eh_return): Always update EH return
8658         address in word_mode.
8660 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
8662         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
8663         TARGET_SPLIT_COMPLEX_ARG is defined.
8665 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8667         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
8669 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8671         PR target/86984
8672         * expr.c (expand_assignment): Assert that bitpos is positive.
8673         (store_field): Likewise
8674         (expand_expr_real_1): Make sure that bitpos is positive.
8675         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
8676         integer overflow.
8678 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8680         * Makefile.in (CPP_ID_DATA_H): Delete.
8681         (CPP_INTERNAL_H): Don't add it.
8682         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
8683         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
8685 2018-08-20  Richard Biener  <rguenther@suse.de>
8687         PR tree-optimization/78655
8688         * tree-vrp.c (extract_range_from_binary_expr_1): Make
8689         pointer + offset nonnull if either operand is nonnull work.
8691 2018-08-20  Tom de Vries  <tdevries@suse.de>
8693         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
8694         unless the referenced die describes the added property using
8695         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
8696         Otherwise, add a DW_AT_location to the referenced die.
8698 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
8700         PR target/86994
8701         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
8702         register_operand when calling ix86_set_reg_reg_cost.
8703         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
8704         Set *total to 0 for operands that satisfy x86_64_immediate_operand
8705         predicate and to 1 otherwise.
8707 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
8709         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
8710         emit a diagnostic that it is not supported and reset the option.
8711         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
8712         supported and consume the option.  (ASM_FINAL_SPEC): New.
8714 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8716         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
8717         a sentence.
8719 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
8721         C-SKY port: Documentation
8723         * doc/extend.texi (C-SKY Function Attributes): New section.
8724         * doc/invoke.texi (Option Summary): Add C-SKY options.
8725         (C-SKY Options): New section.
8726         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
8728 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8729             Huibin Wang  <huibin_wang@c-sky.com>
8730             Sandra Loosemore  <sandra@codesourcery.com>
8731             Chung-Lin Tang  <cltang@codesourcery.com>
8733         C-SKY port: Backend implementation
8735         * config/csky/*: New.
8736         * common/config/csky/*: New.
8738 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8739             Huibin Wang  <huibin_wang@c-sky.com>
8740             Sandra Loosemore  <sandra@codesourcery.com>
8741             Chung-Lin Tang  <cltang@codesourcery.com>
8742             Andrew Jenner  <andrew@codesourcery.com>
8744         C-SKY port: Configury
8746         * config.gcc (csky-*-*): New.
8747         * configure.ac: Add csky to targets for dwarf2 debug_line support.
8748         * configure: Regenerated.
8750 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
8752         * dump-context.h: Include "dumpfile.h".
8753         (dump_context::dump_printf_va): Convert final param from va_list
8754         to va_list *.  Convert from ATTRIBUTE_PRINTF to
8755         ATTRIBUTE_GCC_DUMP_PRINTF.
8756         (dump_context::dump_printf_loc_va): Likewise.
8757         * dumpfile.c: Include "stringpool.h".
8758         (make_item_for_dump_printf_va): Delete.
8759         (make_item_for_dump_printf): Delete.
8760         (class dump_pretty_printer): New class.
8761         (dump_pretty_printer::dump_pretty_printer): New ctor.
8762         (dump_pretty_printer::emit_items): New member function.
8763         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
8764         function.
8765         (dump_pretty_printer::emit_item): New member function.
8766         (dump_pretty_printer::stash_item): New member function.
8767         (dump_pretty_printer::format_decoder_cb): New member function.
8768         (dump_pretty_printer::decode_format): New member function.
8769         (dump_context::dump_printf_va): Reimplement in terms of
8770         dump_pretty_printer.
8771         (dump_context::dump_printf_loc_va): Convert final param from va_list
8772         to va_list *.
8773         (dump_context::begin_scope): Reimplement call to
8774         make_item_for_dump_printf.
8775         (dump_printf): Update for change to dump_printf_va.
8776         (dump_printf_loc): Likewise.
8777         (selftest::test_capture_of_dump_calls): Convert "stmt" from
8778         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
8779         with %T, %E, and %G.
8780         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
8781         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
8782         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
8783         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
8784         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
8785         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
8786         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
8787         within a dump_printf_loc call to "%wu".
8788         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
8789         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
8790         missing space after "=".
8791         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
8792         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
8793         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
8794         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
8795         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
8796         duplicate "vectorized" from message.
8798 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
8800         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
8801         polyNxK_t element's TYPE_STRING_FLAG.
8803 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8805         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
8806         (they were unnamed before).  Fix comments.
8808 2018-08-17  Nathan Sidwell  <nathan@acm.org>
8810         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
8812 2018-08-17  Richard Biener  <rguenther@suse.de>
8814         PR tree-optimization/86841
8815         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
8817 2018-08-17  Martin Liska  <mliska@suse.cz>
8819         * common.opt: Remove Warn, Init and Report for options with
8820         Ignore/Deprecated flag. Warning is done automatically for
8821         Deprecated flags.
8822         * config/i386/i386.opt: Likewise.
8823         * config/ia64/ia64.opt: Likewise.
8824         * config/rs6000/rs6000.opt: Likewise.
8825         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8826         Remove usage of flag_check_pointer_bounds.
8827         * lto-wrapper.c (merge_and_complain): Do not handle
8828         OPT_fcheck_pointer_bounds.
8829         (append_compiler_options): Likewise.
8830         * opt-functions.awk: Do not handle Deprecated.
8831         * optc-gen.awk: Check that Var, Report and Init are not
8832         used for an option with Ignore/Deprecated flag.
8833         * opts-common.c (decode_cmdline_option): Do not report
8834         CL_ERR_DEPRECATED.
8835         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
8836         options.
8837         * opts.h (struct cl_option): Remove cl_deprecated flag.
8838         (CL_ERR_DEPRECATED): Remove error enum value.
8840 2018-08-17  Richard Biener  <rguenther@suse.de>
8842         PR middle-end/86505
8843         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
8844         across a va-arg-pack using call adjust its return value accordingly.
8846 2018-08-16  Martin Sebor  <msebor@redhat.com>
8848         PR tree-optimization/86853
8849         * gimple-ssa-sprintf.c (struct format_result): Rename member.
8850         (struct fmtresult): Add member and initialize it in ctors.
8851         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
8852         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
8853         (format_directive): Set POSUNDER4K when MAYFAIL is set.
8854         (parse_directive): Handle %C same as %c.
8855         (sprintf_dom_walker::compute_format_length): Adjust.
8856         (is_call_safe): Adjust.
8858 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8860         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
8861         for determining how to count the elements.
8862         * builtins.h (c_strlen): Adjust prototype.
8863         * expr.c (string_constant): Add new parameter mem_size.
8864         Set *mem_size appropriately.
8865         * expr.h (string_constant): Adjust protoype.
8866         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
8867         * gimple-fold.h (get_range_strlen): Adjust prototype.
8868         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
8869         (format_string): Call get_string_length with eltsize.
8871 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8873         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
8874         to emit the span, rather than setting it as the prefix.
8876 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8878         * diagnostic-show-locus.c (layout::start_annotation_line): Add
8879         "margin_char" parameter, defaulting to space.  Use it in place
8880         of pp_space for the initial part of the margin.
8881         (layout::print_leading_fixits): Use '+' when filling the margin
8882         of line-insertion fix-it hints.
8884 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8886         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
8887         Delete.
8889 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8891         * config/rs6000/altivec.md: Don't set length attribute to the default
8892         value.
8893         * config/rs6000/darwin.md: Ditto.
8894         * config/rs6000/dfp.md: Ditto.
8895         * config/rs6000/htm.md: Ditto.
8896         * config/rs6000/rs6000.md: Ditto.
8897         * config/rs6000/sync.md: Ditto.
8898         * config/rs6000/vsx.md: Ditto.
8900 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8902         * config/rs6000/altivec.md: Don't set length attribute to the default
8903         value, for branch instructions.
8904         * config/rs6000/darwin.md: Ditto.
8905         * config/rs6000/rs6000.md: Ditto.
8907 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8909         * config/rs6000/rs6000.md (length): Always define as const_int 4.
8910         (unnamed conditional branch define_insn): Set length to 4 or 8
8911         depending on offset.
8912         (<bd>_<mode>): Similar, for alternative 0.
8913         (<bd>tf_<mode>): Ditto.
8915 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8917         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
8919 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
8921         * doc/rtl.texi: Replace old RTX class names with new names.
8924 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
8926         * expmed.h (canonicalize_comparison): New declaration.
8927         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
8928         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
8929         * optabs.c (prepare_cmp_insn): Likewise.
8930         * rtl.h (unsigned_condition_p): New function which checks if a
8931         comparison operator is unsigned.
8933 2018-08-16  Nathan Sidwell  <nathan@acm.org>
8935         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
8936         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
8938 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8940         PR target/84711
8941         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
8942         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
8943         (mov<mov>): ..this and enable unconditionally.
8945 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8947         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
8949 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
8951         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
8952         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
8953         "Common" with "Target".
8955 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8957         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
8958         * doc/invoke.texi (mmitigate-rop): Remove.
8959         * config/i386/i386.c: Do not include "regrename.h".
8960         (ix86_rop_should_change_byte_p, reg_encoded_number)
8961         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
8962         Remove.
8963         (ix86_reorg): Remove call to ix86_mitigate_rop.
8964         * config/i386/i386.md (attr "modrm_class"): Remove.
8965         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
8966         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
8967         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
8969 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
8971         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
8972         allow folding of mergeh() and mergel() for the float and double types.
8973         (fold_mergehl_helper): Rework to handle building a permute tree
8974         for float vectors.
8976 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8978         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
8979         for TARGET_SSE.
8981 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
8983         * common.opt (fdiagnostics-show-labels): New option.
8984         * diagnostic-show-locus.c (class layout_range): Add field
8985         "m_label".
8986         (class layout): Add field "m_show_labels_p".
8987         (layout_range::layout_range): Add param "label" and use it to
8988         initialize m_label.
8989         (make_range): Pass in NULL for new "label" param of layout_range's
8990         ctor.
8991         (layout::layout): Initialize m_show_labels_p.
8992         (layout::maybe_add_location_range): Pass in loc_range->m_label
8993         when constructing layout_range instances.
8994         (struct line_label): New struct.
8995         (layout::print_any_labels): New member function.
8996         (layout::print_line): Call it if label-printing is enabled.
8997         (selftest::test_one_liner_labels): New test.
8998         (selftest::test_diagnostic_show_locus_one_liner): Call it.
8999         * diagnostic.c (diagnostic_initialize): Initialize
9000         context->show_labels_p.
9001         * diagnostic.h (struct diagnostic_context): Add field
9002         "show_labels_p".
9003         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9004         -fno-diagnostics-show-labels.
9005         * dwarf2out.c (gen_producer_string): Add
9006         OPT_fdiagnostics_show_labels to the ignored options.
9007         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9008         param.
9009         (gcc_rich_location::maybe_add_expr): Likewise.
9010         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9011         label" param, defaulting to NULL.
9012         (gcc_rich_location::add_expr): Add "label" param.
9013         (gcc_rich_location::maybe_add_expr): Likewise.
9014         (class text_range_label): New class.
9015         (class range_label_for_type_mismatch): New class.
9016         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9017         of format_warning_va.
9018         (fmtwarn_n): Likewise for new params of format_warning_n_va.
9019         * lto-wrapper.c (merge_and_complain): Add
9020         OPT_fdiagnostics_show_labels to the "pick one setting" options.
9021         (append_compiler_options): Likewise to the dropped options.
9022         (append_diag_options): Likewise to the passed-on options.
9023         * opts.c (common_handle_option): Handle the new option.
9024         * selftest-diagnostic.c
9025         (test_diagnostic_context::test_diagnostic_context): Enable
9026         show_labels_p.
9027         * substring-locations.c: Include "gcc-rich-location.h".
9028         (format_warning_n_va): Add "fmt_label" and "param_label" params
9029         and use them as appropriate.
9030         (format_warning_va): Add "fmt_label" and "param_label" params,
9031         passing them on to format_warning_n_va.
9032         (format_warning_at_substring): Likewise.
9033         (format_warning_at_substring_n): Likewise.
9034         * substring-locations.h (format_warning_va): Add "fmt_label" and
9035         "param_label" params.
9036         (format_warning_n_va): Likewise.
9037         (format_warning_at_substring): Likewise.
9038         (format_warning_at_substring_n): Likewise.
9039         * toplev.c (general_init): Initialize global_dc->show_labels_p.
9041 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
9043         PR testsuite/86519
9044         * builtins.c (expand_builtin_memcmp): Do not expand the call
9045         when overflow is detected.
9047 2018-08-15  Martin Sebor  <msebor@redhat.com>
9049         PR tree-optimization/71625
9050         * config/aarch64/aarch64-builtins.c
9051         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9053 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
9055         * config/s390/s390.c (s390_reorg): Remove loop.
9057 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9059         * config/darwin.c
9060          (darwin_function_switched_text_sections): Delete.
9061         * gcc/config/darwin.h
9062          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9064 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9066         PR target/81685
9067         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9068         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
9069         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9071 2018-08-15  Martin Liska  <mliska@suse.cz>
9073         PR tree-optimization/86925
9074         * predict.c (expr_expected_value_1): When taking
9075         later predictor, assign also probability.
9076         Use fold_build2_initializer_loc in order to fold
9077         the expression in -frounding-math.
9079 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9081         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9082         patterns.
9083         (expand_vec_perm_1): Try the new method.
9085 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9087         PR target/86547
9088         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9089         Check whether lra_live_max_point is 0 before dividing.
9091 2018-08-14  Martin Sebor  <msebor@redhat.com>
9093         PR tree-optimization/86650
9094         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9095         (vrp_prop::check_mem_ref): Same.
9097 2018-08-13  Liu Hao <lh_mouse@126.com>
9099         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9100         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9101         and clear it thereafter, as it only works for DBCS.
9103 2018-08-13  Liu Hao <lh_mouse@126.com>
9105         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9106         handle returned by _get_osf_handle().
9108 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9110         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9111         for folding vec_perm.
9113 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9115         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9116         Add support for gimple-folding of vec_pack() and vec_unpack()
9117         intrinsics.
9119 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9121         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9122         vec_xst variants to the list.
9123         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9124         vector loads and stores.
9126 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9128         * config.gcc (rs6000-ibm-aix4.x): Delete.
9129         (rs6000-ibm-aix5.1): Delete.
9130         (rs6000-ibm-aix5.2): Delete.
9131         (rs6000-ibm-aix5.3): Delete.
9132         * config/rs6000/aix43.h: Delete.
9133         * config/rs6000/aix51.h: Delete.
9134         * config/rs6000/aix52.h: Delete.
9135         * config/rs6000/t-aix43: Delete.
9137 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9139         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9140         New function.
9141         (s390_decompose_address): Factor out constant pool ref
9142         decomposition.
9144 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9146         * config/nds32/nds32-predicates.c
9147         (nds32_can_use_bclr_p): Change return type as bool.
9148         (nds32_can_use_bset_p): Ditto.
9149         (nds32_can_use_btgl_p): Ditto.
9150         (nds32_can_use_bitci_p): Ditto.
9151         * config/nds32/nds32-protos.h
9152         (nds32_can_use_bclr_p): Change declaration.
9153         (nds32_can_use_bset_p): Ditto.
9154         (nds32_can_use_btgl_p): Ditto.
9155         (nds32_can_use_bitci_p): Ditto.
9157 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9159         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9160         Support -msched-prolog-epilog option.
9161         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9163 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9165         * common/config/nds32/nds32-common.c
9166         (nds32_option_optimization_table): Enalbe -malways-align.
9168 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9170         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9171         extra_headers.
9172         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9173         OPT_misr_secure_ case.
9174         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9175         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9176         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9177         secure attribute.
9178         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9179         (nds32_isr_info): New field security_level.
9180         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9181         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9182         * config/nds32/nds32.opt (misr-secure): New option.
9183         * config/nds32/nds32_init.inc: New file.
9184         * config/nds32/nds32_isr.h: New file.
9186 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9188         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9189         Update comment for atomic instructions.
9190         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9191         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9192         Remove.
9193         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9194         load for __ATOMIC_SEQ_CST model.
9195         (atomic_loaddi_1): Remove float register target.
9196         (atomic_storedi): Handle CONST_INT values.
9197         (atomic_storedi_1): Remove float register source.  Add special case
9198         for zero value.
9199         (memory_barrier): New expander and insn.
9201 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9203         PR tree-optimization/86835
9204         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9205         new_stmt after def_gsi, make sure to insert new_square_stmt after
9206         that stmt, not 2 stmts before it.
9208 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9210         PR target/82418
9211         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9212         instead of SWI48.
9214 2018-08-10  Martin Liska  <mliska@suse.cz>
9216         PR target/83610
9217         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9218         function type.
9219         * builtins.c (expand_builtin_expect_with_probability):
9220         New function.
9221         (expand_builtin_expect_with_probability): New function.
9222         (build_builtin_expect_predicate): Add new argumnet probability
9223         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9224         (fold_builtin_expect):
9225         (fold_builtin_2):
9226         (fold_builtin_3):
9227         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9228         * builtins.h (fold_builtin_expect): Set new argument.
9229         * doc/extend.texi: Document __builtin_expect_with_probability.
9230         * doc/invoke.texi: Likewise.
9231         * gimple-fold.c (gimple_fold_call): Pass new argument.
9232         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9233         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9234         * predict.c (get_predictor_value): New function.
9235         (expr_expected_value): Add new argument probability. Assume
9236         that predictor and probability are always non-null.
9237         (expr_expected_value_1): Likewise.  For __builtin_expect and
9238         __builtin_expect_with_probability set probability.  Handle
9239         combination in binary expressions.
9240         (tree_predict_by_opcode): Simplify code by simply calling
9241         get_predictor_value.
9242         (pass_strip_predict_hints::execute): Add handling of
9243         BUILT_IN_EXPECT_WITH_PROBABILITY.
9244         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9245         new predictor.
9246         * tree.h (DECL_BUILT_IN_P): New function.
9248 2018-08-10  Martin Liska  <mliska@suse.cz>
9250         PR tree-optimization/85799
9251         * passes.def: Add argument for pass_strip_predict_hints.
9252         * predict.c (class pass_strip_predict_hints): Add new argument
9253         early_p.
9254         (strip_predictor_early): New function.
9255         (pass_strip_predict_hints::execute): Call the function to
9256         strip predictors.
9257         (strip_predict_hints): New function.
9258         * predict.def: Fix comment.
9260 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9262         * Makefile.in: Clarify which tm.texi to copy over to assert the
9263         right to grant a GFDL license for all.
9265 2018-08-09  Jeff Law  <law@redhat.com>
9267         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9268         unused variable.
9270 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9272         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9273         prototype.
9275 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9277         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9278         reductions for variable-length vectors.
9280 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9282         PR other/84889
9283         * common.opt (fdiagnostics-show-line-numbers): New option.
9284         * diagnostic-show-locus.c (class layout): Add fields
9285         "m_show_line_numbers_p" and "m_linenum_width";
9286         (num_digits): New function.
9287         (test_num_digits): New function.
9288         (layout::layout): Initialize new fields.  Update m_x_offset
9289         logic to handle any left margin.
9290         (layout::print_source_line): Print line number when requested.
9291         (layout::start_annotation_line): New member function.
9292         (layout::print_annotation_line): Call it.
9293         (layout::print_leading_fixits): Likewise.
9294         (layout::print_trailing_fixits): Likewise.  Update calls to
9295         move_to_column for new parameter.
9296         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9297         it to potentially call start_annotation_line.
9298         (layout::show_ruler): Call start_annotation_line.
9299         (selftest::test_line_numbers_multiline_range): New selftest.
9300         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9301         and selftest::test_line_numbers_multiline_range.
9302         * diagnostic.c (diagnostic_initialize): Initialize
9303         show_line_numbers_p.
9304         * diagnostic.h (struct diagnostic_context): Add field
9305         "show_line_numbers_p".
9306         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9307         -fno-diagnostics-show-line-numbers.
9308         * dwarf2out.c (gen_producer_string): Add
9309         OPT_fdiagnostics_show_line_numbers to the ignored options.
9310         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9311         one setting" options.
9312         (append_compiler_options): Likewise to the dropped options.
9313         (append_diag_options): Likewise to the passed-on options.
9314         * opts.c (common_handle_option): Handle the new option.
9315         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9317 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9319         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9320         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9321         third argument of type "const signed char" to descriptions of
9322         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9323         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9324         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9325         __builtin_bcdsub_ov functions.
9327 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9329         PR tree-optimization/86858
9330         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9331         flow_bb_inside_loop_p calls.
9333 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9335         PR tree-optimization/86871
9336         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9337         instead of gimple_assign_lhs.
9339 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9341         PR target/86887
9342         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9343         register constraint to operand 0.
9344         (add<mode>3_carryinC): Likewise.
9345         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9347 2018-08-09  Martin Liska  <mliska@suse.cz>
9349         PR c/86895
9350         * common.opt: Remove extra line.
9352 2018-08-09  Martin Liska  <mliska@suse.cz>
9354         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9355         at the end of a line, make first letter capital and end up
9356         a sentence with a dot.
9357         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9358         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9359         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9360         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9361         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9362         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9363         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9364         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9365         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9366         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9367         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9368         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9369         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9370         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9371         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9372         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9373         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9374         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9376 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9378         PR target/84332
9379         * config/s390/s390.c (s390_option_override_internal): Reduce the
9380         stack-clash-protection-probe-interval param if it would be too big
9381         for z900.
9383 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9385         PR target/46179
9386         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9387         * config/m68k/m68k.c (handle_move_double): Don't call
9388         m68k_final_prescan_insn.
9389         (m68k_adjust_decorated_operand): Renamed from
9390         m68k_final_prescan_insn, remove first and third operand and
9391         simplify.
9392         (print_operand): Call it.
9393         (print_operand_address): Call it.
9395 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9397         * diagnostic.c (diagnostic_report_current_module): Use
9398         linemap_included_from & linemap_included_from_linemap.
9400 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9402         * config/aarch64/aarch64-cores.def: Add phecda core.
9403         * config/aarch64/aarch64-tune.md: Regenerate.
9404         * doc/invoke.texi: Add phecda core.
9406 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9408         PR target/85295
9409         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9410         definitions.
9411         * config/s390/s390.md ("movti"): Add more alternatives for
9412         constant to GPR copies.
9414 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9416         * config/s390/s390.c: Fix whitespace damage throughout the file.
9417         * config/s390/s390.h: Likewise.
9418         * config/s390/tpf.h: Likewise.
9420 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9422         * config/s390/s390.c (s390_loadrelative_operand_p):
9423         Remove TARGET_CPU_ZARCH usages.
9424         (s390_rtx_costs): Likewise.
9425         (s390_legitimate_constant_p): Likewise.
9426         (s390_cannot_force_const_mem): Likewise.
9427         (legitimate_reload_constant_p): Likewise.
9428         (s390_preferred_reload_class): Likewise.
9429         (legitimize_pic_address): Likewise.
9430         (legitimize_tls_address): Likewise.
9431         (s390_split_branches): Removed.
9432         (s390_add_execute): Removed.
9433         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9434         (s390_mainpool_start): Likewise.
9435         (s390_mainpool_finish): Likewise.
9436         (s390_mainpool_cancel): Removed.
9437         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9438         (s390_chunkify_cancel): Likewise.
9439         (s390_return_addr_rtx): Likewise.
9440         (s390_register_info): Remove split_branches_pending_p uages.
9441         (s390_optimize_register_info): Likewise.
9442         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9443         split_branches_pending_p usages.
9444         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9445         (s390_load_got): Likewise.
9446         (s390_expand_split_stack_prologue): Likewise.
9447         (output_asm_nops): Likewise.
9448         (s390_function_profiler): Likewise.
9449         (s390_emit_call): Likewise.
9450         (s390_conditional_register_usage): Likewise.
9451         (s390_optimize_prologue): Likewise.
9452         (s390_reorg): Remove TARGET_CPU_ZARCH and
9453         split_branches_pending_p usages.
9454         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9455         usages.
9456         (s390_output_indirect_thunk_function): Likewise.
9457         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9458         (TARGET_CPU_ZARCH_P): Removed.
9459         (struct machine_function): Remove split_branches_pending_p.
9460         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9462 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9464         * common/config/s390/s390-common.c (processor_flags_table):
9465         Remove flags.
9466         * config.gcc: Remove with_arch/with_tune support.
9467         * config/s390/2064.md: Remove cpu attribute comparisons.
9468         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9469         Remove MTN.
9470         * config/s390/linux.h (ASM_SPEC):
9471         Remove -march support.
9472         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9473         Use a table to get an arch level.
9474         * config/s390/s390-opts.h (enum processor_type):
9475         Remove enum values.
9476         * config/s390/s390.c
9477         (processor_table): Remove entries, add arch_level values.
9478         (s390_issue_rate): Remove cases.
9479         (s390_option_override): Adjust
9480         s390_option_override_internal() call.
9481         (s390_option_override_internal): Remove deprecation warning.
9482         (s390_valid_target_attribute_tree): Adjust
9483         s390_option_override_internal() call.
9484         * config/s390/s390.h (struct s390_processor):
9485         Share with s390-c.c, add arch_level field.
9486         * config/s390/s390.md:
9487         Remove occurrences in cpu attribute.
9488         * config/s390/s390.opt: Remove -march/-mtune support.
9489         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9490         * doc/invoke.texi: Remove deprecation warning.
9492 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9494         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9495         global.
9496         (qdf24xx_tunings): Set vector cost structure to
9497         qdf24xx_vector_cost.
9499         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9500         <register_sextend>: Set to 3.
9502 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9504         PR target/86838
9505         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9506         * config/aarch64/aarch64-simd.md
9507         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9508         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9509         (aarch64_frecpx<mode>): ...this new pattern.
9510         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9511         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9513 2018-08-07  Martin Liska  <mliska@suse.cz>
9515         PR middle-end/83023
9516         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9517         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9518         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9519         * doc/extend.texi: Document that malloc attribute adds
9520         hit to compiler.
9522 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9524         PR target/86785
9525         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9526         Define to speculation_safe_value_not_needed.
9528 2018-08-06  Jeff Law  <law@redhat.com>
9530         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9531         the vr_values instance to cprop_into_stmt.
9532         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9533         (cprop_operand): Also query EVRP to determine if OP is a constant.
9535 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9537         * diagnostic.c (diagnostic_report_current_module): Reroll
9538         included-at loop.  Translate text.
9540 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9542         * function-tests.c (selftest::test_expansion_to_rtl): Call
9543         free_after_compilation.
9545 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9547         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9549 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9551         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9552         loops with memory block operations from getting unrolled.
9554 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9556         PR target/86807
9557         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9558         Define to speculation_safe_value_not_needed.
9560 2018-08-06  Jeff Law  <law@redhat.com>
9562         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9563         assert.
9565 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9567         PR target/86662
9568         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9569         with all enabled __intN types.
9571         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9573 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9575         * alias.c (record_set): Check for clobber high.
9576         * cfgexpand.c (expand_gimple_stmt): Likewise.
9577         * combine-stack-adj.c (single_set_for_csa): Likewise.
9578         * combine.c (find_single_use_1): Likewise.
9579         (set_nonzero_bits_and_sign_copies): Likewise.
9580         (get_combine_src_dest): Likewise.
9581         (is_parallel_of_n_reg_sets): Likewise.
9582         (try_combine): Likewise.
9583         (record_dead_and_set_regs_1): Likewise.
9584         (reg_dead_at_p_1): Likewise.
9585         (reg_dead_at_p): Likewise.
9586         * dce.c (deletable_insn_p): Likewise.
9587         (mark_nonreg_stores_1): Likewise.
9588         (mark_nonreg_stores_2): Likewise.
9589         * df-scan.c (df_find_hard_reg_defs): Likewise.
9590         (df_uses_record): Likewise.
9591         (df_get_call_refs): Likewise.
9592         * dwarf2out.c (mem_loc_descriptor): Likewise.
9593         * haifa-sched.c (haifa_classify_rtx): Likewise.
9594         * ira-build.c (create_insn_allocnos): Likewise.
9595         * ira-costs.c (scan_one_insn): Likewise.
9596         * ira.c (equiv_init_movable_p): Likewise.
9597         (rtx_moveable_p): Likewise.
9598         (interesting_dest_for_shprep): Likewise.
9599         * jump.c (mark_jump_label_1): Likewise.
9600         * postreload-gcse.c (record_opr_changes): Likewise.
9601         * postreload.c (reload_cse_simplify): Likewise.
9602         (struct reg_use): Add source expr.
9603         (reload_combine): Check for clobber high.
9604         (reload_combine_note_use): Likewise.
9605         (reload_cse_move2add): Likewise.
9606         (move2add_note_store): Likewise.
9607         * print-rtl.c (print_pattern): Likewise.
9608         * recog.c (decode_asm_operands): Likewise.
9609         (store_data_bypass_p): Likewise.
9610         (if_test_bypass_p): Likewise.
9611         * regcprop.c (kill_clobbered_value): Likewise.
9612         (kill_set_value): Likewise.
9613         * reginfo.c (reg_scan_mark_refs): Likewise.
9614         * reload1.c (maybe_fix_stack_asms): Likewise.
9615         (eliminate_regs_1): Likewise.
9616         (elimination_effects): Likewise.
9617         (mark_not_eliminable): Likewise.
9618         (scan_paradoxical_subregs): Likewise.
9619         (forget_old_reloads_1): Likewise.
9620         * reorg.c (find_end_label): Likewise.
9621         (try_merge_delay_insns): Likewise.
9622         (redundant_insn): Likewise.
9623         (own_thread_p): Likewise.
9624         (fill_simple_delay_slots): Likewise.
9625         (fill_slots_from_thread): Likewise.
9626         (dbr_schedule): Likewise.
9627         * resource.c (update_live_status): Likewise.
9628         (mark_referenced_resources): Likewise.
9629         (mark_set_resources): Likewise.
9630         * rtl.c (copy_rtx): Likewise.
9631         * rtlanal.c (reg_referenced_p): Likewise.
9632         (single_set_2): Likewise.
9633         (noop_move_p): Likewise.
9634         (note_stores): Likewise.
9635         * sched-deps.c (sched_analyze_reg): Likewise.
9636         (sched_analyze_insn): Likewise.
9638 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9640         * cse.c (invalidate_reg): New function extracted from...
9641         (invalidate): ...here.
9642         (canonicalize_insn): Check for clobber high.
9643         (invalidate_from_clobbers): invalidate clobber highs.
9644         (invalidate_from_sets_and_clobbers): Likewise.
9645         (count_reg_usage): Check for clobber high.
9646         (insn_live_p): Likewise.
9647         * cselib.c (cselib_expand_value_rtx_1):Likewise.
9648         (cselib_invalidate_regno): Check for clobber in setter.
9649         (cselib_invalidate_rtx): Pass through setter.
9650         (cselib_invalidate_rtx_note_stores):
9651         (cselib_process_insn): Check for clobber high.
9652         * cselib.h (cselib_invalidate_rtx): Add operand.
9654 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9656         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
9657         (mark_not_eliminable): Likewise.
9658         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
9659         * lra-lives.c (process_bb_lives): Check for clobber high.
9660         * lra.c (new_insn_reg): Remember clobber highs.
9661         (collect_non_operand_hard_regs): Check for clobber high.
9662         (lra_set_insn_recog_data): Likewise.
9663         (add_regs_to_insn_regno_info): Likewise.
9664         (lra_update_insn_regno_info): Likewise.
9666 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9668         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
9669         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
9671 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9673         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
9674         (copy_insn_1): Likewise.
9675         (gen_hard_reg_clobber_high): New gen function.
9676         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
9677         * genemit.c (gen_exp): Likewise.
9678         (gen_emit_seq): Pass through info.
9679         (gen_insn): Check for CLOBBER_HIGH.
9680         (gen_expand): Pass through info.
9681         (gen_split): Likewise.
9682         (output_add_clobbers): Likewise.
9683         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
9684         (remove_clobbers): Likewise.
9685         * rtl.h (gen_hard_reg_clobber_high): New declaration.
9687 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9689         * doc/rtl.texi (clobber_high): Add.
9690         (parallel): Add in clobber high
9691         * rtl.c (rtl_check_failed_code3): Add function.
9692         * rtl.def (CLOBBER_HIGH): Add expression.
9693         * rtl.h (RTL_CHECKC3): Add macro.
9694         (rtl_check_failed_code3): Add declaration.
9695         (XC3EXP): Add macro.
9697 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
9699         PR target/86386
9700         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
9701         cfun->machine->max_used_stack_alignment if needed.
9703 2018-08-04  Martin Sebor  <msebor@redhat.com>
9705         PR tree-optimization/86571
9706         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
9707         NaN output to 4.
9709 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
9711         PR target/86799
9712         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9713         Define.
9715 2018-08-03  Jeff Law  <law@redhat.com>
9717         PR target/86795
9718         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9719         Define to speculation_safe_value_not_needed.
9721 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
9723         * doc/gcov.texi (-x): Remove duplicate "to".
9724         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
9725         (-Wif-not-aligned): Remove duplicate "is".
9726         (-flto): Remove duplicate "the".
9727         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
9728         duplicate "v5.00.b".
9729         (MSP430 Options): Remove duplicate "and" from the description
9730         of "-mgprel-sec=regexp".
9731         (x86 Options): Remove duplicate copies of "vmldLog102" and
9732         vmlsLog104 from description of "-mveclibabi=type".
9734 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
9736         * internal-fn.h (first_commutative_argument): Declare.
9737         * internal-fn.c (first_commutative_argument): New function.
9738         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
9739         restrictions for pattern statements.  Use first_commutative_argument
9740         to look for commutative operands in calls to internal functions.
9742 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
9744         * Makefile.in (wide-int-range.o): New.
9745         * tree-vrp.c: Move all the wide_int_* functions to...
9746         * wide-int-range.cc: ...here.
9747         * tree-vrp.h: Move all the wide_int_* prototypes to...
9748         * wide-int-range.h: ...here.
9750 2018-08-03  Tom de Vries  <tdevries@suse.de>
9752         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
9753         UI_NONE.
9754         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
9755         * except.c (output_function_exception_table): Do early exit if
9756         targetm_common.except_unwind_info (&global_options) == UI_NONE.
9758 2018-08-03  Martin Liska  <mliska@suse.cz>
9760         * predict.c (dump_prediction): Change to 2 digits
9761         in fraction part.
9763 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9765         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
9766         neon_dup_q to...
9767         (falkor_am_1_gtov_gtov): ... a new insn reservation.
9769 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
9771         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
9772         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
9773         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
9774         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
9775         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
9777 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9779         * diagnostic-show-locus.c (diagnostic_show_locus): Use
9780         pp_take_prefix when saving the existing prefix.
9781         * diagnostic.c (diagnostic_append_note): Likewise.
9782         * langhooks.c (lhd_print_error_function): Likewise.
9783         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
9784         param's type.  Free the existing prefix.
9785         (pp_take_prefix): New function.
9786         (pretty_printer::pretty_printer): Drop the prefix parameter.
9787         Rename the length parameter to match the comment.
9788         (pretty_printer::~pretty_printer): Free the prefix.
9789         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
9790         parameter.
9791         (struct pretty_printer): Drop the "const" from "prefix" field's
9792         type and clarify memory management.
9793         (pp_set_prefix): Drop the "const" from the 2nd param.
9794         (pp_take_prefix): New decl.
9796 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
9798         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
9799         for word_mode_ok here instead of passing as argument.
9800         (expand_block_compare): Change select_block_compare_mode() call.
9801         (expand_strncmp_gpr_sequence): New function.
9802         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
9804 2018-08-02  Jeff Law  <law@redhat.com>
9806         PR target/86790
9807         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9808         Define to speculation_safe_value_not_needed.
9810         PR target/86784
9811         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9812         Define to speculation_safe_value_not_needed.
9814 2018-08-02  Tom de Vries  <tdevries@suse.de>
9816         PR target/86660
9817         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
9818         function.  Return UI_TARGET unconditionally.
9819         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
9820         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
9822 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9824         * genemit.c (print_overload_test): Fix typo.
9826 2018-08-02  Richard Biener  <rguenther@suse.de>
9828         PR tree-optimization/86816
9829         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
9830         which checks for value availability before querying it.
9831         (gvn_uses_equal): Use it.
9832         (same_succ_hash): Likewise.
9833         (gimple_equal_p): Likewise.
9835 2018-08-02  Nick Clifton  <nickc@redhat.com>
9837         PR target/86813
9838         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9839         Define to speculation_safe_value_not_needed.
9841         PR target/86810
9842         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9843         Define to speculation_safe_value_not_needed.
9845         PR target/86810
9846         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9847         Define to speculation_safe_value_not_needed.
9849         PR target/86803
9850         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9851         Define to speculation_safe_value_not_needed.
9853         PR target/86797
9854         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9855         Define to speculation_safe_value_not_needed.
9857         PR target/86791
9858         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9859         Define to speculation_safe_value_not_needed.
9861         PR target/86789
9862         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9863         Define to speculation_safe_value_not_needed.
9865         PR target/86787
9866         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9867         Define to speculation_safe_value_not_needed.
9869         PR target/86782
9870         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9871         speculation_safe_value_not_needed.
9873         PR target/86781
9874         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
9875         to speculation_safe_value_not_needed.
9877 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9879         * doc/md.texi: Expand the documentation of instruction names
9880         to mention port-local uses.  Document '@' in pattern names.
9881         * read-md.h (overloaded_instance, overloaded_name): New structs.
9882         (mapping): Declare.
9883         (md_reader::handle_overloaded_name): New member function.
9884         (md_reader::get_overloads): Likewise.
9885         (md_reader::m_first_overload): New member variable.
9886         (md_reader::m_next_overload_ptr): Likewise.
9887         (md_reader::m_overloads_htab): Likewise.
9888         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
9889         m_next_overload_ptr and m_overloads_htab.
9890         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
9891         (get_mode_token, get_code_token, get_int_token): New functions.
9892         (map_attr_string): Add an optional argument that passes back
9893         the associated iterator.
9894         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
9895         (md_reader::handle_overloaded_name, add_overload_instance): New
9896         functions.
9897         (apply_iterators): Handle '@' names.  Report an error if '@'
9898         is used without iterators.
9899         (initialize_iterators): Initialize the new iterator_group fields.
9900         * genopinit.c (handle_overloaded_code_for)
9901         (handle_overloaded_gen): New functions.
9902         (main): Use them to print declarations of maybe_code_for_* and
9903         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
9904         * genemit.c (print_overload_arguments, print_overload_test)
9905         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
9906         (main): Use it to print definitions of maybe_code_for_* and
9907         maybe_gen_* functions.
9908         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
9909         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
9910         instead of explicit mode checks.
9911         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
9912         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
9913         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
9914         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
9915         (aarch64_expand_compare_and_swap): Likewise
9916         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
9917         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
9918         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
9919         (aarch64_constant_pool_reload_icode): Delete.
9920         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
9921         instead of aarch64_constant_pool_reload_icode.  Use
9922         code_for_aarch64_reload_mov instead of explicit mode checks.
9923         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
9924         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
9925         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
9926         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
9927         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
9928         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
9929         (aarch64_atomic_load_op_code): Delete.
9930         (aarch64_emit_atomic_load_op): Likewise.
9931         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
9932         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
9933         instead of aarch64_emit_atomic_load_op.
9934         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
9935         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
9936         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
9937         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
9938         character before the pattern name.
9939         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
9940         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
9941         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
9942         (aarch64_frecps<mode>): Likewise.
9943         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
9944         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
9945         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
9946         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
9947         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
9949 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9951         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
9952         Allow HFmode constants if TARGET_FP_F16INST.
9954 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
9956         PR target/86014
9957         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
9958         No longer check last store for clobber of address register.
9960 2018-08-02  Martin Liska  <mliska@suse.cz>
9962         PR gcov-profile/86817
9963         * gcov.c (process_all_functions): New function.
9964         (main): Call it.
9965         (process_file): Move functions processing to
9966         process_all_functions.
9968 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9970         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
9971         "const" to the "gimple *" and "rtx_insn *" parameters.
9972         * dumpfile.h (dump_user_location_t::dump_user_location_t):
9973         Likewise.
9974         (dump_location_t::dump_location_t): Likewise.
9976 2018-08-01  Martin Sebor  <msebor@redhat.com>
9978         PR tree-optimization/86650
9979         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
9980         rather than a "gcall *".  Directly pass the data of interest
9981         to percent_K_format, rather than building a temporary CALL_EXPR
9982         to hold it.
9983         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
9984         (gimple_fold_builtin_strncat): Adjust.
9985         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
9986         gcall* argument with gimple*.
9987         * gimple-ssa-warn-restrict.c (check_call): Same.
9988         (wrestrict_dom_walker::before_dom_children): Same.
9989         (builtin_access::builtin_access): Same.
9990         (check_bounds_or_overlap): Same
9991         (maybe_diag_overlap): Same.
9992         (maybe_diag_offset_bounds): Same.
9993         * tree-diagnostic.c (default_tree_printer): Move usage of
9994         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
9995         to this callsite.
9996         * tree-pretty-print.c (percent_K_format): Add argument.
9997         * tree-pretty-print.h: Add argument.
9998         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
9999         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10000         (maybe_diag_stxncpy_trunc): Same.
10001         (handle_builtin_stxncpy): Same.
10002         (handle_builtin_strcat): Same.
10004 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10006         * match.pd: Optimise pointer range checks.
10008 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10010         PR tree-optimization/86758
10011         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10012         to remove pattern statements.
10014 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10016         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10017         result of dfs_enumerate_from when constructing stmt_vec_infos,
10018         instead of additionally calling get_loop_body.
10020 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10022         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10023         parameter.
10024         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10025         When creating an iv, assert that the step is not known to be zero.
10026         (vect_setup_realignment): Update call accordingly.
10027         * tree-vect-stmts.c (vectorizable_store): Likewise.
10028         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
10030 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10032         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10033         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10034         (vectorizable_reduction): Likewise.
10035         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10036         (vect_detect_hybrid_slp_stmts): Likewise.
10037         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10039 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
10041         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10042         (wide_int_set_zero_nonzero_bits): ...this.
10043         (zero_nonzero_bits_from_vr): Rename to...
10044         (vrp_set_zero_nonzero_bits): ...this.
10045         (extract_range_from_multiplicative_op_1): Abstract wide int
10046         code...
10047         (wide_int_range_multiplicative_op): ...here.
10048         (extract_range_from_binary_expr_1): Extract wide int binary
10049         operations into their own functions.
10050         (wide_int_range_lshift): New.
10051         (wide_int_range_can_optimize_bit_op): New.
10052         (wide_int_range_shift_undefined_p): New.
10053         (wide_int_range_bit_xor): New.
10054         (wide_int_range_bit_ior): New.
10055         (wide_int_range_bit_and): New.
10056         (wide_int_range_trunc_mod): New.
10057         (extract_range_into_wide_ints): New.
10058         (vrp_shift_undefined_p): New.
10059         (extract_range_from_multiplicative_op): New.
10060         (vrp_can_optimize_bit_op): New.
10061         * tree-vrp.h (value_range::dump): New.
10062         (wide_int_range_multiplicative_op): New.
10063         (wide_int_range_lshift):New.
10064         (wide_int_range_shift_undefined_p): New.
10065         (wide_int_range_bit_xor): New.
10066         (wide_int_range_bit_ior): New.
10067         (wide_int_range_bit_and): New.
10068         (wide_int_range_trunc_mod): New.
10069         (zero_nonzero_bits_from_bounds): Rename to...
10070         (wide_int_set_zero_nonzero_bits): ...this.
10071         (zero_nonzero_bits_from_vr): Rename to...
10072         (vrp_set_zero_nonzero_bits): ...this.
10073         (range_easy_mask_min_max): Rename to...
10074         (wide_int_range_can_optimize_bit_op): this.
10075         * vr-values.c (simplify_bit_ops_using_ranges): Rename
10076         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10078 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10080         * tree-vectorizer.h (vect_orig_stmt): New function.
10081         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10082         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10083         (vect_create_epilog_for_reduction): Likewise.
10084         (vectorizable_live_operation): Likewise.
10085         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10086         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10087         * tree-vect-stmts.c (vectorizable_call): Likewise.
10088         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10090 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10092         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10093         argument.
10094         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10095         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10096         (vect_transform_loop): Likewise.
10097         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10099 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10101         * tree-vectorizer.h (vect_schedule_slp): Return void.
10102         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10103         (vect_schedule_slp): Likewise.
10105 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10107         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10108         argument.
10109         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10110         instances before the main loop, if any exist.
10112 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10114         PR tree-optimization/86749
10115         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10116         If the lhs is used in a COND_EXPR, check that it is being used
10117         as the "then" or "else" value.
10119 2018-08-01  Tom de Vries  <tdevries@suse.de>
10121         PR target/86800
10122         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10123         speculation_safe_value_not_needed.
10125 2018-08-01  Richard Biener  <rguenther@suse.de>
10127         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10128         as base and offset.
10130 2018-08-01  Martin Liska  <mliska@suse.cz>
10132         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10133         format how successful transformation is dumped.
10134         (gimple_mod_pow2_value_transform): Likewise.
10135         (gimple_mod_subtract_transform): Likewise.
10136         (gimple_stringops_transform): Likewise.
10138 2018-08-01  Martin Liska  <mliska@suse.cz>
10140         PR value-prof/35543
10141         * value-prof.c (interesting_stringop_to_profile_p):
10142         Simplify the code and add BUILT_IN_MEMMOVE.
10143         (gimple_stringops_transform): Likewise.
10145 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10147         * config/aarch64/aarch64-simd.md
10148         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10149         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10150         use GPI iterator instead of SI mode.
10152 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10154         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10155         rs6000_speculation_barrier.
10156         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10157         new barrier pattern name.
10159 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10161         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10162         (speculation_barrier): New insn.
10164 2018-08-01  Richard Biener  <rguenther@suse.de>
10166         PR bootstrap/86724
10167         * graphite.h: Include isl/id.h and isl/space.h to allow build
10168         with ISL 0.20.
10170 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10172         PR target/86651
10173         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10174         mode for COFF targets.
10175         * defaults.h (TARGET_COFF): Define.
10176         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10177         TARGET_COFF): Define.
10178         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10179         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10180         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10182 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10184         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10185         * gimple-streamer-out.c (output_bb): Save it.
10186         * lto-streamer-in.c (input_struct_function_base): Restore
10187         instance discriminator if available.  Create map on demand.
10188         * lto-streamer-out.c (output_struct_function_base): Save it if
10189         available.
10190         * final.c (decl_to_instance_map): Document LTO strategy.
10192 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10193             Olivier Hainque  <hainque@adacore.com>
10195         * debug.h (decl_to_instance_map_t): New type.
10196         (decl_to_instance_map): Declare.
10197         (maybe_create_decl_to_instance_map): New inline function.
10198         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10199         to track basic block discriminators.
10200         (final_start_function_1): Initialize them.
10201         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10202         bb_discriminator.
10203         (decl_to_instance_map): New variable.
10204         (map_decl_to_instance, maybe_set_discriminator): New functions.
10205         (notice_source_line): Set discriminator.
10207 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10209         * targhooks.c (default_have_speculation_safe_value): Add
10210         ATTRIBUTE_UNUSED.
10212 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10214         * dump-context.h: Include "pretty-print.h".
10215         (dump_context::refresh_dumps_are_enabled): New decl.
10216         (dump_context::emit_item): New decl.
10217         (class dump_context): Add fields "m_test_pp" and
10218         "m_test_pp_flags".
10219         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10220         (temp_dump_context::get_dumped_text): New decl.
10221         (class temp_dump_context): Add field "m_pp".
10222         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10223         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10224         m_test_pp.
10225         (set_dump_file): Update for above change.
10226         (set_alt_dump_file): Likewise.
10227         (dump_loc): New overload, taking a pretty_printer *.
10228         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10229         to any test pretty-printer.
10230         (make_item_for_dump_gimple_stmt): New function, adapted from
10231         optinfo::add_gimple_stmt.
10232         (dump_context::dump_gimple_stmt): Call it, and use the result,
10233         eliminating the direct usage of dump_file and alt_dump_file in
10234         favor of indirectly using them via emit_item.
10235         (make_item_for_dump_gimple_expr): New function, adapted from
10236         optinfo::add_gimple_expr.
10237         (dump_context::dump_gimple_expr): Call it, and use the result,
10238         eliminating the direct usage of dump_file and alt_dump_file in
10239         favor of indirectly using them via emit_item.
10240         (make_item_for_dump_generic_expr): New function, adapted from
10241         optinfo::add_tree.
10242         (dump_context::dump_generic_expr): Call it, and use the result,
10243         eliminating the direct usage of dump_file and alt_dump_file in
10244         favor of indirectly using them via emit_item.
10245         (make_item_for_dump_printf_va): New function, adapted from
10246         optinfo::add_printf_va.
10247         (make_item_for_dump_printf): New function.
10248         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10249         and use the result, eliminating the direct usage of dump_file and
10250         alt_dump_file in favor of indirectly using them via emit_item.
10251         (make_item_for_dump_dec): New function.
10252         (dump_context::dump_dec): Call it, and use the result,
10253         eliminating the direct usage of dump_file and alt_dump_file in
10254         favor of indirectly using them via emit_item.
10255         (make_item_for_dump_symtab_node): New function, adapted from
10256         optinfo::add_symtab_node.
10257         (dump_context::dump_symtab_node): Call it, and use the result,
10258         eliminating the direct usage of dump_file and alt_dump_file in
10259         favor of indirectly using them via emit_item.
10260         (dump_context::begin_scope): Reimplement, avoiding direct usage
10261         of dump_file and alt_dump_file in favor of indirectly using them
10262         via emit_item.
10263         (dump_context::emit_item): New member function.
10264         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10265         Set up test pretty-printer on the underlying context.  Call
10266         refresh_dumps_are_enabled.
10267         (temp_dump_context::~temp_dump_context): Call
10268         refresh_dumps_are_enabled.
10269         (temp_dump_context::get_dumped_text): New member function.
10270         (selftest::verify_dumped_text): New function.
10271         (ASSERT_DUMPED_TEXT_EQ): New macro.
10272         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10273         and then without optinfo enabled.  Add uses of
10274         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10275         * dumpfile.h: Update comment for the dump_* API.
10276         * optinfo-emit-json.cc
10277         (selftest::test_building_json_from_dump_calls): Update for new
10278         param for temp_dump_context ctor.
10279         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10280         and "m_owned" field.
10281         (optinfo_item::~optinfo_item): Likewise.
10282         (optinfo::add_item): New member function.
10283         (optinfo::emit): Update comment.
10284         (optinfo::add_string): Delete.
10285         (optinfo::add_printf): Delete.
10286         (optinfo::add_printf_va): Delete.
10287         (optinfo::add_gimple_stmt): Delete.
10288         (optinfo::add_gimple_expr): Delete.
10289         (optinfo::add_tree): Delete.
10290         (optinfo::add_symtab_node): Delete.
10291         (optinfo::add_dec): Delete.
10292         * optinfo.h (class dump_context): New forward decl.
10293         (optinfo::add_item): New decl.
10294         (optinfo::add_string): Delete.
10295         (optinfo::add_printf): Delete.
10296         (optinfo::add_printf_va): Delete.
10297         (optinfo::add_gimple_stmt): Delete.
10298         (optinfo::add_gimple_expr): Delete.
10299         (optinfo::add_tree): Delete.
10300         (optinfo::add_symtab_node): Delete.
10301         (optinfo::add_dec): Delete.
10302         (optinfo::add_poly_int): Delete.
10303         (optinfo_item::optinfo_item): Remove "owned" param.
10304         (class optinfo_item): Remove field "m_owned".
10306 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10308         PR middle-end/86705
10309         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10310         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10311         requested variable alignment.
10312         (expand_one_ssa_partition): Likewise.
10313         (expand_one_var): Likewise.
10315 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10317         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10318         to speculation_safe_value_not_needed.
10320 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10322         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10323         * targhooks.c (speculation_safe_value_not_needed): New function.
10324         * target.def (have_speculation_safe_value): Update documentation.
10325         * doc/tm.texi: Regenerated.
10327 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10329         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10330         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10331         expand.
10332         (despeculate_copy<ALLI:mode>_insn): New insn.
10333         (despeculate_copyti_insn): New insn.
10334         (despeculate_simple<ALLI:mode>): New insn
10335         (despeculate_simpleti): New insn.
10336         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10337         function.
10338         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10339         aarch64_speculation_safe_value.
10340         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10342 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10344         * config/aarch64/aarch64-speculation.cc: New file.
10345         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10346         before pass_reorder_blocks.
10347         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10348         prototype.
10349         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10350         X14 and X15 when tracking speculation.
10351         * config/aarch64/aarch64.md (register name constants): Add
10352         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10353         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10354         (speculation_barrier): New insn attribute.
10355         (cmp<mode>): Allow SP in comparisons.
10356         (speculation_tracker): New insn.
10357         (speculation_barrier): Add speculation_barrier attribute.
10358         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10359         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10360         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10362 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10364         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10365         aarch64_track_speculation is true.
10366         (tb<optab><mode>1): Likewise.
10367         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10368         generate CB[N]Z when tracking speculation.
10369         (aarch64_split_compare_and_swap): Likewise.
10370         (aarch64_split_atomic_op): Likewise.
10372 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10374         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10376 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10378         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10379         (speculation_barrier): New insn.
10381 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10383         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10384         * config/arm/arm.md (speculation_barrier): New expand.
10385         (speculation_barrier_insn): New pattern.
10387 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10389         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10390         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10391         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10392         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10393         list.
10394         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10395         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10396         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10397         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10398         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10399         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10400         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10401         * builtins.c (expand_speculation_safe_value): New function.
10402         (expand_builtin): Call it.
10403         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10404         * doc/extend.texi: Document __builtin_speculation_safe_value.
10405         * doc/md.texi: Document "speculation_barrier" pattern.
10406         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10407         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10408         * doc/tm.texi: Regenerated.
10409         * target.def (have_speculation_safe_value, speculation_safe_value): New
10410         hooks.
10411         * targhooks.c (default_have_speculation_safe_value): New function.
10412         (default_speculation_safe_value): New function.
10413         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10414         (default_speculation_safe_value): Add prototype.
10416 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10418         * dump-context.h (dump_context::dump_loc): New decl.
10419         * dumpfile.c (dump_context::dump_loc): New member function.
10420         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10421         and dump_gimple_stmt.
10422         (dump_context::dump_gimple_expr_loc): Likewise, using
10423         dump_gimple_expr.
10424         (dump_context::dump_generic_expr_loc): Likewise, using
10425         dump_generic_expr.
10426         (dump_context::dump_printf_loc_va): Likewise, using
10427         dump_printf_va.
10428         (dump_context::begin_scope): Explicitly using the global function
10429         "dump_loc", rather than the member function.
10431 2018-07-31  Martin Sebor  <msebor@redhat.com>
10433         PR tree-optimization/86741
10434         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10436 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10438         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10439         depend on whether prefetch instructions will be emitted or not.
10440         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10441         will be emitted or not.
10442         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10443         (TARGET_SETMEM_PFD): New macros.
10445 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10447         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10448         (NULL_STMT_VEC_INFO): Delete.
10449         (stmt_vec_info::operator*): Likewise.
10450         (stmt_vec_info::operator gimple *): Likewise.
10451         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10452         of NULL_STMT_VEC_INFO.
10453         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10454         (vect_reassociating_reduction_p): Likewise.
10455         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10456         (vectorizable_store): Likewise.
10457         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10458         (vec_info::free_stmt_vec_infos): Likewise.
10460 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10462         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10463         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10465 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10467         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10468         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10469         (vec_info::free_stmt_vec_info): New private member functions.
10470         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10471         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10472         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10473         set_stmt_vec_info_vec and free_stmt_vec_infos.
10474         * tree-vect-loop.c (_loop_vec_info): Remove call to
10475         set_stmt_vec_info_vec.
10476         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10477         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10478         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10479         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10480         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10481         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10482         with the clearing in the xcalloc.
10483         (stmt_vec_info_vec): Delete.
10484         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10485         (vectorize_loops): Likewise.
10486         (vec_info::~vec_info): Remove argument from call to
10487         free_stmt_vec_infos.
10488         (vec_info::add_stmt): Remove vinfo argument from call to
10489         new_stmt_vec_info.
10491 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10493         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10494         rather than a gimple stmt.
10495         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10496         information for pattern statements when passed the original
10497         statement; instead wait to be passed the pattern statement itself.
10498         Don't call set_vinfo_for_stmt here.
10499         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10500         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10501         stmt_vec_infos here.
10502         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10503         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10504         stmt_vec_infos entry.
10506 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10508         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10509         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10510         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10511         * tree-vect-stmts.c (vectorizable_call): Likewise.
10512         (vectorizable_simd_clone_call): Likewise.
10514 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10516         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10517         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10518         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10519         * tree-vect-loop.c (vect_transform_loop): Likewise.
10520         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10521         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10523 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10525         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10526         (vect_dr_stmt): Delete.
10527         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10528         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10529         of DR_VECT_AUX.
10530         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10531         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10532         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10533         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10534         (vect_analyze_data_refs): Likewise.
10535         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10536         argument.
10537         (vect_find_same_alignment_drs): Likewise.
10538         (vect_slp_analyze_node_dependences): Update calls accordingly.
10539         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10540         instead of DR_VECT_AUX.
10541         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10542         of a vector data references.  Use vec_info::lookup_dr instead of
10543         DR_VECT_AUX.
10544         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10545         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10546         instead of DR_VECT_AUX.
10548 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10550         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10551         dr_vec_info.
10552         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10553         accordingly.
10554         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10555         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10556         (vect_gen_prolog_loop_niters): Likewise.
10558 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10560         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10561         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10562         (vect_known_alignment_in_bytes, vect_dr_behavior)
10563         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10564         instead of data_references.  Update calls to other routines for
10565         which the same change has been made.
10566         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10567         dr_vec_infos instead of stmt_vec_infos.
10568         (vect_analyze_data_ref_dependence): Update call accordingly.
10569         (vect_slp_analyze_data_ref_dependence)
10570         (vect_record_base_alignments): Use DR_VECT_AUX.
10571         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10572         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10573         (vector_alignment_reachable_p, vect_get_data_access_cost)
10574         (vect_peeling_supportable, vect_analyze_group_access_1)
10575         (vect_analyze_group_access, vect_analyze_data_ref_access)
10576         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10577         (vect_compile_time_alias, vect_small_gap_p)
10578         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
10579         (vect_supportable_dr_alignment): Take references as dr_vec_infos
10580         instead of data_references.  Update calls to other routines for
10581         which the same change has been made.
10582         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
10583         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10584         (vect_slp_analyze_and_verify_node_alignment)
10585         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10586         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10587         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
10588         above changes.
10589         (_vect_peel_info::dr): Replace with...
10590         (_vect_peel_info::dr_info): ...this new field.
10591         (vect_peeling_hash_get_most_frequent)
10592         (vect_peeling_hash_choose_best_peeling): Update accordingly.
10593         (vect_peeling_hash_get_lowest_cost):
10594         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
10595         routines for which the same change has been made.
10596         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
10597         data_reference.
10598         * tree-vect-loop-manip.c (get_misalign_in_elems)
10599         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
10600         above changes.
10601         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10602         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
10603         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
10604         (get_group_load_store_type, get_negative_load_store_type)
10605         (vect_get_data_ptr_increment, vectorizable_store)
10606         (vectorizable_load): Likewise.
10607         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
10608         Update calls to other routines for which the same change has been made.
10610 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10612         * tree-vectorizer.h (vec_info::move_dr): New member function.
10613         (dataref_aux): Rename to...
10614         (dr_vec_info): ...this and add "dr" and "stmt" fields.
10615         (_stmt_vec_info::dr_aux): Update accordingly.
10616         (_stmt_vec_info::data_ref_info): Delete.
10617         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
10618         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
10619         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
10620         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
10621         of data_ref.
10622         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
10623         (STMT_VINFO_DR_INFO): New macro.
10624         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
10625         (set_dr_misalignment): Update after rename of dataref_aux.
10626         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
10627         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
10628         initialization of STMT_VINFO_DATA_REF.
10629         * tree-vectorizer.c (vec_info::move_dr): New function.
10630         * tree-vect-patterns.c (vect_recog_bool_pattern)
10631         (vect_recog_mask_conversion_pattern)
10632         (vect_recog_gather_scatter_pattern): Use it.
10633         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
10634         the "dr" and "stmt" fields of dr_vec_info instead of
10635         STMT_VINFO_DATA_REF.
10637 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10639         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
10640         (is_pattern_stmt_p): Use it.
10641         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
10642         on pattern statements.
10644 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10646         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
10647         original stmt as a stmt_vec_info rather than a gimple stmt.
10648         (vect_pattern_recog_1): Take the statement directly as a
10649         stmt_vec_info, rather than via a gimple_stmt_iterator.
10650         Update call to vect_mark_pattern_stmts.
10651         (vect_pattern_recog): Update calls accordingly.
10653 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10655         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
10656         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
10657         a vect_def_type for the first argument.
10658         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
10659         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
10660         operand if it isn't defined by a vectorized statement.
10661         (vect_build_gather_load_calls): Remove the mask_dt argument and
10662         update calls to vect_get_vec_def_for_stmt_copy.
10663         (vectorizable_bswap): Likewise the dt argument.
10664         (vectorizable_call): Update calls to vectorizable_bswap and
10665         vect_get_vec_def_for_stmt_copy.
10666         (vectorizable_simd_clone_call, vectorizable_assignment)
10667         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
10668         (vectorizable_comparison): Update calls to
10669         vect_get_vec_def_for_stmt_copy.
10670         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
10671         vect_is_simple_use.
10672         (vect_get_loop_based_defs): Remove dt argument and update call
10673         to vect_get_vec_def_for_stmt_copy.
10674         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
10675         and vect_get_vec_def_for_stmt_copy.
10676         (vectorizable_load): Update calls to vect_build_gather_load_calls
10677         and vect_get_vec_def_for_stmt_copy.
10678         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10679         (vectorizable_reduction, vectorizable_live_operation): Update calls
10680         to vect_get_vec_def_for_stmt_copy.
10682 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10684         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
10685         and gimple stmt arguments with a stmt_vec_info.
10686         (vect_record_base_alignments): Update calls accordingly.
10687         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
10688         and gimple stmt arguments with a stmt_vec_info.
10689         (vect_build_slp_tree_1): Remove vinfo argument and update call
10690         to vect_record_max_nunits.
10691         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
10692         and vect_record_max_nunits.
10694 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10696         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
10697         file and take a stmt_vec_info instead of a gimple stmt.
10698         (supportable_widening_operation, vect_finish_replace_stmt)
10699         (vect_finish_stmt_generation, vect_get_store_rhs)
10700         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
10701         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
10702         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
10703         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
10704         (vect_create_data_ref_ptr, bump_vector_ptr)
10705         (vect_permute_store_chain, vect_setup_realignment)
10706         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
10707         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
10708         (vectorizable_reduction, vectorizable_induction)
10709         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
10710         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
10711         than gimple stmts as arguments.
10712         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
10713         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
10714         (can_group_stmts_p, vect_check_gather_scatter)
10715         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10716         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
10717         (vect_permute_load_chain, vect_shift_permute_load_chain)
10718         (vect_transform_grouped_load)
10719         (vect_record_grouped_load_vectors): Likewise.
10720         * tree-vect-loop.c (vect_fixup_reduc_chain)
10721         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
10722         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
10723         (vectorizable_reduction, vectorizable_induction)
10724         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
10725         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
10726         (vect_get_load_store_mask): Likewise.
10727         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
10728         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
10729         * tree-vect-stmts.c (vect_mark_relevant)
10730         (is_simple_and_all_uses_invariant)
10731         (exist_non_indexing_operands_for_use_p, process_use)
10732         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
10733         (vect_get_vec_def_for_operand, vect_get_vec_defs)
10734         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
10735         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
10736         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
10737         (get_negative_load_store_type, get_load_store_type)
10738         (vect_check_load_store_mask, vect_check_store_rhs)
10739         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
10740         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10741         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10742         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10743         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
10744         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
10745         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
10746         (supportable_widening_operation): Likewise.
10748 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10750         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
10751         a stmt_vec_info instead of a gcall.
10752         (vect_check_gather_scatter): Update call accordingly.
10753         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
10754         of a gphi.
10755         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
10756         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
10757         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
10758         instead of a gimple stmt.
10759         (vect_transform_loop): Update calls accordingly.
10760         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
10761         stmt_vec_infos instead of gimple stmts.
10762         (vect_analyze_slp_instance): Update use accordingly.
10763         * tree-vect-stmts.c (read_vector_array, write_vector_array)
10764         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
10765         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
10766         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
10767         (vect_gen_widened_results_half, vect_get_loop_based_defs)
10768         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
10769         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
10770         down to subroutines.
10772 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10774         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
10775         of the worklist from a vector of gimple stmts to a vector of
10776         stmt_vec_infos.
10777         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10778         (vect_mark_stmts_to_be_vectorized): Likewise
10780 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10782         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
10783         statement before passing it to vect_analyze_stmt.
10784         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
10785         the chain of phi vector definitions.  Track the exit phi via its
10786         stmt_vec_info.
10787         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
10788         STMT_VINFO_REDUC_DEF.
10789         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
10790         stmt_vec_infos to handle the statement chains.
10791         (vect_get_slp_defs): Record the first statement in the node
10792         using a stmt_vec_info.
10793         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
10794         statements here and pass their stmt_vec_info down to subroutines.
10795         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
10796         down to vect_finish_stmt_generation.
10797         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
10798         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
10799         stmt_vec_infos to subroutines.
10800         (vect_remove_stores): Use stmt_vec_infos to handle the statement
10801         chains.
10803 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10805         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
10806         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
10807         (vect_permute_store_chain, vect_setup_realignment)
10808         (vect_permute_load_chain, vect_shift_permute_load_chain)
10809         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
10810         stmts internally, and when passing values to other vectorizer routines.
10811         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10812         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
10813         (vect_analyze_loop_operations, get_initial_def_for_reduction)
10814         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
10815         (vectorizable_reduction, vectorizable_induction)
10816         (vectorizable_live_operation, vect_transform_loop_stmt)
10817         (vect_transform_loop): Likewise.
10818         * tree-vect-patterns.c (vect_reassociating_reduction_p)
10819         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
10820         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
10821         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10822         (vect_slp_analyze_node_operations_1): Likewise.
10823         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10824         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
10825         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
10826         (vect_finish_stmt_generation_1, get_group_load_store_type)
10827         (get_load_store_type, vect_build_gather_load_calls)
10828         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10829         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10830         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10831         (vectorizable_store, vectorizable_load, vectorizable_condition)
10832         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
10833         (supportable_widening_operation): Likewise.
10834         (vect_get_vector_types_for_stmt): Likewise.
10835         * tree-vectorizer.h (vect_dr_behavior): Likewise.
10837 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10839         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10840         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
10841         (vect_permute_store_chain, vect_permute_load_chain)
10842         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
10843         repeated stmt_vec_info lookups.
10844         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10845         (vect_update_ivs_after_vectorizer): Likewise.
10846         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
10847         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10848         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
10849         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10850         (vect_bb_slp_scalar_cost): Likewise.
10851         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
10853 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10855         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
10856         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
10857         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
10858         to get gassigns and gcalls, rather than operating on generc gimple
10859         stmts.
10860         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
10861         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
10862         (vectorizable_load, vect_analyze_stmt): Likewise.
10863         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
10865 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10867         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
10868         return stmt_vec_infos rather than gimple stmts.  Do not accept
10869         null arguments.
10870         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
10871         of a gimple stmt.
10872         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10873         Update use of get_later_stmt.
10874         (vect_get_constant_vectors): Update call accordingly.
10875         (vect_schedule_slp_instance): Likewise
10876         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
10877         (vect_slp_analyze_instance_dependence): Likewise.
10878         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
10880 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10882         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
10883         (stmt_info_for_cost::stmt_info): ...this new field.
10884         (add_stmt_costs): Update accordingly.
10885         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10886         (vect_get_known_peeling_cost): Likewise.
10887         (vect_estimate_min_profitable_iters): Likewise.
10888         * tree-vect-stmts.c (record_stmt_cost): Likewise.
10890 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10892         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
10893         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10894         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10895         accordingly.
10896         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
10898 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10900         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
10901         a gimple stmt to a stmt_vec_info.
10902         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10904 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10906         * tree-vectorizer.h (vec_info::grouped_stores): Change from
10907         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10908         (_loop_vec_info::reduction_chains): Likewise.
10909         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
10910         accordingly.
10911         * tree-vect-slp.c (vect_analyze_slp): Likewise.
10913 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10915         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
10916         a gimple stmt to a stmt_vec_info.
10917         (_stmt_vec_info::next_element): Likewise.
10918         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
10919         (vect_slp_analyze_and_verify_node_alignment)
10920         (vect_analyze_group_access_1, vect_analyze_group_access)
10921         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
10922         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
10923         (vect_supportable_dr_alignment): Update accordingly.
10924         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
10925         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
10926         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
10927         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10928         * tree-vect-slp.c (vect_build_slp_tree_1)
10929         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
10930         (vect_split_slp_store_group, vect_analyze_slp_instance)
10931         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
10932         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
10933         (get_group_load_store_type, get_load_store_type)
10934         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
10935         (vect_transform_stmt, vect_remove_stores): Likewise.
10937 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10939         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
10940         than a gimple stmt.
10941         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10942         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
10943         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
10944         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
10945         (vector_alignment_reachable_p, vect_get_data_access_cost)
10946         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
10947         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
10948         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10949         (vect_analyze_group_access_1, vect_analyze_group_access)
10950         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
10951         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
10952         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
10953         result of vect_dr_stmt and use the stmt_vec_info instead of
10954         the associated gimple stmt.
10955         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10956         (vect_gen_prolog_loop_niters): Likewise.
10957         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10959 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10961         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
10962         to a vec<stmt_vec_info>.
10963         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
10964         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
10965         vec<stmt_vec_info>.
10966         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
10967         to a vec<stmt_vec_info>.
10968         (bst_traits::value_type, bst_traits::value_type): Likewise.
10969         (bst_traits::hash): Update accordingly.
10970         (vect_get_and_check_slp_defs): Change the stmts parameter from
10971         a vec<gimple *> to a vec<stmt_vec_info>.
10972         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
10973         (vect_build_slp_tree): Likewise.
10974         (vect_build_slp_tree_2): Likewise.  Update uses of
10975         SLP_TREE_SCALAR_STMTS.
10976         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
10977         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
10978         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
10979         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
10980         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
10981         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
10982         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
10983         (vect_get_constant_vectors, vect_get_slp_defs)
10984         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
10985         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
10986         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
10987         instead of gimple stmts.
10988         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
10989         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
10990         (vect_slp_analyze_instance_dependence): Update uses of
10991         SLP_TREE_SCALAR_STMTS.
10992         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10993         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
10994         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
10995         (get_initial_defs_for_reduction): Likewise.
10996         (vect_create_epilog_for_reduction): Likewise.
10997         (vectorize_fold_left_reduction): Likewise.
10998         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
10999         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11000         (can_vectorize_live_stmts): Likewise.
11002 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11004         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11005         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11006         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11007         than gimple stmts.
11008         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11009         of a gimple stmt.
11010         (gather_scalar_reductions): Update after above interface changes.
11011         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11012         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11013         than gimple stmts.
11014         (vect_force_simple_reduction): Likewise.
11015         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11016         LOOP_VINFO_REDUCTIONS.
11017         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11019 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11021         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11022         a gimple stmt to a stmt_vec_info.
11023         * tree-vect-loop.c (vect_active_double_reduction_p)
11024         (vect_force_simple_reduction, vectorizable_reduction): Update
11025         accordingly.
11027 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11029         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11030         vec<gimple *> to a vec<stmt_vec_info>.
11031         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11032         the reduction_phis argument from a vec<gimple *> to a
11033         vec<stmt_vec_info>.
11034         (vectorizable_reduction): Likewise the phis local variable that
11035         is passed to vect_create_epilog_for_reduction.  Update for new type
11036         of SLP_TREE_VEC_STMTS.
11037         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11038         (vectorizable_live_operation): Likewise.
11039         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11040         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11042 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11044         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11045         a gimple stmt to a stmt_vec_info.
11046         (vectorizable_condition, vectorizable_live_operation)
11047         (vectorizable_reduction, vectorizable_induction): Pass back the
11048         vectorized statement as a stmt_vec_info.
11049         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11050         use of STMT_VINFO_VEC_STMT.
11051         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11052         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11053         as stmt_vec_infos rather than gimple stmts.
11054         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11055         to a stmt_vec_info.
11056         (vectorizable_live_operation): Likewise.
11057         (vectorizable_reduction, vectorizable_induction): Likewise,
11058         updating use of STMT_VINFO_VEC_STMT.
11059         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11060         of STMT_VINFO_VEC_STMT.
11061         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11062         (vectorizable_simd_clone_call, vectorizable_conversion)
11063         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11064         (vectorizable_store, vectorizable_load, vectorizable_condition)
11065         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11066         from a gimple stmt to a stmt_vec_info.
11067         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
11068         pointer to a stmt_vec_info to the vectorizable_* routines.
11070 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11072         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11073         a gimple stmt to a stmt_vec_info.
11074         (is_pattern_stmt_p): Update accordingly.
11075         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11076         (vect_record_grouped_load_vectors): Likewise.
11077         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11078         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11079         (vect_model_reduction_cost): Likewise.
11080         (vect_create_epilog_for_reduction): Likewise.
11081         (vectorizable_reduction, vectorizable_induction): Likewise.
11082         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11083         Return the stmt_vec_info for the pattern statement.
11084         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11085         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11086         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11087         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11088         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11089         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11090         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11091         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11093 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11095         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11096         (vect_finish_stmt_generation): Likewise.
11097         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11098         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11099         (vect_build_gather_load_calls): Use the return value of the above
11100         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11101         scopes for the input gimple stmt and wider scopes for the associated
11102         stmt_vec_info.  Use vec_info::lookup_def when setting these
11103         stmt_vec_infos from an SSA_NAME definition.
11104         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11105         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11106         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11107         (vectorizable_store, vectorizable_load, vectorizable_condition)
11108         (vectorizable_comparison): Likewise.
11109         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11110         (vectorizable_reduction): Likewise.
11112 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11114         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11115         stmt_vec_info * parameter before the optional gimple **.
11116         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11117         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11118         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11119         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11120         (vectorizable_live_operation): Likewise.
11121         * tree-vect-patterns.c (type_conversion_p): Likewise.
11122         (vect_look_through_possible_promotion): Likewise.
11123         (vect_recog_rotate_pattern): Likewise.
11124         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11126 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11128         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11129         a typedef to a wrapper class.
11130         (NULL_STMT_VEC_INFO): New macro.
11131         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11132         (stmt_vec_info::operator*): New function.
11133         (stmt_vec_info::operator gimple *): Likewise.
11134         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11135         (add_stmt_costs): Likewise.
11136         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11137         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11138         (vect_get_known_peeling_cost): Likewise.
11139         (vect_estimate_min_profitable_iters): Likewise.
11140         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11141         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11142         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11143         (vectorizable_store, free_stmt_vec_infos): Likewise.
11144         (new_stmt_vec_info): Change return type of xcalloc to
11145         _stmt_vec_info *.
11147 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11149         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11150         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11151         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11152         a single_imm_use-based sequence.
11153         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11155 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11157         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11158         * tree-vectorizer.c (vec_info::lookup_def): New function.
11159         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11160         (vect_widened_op_tree): Likewise.
11161         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11162         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11163         (vectorizable_reduction): Likewise.
11164         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11165         of a gimple *.
11166         (vect_is_slp_reduction): Update calls accordingly.  Use
11167         vec_info::lookup_def.
11168         (vect_is_simple_reduction): Likewise
11169         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11171 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11173         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11174         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11175         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11176         of vinfo_for_stmt.
11177         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11178         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11179         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11180         (vect_is_slp_reduction, vectorizable_induction)
11181         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11182         * tree-vect-patterns.c (vect_init_pattern_stmt):
11183         (vect_determine_min_output_precision_1, vect_determine_precisions)
11184         (vect_pattern_recog): Likewise.
11185         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11186         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11187         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11188         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11189         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11190         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11191         info field from a loop to a loop_vec_info.
11193 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11195         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11196         (vec_info::add_stmt): Declare.
11197         * tree-vectorizer.c (vec_info::add_stmt): New function.
11198         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11199         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11200         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11201         (vectorizable_induction): Likewise.
11202         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11203         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11204         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11205         (vectorizable_load): Likewise.
11206         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11207         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11208         (vect_recog_gather_scatter_pattern): Likewise.
11209         (append_pattern_def_seq): Likewise.  Remove a check that is
11210         performed by add_stmt itself.
11212 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11214         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11215         which make_ssa_name was called with new_stmt before new_stmt
11216         had been created.
11218 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11220         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11221         split out from...
11222         (vect_is_slp_reduction): ...here...
11223         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11224         that are already known to be false.
11226 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11228         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11229         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11230         STMT_VINFO_NUM_SLP_USES when it's true.
11231         (vect_free_slp_instance): Add a final_p parameter and pass it to
11232         vect_free_slp_tree.
11233         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11234         (vect_analyze_slp_instance): Likewise.
11235         (vect_slp_analyze_operations): Likewise.
11236         (vect_slp_analyze_bb_1): Likewise.
11237         * tree-vectorizer.c (vec_info): Likewise.
11238         * tree-vect-loop.c (vect_transform_loop): Likewise.
11240 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11242         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11243         function is not called for second and subsequent members of
11244         a reduction group.
11246 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11248         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11249         cases for nested loops from here to ...
11250         (vect_create_epilog_for_reduction): ...here.  Only call
11251         vect_is_simple_use for inner-loop reductions.
11253 2018-07-31  Martin Liska  <mliska@suse.cz>
11255         PR gcov-profile/85338
11256         PR gcov-profile/85350
11257         PR gcov-profile/85372
11258         * profile.c (struct location_triplet): New.
11259         (struct location_triplet_hash): Likewise.
11260         (output_location): Do not output a BB that
11261         is already recorded for a line.
11262         (branch_prob): Use streamed_locations.
11264 2018-07-31  Martin Liska  <mliska@suse.cz>
11266         PR gcov-profile/85370
11267         * coverage.c (coverage_begin_function): Do not mark target
11268         clones as artificial functions.
11270 2018-07-31  Martin Liska  <mliska@suse.cz>
11272         PR gcov-profile/83813
11273         PR gcov-profile/84758
11274         PR gcov-profile/85217
11275         PR gcov-profile/85332
11276         * profile.c (branch_prob): Do not record GOTO expressions
11277         for GIMPLE statements which locations are already streamed.
11279 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11281         * gcc.c (handle_spec_function): Accept a soft_matched_part
11282         argument, as do_spec_1.  Pass it down to ...
11283         (eval_spec_function): Accept a soft_matched_part argument,
11284         and pass it down to ...
11285         (do_spec_2): Accept a soft_matched_part argument, and pass
11286         it down to do_spec_1.
11287         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11288         (handle_braces): Update call to handle_spec_function.
11289         (driver::set_up_specs): Update calls to do_spec_2.
11290         (compare_debug_dump_opt_spec_function): Likewise.
11291         (compare_debug_self_opt_spec_function): Likewise.
11293 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11295         * common.opt (nolibc): New option.
11296         * doc/invoke.texi (Link Options): Document it.
11297         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11298         * config/alpha/linux.h: Likewise.
11299         * config/arc/elf.h: Likewise.
11300         * config/arm/uclinux-elf.h: Likewise.
11301         * config/arm/unknown-elf.h: Likewise.
11302         * config/avr/avrlibc.h: Likewise.
11303         * config/bfin/bfin.h: Likewise.
11304         * config/bfin/linux.h: Likewise.
11305         * config/bfin/uclinux.h: Likewise.
11306         * config/darwin.h: Likewise.
11307         * config/darwin10.h: Likewise.
11308         * config/darwin12.h: Likewise.
11309         * config/gnu-user.h: Likewise.
11310         * config/lm32/uclinux-elf.h: Likewise.
11311         * config/pa/pa-hpux11.h: Likewise.
11312         * config/pa/pa64-hpux.h: Likewise.
11313         * config/sparc/sparc.h: Likewise.
11315 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11317         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11318         undefined variables.
11320 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11322         PR target/86640
11323         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11324         instead of GEN_INT.
11326 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11328         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11329         terminated string literal.
11331 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11333         PR rtl-optimization/85160
11334         * combine.c (is_just_move): New function.
11335         (try_combine): Allow combining two instructions into two if neither of
11336         the original instructions was a move.
11338 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11340         PR target/86673
11341         * doc/extend.texi (Global Register Variables): Discourage use of type
11342         qualifiers.
11343         (Local Register Variables): Likewise.
11345 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11347         PR tree-optimization/86506
11348         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11350 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11352         PR target/86547
11353         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11354         hard_regno, make sure no insn between `from` and `to` clobbers it.
11356 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11357             Tom de Vries  <tdevries@suse.de>
11359         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11360         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11361         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11362         PTX_DEFAULT_RUNTIME_DIM.
11363         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11365 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11367         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11368         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11369         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11371 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11373         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11374         used.
11375         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11376         predicate to reflect toc fusion has been deleted.
11377         (toc_fusion_mem_raw): Delete, no longer used.
11378         (toc_fusion_mem_wrapped): Likewise.
11379         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11380         fusion mask bit.
11381         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11382         Delete, no longer used.
11383         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11384         meant to be used for toc fusion.
11385         (rs6000_debug_print_mode): Delete toc fusion debugging.
11386         (rs6000_debug_reg_global): Likewise.
11387         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11388         fusion and secondary reload support that were never used.
11389         (rs6000_option_override_internal): Delete TOC fusion, that was only
11390         partially defined, and it did not work unless you also used the
11391         -mcmodel= switch.
11392         (rs6000_legitimate_address_p): Delete TOC fusion support.
11393         (rs6000_opt_masks): Likewise.
11394         (fusion_wrap_memory_address): Delete function, no longer used.
11395         (fusion_split_address); Delete TOC fusion support.
11396         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11397         longer used with toc fusion being deleted.
11398         (TARGET_TOC_FUSION_FP): Likewise.
11399         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11400         UNSPEC.
11401         (toc fusion spliter): Delete TOC fusion support.
11402         (toc_fusionload_<mode>): Likewise.
11403         (toc_fusionload_di): Likewise.
11404         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11405         longer needs to be named.  Rename predicate to delete TOC fusion.
11406         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11407         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11408         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11409         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11410         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11412 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11414         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11415         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11416         __int128 in built-in function prototypes.
11417         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11418         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11420 2018-07-27  Martin Sebor  <msebor@redhat.com>
11422         PR tree-optimization/86696
11423         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11424         types, including enums.
11425         (handle_char_store): Be prepared for the above function to fail.
11427 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11429         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11430         when optimization level is lower than 2 or optimize for size.
11432 2018-07-26  Martin Sebor  <msebor@redhat.com>
11434         PR tree-optimization/86043
11435         PR tree-optimization/86042
11436         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11437         (get_string_cst_length): Rename...
11438         (get_min_string_length): ...to this.  Add argument.
11439         (handle_char_store): Extend to handle multi-character stores by
11440         MEM_REF.
11441         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11442         * tree.h (initializer_zerop): Add argument.
11444 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11446         PR middle-end/86660
11447         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11448         declare target to variables if they have always,{to,from,tofrom} map
11449         kinds.
11451 2018-07-26  Martin Liska  <mliska@suse.cz>
11453         PR lto/86548
11454         * lto-wrapper.c: Add linker_output as prefix
11455         for ltrans_output_file.
11457 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11459         PR rtl-optimization/85805
11460         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11461         value for hard registers if that was written in the same mode.
11463 2018-07-26  Martin Liska  <mliska@suse.cz>
11465         PR gcov-profile/86536
11466         * gcov.c (format_gcov): Use printf format %.*f directly
11467         and do not handle special values.
11469 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11471         * common/config/arc/arc-common.c (arc_option_optimization_table):
11472         Update default optimizations for size.
11474 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11476         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11477         * config/arc/constraints.md (CfZ): New constraint.
11478         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11479         (subsf3_fpu): Likewise.
11480         (cmpsf_fpu): Likewise.
11481         (cmpsf_fpu_uneq): Likewise.
11483 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11485         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11486         accesses as well.
11487         (arc_is_uncached_mem_p): uncached applies to both the variable and
11488         the pointer.
11490 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11492         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11493         register names.
11495 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11497         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11498         field "m_scopes" from vec to auto_vec.
11500 2018-07-25  Martin Liska  <mliska@suse.cz>
11502         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11503         return type.
11505 2018-07-25  Richard Biener  <rguenther@suse.de>
11507         PR debug/86654
11508         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11509         special wrt context_die late.
11510         (gen_subprogram_die): Re-use DIEs in local scope.
11512 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11514         PR tree-optimization/86644
11515         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11517 2018-07-25  Martin Liska  <mliska@suse.cz>
11519         PR middle-end/86645
11520         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11521         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11523 2018-07-25  Martin Liska  <mliska@suse.cz>
11525         PR sanitizer/79635
11526         * params.def: Explain ASan abbreviation and provide
11527         a documentation link.
11529 2018-07-24  Martin Sebor  <msebor@redhat.com>
11531         PR tree-optimization/86622
11532         PR tree-optimization/86532
11533         * builtins.h (string_length): Declare.
11534         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11535         (check_access): Be prepared for non-constant length ranges.
11536         (string_length): Make extern.
11537         * expr.c (string_constant): Only handle the minor non-constant
11538         array index.  Use string_constant to compute the length of
11539         a generic string constant.
11541 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11543         PR tree-optimization/86618
11544         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11545         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11547 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11549         PR tree-optimization/86636
11550         * json.cc (json::object::set): Fix comment.  Add assertions.
11551         (json::array::append): Move here from json.h.  Add comment and an
11552         assertion.
11553         (json::string::string): Likewise.
11554         * json.h (json::array::append): Move to json.cc.
11555         (json::string::string): Likewise.
11556         * optinfo-emit-json.cc
11557         (optrecord_json_writer::impl_location_to_json): Assert that we
11558         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11559         wrapper around it.  Expand the location once, rather than three
11560         times.
11561         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11562         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11563         wrappers.
11564         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11565         places.  Fix some overlong lines.
11567 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11569         * config/aarch64/aarch64-simd.md
11570         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11571         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11572         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11573         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11574         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11575         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11576         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11577         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
11578         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
11580 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
11582         PR middle-end/86627
11583         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
11584         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
11585         and abs_d == d, do the power of two handling if profitable.
11587 2018-07-24  Richard Biener  <rguenther@suse.de>
11589         * match.pd: Add BIT_FIELD_REF canonicalizations.
11591 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11593         PR c/86617
11594         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
11596 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11598         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
11599         terminated STRING_CST object.
11601 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11603         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
11605 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11607         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
11608         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
11609         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
11610         the elements into a register.
11611         (rs6000_split_v4si_init_di_reg): Delete.
11612         (rs6000_split_v4si_init): Delete.
11613         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
11614         (vsx_init_v4si): Rewrite as a define_expand.
11616 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11618         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
11619         zero_extend argument from memory): New.
11621 2018-07-22  Martin Sebor  <msebor@redhat.com>
11623         PR bootstrap/86621
11624         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
11625         diagnosing calls with unknown arguments unless -Walloca-larger-than
11626         is restricted to less than PTRDIFF_MAX bytes.
11628 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
11630         * doc/gcov.texi (Invoking Gcov): Editorial changes.
11632 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11634         * pretty-print.c (text_info::set_location): Remove redundant
11635         "line_table" parameter from call to rich_location::set_range.
11637 2018-07-20  Martin Sebor  <msebor@redhat.com>
11639         PR middle-end/82063
11640         * builtins.c (expand_builtin_alloca): Adjust.
11641         * calls.c (alloc_max_size): Simplify.
11642         * cgraphunit.c (cgraph_node::expand): Adjust.
11643         * common.opt (larger_than_size, warn_frame_larger_than): Remove
11644         variables.
11645         (frame_larger_than_size): Same.
11646         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
11647         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
11648         Initialize.
11649         * doc/invoke.texi (GCC Command Options): Document option arguments.
11650         Explain byte-size arguments and suffixes.
11651         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
11652         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
11653         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
11654         * doc/options.texi (UInteger): Expand.
11655         (Host_Wide_Int, ByteSize): Document new properties.
11656         * final.c (final_start_function_1): Include sizes in an error message.
11657         * function.c (frame_offset_overflow): Same.
11658         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
11659         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
11660         Diagnose unbounded alloca calls only for limits of less than
11661         PTRDIFF_MAX.
11662         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
11663         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
11664         for alloca(0).
11665         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
11666         only for limits of less than PTRDIFF_MAX.
11667         * langhooks-def.h (lhd_handle_option): Change function argument
11668         to HOST_WIDE_INT.
11669         * langhooks.c (lhd_handle_option): Same.
11670         * langhooks.h (handle_option): Same.
11671         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
11672         ByteSize flags.
11673         (var_type, var_type_struct): Same.
11674         (var_set): Handle ByteSize flag.
11675         * optc-gen.awk: Add comments to output to ease debugging.  Make
11676         use of HOST_WIDE_INT where appropriate.
11677         * opts-gen-save.awk:  Use %lx to format unsigned long.
11678         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
11679         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
11680         arguments.  Parse bytes-size suffixes.
11681         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
11682         (enum_value_to_arg): Same.
11683         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
11684         (handle_option): Adjust.
11685         (generate_option): Change function argument to HOST_WIDE_INT.
11686         (cmdline_handle_error): Adjust.
11687         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
11688         (set_option): Change function argument to HOST_WIDE_INT.
11689         (option_enabled): Handle cl_host_wide_int.
11690         (get_option_state): Handle CLVC_SIZE.
11691         (control_warning_option): Same.
11692         * opts.c (common_handle_option): Change function argument to
11693         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
11694         OPT_Wvla_larger_than_.
11695         * opts.h (enum cl_var_type): Add an enumerator.
11696         * stor-layout.c (layout_decl): Print a more meaningful warning.
11697         * toplev.c (output_stack_usage): Adjust.
11699 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
11701         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
11702         call to inline_expand_builtin_string_cmp.
11703         (expand_builtin_strcmp): Likewise.
11704         (expand_builtin_strncmp): Likewise.
11705         (inline_string_cmp): Delete the last parameter, change char_type_node
11706         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
11707         two operands.
11708         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
11709         the inlining expansion on target where the type of the call has same or
11710         narrower precision than unsigned char.
11712 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11714         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
11715         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
11716         * common.opt (fsave-optimization-record): New option.
11717         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
11718         * doc/invoke.texi (-fsave-optimization-record): New option.
11719         * dumpfile.c: Include "optinfo-emit-json.h".
11720         (struct kv_pair): Move to coretypes.h.
11721         (optgroup_options): Make non-static.
11722         (dump_context::end_scope): Call
11723         optimization_records_maybe_pop_dump_scope.
11724         * dumpfile.h (optgroup_options): New decl.
11725         * json.cc: New file.
11726         * json.h: New file.
11727         * optinfo-emit-json.cc: New file.
11728         * optinfo-emit-json.h: New file.
11729         * optinfo.cc: Include "optinfo-emit-json.h".
11730         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
11731         (optinfo_enabled_p): Check optimization_records_enabled_p.
11732         (optinfo_wants_inlining_info_p): Likewise.
11733         * optinfo.h: Update comment.
11734         * profile-count.c (profile_quality_as_string): New function.
11735         * profile-count.h (profile_quality_as_string): New decl.
11736         (profile_count::quality): New accessor.
11737         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
11738         and optinfo_emit_json_cc_tests.
11739         * selftest.h (selftest::json_cc_tests): New decl.
11740         (selftest::optinfo_emit_json_cc_tests): New decl.
11741         * toplev.c: Include "optinfo-emit-json.h".
11742         (compile_file): Call optimization_records_finish.
11743         (do_compile): Call optimization_records_start.
11744         * tree-ssa-live.c: Include optinfo.h.
11745         (remove_unused_scope_block_p): Retain inlining information if
11746         optinfo_wants_inlining_info_p returns true.
11748 2018-07-20  Richard Biener  <rguenther@suse.de>
11750         PR debug/86585
11751         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
11752         to cover -flto-partition=none.
11754 2018-07-20  Martin Liska  <mliska@suse.cz>
11756         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
11757         (get_decl_source_range): Remove unused function.
11759 2018-07-20  Richard Biener  <rguenther@suse.de>
11761         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
11762         (struct vn_phi_s): Likewise.
11763         (struct vn_reference_s): Likewise.
11764         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
11765         for searching the slot of an entry known to be in the hash itself.
11766         (vn_phi_hasher::equal): Likewise.
11767         (vn_reference_hasher::equal): Likewise.
11768         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
11769         globals.
11770         (optimistic_info, current_info): Remove, keeping only valid_info.
11771         (vn_reference_lookup_1): Remove fallback lookup.
11772         (vn_reference_lookup_2): Likewise.
11773         (vn_nary_op_lookup_1): Likewise.
11774         (vn_phi_lookup): Likewise.
11775         (vn_nary_build_or_lookup_1): Make sure to not chain the built
11776         hash element.
11777         (vn_reference_insert): Adjust, chain the inserted hash element
11778         at last_inserted_ref.
11779         (vn_reference_insert_pieces): Likewise.
11780         (visit_reference_op_call): Likewise.
11781         (vn_nary_op_insert_into): Chain the inserted hash element at
11782         last_inserted_nary.
11783         (vn_nary_op_insert_pieces): Adjust.
11784         (vn_nary_op_insert): Likewise.
11785         (vn_nary_op_insert_stmt): Likewise.
11786         (vn_phi_insert): Adjust, chain the inserted hash element at
11787         last_inserted_phi.
11788         (process_scc): Remove clearing and copying the optimistic
11789         table.  Instead remove elements inserted during an optimistic
11790         iteration from the single table we maintain.
11791         (init_scc_vn): Adjust.
11792         (free_scc_vn): Likewise.
11793         (sccvn_dom_walker::record_cond): Likewise.
11794         (sccvn_dom_walker::after_dom_children): Likewise.
11796 2018-07-19  Martin Sebor  <msebor@redhat.com>
11798         PR tree-optimization/84047
11799         PR tree-optimization/83776
11800         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
11801         (check_array_bounds): Call it.
11803 2018-07-19  Martin Sebor  <msebor@redhat.com>
11805         * align.h (align_flags): Use member initialization.
11807 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
11809         * Makefile.in (OBJS): Add optinfo.o.
11810         * coretypes.h (class symtab_node): New forward decl.
11811         (struct cgraph_node): New forward decl.
11812         (class varpool_node): New forward decl.
11813         * dump-context.h: New file.
11814         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
11815         "tree-pass.h".
11816         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
11817         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
11818         (set_alt_dump_file): Likewise.
11819         (dump_context::~dump_context): New dtor.
11820         (dump_gimple_stmt): Move implementation to...
11821         (dump_context::dump_gimple_stmt): ...this new member function.
11822         Add the stmt to any pending optinfo, creating one if need be.
11823         (dump_gimple_stmt_loc): Move implementation to...
11824         (dump_context::dump_gimple_stmt_loc): ...this new member function.
11825         Start a new optinfo and add the stmt to it.
11826         (dump_gimple_expr): Move implementation to...
11827         (dump_context::dump_gimple_expr): ...this new member function.
11828         Add the stmt to any pending optinfo, creating one if need be.
11829         (dump_gimple_expr_loc): Move implementation to...
11830         (dump_context::dump_gimple_expr_loc): ...this new member function.
11831         Start a new optinfo and add the stmt to it.
11832         (dump_generic_expr): Move implementation to...
11833         (dump_context::dump_generic_expr): ...this new member function.
11834         Add the tree to any pending optinfo, creating one if need be.
11835         (dump_generic_expr_loc): Move implementation to...
11836         (dump_context::dump_generic_expr_loc): ...this new member
11837         function.  Add the tree to any pending optinfo, creating one if
11838         need be.
11839         (dump_printf): Move implementation to...
11840         (dump_context::dump_printf_va): ...this new member function.  Add
11841         the text to any pending optinfo, creating one if need be.
11842         (dump_printf_loc): Move implementation to...
11843         (dump_context::dump_printf_loc_va): ...this new member function.
11844         Start a new optinfo and add the stmt to it.
11845         (dump_dec): Move implementation to...
11846         (dump_context::dump_dec): ...this new member function.  Add the
11847         value to any pending optinfo, creating one if need be.
11848         (dump_context::dump_symtab_node): New member function.
11849         (dump_context::get_scope_depth): New member function.
11850         (dump_context::begin_scope): New member function.
11851         (dump_context::end_scope): New member function.
11852         (dump_context::ensure_pending_optinfo): New member function.
11853         (dump_context::begin_next_optinfo): New member function.
11854         (dump_context::end_any_optinfo): New member function.
11855         (dump_context::s_current): New global.
11856         (dump_context::s_default): New global.
11857         (dump_scope_depth): Delete global.
11858         (dumpfile_ensure_any_optinfo_are_flushed): New function.
11859         (dump_symtab_node): New function.
11860         (get_dump_scope_depth): Reimplement in terms of dump_context.
11861         (dump_begin_scope): Likewise.
11862         (dump_end_scope): Likewise.
11863         (selftest::temp_dump_context::temp_dump_context): New ctor.
11864         (selftest::temp_dump_context::~temp_dump_context): New dtor.
11865         (selftest::verify_item): New function.
11866         (ASSERT_IS_TEXT): New macro.
11867         (ASSERT_IS_TREE): New macro.
11868         (ASSERT_IS_GIMPLE): New macro.
11869         (selftest::test_capture_of_dump_calls): New test.
11870         (selftest::dumpfile_c_tests): Call it.
11871         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
11872         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
11873         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
11874         descriptive comment.
11875         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
11876         (dump_node, dump_bb): Move these unrelated decls.
11877         (class dump_manager): Add leading comment.
11878         * optinfo.cc: New file.
11879         * optinfo.h: New file.
11881 2018-07-19  Michael Collison  <michael.collison@arm.com>
11882             Richard Henderson <rth@redhat.com>
11884         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
11885         (subti): Handle op1 zero.
11886         (subvti4, usub4ti4): New.
11887         (*sub<GPI>3_compare1_imm): New.
11888         (sub<GPI>3_carryinCV): New.
11889         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
11890         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
11892 2018-07-19  Michael Collison  <michael.collison@arm.com>
11893             Richard Henderson <rth@redhat.com>
11895         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
11896         (addti3): Create simpler code if low part is already known to be 0.
11897         (addvti4, uaddvti4): New.
11898         (*add<GPI>3_compareC_cconly_imm): New.
11899         (*add<GPI>3_compareC_cconly): New.
11900         (*add<GPI>3_compareC_imm): New.
11901         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
11902         handle constants within this pattern..
11903         (*add<GPI>3_compareV_cconly_imm): New.
11904         (*add<GPI>3_compareV_cconly): New.
11905         (*add<GPI>3_compareV_imm): New.
11906         (add<GPI>3_compareV): New.
11907         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
11908         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
11909         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
11910         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
11911         with 'comparison' operator.
11912         (*add<GPI>3_compareV_cconly_imm): Ditto.
11913         (*add<GPI>3_compareV_cconly): Ditto.
11914         (*add<GPI>3_compareV_imm): Ditto.
11915         (add<GPI>3_compareV): Ditto.
11916         (add<mode>3_carryinC): Ditto.
11917         (*add<mode>3_carryinC_zero): Ditto.
11918         (*add<mode>3_carryinC): Ditto.
11919         (add<mode>3_carryinV): Ditto.
11920         (*add<mode>3_carryinV_zero): Ditto.
11921         (*add<mode>3_carryinV): Ditto.
11923 2018-07-19  Michael Collison  <michael.collison@arm.com>
11924             Richard Henderson <rth@redhat.com>
11926         * config/aarch64/aarch64-modes.def (CC_V): New.
11927         * config/aarch64/aarch64-protos.h
11928         (aarch64_addti_scratch_regs): Declare
11929         (aarch64_subvti_scratch_regs): Declare.
11930         (aarch64_expand_subvti): Declare.
11931         (aarch64_gen_unlikely_cbranch): Declare
11932         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
11933         for signed overflow using CC_Vmode.
11934         (aarch64_get_condition_code_1): Handle CC_Vmode.
11935         (aarch64_gen_unlikely_cbranch): New function.
11936         (aarch64_addti_scratch_regs): New function.
11937         (aarch64_subvti_scratch_regs): New function.
11938         (aarch64_expand_subvti): New function.
11940 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11942         * config/aarch64/aarch64-option-extensions.def: New entry for profile
11943         extension.
11944         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
11945         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
11946         extension.
11948 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11950         PR target/83009
11951         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11952         address check not strict.
11954 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11956         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
11957         Umq with Umn.
11958         (store_pair_lanes<mode>): Likewise.
11959         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
11960         enum value 'ADDR_QUERY_LDP_STP_N'.
11961         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
11962         (aarch64_print_address_internal): Add declaration.
11963         (aarch64_print_ldpstp_address): Remove.
11964         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
11965         (aarch64_print_operand): Change printing of 'y'.
11966         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
11967         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
11968         'true' rather than '1'.
11969         * config/aarch64/constraints.md (Uml): Likewise.
11970         (Uml): Rename to Umn.
11971         (Umq): Remove.
11973 2018-07-19  Richard Biener  <rguenther@suse.de>
11975         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
11976         a trailing array.
11977         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
11978         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
11979         (vn_reference_hasher): Likewise.
11980         (struct vn_tables_s): Remove obstack and alloc-pool members.
11981         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
11982         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
11983         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
11984         (vn_reference_insert_pieces): Likewise.
11985         (alloc_vn_nary_op_noinit): Adjust.
11986         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
11987         (vn_phi_eq): Adjust.
11988         (shared_lookup_phiargs): Remove.
11989         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
11990         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
11991         (visit_reference_op_call): Likewise.
11992         (copy_nary, copy_phi, copy_reference): Remove.
11993         (process_scc): Rewind the obstack when iterating.  Do not
11994         copy the elements to valid_info but just move them from one
11995         hashtable to the other.
11996         (allocate_vn_table): Adjust.
11997         (free_vn_table): Likewise.
11998         (init_scc_vn): Likewise.
11999         (free_scc_vn): Likewise.
12001 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
12003         PR target/86560
12004         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12005         indirect_return as function type attribute.
12006         (ix86_attribute_table): Change indirect_return to function
12007         type attribute.
12008         * doc/extend.texi: Update indirect_return attribute.
12010 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
12012         * wide-int.h (widest2_int): New.
12013         * gimple-fold.c (arith_overflowed_p): Use it.
12014         * tree.h (widest2_int_cst): New.
12015         * tree-vrp.c (wide_int_binop_overflow): Rename from
12016         vrp_int_const_binop.
12017         Rewrite to work on trees.
12018         (extract_range_from_multiplicative_op_1): Abstract code to...
12019         (wide_int_range_min_max): ...here.
12020         (wide_int_range_cross_product): ...and here.
12021         (extract_range_from_binary_expr_1): Abstract overflow code to...
12022         (wide_int_range_mult_wrapping): ...here.
12023         * tree-vrp.h (wide_int_range_cross_product): New.
12024         (wide_int_range_mult_wrapping): New.
12026 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
12027             Julia Koval  <julia.koval@intel.com>
12029         * config/i386/x86-tune-costs.h (skylake_memcpy,
12030         skylake_memset): Replace rep_prefix with unrolling for size 512.
12032 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
12034         PR middle-end/86544
12035         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12036         comparision with EQ_EXPR in last stmt.
12038 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12040         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12041         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12042         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12043         previously known as "PowerPC AltiVec Built-in Functions".  Move
12044         some material to new subsubsections "PowerPC AltiVec Built-in
12045         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12046         ISA 2.07".
12047         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12048         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12049         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12050         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12052 2018-07-18  Richard Biener  <rguenther@suse.de>
12054         PR tree-optimization/86557
12055         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12056         EXACT_DIV_EXPR.
12058 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
12060         * config/s390/s390.c (s390_function_profiler): Generate CFI.
12062 2018-07-17  Jeff Law  <law@redhat.com>
12064         * config/arm/arm.c (get_label_padding): Update for recent
12065         changes to label_to_alignment.
12067         PR tree-optimization/86010
12068         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12070         * config/mips/mips.c (vr4130_align_insns): Update for recent
12071         changes to label_to_alignment.
12073         * config/frv/frv.c (frv_label_align): Update for recent changes
12074         to label_to_alignment.
12076         * config/nios2/nios2.c (nios2_label_align): Update for recent
12077         changes which dropped ALIGN_LABELS_LOG.
12079 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12081         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12082         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12084 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12086         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12087         deprecated align_labels_log.
12089 2018-07-17  Richard Biener  <rguenther@suse.de>
12091         PR lto/86456
12092         * dwarf2out.c (init_sections_and_labels): Always generate
12093         a debug_line_str_section for early LTO debug.
12094         (dwarf2out_finish): Reset debug_line_str_hash output early.
12095         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12096         with fat LTO part.
12097         (dwarf2out_early_finish): Output debug_line_str.
12099 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12101         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12102         index register on z196 or later.
12104 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12106         * config/s390/s390.c (s390_default_align): Set default function
12107         alignment to 16.
12108         (s390_override_options_after_change): Call s390_default align.
12109         (s390_option_override_internal): Call s390_default_align.
12110         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12112 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12114         PR middle-end/86542
12115         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12116         to _looptemp_ clauses, other than the first two.
12118 2018-07-17  Martin Liska  <mliska@suse.cz>
12120         * opts.c: Do not enable OPT_falign_* for -Os.
12122 2018-07-17  Martin Liska  <mliska@suse.cz>
12124         * align.h (MAX_CODE_ALIGN): New.
12125         (MAX_CODE_ALIGN_VALUE): New.
12126         * common/config/i386/i386-common.c (ix86_handle_option):
12127         (MAX_CODE_ALIGN): Moved to align.h.
12128         * final.c (MAX_CODE_ALIGN): Likewise.
12129         * opts.c (parse_and_check_align_values):
12130         (MAX_CODE_ALIGN): Likewise.
12131         (MAX_CODE_ALIGN_VALUE): Likewise.
12133 2018-07-17  Martin Liska  <mliska@suse.cz>
12135         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12136         in order to fulfil coding style.
12137         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12138         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12139         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12140         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12141         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12142         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12143         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12144         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12145         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12147 2018-07-17  Martin Liska  <mliska@suse.cz>
12149         * align.h: New file.
12150         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12151         directly.
12152         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12153         align_flags of label_to_alignment.
12154         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12155         align_flags class.
12156         * config/m68k/m68k.c: Do not use removed align_labels_value and
12157         align_loops_value.
12158         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12159         (LOOP_ALIGN): Likewise.
12160         (LABEL_ALIGN): Likewise.
12161         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12162         Remove not used macro.
12163         (rs6000_loop_align): Change return type to align_flags.
12164         (rs6000_loop_align_max_skip): Remove.
12165         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12166         Change return type to align_flags.
12167         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12168         Remove not used macro.
12169         (rs6000_loop_align):  Change return type to align_flags.
12170         (rs6000_loop_align_max_skip): Remove.
12171         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12172         * config/rx/rx-protos.h (rx_align_for_label): Make it
12173         static function.
12174         * config/rx/rx.c (rx_align_for_label): Change return type
12175         to align_flags.
12176         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12177         macro definitions.
12178         into align_flags class.
12179         (LABEL_ALIGN): Likewise.
12180         (LOOP_ALIGN): Likewise.
12181         * config/s390/s390.c (s390_label_align): Use align_flags
12182         class member.
12183         (s390_asm_output_function_label): Likewise.
12184         * config/sh/sh.c (sh_override_options_after_change):
12185         Use align_flags class directly without macros.
12186         (find_barrier): Likewise.
12187         (barrier_align): Likewise.
12188         (sh_loop_align): Likewise.
12189         * config/spu/spu.c (spu_option_override):
12190         Use align_flags_tuple::get_value instead of removed macros.
12191         (spu_sched_init): Likewise.
12192         * config/spu/spu.h (GTY): Likewise.
12193         * config/visium/visium.c (visium_option_override):
12194         Set "8" as default secondary alignment.
12195         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12196         in order to guarantee secondary alignment of 8.
12197         * coretypes.h: Include align.h header file.
12198         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12199         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12200         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12201         * doc/tm.texi.in: Likewise.
12202         * final.c (struct label_alignment): Remove not used structure.
12203         (LABEL_ALIGN): Change type to align_flags.
12204         (LOOP_ALIGN): Likewise.
12205         (JUMP_ALIGN): Likewise.
12206         (default_loop_align_max_skip): Remove.
12207         (default_label_align_max_skip): Likewise.
12208         (default_jump_align_max_skip): Likewise.
12209         (default_label_align_after_barrier_max_skip):
12210         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12211         (LABEL_TO_MAX_SKIP): Remove.
12212         (label_to_alignment): Return align_flags type instead of integer.
12213         (label_to_max_skip): Remove.
12214         (align_fuzz): Use align_flags type.
12215         (compute_alignments): Use align_flags type and use align_flags::max
12216         to combine multiple alignments.
12217         (grow_label_align): Grow vec instead of C array.
12218         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12219         (shorten_branches):  Use align_flags type and use align_flags::max
12220         to combine multiple alignments.
12221         (final_scan_insn_1): Remove usage of secondary alignment that comes
12222         from label alignment, but instead use proper secondary alignment
12223         which is computed in grow_label_align.
12224         * flags.h (struct align_flags_tuple): Move to align.h.
12225         (struct align_flags): Likewise.
12226         (state_align_loops): Rename to align_loops.
12227         (state_align_jumps): Rename to align_jumps.
12228         (state_align_labels): Rename to align_labels.
12229         (state_align_functions): Rename to align_functions.
12230         (align_loops_log): Remove.
12231         (align_jumps_log): Remove.
12232         (align_labels_log): Remove.
12233         (align_functions_log): Remove.
12234         (align_loops_max_skip): Remove.
12235         (align_jumps_max_skip): Remove.
12236         (align_labels_max_skip): Remove.
12237         (align_functions_max_skip): Remove.
12238         (align_loops_value): Remove.
12239         (align_jumps_value): Remove.
12240         (align_labels_value): Remove.
12241         (align_functions_value): Remove.
12242         * output.h (label_to_alignment): Change return type to align_flags.
12243         (label_to_max_skip): Remove.
12244         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12245         jump_align_max_skip macros.
12246         * targhooks.h (default_loop_align_max_skip): Remove.
12247         (default_label_align_max_skip): Likewise.
12248         (default_jump_align_max_skip): Likewise.
12249         (default_label_align_after_barrier_max_skip): Remove.
12250         * toplev.c (read_log_maxskip): Use ::normalize function.
12251         (parse_N_M): Remove not used argument and also call ::normalize.
12252         (parse_alignment_opts): Do not pass unused arguments.
12253         * varasm.c (assemble_start_function): Use directly align_functions
12254         instead of removed macros.
12255         * system.h: Do not poison removed macros.
12257 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12259         PR middle-end/86539
12260         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12261         and cond temporaries don't have reference type if iterator has
12262         pointer type.  For init use &for_pre_body instead of pre_p if
12263         for_pre_body is non-empty.
12265 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12267         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12268         double-double modes to SFmode directly directly.
12269         (trunc<mode>sf2_fprs): Delete.
12271 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12273         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12274         for conversions between IFmode and the decimal floating point modes.
12275         (init_float128_ieee): Use the correct names for conversions between
12276         KFmode and the decimal floating point modes.
12278 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12280         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12281         for the conversions between TDmode and IFmode.
12282         (init_float128_ieee): Use more correct names for the conversions
12283         between TDmode and KFmode.
12285 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12287         PR tree-optimization/86526
12288         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12289         (inline_expand_builtin_string_cmp): Likewise.
12290         (inline_string_cmp): Likewise.  Use c_readstr instead of
12291         builtin_memcpy_read_str.  Add unit_mode temporary.
12293 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12295         PR middle-end/86528
12296         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12297         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12299 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12301         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12302         Alphabetize prototypes of built-in functions, separating out
12303         built-in functions that are listed in this section but should be
12304         described elsewhere.
12306 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12308         PR target/86511
12309         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12310         conditional move when trapping comparison was split to a
12311         non-trapping one (and vice versa).
12313 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12315         * config/s390/s390.c (s390_function_profiler): Generate nops
12316         instead of profiler call sequences.
12317         * config/s390/s390.opt: Add the new option.
12319 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12321         * config/s390/s390.c (s390_function_profiler): Generate
12322         __mcount_loc section.
12323         * config/s390/s390.opt: Add the new option.
12325 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12327         * common.opt: Add the new warning.
12328         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12329         %r0,__fentry__" when -mfentry is specified.
12330         (s390_option_override_internal): Disallow -mfentry for 31-bit
12331         CPUs.
12332         * config/s390/s390.opt: Add the new option.
12334 2018-07-16  Richard Biener  <rguenther@suse.de>
12336         PR lto/86523
12337         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12338         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12340 2018-07-16  Martin Liska  <mliska@suse.cz>
12342         PR ipa/86529
12343         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12344         to ::get_create.
12346 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12348         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12350 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12352         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12353         for ARCHS4x.
12354         * config/arc/arc-cpus.def (hs4x): New cpu.
12355         (hs4xd): Likewise.
12356         * config/arc/arc-tables.opt: Regenerate.
12357         * config/arc/arc.c (arc_sched_issue_rate): New function.
12358         (TARGET_SCHED_ISSUE_RATE): Define.
12359         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12360         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12361         fpu_cvt.
12362         (attr tune): Add ARCHS4x tune values.
12363         (attr tune_dspmpy): Define.
12364         (*tst): Correct instruction type.
12365         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12366         * config/arc/arcHS4x.md: New file.
12367         * config/arc/fpu.md: Update instruction type attributes.
12368         * config/arc/t-multilib: Regenerate.
12370 2018-07-16  Tom de Vries  <tdevries@suse.de>
12372         PR debug/86455
12373         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12375 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12377         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12378         early clobber.
12380 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12382         PR tree-optimization/86514
12383         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12384         conversion to a boolean type from a type with greater precision.
12386 2018-07-16  Tom de Vries  <tdevries@suse.de>
12388         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12389         dump_flags request TDF_SLIM.
12391 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12393         * fold-const.c (int_const_binop_1): Abstract...
12394         (wide_int_binop): ...wide int code here.
12395         (poly_int_binop): ...poly int code here.
12396         Abstract the rest of int_const_binop_1 into int_const_binop.
12397         * fold-const.h (wide_int_binop): New.
12398         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12399         Remove useless PLUS/MINUS_EXPR case.
12400         (zero_nonzero_bits_from_vr): Move wide int code...
12401         (zero_nonzero_bits_from_bounds): ...here.
12402         (extract_range_from_binary_expr_1): Move mask optimization code...
12403         (range_easy_mask_min_max): ...here.
12404         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12405         (range_easy_mask_min_max): New.
12407 2018-07-15  Jeff Law  <law@redhat.com>
12409         PR target/85993
12410         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12411         block.
12413 2018-07-14  Jim Wilson  <jimw@sifive.com>
12415         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12417 2018-07-14  Paul Koning  <ni1d@arrl.net>
12419         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12421 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12423         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12424         copying section.
12426 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12427             Steve Munroe  <munroesj52@gmail.com>
12429         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12430         (_mm_andnot_si128): Likewise.
12431         (_mm_or_si128): Likewise.
12432         (_mm_xor_si128): Likewise.
12434 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12436         PR middle-end/78809
12437         * builtins.c (expand_builtin_memcmp): Inline the calls first
12438         when result_eq is false.
12439         (expand_builtin_strcmp): Inline the calls first.
12440         (expand_builtin_strncmp): Likewise.
12441         (inline_string_cmp): New routine. Expand a string compare
12442         call by using a sequence of char comparison.
12443         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12444         a call to str(n)cmp/memcmp.
12445         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12446         New option.
12447         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12449 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12451         * config/arm/driver-arm.c: Include arm-native.h.
12452         (host_detect_local_cpu): Use auto-generated data tables.
12453         (vendors, arm_cpu_table): Delete.  Move part information to ...
12454         * config/arm/arm-cpus.in: ... here.
12455         * config/arm/parsecpu.awk (gen_native): New function.
12456         (vendor, part): New CPU fields.
12457         (END): Add support for building the native CPU detection tables.
12458         * config/arm/t-arm (arm-native.h): Add build rule.
12459         (driver-arm.o): Add dependency on arm-native.h.
12461 2018-07-13  Richard Biener  <rguenther@suse.de>
12463         PR middle-end/85974
12464         * match.pd (addr1 - addr2): Allow either of the operand to
12465         have a conversion.
12467 2018-07-13  Tom de Vries  <tdevries@suse.de>
12469         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12470         in remap_ssa_name.
12472 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12474         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12475         arrays instead of numbered variables.
12477 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12479         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12480         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12482 2018-07-13  Richard Biener  <rguenther@suse.de>
12484         PR debug/86452
12485         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12486         instead of get_context_die.
12488 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12489             Richard Biener  <rguenther@suse.de>
12491         PR middle-end/86489
12492         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12493         that the loop latch destination where phi is defined.
12495 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12497         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12498         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12499         (riscv_set_current_function): Extract getting interrupt type to new
12500         function.
12501         (riscv_get_interrupt_type): New function.
12502         (riscv_merge_decl_attributes): New function, checking interrupt type is
12503         same.
12504         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12506 2018-07-12  Paul Koning  <ni1d@arrl.net>
12508         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12509         directive.
12511 2018-07-12  Paul Koning  <ni1d@arrl.net>
12513         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12514         zero reference, add doloop_end instead.
12515         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12516         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12517         detail for doloop_end.
12519 2018-07-12  Martin Sebor  <msebor@redhat.com>
12521         PR c/86453
12522         * attribs.c (decl_attributes): Reject conflicting attributes before
12523         calling attribute handlers.
12525 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12527         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12528          parameter.
12529         (gcc::dump_manager::get_dump_file_name): likewise.
12530         (dump_begin): Likewise.
12531         * dumpfile.h (dump_begin): Update prototype.
12532         (gcc::dump_manager::get_dump_file_name,
12533         gcc::dump_manager::get_dump_file_name): Update prototype.
12535 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12537         * internal-fn.h (vectorizable_internal_fn_p): New function.
12538         * tree-vect-slp.c (compatible_calls_p): Likewise.
12539         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12540         to internal functions.
12541         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12543 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12545         * fold-const.h (inverse_conditions_p): Declare.
12546         * fold-const.c (inverse_conditions_p): New function.
12547         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12548         that test the inverse condition of a conditional internal function.
12549         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12550         * internal-fn.c (internal_fn_mask_index): Handle conditional
12551         internal functions.
12552         (vectorized_internal_fn_supported_p): New function.
12553         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12554         (any_pred_load_store): Replace with...
12555         (need_to_predicate): ...this new variable.
12556         (redundant_ssa_names): New variable.
12557         (ifcvt_can_use_mask_load_store): Move initial checks to...
12558         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12559         for which a conditional internal function exists.
12560         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12561         instead of ifcvt_can_use_mask_load_store.  Update after variable
12562         name change.
12563         (predicate_load_or_store): New function, split out from
12564         predicate_mem_writes.
12565         (check_redundant_cond_expr): New function.
12566         (value_available_p): Likewise.
12567         (predicate_rhs_code): Likewise.
12568         (predicate_mem_writes): Rename to...
12569         (predicate_statements): ...this.  Use predicate_load_or_store
12570         and predicate_rhs_code.
12571         (combine_blocks, tree_if_conversion): Update after above name changes.
12572         (ifcvt_local_dce): Handle redundant_ssa_names.
12573         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12574         general conditional functions.
12575         * tree-vect-stmts.c (vectorizable_call): Likewise.
12577 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12578             Alan Hayward  <alan.hayward@arm.com>
12579             David Sherwood  <david.sherwood@arm.com>
12581         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
12582         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
12583         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
12584         plus and minus and convert them into IFN_COND_FMA-based sequences.
12585         (convert_mult_to_fma): Handle conditional plus and minus.
12587 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12589         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
12590         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
12591         (cond_fnms_optab): New optabs.
12592         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
12593         internal functions.
12594         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
12595         * internal-fn.h (get_conditional_internal_fn): Declare.
12596         (get_unconditional_internal_fn): Likewise.
12597         * internal-fn.c (cond_ternary_direct): New macro.
12598         (expand_cond_ternary_optab_fn): Likewise.
12599         (direct_cond_ternary_optab_supported_p): Likewise.
12600         (FOR_EACH_COND_FN_PAIR): Likewise.
12601         (get_conditional_internal_fn): New function.
12602         (get_unconditional_internal_fn): Likewise.
12603         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
12604         (gimple_match_op::gimple_match_op): Add a new overload for 5
12605         operands.
12606         (gimple_match_op::set_op): Likewise.
12607         (gimple_resimplify5): Declare.
12608         * genmatch.c (decision_tree::gen): Generate simplifications for
12609         5 operands.
12610         * gimple-match-head.c (gimple_simplify): Define an overload for
12611         5 operands.  Handle calls with 5 arguments in the top-level overload.
12612         (convert_conditional_op): Handle conversions from unconditional
12613         internal functions to conditional ones.
12614         (gimple_resimplify5): New function.
12615         (build_call_internal): Pass a fifth operand.
12616         (maybe_push_res_to_seq): Likewise.
12617         (try_conditional_simplification): Try converting conditional
12618         internal functions to unconditional internal functions.
12619         Handle 3-operand unconditional forms.
12620         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
12621         Define ternary equivalents of the current rules for binary conditional
12622         internal functions.
12623         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
12624         ternary operations.
12625         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
12626         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
12627         (optab): Handle them.
12628         (SVE_COND_FP_TERNARY): New int iterator.
12629         (sve_fmla_op, sve_fmad_op): New int attributes.
12630         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
12631         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
12632         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
12634 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12636         * target.def (preferred_else_value): New target hook.
12637         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
12638         * doc/tm.texi: Regenerate.
12639         * targhooks.h (default_preferred_else_value): Declare.
12640         * targhooks.c (default_preferred_else_value): New function.
12641         * internal-fn.h (conditional_internal_fn_code): Declare.
12642         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
12643         (get_conditional_internal_fn): Use it.
12644         (conditional_internal_fn_code): New function.
12645         * gimple-match.h (gimple_match_cond): New struct.
12646         (gimple_match_op): Add a cond member function.
12647         (gimple_match_op::gimple_match_op): Update all forms to take a
12648         gimple_match_cond.
12649         * genmatch.c (expr::gen_transform): Use the same condition as res_op
12650         for the suboperation, but don't specify a particular else_value.
12651         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
12652         (visit_nary_op, visit_reference_op_load): Pass
12653         gimple_match_cond::UNCOND to the gimple_match_op constructor.
12654         * gimple-match-head.c: Include tree-eh.h
12655         (convert_conditional_op): New function.
12656         (maybe_resimplify_conditional_op): Likewise.
12657         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
12658         (gimple_resimplify2): Likewise.
12659         (gimple_resimplify3): Likewise.
12660         (gimple_resimplify4): Likewise.
12661         (maybe_push_res_to_seq): Return null for conditional operations.
12662         (try_conditional_simplification): New function.
12663         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
12664         constructor.
12665         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
12666         IFN_COND_* call.
12667         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
12668         function.
12669         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
12671 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12673         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12674         DECL_FCONTEXT
12675         (hash_tree): Do not hash DECL_FCONTEXT
12676         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
12677         Do not stream DECL_FCONTEXT.
12678         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
12679         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
12681 2018-07-12  Richard Biener  <rguenther@suse.de>
12683         PR debug/86462
12684         * dwarf2out.c (gen_block_die): Only output blocks when they have
12685         at least one !DECL_IGNORED_P variable.
12687 2018-07-12  Richard Biener  <rguenther@suse.de>
12689         PR target/84829
12690         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
12691         Remove -mieee-fp handling.
12693 2018-07-12  Richard Biener  <rguenther@suse.de>
12695         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
12696         left-over from last patch.
12698 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
12700         PR tree-optimization/86492
12701         * gimple-ssa-store-merging.c
12702         (imm_store_chain_info::coalesce_immediate_stores): Call
12703         check_no_overlap even for the merge_overlapping case.  Formatting fix.
12705 2018-07-12  Richard Biener  <rguenther@suse.de>
12707         PR middle-end/86479
12708         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
12709         move possibly trapping operations into the conditional.
12711 2018-07-12  Richard Biener  <rguenther@suse.de>
12713         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
12714         (vn_lookup_simplify_result): Remove recursion limit applied
12715         here.
12716         (vn_nary_build_or_lookup_1): Adjust.
12717         (try_to_simplify): Likewise.
12718         * gimple-match-head.c (gimple_resimplify1): Instead apply one
12719         here.
12720         (gimple_resimplify2): Likewise.
12721         (gimple_resimplify3): Likewise.
12722         (gimple_resimplify4): Likewise.
12724 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
12726         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
12727         Use __mmask64 type instead of __mmask8 for __M argument.
12728         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
12729         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
12730         __U argument.
12731         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
12732         __mmask16 for __M argument.
12733         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
12734         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12735         to __mmask16 instead of __mmask8.
12736         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
12737         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
12738         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
12739         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
12740         instead of __mmask16 for __U argument.
12741         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
12742         __mmask16 instead of __mmask8 for __U argument.
12743         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
12744         __U argument.
12745         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
12746         __mmask16.
12747         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
12748         argument.
12749         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
12750         __U argument.
12751         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
12752         __mmask16.
12753         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
12754         of __mmask16.
12755         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
12756         __U argument.
12757         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
12758         __U argument.
12759         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
12760         __U argument.
12761         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
12762         __U argument.
12763         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
12764         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
12765         return type as well as __M argument type and all casts from __mmask8
12766         to __mmask32.
12767         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
12768         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
12769         return type as well as __M argument type and all casts from __mmask8
12770         to __mmask16.
12771         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
12772         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
12773         return type as well as __M argument type and all casts from __mmask8
12774         to __mmask32.
12775         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
12776         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
12777         return type as well as __M argument type and all casts from __mmask8
12778         to __mmask16.
12779         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
12780         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
12781         __mmask16.
12783 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
12785         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
12786         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
12787         for __U argument.
12789 2018-07-11  Paul Koning  <ni1d@arrl.net>
12791         * doc/md.texi (define_subst): Document how multiple occurrences of
12792         the same argument in the replacement pattern are handled.
12794 2018-07-11  Paul Koning  <ni1d@arrl.net>
12796         * doc/extend.texi (Common Variable Attributes): Move "mode" into
12797         alphabetical order.
12798         (Common Type Attributes): Add "mode" attribute.
12800 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
12802         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
12803         stream DECL_ORIGINAL_TYPE.
12804         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
12805         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
12806         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
12807         Do not walk original type.
12808         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
12809         external decls.
12810         (write_ts_decl_non_common_tree_pointers): Do not stream
12811         DECL_ORIGINAL_TYPE
12812         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
12813         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
12815 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12817         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
12818         thread twice from the same starting edge.
12820 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12822         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
12823         * gimple.c (gimple_call_nonnull_result_p): ...here...
12824         (gimple_call_nonnull_arg): ...and here.
12825         * gimple.h (gimple_call_nonnull_result_p): New.
12826         (gimple_call_nonnull_arg): New.
12828 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
12830         * config/arm/arm-cpus.in: Move information from fpu field of each
12831         cpu definition to the isa field.
12832         * config/arm/parsecpu.awk (fpu): Delete match rule.
12833         (gen_comm_data): Don't add bits from the CPU's FPU entry.
12835 2018-07-11  Richard Biener  <rguenther@suse.de>
12837         PR debug/86457
12838         * dwarf2out.c (init_sections_and_labels): Use
12839         output_asm_line_debug_info consistently.
12840         (dwarf2out_early_finish): Likewise.
12841         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
12842         type units.
12844 2018-07-11  Richard Biener  <rguenther@suse.de>
12846         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
12847         Rework father_bb setting in a way to avoid propagating constants
12848         multiple times on a loop body.
12850 2018-07-10  Mark Wielaard  <mark@klomp.org>
12852         PR debug/86459
12853         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
12855 2018-07-10  Richard Biener  <rguenther@suse.de>
12857         * hash-map.h (hash_map::iterator::operator*): Return
12858         references to key and value.
12860 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
12862         PR c++/86443
12863         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
12864         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
12865         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
12866         (gimplify_omp_for): For composite loops, move outer
12867         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
12868         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
12869         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
12870         TREE_LIST for both the original class iterator and the "last" helper
12871         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
12872         loop, remember has_decl_expr from outer composite loops for the
12873         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
12875 2018-07-09  Martin Sebor  <msebor@redhat.com>
12877         PR middle-end/77357
12878         PR middle-end/86428
12879         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
12880         accessing implicitly initialized array elements.
12881         * expr.c (string_constant): Handle string initializers of
12882         character arrays within aggregates.
12883         * gimple-fold.c (fold_array_ctor_reference): Add argument.
12884         Store element offset.  As a special case, handle zero size.
12885         (fold_nonarray_ctor_reference): Same.
12886         (fold_ctor_reference): Add argument.  Store subobject offset.
12887         * gimple-fold.h (fold_ctor_reference): Add argument.
12889 2018-07-09  Paul Koning  <ni1d@arrl.net>
12891         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
12892         (pdp11_insn_cost): New function.
12893         (pdp11_md_asm_adjust): New function.
12894         (TARGET_INVALID_WITHIN_DOLOOP): Define.
12895         (pdp11_rtx_costs): Update to match machine better.
12896         (output_addr_const_pdp11): Correct format mismatch warnings.
12897         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
12898         * config/pdp11/pdp11.md: General change to add base_cost and/or
12899         length attributes for use by new pdp11_insn_cost function.
12900         (MIN_BRANCH): Correct definition.
12901         (MIN_SOB): Ditto.
12902         (doloop_end): Use standard pattern name for looping pattern.
12903         (doloop_end_nocc): New.
12904         (movsf): Add another constraint alternative.
12905         (zero_extendqihi2): Add constraint alternatives for not in place
12906         extend.
12907         (zero_extendhisi2): Remove.
12908         (shift patterns): Add CC handling variants.
12909         (bswaphi2): New.
12910         (bswapsi2): New.
12911         (rothi3): New.
12912         (define_peephole2): New peephole to recognize mov that sets CC for
12913         subsequent test.
12915 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12917         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
12918         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
12919         wi::add.
12921 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
12923         PR c/86420
12924         * real.c (real_nextafter): Return true if result is denormal.
12926 2018-07-09  Martin Liska  <mliska@suse.cz>
12928         * common.opt: Add back wrongly removed attribute.
12930 2018-07-09  Richard Biener  <rguenther@suse.de>
12932         PR debug/86413
12933         * dwarf2out.c (gen_block_die): For an early generated DIE
12934         always output high/low PC attributes.
12936 2018-07-09  Tom de Vries  <tdevries@suse.de>
12938         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
12939         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
12940         onto VAR_DECL with abstract origin.
12942 2018-07-07  Jim Wilson  <jimw@sifive.com>
12944         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
12946 2018-07-07  Tom de Vries  <tdevries@suse.de>
12948         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
12950 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
12952         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
12953         overflow_type.
12954         (combine_bound): Use wide-int overflow calculation instead of
12955         rolling our own.
12956         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
12957         overflow_type.
12958         * fold-const.c (int_const_binop_2): Same.
12959         (extract_muldiv_1): Same.
12960         (fold_div_compare): Same.
12961         (fold_abs_const): Same.
12962         * match.pd: Same.
12963         * poly-int.h (add): Same.
12964         (sub): Same.
12965         (neg): Same.
12966         (mul): Same.
12967         * predict.c (predict_iv_comparison): Same.
12968         * profile-count.c (slow_safe_scale_64bit): Same.
12969         * simplify-rtx.c (simplify_const_binary_operation): Same.
12970         * tree-chrec.c (tree_fold_binomial): Same.
12971         * tree-data-ref.c (split_constant_offset_1): Same.
12972         * tree-if-conv.c (idx_within_array_bound): Same.
12973         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
12974         * tree-ssa-phiopt.c (minmax_replacement): Same.
12975         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
12976         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
12977         * vr-values.c (vr_values::adjust_range_with_scev): Same.
12978         * wide-int.cc (wi::add_large): Same.
12979         (wi::mul_internal): Same.
12980         (wi::sub_large): Same.
12981         (wi::divmod_internal): Same.
12982         * wide-int.h: Change overflow type to overflow_type for neg, add,
12983         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
12984         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
12985         mul_internal, divmod_internal.
12986         (overflow_type): New enum.
12987         (accumulate_overflow): New.
12989 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
12991         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
12992         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
12994 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12996         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
12997         argument is checked for zero before entering loop, avoid checking again.
12999 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13001         * gimplify.h (generic_expr_could_trap_p): Set as global function.
13002         * gimplify.h (generic_expr_could_trap_p): Likwise.
13003         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13005 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
13007         PR tree-optimization/86401
13008         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13009         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13010         (fold_bit_and_mask): ... here.  New helper function for match.pd.
13011         * fold-const.h (fold_bit_and_mask): Declare.
13012         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13014 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
13016         PR target/86324
13017         * target.def (translate_mode_attribute): New hook.
13018         * targhooks.h (default_translate_mode_attribute): Declare.
13019         * targhooks.c (default_translate_mode_attribute): New function.
13020         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13021         * doc/tm.texi: Regenerate.
13022         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13023         (rs6000_translate_mode_attribute): New function.
13025 2018-07-06  Paul Koning  <ni1d@arrl.net>
13027         * doc/md.texi (define_split): Document DONE and FAIL.
13028         (define_peephole2): Ditto.
13030 2018-07-05  Jeff Law  <law@redhat.com>
13032         PR tree-optimization/86010
13033         * tree-ssa-dse.c (compute_trims): More aggressively trim at
13034         both the head and tail of mem* and str* calls.
13036 2018-07-05  Jim Wilson  <jimw@sifive.com>
13038         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13040 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
13042         * config/aarch64/aarch64-simd.md: correct flags text for
13043         MIN_EXPR replacement.
13045 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
13047         * configure: Regenerated.
13049 2018-07-05  Carl Love  <cel@us.ibm.com>
13051         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13052         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13053         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13054         VSX_BUILTIN_DOUBLEL_V4SF.
13056 2018-07-05  Martin Sebor  <msebor@redhat.com>
13058         PR c++/86400
13059         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13060         than its domain to compute its the upper bound of a char array.
13062 2018-07-05  Nathan Sidwell  <nathan@acm.org>
13064         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13065         * doc/cpp.texi: Update comment.
13066         * doc/tm.texi: Rebuilt.
13067         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13068         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13069         * doc/extend.texi (Backwards Compatibility): Clarify it is system
13070         headers affected by extern "C".
13071         * system.h: Poison NO_IMPLICIT_EXTERN_C.
13072         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13073         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13074         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13075         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13076         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13077         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13078         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13079         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13080         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13081         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13082         config/sparc/sp64-elf.h, config/spu/spu.h,
13083         config/stormy16/stormy16.h, config/v850/v850.h,
13084         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13085         define NO_IMPLICIT_EXTERN_C.
13086         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13088 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13090         PR target/84711
13091         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13092         instead of GET_MODE_SIZE when comparing Units.
13094 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13096         PR target/84711
13097         * rtlanal.c (set_noop_p): Constrain on mode change,
13098         include hard-reg-set.h
13100 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13102         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13104 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13106         Revert
13107         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13109         PR sanitizer/84250
13110         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13111         libasan.
13112         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13114 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13116         PR sanitizer/84250
13117         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13118         libasan.
13119         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13121 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13123         PR middle-end/86380
13124         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13126 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13128         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13129         neg_*_op* variables.
13131 2018-07-04  Martin Liska  <mliska@suse.cz>
13133         * tree-switch-conversion.c: Define
13134         max_ratio_for_speed and max_ratio_for_size constants.
13136 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13137             Martin Liska  <mliska@suse.cz>
13139         PR middle-end/66240
13140         PR target/45996
13141         PR c/84100
13142         * common.opt: Rename align options with 'str_' prefix.
13143         * common/config/i386/i386-common.c (set_malign_value): New
13144         function.
13145         (ix86_handle_option): Use it to set -falign-* options/
13146         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13147         type from int to string.
13148         * config/aarch64/aarch64.c: Update default values from int
13149         to string.
13150         * config/alpha/alpha.c (alpha_override_options_after_change):
13151         Likewise.
13152         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13153         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13154         max skip conditionally.
13155         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13156         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13157         max skip conditionally.
13158         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13159         max skip conditionally.
13160         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13161         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13162         max skip conditionally.
13163         * config/i386/i386.c (struct ptt): Change type from int to
13164         string.
13165         (ix86_default_align): Set default values.
13166         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13167         max skip conditionally.
13168         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13169         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13170         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13171         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13172         max skip conditionally.
13173         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13174         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13175         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13176         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13177         max skip conditionally.
13178         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13179         * config/ia64/ia64.c (ia64_option_override): Set default values
13180         for alignment options.
13181         * config/m68k/m68k.c: Handle new str_align_* options.
13182         * config/mips/mips.c (mips_set_compression_mode): Change
13183         type of constants.
13184         (mips_option_override): Set default values for options.
13185         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13186         Likewise.
13187         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13188         Likewise.
13189         * config/rx/rx.c (rx_option_override): Likewise.
13190         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13191         (LABEL_ALIGN): Use align_labels_log.
13192         (LOOP_ALIGN): Use align_loops_align.
13193         * config/s390/s390.c (s390_asm_output_function_label): Use new
13194         macros.
13195         * config/sh/sh.c (sh_override_options_after_change):
13196         Change type of constants.
13197         * config/spu/spu.c (spu_sched_init): Likewise.
13198         * config/sparc/sparc.c (sparc_option_override): Set default
13199         values for options.
13200         * config/visium/visium.c (visium_option_override): Likewise.
13201         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13202         emit p2align format with last argument if it's not needed.
13203         * doc/invoke.texi: Document extended format of -falign-*.
13204         * final.c: Use align_labels alignment.
13205         * flags.h (struct target_flag_state): Change type to use
13206         align_flags.
13207         (struct align_flags_tuple): New.
13208         (struct align_flags): Likewise.
13209         (align_loops_log): Redefine macro to use new types.
13210         (align_loops_max_skip): Redefine macro to use new types.
13211         (align_jumps_log): Redefine macro to use new types.
13212         (align_jumps_max_skip): Redefine macro to use new types.
13213         (align_labels_log): Redefine macro to use new types.
13214         (align_labels_max_skip): Redefine macro to use new types.
13215         (align_functions_log): Redefine macro to use new types.
13216         (align_loops): Redefine macro to use new types.
13217         (align_jumps): Redefine macro to use new types.
13218         (align_labels): Redefine macro to use new types.
13219         (align_functions): Redefine macro to use new types.
13220         (align_functions_max_skip): Redefine macro to use new types.
13221         (align_loops_value): New macro.
13222         (align_jumps_value): New macro.
13223         (align_labels_value): New macro.
13224         (align_functions_value): New macro.
13225         * function.c (invoke_set_current_function_hook): Propagate
13226         alignment values from flags to global variables default in
13227         topleev.h.
13228         * ipa-icf.c (sem_function::equals_wpa): Use
13229         cl_optimization_option_eq instead of memcmp.
13230         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13231         of string types.
13232         (cl_optimization_stream_in): Likewise.
13233         * optc-save-gen.awk: Support strings in cl_optimization.
13234         * opth-gen.awk: Likewise.
13235         * opts.c (finish_options): Remove error checking of invalid
13236         value ranges.
13237         (MAX_CODE_ALIGN): Remove.
13238         (MAX_CODE_ALIGN_VALUE): Likewise.
13239         (parse_and_check_align_values): New function.
13240         (check_alignment_argument): Likewise.
13241         (common_handle_option): Use check_alignment_argument.
13242         * opts.h (parse_and_check_align_values): Declare.
13243         * toplev.c (init_alignments): Remove.
13244         (read_log_maxskip): New.
13245         (parse_N_M): Likewise.
13246         (parse_alignment_opts): Likewise.
13247         (backend_init_target): Remove usage of init_alignments.
13248         * toplev.h (parse_alignment_opts): Declare.
13249         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13250         argument.
13251         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13252         * tree.c (cl_option_hasher::equal): New.
13253         * varasm.c: Use new global macros.
13255 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13257         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13258         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13259         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13260         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13261         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13262         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13263         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13264         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13265         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13266         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13267         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13268         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13269         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13271 2018-07-04  Martin Liska  <mliska@suse.cz>
13272             Jonathan Wakely  <jwakely@redhat.com>
13274         * coverage.c: Use correct type.
13275         * doc/invoke.texi: Language correction.
13277 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13279         PR target/85620
13280         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13281         ENDBRANCH for non-tail call which may return via indirect branch.
13282         * doc/extend.texi: Document indirect_return attribute.
13284 2018-07-03  Martin Sebor  <msebor@redhat.com>
13286         PR tree-optimization/86274
13287         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13288         precondition.
13289         (format_floating): Correct handling of infinities and NaNs.
13291 2018-07-03  Martin Sebor  <msebor@redhat.com>
13293         * print-tree.c (print_real_cst): New function.
13294         (print_node_brief): Call it.
13295         (print_node): Ditto.
13297 2018-07-03  Jeff Law  <law@redhat.com>
13299         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13300         into a single pattern.
13302         * config/h8300/h8300.md (ors code_iterator): New.
13303         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13304         a single pattern and single splitter.
13305         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13306         (iorqi3_1, xorqi3_1): Likewise.
13307         (iorqi3, xorqi3 expanders): Similarly.
13309         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13310         (movmd_internal) into a single pattern using the P mode iterator.
13311         (movmd splitters): Similarly.
13312         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13313         (movsd splitters): Similarly.
13315         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13316         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13317         right table based on the size of the operand.
13318         * config/h8300/h8300.md (length_table): Corresponding changes. All
13319         references to "addb", "addw" and "addl" changed to "add".
13320         (btst patterns): Merge two variants into a single pattern.
13321         (tstqi, tsthi): Likewise.
13322         (addhi3_incdec, addsi3_incdec): Likewise.
13323         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13324         (mulhi3, mulsi3): Likewise.
13325         (udivhi3, udivsi3): Likewise.
13326         (divhi3, divsi3): Likewise.
13327         (andorqi3, andorhi3, andorsi3): Likewise.
13329 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13331         PR target/85694
13332         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13333         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13335 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13337         PR tree-optimization/85694
13338         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13339         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13340         UNSPEC_URHADD.
13341         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13342         (<u>avg<mode>3_ceil): New patterns.
13344 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13346         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13347         scan-tree-dump directive.
13348         * gcc.dg/vect/slp-perm-2.c: Likewise.
13349         * gcc.dg/vect/slp-perm-3.c: Likewise.
13350         * gcc.dg/vect/slp-perm-5.c: Likewise.
13351         * gcc.dg/vect/slp-perm-6.c: Likewise.
13352         * gcc.dg/vect/slp-perm-7.c: Likewise.
13353         * gcc.dg/vect/slp-perm-8.c: Likewise.
13355 2018-07-03  Marek Polacek  <polacek@redhat.com>
13357         PR middle-end/86202
13358         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13360 2018-07-03  Richard Biener  <rguenther@suse.de>
13362         PR ipa/86389
13363         * tree-ssa-structalias.c (find_func_clobbers): Properly
13364         handle indirect calls.
13366 2018-07-03  Jeff Law  <law@redhat.com>
13368         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13369         (shifts): New code iterator.
13370         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13371         expander.  Fix HImode handling on H8/SX.
13372         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13373         (subqi3, subhi3, subsi3 expanders): Likewise.
13374         (andqi3, andhi3, andsi3 expanders): Likewise.
13375         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13376         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13377         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13378         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13379         (zero_extendqihi2, zero_extendqisi2): Likewise.
13380         (extendqihi2, extendqisi2): Likewise.
13381         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13382         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13383         (rotlqi3_1, rotlhi3_1): Likewise.
13384         (logicalhi3_sn, logicalsi3_sn): Likewise.
13385         (logicalhi3, logicalsi3): Likewise.
13387 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13389         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13390         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13391         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13392         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13393         type to append_pattern_def_seq instead of creating a stmt_vec_info
13394         directly.
13395         (build_mask_conversion): Likewise.  Remove vinfo argument.
13396         (vect_add_conversion_to_patterm): Likewise, renaming to...
13397         (vect_add_conversion_to_pattern): ...this.
13398         (vect_recog_mask_conversion_pattern): Update call to
13399         build_mask_conversion.  Pass the vector type to
13400         append_pattern_def_seq here too.
13401         (vect_recog_gather_scatter_pattern): Update call to
13402         vect_add_conversion_to_pattern.
13404 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13406         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13407         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13408         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13409         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13410         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13411         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13412         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13413         append_pattern_def_seq instead of new_pattern_def_seq.
13414         (vect_recog_divmod_pattern): Do both of the above.
13415         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13416         is null.
13418 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13420         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13421         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13422         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13423         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13424         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13425         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13426         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13427         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13428         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13429         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13430         parameter with a single stmt_vec_info.
13431         (vect_recog_func_ptr): Likewise.
13432         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13433         (vect_try_gather_scatter_pattern): ...this.
13434         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13435         the stmt_vec_info of the statement to be matched.  Don't clear
13436         STMT_VINFO_RELATED_STMT.
13437         (vect_pattern_recog): Update call accordingly.
13439 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13441         PR tree-optimization/85694
13442         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13443         (uavgM3_ceil): Document new optabs.
13444         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13445         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13446         functions.
13447         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13448         (savg_ceil_optab): New optabs.
13449         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13450         (vect_vect_recog_func_ptrs): Add it.
13451         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13452         constant directly from the associated lhs.
13454 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13456         * tree-vect-patterns.c (vect_split_statement): New function.
13457         (vect_convert_input): Use it to try to split an existing cast.
13459 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13461         * poly-int.h (print_hex): New function.
13462         * dumpfile.h (dump_dec, dump_hex): Declare.
13463         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13464         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13465         min_input_precision, operation_precision and operation_sign.
13466         * tree-vect-patterns.c (vect_get_range_info): New function.
13467         (vect_same_loop_or_bb_p, vect_single_imm_use)
13468         (vect_operation_fits_smaller_type): Delete.
13469         (vect_look_through_possible_promotion): Add an optional
13470         single_use_p parameter.
13471         (vect_recog_over_widening_pattern): Rewrite to use new
13472         stmt_vec_info infomration.  Handle one operation at a time.
13473         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13474         (vect_truncatable_operation_p, vect_set_operation_type)
13475         (vect_set_min_input_precision): New functions.
13476         (vect_determine_min_output_precision_1): Likewise.
13477         (vect_determine_min_output_precision): Likewise.
13478         (vect_determine_precisions_from_range): Likewise.
13479         (vect_determine_precisions_from_users): Likewise.
13480         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13481         (vect_vect_recog_func_ptrs): Put over_widening first.
13482         Add cast_forwprop.
13483         (vect_pattern_recog): Call vect_determine_precisions.
13485 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13487         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13488         statements that have been replaced by further pattern statements.
13489         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13491 2018-07-03  Richard Biener  <rguenther@suse.de>
13493         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13494         always set *dt.  Dump vectype in vectype overload.
13495         * dumpfile.h (dump_gimple_expr): New function.
13496         (dump_gimple_expr_loc): Likewise.
13497         * dumpfile.c (dump_gimple_expr): New function.
13498         (dump_gimple_expr_loc): Likewise.
13500 2018-07-02  Jeff Law  <law@redhat.com>
13502         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13503         the H8/300, H8/300H and H8/S variants into a single pattern.
13504         (movhi_h8300, movqi_h8300hs): Similarly.
13505         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13506         (QHI mode iterator): New.
13508         * config/h8300/h8300.md: Remove trailing whitespace.
13510 2018-07-02  Jim Wilson  <jimw@sifive.com>
13512         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13513         instead of emit_insn for interrupt returns.
13514         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13515         (riscv_sret, riscv_uret): Likewise.
13517 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13519         * pretty-print.c (selftest::test_pp_format): Move save and restore
13520         of quotes to class auto_fix_quotes, and add an instance.
13521         * selftest.c: Include "intl.h".
13522         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13523         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13524         * selftest.h (selftest::auto_fix_quotes): New class.
13526 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13528         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13529         (aarch64_sve_prepare_conditional_op): Remove.
13530         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13531         Allow aarch64_simd_reg_or_zero as select operand; remove
13532         the aarch64_sve_prepare_conditional_op call.
13533         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13534         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13535         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13536         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13537         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13538         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13539         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13540         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13541         and a splitters to match all of the *_any patterns.
13542         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13544         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13545         (SVE_COND_FP_BINARY_REV): Remove.
13546         (sve_int_op_rev, sve_fp_op_rev): New.
13547         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13548         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13549         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13550         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13551         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13552         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13553         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13554         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13555         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13557         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13558         Remove match_dup 1 from the inner unspec.
13559         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13561         * config/aarch64/aarch64.md (movprfx): New attr.
13562         (length): Default movprfx to 8.
13563         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13564         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13565         (*<su>mul<SVE_I>3_highpart): Likewise.
13566         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13567         (*v<ASHIFT><SVE_I>3): Likewise.
13568         (*<su><MAXMIN><SVE_I>3): Likewise.
13569         (*<su><MAXMIN><SVE_F>3): Likewise.
13570         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13571         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13572         (*div<SVE_F>4): Likewise.
13574 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
13576         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13577         in dump string.
13579 2018-07-02  Richard Biener  <rguenther@suse.de>
13581         PR tree-optimization/86363
13582         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
13583         memset argument refers to a non-variable address.
13585 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13587         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
13588         {PLUS,MINUS}_EXPR code to...
13589         (adjust_symbolic_bound): ...here,
13590         (combine_bound): ...here,
13591         (set_value_range_with_overflow): ...and here.
13593 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13595         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
13596         code...
13597         (extract_range_from_abs_expr): ...here.
13599 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
13601         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
13602         -fno-omit-frame-pointer when not optimizing.
13604 2018-07-02  Martin Liska  <mliska@suse.cz>
13606         PR ipa/86279
13607         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
13608         (propagate_nothrow): Likewise.
13610 2018-07-02  Martin Liska  <mliska@suse.cz>
13612         PR ipa/86323
13613         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
13615 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13617         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
13618         function in r262149, changing "loc" param from source_location to
13619         const dump_location_t &.
13620         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
13621         declaration, as above.
13623 2018-07-01  Paul Koning  <ni1d@arrl.net>
13625         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13626         -munit-asm, -mgnu-asm, -mdec-asm.
13627         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
13628         (pdp11_output_labelref): New.
13629         (pdp11_output_def): New.
13630         (pdp11_output_addr_vec_elt): New.
13631         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
13632         %# and %@ format codes.
13633         (pdp11_option_override): New.
13634         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
13635         (pdp11_output_ident): New.
13636         (pdp11_asm_named_section): New.
13637         (pdp11_asm_init_sections): New.
13638         (pdp11_file_start): New.
13639         (pdp11_file_end): New.
13640         (output_ascii): Use .ascii/.asciz for -mdec-asm.
13641         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
13642         %o, like %c but octal.
13643         (pdp11_option_override): New.
13644         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
13645         -mdec-asm.
13646         (DATA_SECTION_ASM_OP): Ditto.
13647         (READONLY_DATA_SECTION_ASM_OP): New.
13648         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
13649         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
13650         (ASM_OUTPUT_LABELREF): Ditto.
13651         (ASM_OUTPUT_DEF): Ditto.
13652         (ASM_OUTPUT_EXTERNAL): New.
13653         (ASM_OUTPUT_SOURCE_FILENAME): New.
13654         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
13655         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
13656         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
13657         %# and %@ format codes.
13658         * config/pdp11/pdp11.opt (mgnu-asm): New.
13659         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
13660         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
13661         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
13663 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
13665         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
13666         dereferencing path[] beyond its length.
13667         (debug_path): New.
13668         (debug_all_paths): New.
13669         (rewire_first_differing_edge): New.
13670         (adjust_paths_after_duplication): New.
13671         (duplicate_thread_path): Call adjust_paths_after_duplication.
13672         Add new argument.
13673         (thread_through_all_blocks): Add new argument to
13674         duplicate_thread_path.
13676 2018-06-30  Jim Wilson  <jimw@sifive.com>
13678         * config/riscv/predicates.md (p2m1_shift_operand): New.
13679         (high_mask_shift_operand): New.
13680         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
13681         pattern using p2m1_shift_operand.
13682         (lshsi3_zero_extend_3+2): New combiner pattern using
13683         high_mask_shift_operand.
13685 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13687         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
13688         split out from...
13689         (vect_recog_rotate_pattern): ...here.
13690         (vect_convert_input): Try to insert casts of invariants in the
13691         preheader.
13692         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
13693         preheader to be empty.
13695 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13697         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
13698         vector type.  If given, install it in the new statement's
13699         STMT_VINFO_VECTYPE.
13700         (vect_element_precision): New function.
13701         (vect_unpromoted_value): New struct.
13702         (vect_unpromoted_value::vect_unpromoted_value): New function.
13703         (vect_unpromoted_value::set_op): Likewise.
13704         (vect_look_through_possible_promotion): Likewise.
13705         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
13706         (vect_widened_op_tree, vect_convert_input): Likewise.
13707         (vect_convert_inputs, vect_convert_output): Likewise.
13708         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
13709         to handle the optional cast of the multiplication result and
13710         vect_widened_op_tree to detect the widened multiplication itself.
13711         Do not require the input and output of promotion casts to have
13712         the same sign, but base the signedness of the operation on the
13713         input rather than the result.  If the pattern includes two
13714         promotions, check that those promotions have the same sign.
13715         Do not restrict the MULT_EXPR handling to a double-width result;
13716         handle quadruple-width results and wider.  Use vect_convert_inputs
13717         to convert the inputs to the common type.
13718         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
13719         to handle the optional cast of the ABS result.  Also allow a sign
13720         change or a sign extension between the ABS and MINUS.
13721         Use vect_widened_op_tree to detect the widened subtraction and use
13722         vect_convert_inputs to convert the inputs to the common type.
13723         (vect_handle_widen_op_by_const): Delete.
13724         (vect_recog_widen_op_pattern): New function.
13725         (vect_recog_widen_mult_pattern): Use it.
13726         (vect_recog_widen_shift_pattern): Likewise.
13727         (vect_recog_widen_sum_pattern): Use
13728         vect_look_through_possible_promotion to handle the promoted
13729         PLUS_EXPR operand.
13731 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13733         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
13734         the containing gimple_seq *.
13735         * gimple-iterator.h (gsi_for_stmt): Declare it.
13736         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
13737         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
13738         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13739         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13740         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
13741         checks.
13742         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
13743         split out from...
13744         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
13745         statement being replaced is part of an existing pattern
13746         definition sequence, inserting the new pattern statements before
13747         the original one.
13748         (vect_pattern_recog_1): Don't return a bool.  If the statement
13749         is already part of a pattern, instead apply pattern matching
13750         to the pattern definition statements.  Don't clear the
13751         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
13752         (vect_pattern_recog): Don't break after the first match;
13753         continue processing the pattern definition statements instead.
13754         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
13756 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13758         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13759         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13760         (vect_recog_widen_sum_pattern): Use it.
13762 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13764         * tree-vect-loop.c (vectorizable_reduction): Assert that the
13765         phi is not a pattern statement and has not been replaced by
13766         a pattern statement.
13767         * tree-vect-patterns.c (type_conversion_p): Don't check
13768         STMT_VINFO_IN_PATTERN_P.
13769         (vect_recog_vector_vector_shift_pattern): Likewise.
13770         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
13771         the pattern statement rather than the original statement; check
13772         directly for a WIDEN_MULT_EXPR here.
13773         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
13774         vect_is_simple_use to return the pattern statement rather
13775         than the original statement; use is_pattern_stmt_p to check
13776         for such a pattern statement.
13777         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
13778         to return the pattern statement rather than the original statement;
13779         don't do the same transformation here.
13780         (vect_is_simple_use): If the defining statement has been replaced
13781         by a pattern statement, return the pattern statement instead.
13782         Remove the corresponding (local) transformation from the vectype
13783         overload.
13785 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13787         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
13788         end and default to null.
13789         * tree-vect-loop.c (vect_create_epilog_for_reduction)
13790         (vectorizable_reduction): Update calls accordingly, dropping the
13791         gimple ** argument if the passed-back statement isn't needed.
13792         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
13793         (vect_recog_rotate_pattern): Likewise.
13794         (vect_recog_mask_conversion_pattern): Likewise.
13795         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13796         (vect_mask_constant_operand_p): Likewise.
13797         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
13798         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
13799         (get_group_load_store_type, get_load_store_type): Likewise.
13800         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
13801         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
13802         (vectorizable_conversion, vectorizable_assignment): Likewise.
13803         (vectorizable_shift, vectorizable_operation): Likewise.
13804         (vectorizable_store, vect_is_simple_cond): Likewise.
13805         (vectorizable_condition, vectorizable_comparison): Likewise.
13806         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
13807         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
13808         and move it to the end.  Cope with null def_stmt_outs.
13810 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13812         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
13814 2018-06-29  Jeff Law  <law@redhat.com>
13816         * config/v850/v850.c (v850_legitimate_address_p): Handle large
13817         displacements for TARGET_V850E2V3 and newer.
13818         (TARGET_LRA_P): Remove.  Defaults to LRA now.
13819         * config/v850/v850.md (sign23byte_load): Remove.
13820         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
13821         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
13823 2018-06-29  Martin Liska  <mliska@suse.cz>
13825         PR lto/85759
13826         * coverage.c (coverage_init): Mangle full path name.
13827         * doc/invoke.texi: Document the change.
13828         * gcov-io.c (mangle_path): New.
13829         * gcov-io.h (mangle_path): Likewise.
13830         * gcov.c (mangle_name): Use mangle_path for path mangling.
13832 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13834         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
13835         if starting source register is not even.
13837 2018-06-29  Martin Liska  <mliska@suse.cz>
13839         PR tree-optimization/86263
13840         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
13841         Make edge redirection.
13843 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
13845         * dumpfile.c (dump_loc): Add indentation based on scope depth.
13846         (dump_scope_depth): New variable.
13847         (get_dump_scope_depth): New function.
13848         (dump_begin_scope): New function.
13849         (dump_end_scope): New function.
13850         * dumpfile.h (get_dump_scope_depth): New declaration.
13851         (dump_begin_scope): New declaration.
13852         (dump_end_scope): New declaration.
13853         (class auto_dump_scope): New class.
13854         (AUTO_DUMP_SCOPE): New macro.
13855         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
13856         AUTO_DUMP_SCOPE.
13858 2018-06-29  Richard Biener  <rguenther@suse.de>
13860         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
13861         compute_all_dependences succeeds.
13862         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
13863         exceed --param loop-max-datarefs-for-datadeps.
13865 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
13867         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
13869 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
13871         PR target/86348
13872         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
13873         alternative 0 in preferred_for_speed attribute.
13875 2018-06-28  Paul Koning  <ni1d@arrl.net>
13877         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
13878         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
13879         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
13880         * config/pdp11/pdp11.md: Correct "length" attribute calculation
13881         for shift insn patterns.
13883 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
13885         * cgraph.c (cgraph_node::get_body): Replace assignments to
13886         "dump_file" with calls to set_dump_file.
13887         * dumpfile.c (alt_dump_file): Make static, and group with...
13888         (alt_flags): ...this definition.
13889         (dumps_are_enabled): New variable.
13890         (refresh_dumps_are_enabled): New function.
13891         (set_dump_file): New function.
13892         (set_alt_dump_file): New function.
13893         (gcc::dump_manager::dump_start): Replace assignments to
13894         "dump_file" and "alt_dump_file" with calls to set_dump_file and
13895         set_alt_dump_file.
13896         (gcc::dump_manager::dump_finish): Likewise.
13897         * dumpfile.h (alt_dump_file): Delete decl.
13898         (dumps_are_enabled): New variable decl.
13899         (set_dump_file): New function decl.
13900         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
13901         global.
13902         * tree-nested.c (lower_nested_functions): Replace assignments to
13903         "dump_file" with calls to set_dump_file.
13905 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13907         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
13908         goto_locus of each outgoing edge of each basic block.
13910 2018-06-28  Richard Biener  <rguenther@suse.de>
13912         * dwarf2out.c (decl_scope_table): Remove.
13913         (push_decl_scope): Likewise.
13914         (pop_decl_scope): Likewise.
13915         (gen_type_die_for_member): Do not call push/pop_decl_scope.
13916         (gen_struct_or_union_type_die): Likewise.
13917         (gen_tagged_type_die): Likewise.
13918         (dwarf2out_init): Do not initialize decl_scope_table.
13919         (dwarf2out_c_finalize): Do not free it.
13921 2018-06-28  Richard Biener  <rguenther@suse.de>
13923         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
13924         deciding whether to not re-use a DIE.
13926 2018-06-28  Richard Biener  <rguenther@suse.de>
13928         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
13929         DW_AT_abstract_origin attribute.
13931 2018-06-28  Martin Liska  <mliska@suse.cz>
13933         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13934         Use newly introduced constants.
13935         * tree-switch-conversion.h (struct jump_table_cluster):
13936         Define max_ratio_for_size and max_ratio_for_speed.
13938 2018-06-28  Martin Liska  <mliska@suse.cz>
13940         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13941         Add new checking assert to catch invalid state.
13942         (jump_table_cluster::can_be_handled): Handle single case
13943         clusters.
13944         (jump_table_cluster::is_beneficial): Bail out for such case.
13945         (bit_test_cluster::find_bit_tests):
13946         Add new checking assert to catch invalid state.
13947         (bit_test_cluster::can_be_handled): Handle single case
13948         clusters.
13949         (bit_test_cluster::is_beneficial): Bail out for such case.
13950         (switch_decision_tree::analyze_switch_statement):
13951         Fix comment.
13953 2018-06-28  Martin Liska  <mliska@suse.cz>
13955         * common.opt: Introduce -completion option.
13956         * gcc.c (driver_handle_option): Handle it.
13957         (driver::main): Print completions if completion
13958         is set.
13959         * opt-suggestions.c (option_proposer::get_completions):
13960         New function.
13961         (option_proposer::suggest_completion): Likewise.
13962         (option_proposer::find_param_completions): Likewise.
13963         (verify_autocompletions): Likewise.
13964         (test_completion_valid_options): Likewise.
13965         (test_completion_valid_params): Likewise.
13966         (in_completion_p): Likewise.
13967         (empty_completion_p): Likewise.
13968         (test_completion_partial_match): Likewise.
13969         (test_completion_garbage): Likewise.
13970         (opt_proposer_c_tests): Likewise.
13971         * opt-suggestions.h: Declare new functions.
13972         * opts.c (common_handle_option): Handle OPT__completion_.
13973         * selftest-run-tests.c (selftest::run_tests): Add
13974         opt_proposer_c_tests.
13975         * selftest.c (assert_str_startswith): New.
13976         * selftest.h (assert_str_startswith): Likewise.
13977         (opt_proposer_c_tests): New.
13978         (ASSERT_STR_STARTSWITH): Likewise.
13980 2018-06-28  Martin Liska  <mliska@suse.cz>
13982         * Makefile.in: Add opt-suggestions.o.
13983         * gcc-main.c: Include opt-suggestions.h.
13984         * gcc.c (driver::driver): Likewise.
13985         (driver::~driver): Remove m_option_suggestions.
13986         (driver::build_option_suggestions): Moved to option_proposer.
13987         (driver::suggest_option): Likewise.
13988         (driver::handle_unrecognized_options): Use option_proposer.
13989         * gcc.h (class driver): Add new memver m_option_proposer.
13990         * opt-suggestions.c: New file.
13991         * opt-suggestions.h: New file.
13993 2018-06-28  Martin Liska  <mliska@suse.cz>
13995         * vec.h (class auto_string_vec): New (moved from auto_argvec).
13996         (auto_string_vec::~auto_string_vec): Likewise.
13998 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14000         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14001         prevent_decl_creation_for_types fields up and add reset_location field.
14002         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14003         statement if id->reset_location is true.
14004         (copy_edges_for_bb): Do not set goto_locus on the new edges if
14005         id->reset_location is true.
14006         (copy_phis_for_bb): Force input_location on the arguments if
14007         id->reset_location is true.
14008         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14009         is set on the function to be inlined.
14011 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
14013         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14015 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
14017         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14018         registers for Pmode.
14019         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14020         hard registers for the clobbered pseudo.
14022 2018-06-27  Paul Koning  <ni1d@arrl.net>
14024         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14025         mutually exclusive options.
14026         * config/pdp11/constraints.md (h): New constraint.
14027         (O): Update definition to match shift code generation.
14028         (D): New constraint.
14029         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14030         (CCFP): Remove.
14031         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14032         function.
14033         (output_jump): Change arguments.
14034         (pdp11_fixed_cc_regs): New function.
14035         (pdp11_cc_mode): Ditto.
14036         (pdp11_expand_shift): Ditto.
14037         (pdp11_assemble_shift): Ditto.
14038         (pdp11_small_shift): Ditto.
14039         (pdp11_branch_cost): Remove.
14040         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14041         from output.
14042         (pdp11_register_move_cost): Update for CC registers.
14043         (pdp11_rtx_costs): Add case for LSHIFTRT.
14044         (pdp11_output_jump): Add CCNZ mode conditional branches.
14045         (notice_update_cc_on_set): Remove.
14046         (pdp11_cc_mode): New function.
14047         (simple_memory_operand): Correct pre/post decrement case.
14048         (no_side_effect_operand): New function.
14049         (pdp11_regno_reg_class): Add CC_REGS class.
14050         (pdp11_fixed_cc_regs): New function.
14051         (pdp11_small_shift): New function.
14052         (pdp11_expand_shift): New function to expand shift insns.
14053         (pdp11_assemble_shift): New function to output shifts.
14054         (pdp11_branch_cost): Remove.
14055         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14056         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14057         (WCHAR_TYPE): Ditto.
14058         (PTRDIFF_TYPE): Ditto.
14059         (ADJUST_INSN_LENGTH): New macro.
14060         (FIXED_REGISTERS): Add CC registers.
14061         (CALL_USED_REGISTERS): Ditto.
14062         (reg_class): Ditto.
14063         (REG_CLASS_NAMES): Ditto.
14064         (REG_CLASS_CONTENTS): Ditto.
14065         (SELECT_CC_MODE): Use new function.
14066         (TARGET_FLAGS_REGNUM): New macro.
14067         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14068         (cc0_reg_rtx): Remove.
14069         (CC_STATUS_MDEP): Remove.
14070         (CC_STATUS_MDEFP_INIT): Remove.
14071         (CC_IN_FPU): Remove.
14072         (NOTICE_UPDATE_CC): Remove.
14073         (REGISTER_NAMES): Add CC registers.
14074         (BRANCH_COST): Change to constant 1.
14075         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14076         handling.
14077         * config/pdp11/pdp11.opt (mbcopy): Remove.
14078         (mbcopy-builtin): Remove.
14079         (mbranch-cheap): Remove.
14080         (mbranch-expensive): Remove.
14081         * config/pdp11/predicates.md (expand_shift_operand): Update to
14082         match shift code generation.
14083         (ccnz_operator): New predicate.
14084         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14085         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14086         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14087         exclusive options.
14088         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14089         description of O constraint.
14091 2018-06-27  Jeff Law  <law@redhat.com>
14092             Austin Law  <austinklaw@gmail.com>
14094         * config/v850/v850.md (addsi3_set_flags): New pattern.
14095         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14096         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14097         (zero_extendhisi2_v850_set_flags): Likewise.
14098         (zero_extendqisi2_v850_set_flags): Likewise.
14099         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14100         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14101         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14103         * config/v850/v850-protos.h (notice_update_cc): Remove.
14104         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14105         (v850_print_operand): Handle 'D' and "d".
14106         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14107         Add handling of arithmetic/logical operations compared against zero.
14108         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14109         Do not look at v850_compare_op, instead get mode from last argument.
14110         (v850_gen_compare): Remove
14111         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14112         after reload for prologue insns.
14113         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14114         patterns.
14115         (construct_save_jarl): Likewise.
14116         (TARGET_FLAGS_REGNUM): Define.
14117         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14118         (NOTICE_UPDATE_CC): Remove.
14119         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14120         than cc0.  Conditionalize on reload_completed.
14121         (cmpsi_insn, setfcc_insn): Likewise.
14122         (tst1 splitter): Turn into define_and_split which sets the flags
14123         after reload.
14124         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14125         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14126         (cstoresf4, cstoredf4): Clobber the flags.
14127         (cmpsi, cmpsf, cmpdf): Remove expanders.
14128         (setf_insn): Remove pattern.
14129         (addsi3): Turn into define_and_split which clobbers the flags after
14130         reload and a suitable pattern (addsi3_clobber_flags) for use after
14131         reload.
14132         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14133         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14134         (ashrsi3, ashrsi3_v850e2): Likewise.
14135         (bins): Clobber the flags.
14136         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14137         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14138         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14139         (call_value_internal_short, call_value_internal_long): Likewise.
14140         (callt_save_interrupt, callt_return_interrupt): Likewise.
14141         (save_interrupt, return_interrupt): Likewise.
14142         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14143         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14144         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14145         (All FP comparisons): Only allow after reload has completed.
14146         (trfsr): Likewise.
14147         (divh, divhu): Tweak output template.
14148         (branch_z_normal, branch_z_invert): Remove
14149         (branch_nz_normal, branch_nz_invert): Likewise.
14150         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14152         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14153         * config/v850/v850.c (notice_update_cc): Remove.
14154         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14155         (CC_NO_CARRY): Likewise.
14156         (NOTICE_UPDATE_CC): Define to nothing.
14157         * config/v850/v850.md: Remove block comment on cc0 handling
14158         Remove "cc" attribute from all patterns.  Remove cc_status handling
14159         from all patterns.  Minor formatting fixes.
14161 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14163         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14164         (cortex-a76.cortex-a55): Likewise.
14165         * config/aarch64/aarch64-tune.md: Regenerate.
14166         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14167         cortex-a76.cortex-a55.
14169 2018-06-27  Jeff Law  <law@redhat.com>
14171         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14172         (MULTILIB_DIRNAMES): Similarly.
14174 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14176         * gimple.h (gimple_return_retbnd): Delete.
14177         (gimple_return_set_retbnd): Likewise.
14178         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14179         gimple_return_set_retbnd.
14180         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14181         gimple_return_retbnd and adjust.
14182         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14183         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14184         Explicitly return NULL in a couple more cases.  Move assertion
14185         on debug statements and remove unreachable code.
14186         (reset_debug_binding): Do not test id->retbnd.
14187         (expand_call_inline): Do not set it.
14189 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14191         * configure.ac: Add --disable-gcov option.
14192         * configure: Regenerate.
14193         * Makefile.in: Honour @enable_gcov@.
14194         * doc/install.texi: Document --disable-gcov.
14196 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14198         * config/arm/arm-cpus.in (cortex-a76): New entry.
14199         (cortex-a76.cortex-a55): Likewise.
14200         * config/arm/arm-tables.opt: Regenerate.
14201         * config/arm/arm-tune.md: Likewise.
14202         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14203         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14204         cortex-a76.cortex-a55.
14206 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14208         PR target/85769
14209         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14211 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14213         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14214         comment.
14215         (EPILOGUE_USES): Likewise.
14217 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14219         * tree-inline.c (remap_location): New function extracted from...
14220         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14221         (copy_phis_for_bb): ...here.  Call remap_location.
14222         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14224 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14226         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14227         unaligned vsx for 16B memset.
14229 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14231         PR target/86285
14232         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14233         ieee128_float_type_node to long_double_type_node unless
14234         TARGET_LONG_DOUBLE_128 is set.
14236 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14238         * cfgloop.c (get_loop_location): Convert return type from
14239         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14240         by implicit construction from rtx_insn *, and using
14241         dump_user_location_t::from_function_decl for the fallback case.
14242         * cfgloop.h (get_loop_location): Convert return type from
14243         location_t to dump_user_location_t.
14244         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14245         dump_printf_loc to pass in a dump_location_t rather than a
14246         location_t, via the gimple stmt.
14247         * coverage.c (get_coverage_counts): Update calls to
14248         dump_printf_loc to pass in dump_location_t rather than a
14249         location_t.
14250         * doc/optinfo.texi (Dump types): Convert example of
14251         dump_printf_loc from taking "locus" to taking "insn".  Update
14252         description of the "_loc" calls to cover dump_location_t.
14253         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14254         "selftest.h".
14255         (dump_user_location_t::dump_user_location_t): New constructors,
14256         from gimple *stmt and rtx_insn *.
14257         (dump_user_location_t::from_function_decl): New function.
14258         (dump_loc): Make static.
14259         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14260         const dump_location_t &.
14261         (dump_generic_expr_loc): Delete.
14262         (dump_printf_loc): Convert param "loc" from location_t to
14263         const dump_location_t &.
14264         (selftest::test_impl_location): New function.
14265         (selftest::dumpfile_c_tests): New function.
14266         * dumpfile.h: Include "profile-count.h".
14267         (class dump_user_location_t): New class.
14268         (struct dump_impl_location_t): New struct.
14269         (class dump_location_t): New class.
14270         (dump_printf_loc): Convert 2nd param from source_location to
14271         const dump_location_t &.
14272         (dump_generic_expr_loc): Delete.
14273         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14274         const dump_location_t &.
14275         * gimple-fold.c (fold_gimple_assign): Update call to
14276         dump_printf_loc to pass in a dump_location_t rather than a
14277         location_t, via the gimple stmt.
14278         (gimple_fold_call): Likewise.
14279         * gimple-loop-interchange.cc
14280         (loop_cand::analyze_iloop_reduction_var): Update for change to
14281         check_reduction_path.
14282         (tree_loop_interchange::interchange): Update for change to
14283         find_loop_location.
14284         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14285         change in return-type of find_loop_location.
14286         (graphite_regenerate_ast_isl): Likewise.
14287         * graphite-optimize-isl.c (optimize_isl): Likewise.
14288         * graphite.c (graphite_transform_loops): Likewise.
14289         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14290         pass in a dump_location_t rather than a location_t, via the
14291         gimple stmt.
14292         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14293         * ipa.c (walk_polymorphic_call_targets): Likewise.
14294         * loop-unroll.c (report_unroll): Convert "locus" param from
14295         location_t to dump_location_t.
14296         (decide_unrolling): Update for change to get_loop_location's
14297         return type.
14298         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14299         location_t to dump_user_location_t.
14300         (grid_find_single_omp_among_assignments_1): Updates calls to
14301         dump_printf_loc to pass in a dump_location_t rather than a
14302         location_t, via the gimple stmt.
14303         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14304         location_t to dump_location_t.  Updates calls to dump_printf_loc
14305         to pass in a dump_location_t rather than a location_t, via the
14306         gimple stmt.
14307         (grid_inner_loop_gridifiable_p): Likewise.
14308         (grid_dist_follows_simple_pattern): Likewise.
14309         (grid_gfor_follows_tiling_pattern): Likewise.
14310         (grid_target_follows_gridifiable_pattern): Likewise.
14311         (grid_attempt_target_gridification): Convert initialization
14312         of local "grid" from memset to zero-initialization; FIXME: does
14313         this require C++11?  Update call to dump_printf_loc to pass in a
14314         optinfo_location rather than a location_t, via the gimple stmt.
14315         * profile.c (read_profile_edge_counts): Updates call to
14316         dump_printf_loc to pass in a dump_location_t rather than a
14317         location_t
14318         (compute_branch_probabilities): Likewise.
14319         * selftest-run-tests.c (selftest::run_tests): Call
14320         dumpfile_c_tests.
14321         * selftest.h (dumpfile_c_tests): New decl.
14322         * tree-loop-distribution.c (pass_loop_distribution::execute):
14323         Update for change in return type of find_loop_location.
14324         * tree-parloops.c (parallelize_loops): Likewise.
14325         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14326         "locus" from location_t to dump_user_location_t.
14327         (canonicalize_loop_induction_variables): Likewise.
14328         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14329         for change in return type of find_loop_location.
14330         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14331         to dump_printf_loc to pass in a dump_location_t rather than a
14332         location_t, via the stmt.
14333         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14334         Likewise.
14335         * tree-vect-loop-manip.c (find_loop_location): Convert return
14336         type from source_location to dump_user_location_t.
14337         (vect_do_peeling): Update for above change.
14338         (vect_loop_versioning): Update for change in type of
14339         vect_location.
14340         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14341         from location_t to dump_user_location_t.
14342         (vect_estimate_min_profitable_iters): Update for change in type
14343         of vect_location.
14344         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14345         location_t to dump_location_t.
14346         (vect_slp_bb): Update for change in type of vect_location.
14347         * tree-vectorizer.c (vect_location): Convert from source_location
14348         to dump_user_location_t.
14349         (try_vectorize_loop_1): Update for change in vect_location's type.
14350         (vectorize_loops): Likewise.
14351         (increase_alignment): Likewise.
14352         * tree-vectorizer.h (vect_location): Convert from source_location
14353         to dump_user_location_t.
14354         (find_loop_location): Convert return type from source_location to
14355         dump_user_location_t.
14356         (check_reduction_path): Convert 1st param from location_t to
14357         dump_user_location_t.
14358         * value-prof.c (check_counter): Update call to dump_printf_loc to
14359         pass in a dump_user_location_t rather than a location_t; update
14360         call to error_at for change in type of "locus".
14361         (check_ic_target): Update call to dump_printf_loc to
14362         pass in a dump_user_location_t rather than a location_t, via the
14363         call_stmt.
14365 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14367         * config/s390/s390.h (enum processor_flags): Do not use
14368         default tune parameter when -march was specified.
14370 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14372         PR target/86314
14373         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14374         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14375         operands.
14377 2018-06-26  Richard Biener  <rguenther@suse.de>
14379         PR tree-optimization/86287
14380         PR bootstrap/86316
14381         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14382         (vect_analyze_loop): Initialize n_stmts.
14384 2018-06-26  Richard Biener  <rguenther@suse.de>
14386         PR middle-end/86271
14387         * fold-const.c (fold_convertible_p): Pointer extension
14388         isn't valid.
14390 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14392         PR debug/86064
14393         * dwarf2out.c (loc_list_has_views): Adjust comments.
14394         (dw_loc_list): Split single cross-partition range with
14395         nonzero locview.
14397 2018-06-25  Jeff Law  <law@redhat.com>
14399         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14400         on -mbig-switch by default.
14402         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14403         test.
14404         (const_float_0_operand): Remove unused predicate.
14405         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14406         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14407         (recipsf2): New expander.  Original pattern now called
14408         (recipsf2_insn).
14409         (recipdf2, recipdf2_insn): Similarly.
14410         (rsqrtsf2, rsqrtsf2_insn): Similarly
14411         (rsqrtdf2, rsqrtdf2_insn): Similarly
14413 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14415         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14416         Simplify logic for FreeBSD (twice).
14418 2018-06-25  Martin Sebor  <msebor@redhat.com>
14420         PR tree-optimization/86204
14421         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14422         a strnlen result if it's less than the length of the string.
14424 2018-06-25  Martin Sebor  <msebor@redhat.com>
14426         PR tree-optimization/85700
14427         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14428         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14429         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14431 2018-06-25  Martin Sebor  <msebor@redhat.com>
14433         * doc/extend.texi (Zero-length arrays): Update and clarify.
14435 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14437         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14438         added IEEE/IBM long double multilib support on PowerPC little
14439         endian Linux systems.
14440         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14441         (MULTILIB_DEFAULTS): Likewise.
14442         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14443         Likewise.
14444         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14445         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14446         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14448 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14450         PR middle-end/86311
14451         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14452         (REORDER_45): Likewise.
14454 2018-06-25  Jeff Law  <law@redhat.com>
14456         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14457         dividend to 32 bits.  Adjust length.
14458         (udivmodhi4): Cleanup output template.  Fix length.
14460 2018-06-25  Carl Love  <cel@us.ibm.com>
14462         * config/rs6000/vsx.md: Change word selector to prefered location.
14464 2018-06-25  Richard Biener  <rguenther@suse.de>
14466         PR tree-optimization/86304
14467         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14468         epilogue-if-converted loops as well.
14470 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14472         * lto-section-out.c (lto_begin_section): Do not print section
14473         name for noaddr and unnumbered dumps.
14475 2018-06-25  Richard Biener  <rguenther@suse.de>
14477         * tree-vectorizer.h (struct vec_info_shared): New structure
14478         with parts split out from struct vec_info and loop_nest from
14479         struct _loop_vec_info.
14480         (struct vec_info): Adjust accordingly.
14481         (struct _loop_vec_info): Likewise.
14482         (LOOP_VINFO_LOOP_NEST): Adjust.
14483         (LOOP_VINFO_DATAREFS): Likewise.
14484         (LOOP_VINFO_DDRS): Likewise.
14485         (struct _bb_vec_info): Likewise.
14486         (BB_VINFO_DATAREFS): Likewise.
14487         (BB_VINFO_DDRS): Likewise.
14488         (struct _stmt_vec_info): Add dr_aux member.
14489         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14490         (DR_MISALIGNMENT_UNINITIALIZED): New.
14491         (set_dr_misalignment): Adjust.
14492         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14493         (vect_analyze_loop): Adjust prototype.
14494         (vect_analyze_loop_form): Likewise.
14495         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14496         Compute dependences lazily.
14497         (vect_record_base_alignments): Use shared datarefs/ddrs.
14498         (vect_verify_datarefs_alignment): Likewise.
14499         (vect_analyze_data_refs_alignment): Likewise.
14500         (vect_analyze_data_ref_accesses): Likewise.
14501         (vect_analyze_data_refs): Likewise.
14502         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14503         constructor parameter for shared part.
14504         (vect_analyze_loop_form): Pass in shared part and adjust.
14505         (vect_analyze_loop_2): Pass in storage for the number of
14506         stmts.  Move loop nest finding to the caller.  Compute
14507         datarefs lazily.
14508         (vect_analyze_loop): Pass in shared part.
14509         (vect_transform_loop): Verify shared datarefs are unchanged.
14510         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14511         constructor parameter for shared part.
14512         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14513         (vect_slp_bb): Verify shared datarefs are unchanged before
14514         transform.
14515         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14516         change.
14517         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14518         DR_MISALIGNMENT_UNINITIALIZED.
14519         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14520         parameter for shared part.
14521         (vec_info::~vec_info): Adjust.
14522         (vec_info_shared::vec_info_shared): New.
14523         (vec_info_shared::~vec_info_shared): Likewise.
14524         (vec_info_shared::save_datarefs): Likewise.
14525         (vec_info_shared::check_datarefs): Likewise.
14526         (try_vectorize_loop_1): Construct shared part live for analyses
14527         of a single loop for multiple vector sizes.
14528         * tree-parloops.c (gather_scalar_reductions): Adjust.
14530 2018-06-25  Richard Biener  <rguenther@suse.de>
14532         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14533         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14534         (vect_analyze_data_refs): Remove similar code from here and
14535         simplify accordingly.
14537 2018-06-25  Richard Biener  <rguenther@suse.de>
14539         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14540         for reverse storage order accesses rather than asserting
14541         they cannot happen here.
14543 2018-06-25  Tom de Vries  <tdevries@suse.de>
14545         PR debug/86257
14546         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14547         Use data16 instead of .byte for insn prefix.
14549 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14551         PR C++/86082
14552         * parser.c (make_char_string_pack): Pass this literal chars
14553         through cpp_interpret_string.
14554         (cp_parser_userdef_numeric_literal): Check the result of
14555         make_char_string_pack.
14557 2018-06-24  Maya Rashish  <coypu@sdf.org>
14559         * ginclude/stddef.h: Simplify conditions around avoiding
14560         re-definition of __size_t.
14562 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14564         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14565         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14567 2018-06-22  Maya Rashish  <coypu@sdf.org>
14569         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14570         made to OpenBSD and NetBSD through the years.
14572 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14574         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14575         behavior of vec_pack (vector double, vector double) to match
14576         behavior of vec_float2 (vector double, vector double).
14578 2018-06-22  Olivier Hainque  <hainque@adacore.com>
14580         * gimplify.c (gimplify_function_tree): Prevent creation
14581         of a trampoline for the address of the current function
14582         passed to entry/exit instrumentation hooks.
14584 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
14586         PR target/86222
14587         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
14588         correctly.
14590 2018-06-22  Martin Liska  <mliska@suse.cz>
14592         PR tree-optimization/86263
14593         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14594         Bail out if is_enabled is false.
14595         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
14596         New declaration.
14597         (jump_table_cluster::is_enabled): New function.
14599 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14601         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14602         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
14603         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
14604         (lto_input_ts_binfo_tree_pointers): Likewise.
14605         * tree-streamer-out.c (streamer_write_tree_bitfields,
14606         write_ts_binfo_tree_pointers): Likewise.
14607         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
14609 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14611         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
14613 2018-06-22  Martin Liska  <mliska@suse.cz>
14615         * symbol-summary.h (get): Make it pure and inline move
14616         functionality from ::get function.
14617         (get): Remove and inline into ::get and ::get_create.
14618         (get_create): Move code from ::get function.
14620 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14622         PR target/85994
14623         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
14624         -x assembler-with-cpp.
14626 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14628         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
14629         _FILE_OFFSET_BITS=64 for C++.
14631 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
14633         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
14634         conversion insn that shows up when pr85657-3.c is compiled using
14635         IEEE 128-bit long double.
14636         (neg<mode>2_internal): Use the correct mode to check whether the
14637         mode is IBM extended.
14638         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
14639         multiply and divide external functions from being created more
14640         than once.
14642 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14644         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
14645         functions.
14646         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
14647         the edge can be forwarded.
14648         (cfg_layout_merge_blocks): Likewise.
14650 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14652         * except.c (finish_eh_generation): Commit edge insertions only after
14653         the EH edges have been redirected from post-landing to landing pads.
14655 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14657         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
14658         create_tmp_var_for to create the FRAME decl.
14659         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
14661 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14663         * tree-inline.c (copy_edges_for_bb): Minor tweak.
14664         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
14665         debug statement when resetting its value.
14666         (expand_call_inline): Copy the locus of the call onto the assignment
14667         of the return value, if any.  Use local variable in more cases.
14669 2018-06-21  Martin Liska  <mliska@suse.cz>
14671         * ipa-pure-const.c (propagate_nothrow): Use
14672         funct_state_summaries->get.
14673         (dump_malloc_lattice): Likewise.
14674         (propagate_malloc): Likewise.
14676 2018-06-21  Richard Biener  <rguenther@suse.de>
14678         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
14679         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
14680         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
14681         comment.
14682         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
14683         BLOCK_ABSTRACT_ORIGIN unconditionally.
14685 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
14687         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
14688         deleting it.
14689         * ipa-reference.c (ipa_reference_c_finalize): Delete
14690         ipa_ref_opt_sum_summaries and set it to NULL.
14692 2018-06-21  Tom de Vries  <tdevries@suse.de>
14694         PR tree-optimization/85859
14695         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
14696         test with comment from bb_no_side_effects_p.
14698 2018-06-21  Richard Biener  <rguenther@suse.de>
14700         PR tree-optimization/86232
14701         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
14702         max for constant niter.
14704 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14706         * config/aarch64/aarch64-simd.md
14707         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
14709 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14711         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
14712         Make opernads of the unspec commutative.
14714 2018-06-21  Richard Biener  <rguenther@suse.de>
14716         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
14717         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14718         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
14719         (vect_analyze_data_ref_dependence): Re-order checks to deal with
14720         NULL DR_STEP.
14721         (vect_record_base_alignments): Do not record base alignment
14722         for gathers or scatters.
14723         (vect_compute_data_ref_alignment): Drop return value that is always
14724         true.  Bail out early for gathers or scatters.
14725         (vect_enhance_data_refs_alignment): Bail out early for gathers
14726         or scatters.
14727         (vect_find_same_alignment_drs): Likewise.
14728         (vect_analyze_data_refs_alignment): Remove dead code.
14729         (vect_slp_analyze_and_verify_node_alignment): Likewise.
14730         (vect_analyze_data_refs): For possible gathers or scatters do
14731         not create an alternate DR, just check their possible validity
14732         and mark them.  Adjust DECL_NONALIASED handling to not rely
14733         on DR_BASE_ADDRESS.
14734         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
14735         update inits of gathers or scatters.
14736         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
14737         Also copy gather/scatter flag to pattern vinfo.
14739 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14741         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14742         behavior of vec_packsu (vector unsigned long long, vector unsigned
14743         long long) to match behavior of vec_packs with same signature.
14745 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
14746             Thomas Schwinge <thomas@codesourcery.com>
14747             Cesar Philippidis  <cesar@codesourcery.com>
14749         * gimplify.c (gimplify_scan_omp_clauses): Add support for
14750         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14751         (gimplify_adjust_omp_clauses): Likewise.
14752         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
14753         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
14754         (gimplify_omp_target_update): Update handling of acc update and
14755         enter/exit data.
14756         * omp-low.c (install_var_field): Remove unused parameter
14757         base_pointers_restrict.
14758         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
14759         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
14760         FINALIZE}
14761         (omp_target_base_pointers_restrict_p): Delete.
14762         (scan_omp_target): Update call to scan_sharing_clauses.
14763         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
14764         FINALIZE}.
14765         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
14766         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14767         (convert_local_omp_clauses): Likewise.
14768         * tree-pretty-print.c (dump_omp_clause): Likewise.
14769         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
14770         FINALIZE}.
14771         (omp_clause_code_name): Likewise.
14773 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14775         PR debug/86194
14776         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
14777         be narrowed.
14779         PR tree-optimization/86231
14780         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
14781         anti-range don't overwrite *vr0min before using it to compute *vr0max.
14783 2018-06-20  Tom de Vries  <tdevries@suse.de>
14785         PR tree-optimization/86097
14786         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
14787         iv type if signedness of iv type is not the same as that of *nit.
14789 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14791         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
14792         EDGE_EH edges, verify they are all EDGE_EH.
14794 2018-06-20  Maya Rashish  <coypu@sdf.org>
14796         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
14798 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14800         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
14801         * config/aarch64/aarch64.c (xgene1_tunings): Add
14802         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
14803         (aarch64_mode_valid_for_sched_fusion_p):
14804         Allow 16-byte modes.
14805         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
14806         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
14807         128-bit modes.
14808         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
14809         New pattern.
14810         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
14811         * config/aarch64/iterators.md (VQ2): New mode iterator.
14813 2018-06-20  Martin Liska  <mliska@suse.cz>
14815         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14816         Change default ratio from 10 to 8.
14818 2018-06-20  Martin Liska  <mliska@suse.cz>
14820         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14821         New.
14822         (bit_test_cluster::find_bit_tests): Likewise.
14823         (switch_decision_tree::analyze_switch_statement): Find clusters.
14824         * tree-switch-conversion.h (struct jump_table_cluster): Document
14825         hierarchy.
14827 2018-06-20  Martin Liska  <mliska@suse.cz>
14829         * tree-switch-conversion.c (switch_conversion::collect):
14830         Record m_uniq property.
14831         (switch_conversion::expand): Bail out for special conditions.
14832         (group_cluster::~group_cluster): New.
14833         (group_cluster::group_cluster): Likewise.
14834         (group_cluster::dump): Likewise.
14835         (jump_table_cluster::emit): New.
14836         (switch_decision_tree::fix_phi_operands_for_edges): New.
14837         (struct case_node): Remove struct.
14838         (jump_table_cluster::can_be_handled): New.
14839         (case_values_threshold): Moved to header.
14840         (reset_out_edges_aux): Likewise.
14841         (jump_table_cluster::is_beneficial): New.
14842         (bit_test_cluster::can_be_handled): Likewise.
14843         (add_case_node): Remove.
14844         (bit_test_cluster::is_beneficial): New.
14845         (case_bit_test::cmp): New.
14846         (bit_test_cluster::emit): New.
14847         (expand_switch_as_decision_tree_p): Remove.
14848         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
14849         (fix_phi_operands_for_edge): Likewise.
14850         (switch_decision_tree::analyze_switch_statement): New.
14851         (compute_cases_per_edge): Move ...
14852         (switch_decision_tree::compute_cases_per_edge): ... here.
14853         (try_switch_expansion): Likewise.
14854         (switch_decision_tree::try_switch_expansion): Likewise.
14855         (record_phi_operand_mapping): Likewise.
14856         (switch_decision_tree::record_phi_operand_mapping): Likewise.
14857         (emit_case_decision_tree): Likewise.
14858         (switch_decision_tree::emit): Likewise.
14859         (balance_case_nodes): Likewise.
14860         (switch_decision_tree::balance_case_nodes): Likewise.
14861         (dump_case_nodes): Likewise.
14862         (switch_decision_tree::dump_case_nodes): Likewise.
14863         (emit_jump): Likewise.
14864         (switch_decision_tree::emit_jump): Likewise.
14865         (emit_cmp_and_jump_insns): Likewise.
14866         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
14867         (emit_case_nodes): Likewise.
14868         (switch_decision_tree::emit_case_nodes): Likewise.
14869         (conditional_probability): Remove.
14870         * tree-switch-conversion.h (enum cluster_type): New.
14871         (PRINT_CASE): New.
14872         (struct cluster): Likewise.
14873         (cluster::cluster): Likewise.
14874         (struct simple_cluster): Likewise.
14875         (simple_cluster::simple_cluster): Likewise.
14876         (struct group_cluster): Likewise.
14877         (struct jump_table_cluster): Likewise.
14878         (struct bit_test_cluster): Likewise.
14879         (struct min_cluster_item): Likewise.
14880         (struct case_tree_node): Likewise.
14881         (case_tree_node::case_tree_node): Likewise.
14882         (jump_table_cluster::case_values_threshold): Likewise.
14883         (struct case_bit_test): Likewise.
14884         (struct switch_decision_tree): Likewise.
14885         (struct switch_conversion): Likewise.
14886         (switch_decision_tree::reset_out_edges_aux): Likewise.
14888 2018-06-20  Martin Liska  <mliska@suse.cz>
14890         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
14891         (hoist_edge_and_branch_if_true): Likewise.
14892         (expand_switch_using_bit_tests_p): Likewise.
14893         (struct case_bit_test): Likewise.
14894         (case_bit_test_cmp): Likewise.
14895         (emit_case_bit_tests): Likewise.
14896         (switch_conversion::switch_conversion): New class.
14897         (struct switch_conv_info): Remove old struct.
14898         (collect_switch_conv_info): More to ...
14899         (switch_conversion::collect): ... this.
14900         (check_range): Likewise.
14901         (switch_conversion::check_range): Likewise.
14902         (check_all_empty_except_final): Likewise.
14903         (switch_conversion::check_all_empty_except_final): Likewise.
14904         (check_final_bb): Likewise.
14905         (switch_conversion::check_final_bb): Likewise.
14906         (create_temp_arrays): Likewise.
14907         (switch_conversion::create_temp_arrays): Likewise.
14908         (free_temp_arrays): Likewise.
14909         (gather_default_values): Likewise.
14910         (switch_conversion::gather_default_values): Likewise.
14911         (build_constructors): Likewise.
14912         (switch_conversion::build_constructors): Likewise.
14913         (constructor_contains_same_values_p): Likewise.
14914         (switch_conversion::contains_same_values_p): Likewise.
14915         (array_value_type): Likewise.
14916         (switch_conversion::array_value_type): Likewise.
14917         (build_one_array): Likewise.
14918         (switch_conversion::build_one_array): Likewise.
14919         (build_arrays): Likewise.
14920         (switch_conversion::build_arrays): Likewise.
14921         (gen_def_assigns): Likewise.
14922         (switch_conversion::gen_def_assigns): Likewise.
14923         (prune_bbs): Likewise.
14924         (switch_conversion::prune_bbs): Likewise.
14925         (fix_phi_nodes): Likewise.
14926         (switch_conversion::fix_phi_nodes): Likewise.
14927         (gen_inbound_check): Likewise.
14928         (switch_conversion::gen_inbound_check): Likewise.
14929         (process_switch): Use the newly created class.
14930         (switch_conversion::expand): New.
14931         (switch_conversion::~switch_conversion): New.
14932         * tree-switch-conversion.h: New file.
14934 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14936         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
14937         tree-vect-patterns.c.
14938         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
14939         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
14940         (vect_recog_sad_pattern): Likewise.
14941         (vect_recog_widen_sum_pattern): Likewise.
14942         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
14943         (vect_recog_widen_shift_pattern): Remove the type_in argument.
14944         (vect_recog_rotate_pattern): Likewise.
14945         (vect_recog_mult_pattern): Likewise.
14946         (vect_recog_vector_vector_shift_pattern): Likewise.
14947         (vect_recog_divmod_pattern): Likewise.
14948         (vect_recog_mixed_size_cond_pattern): Likewise.
14949         (vect_recog_bool_pattern): Likewise.
14950         (vect_recog_mask_conversion_pattern): Likewise.
14951         (vect_try_gather_scatter_pattern): Likewise.
14952         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
14953         (vect_recog_over_widening_pattern): Likewise.
14954         (vect_recog_gather_scatter_pattern): Likewise.
14955         (vect_recog_func_ptr): Move from tree-vectorizer.h
14956         (vect_vect_recog_func_ptrs): Move further down the file.
14957         (vect_recog_func): Likewise.  Remove the third argument.
14958         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
14959         (vect_pattern_recog_1): Expect the pattern function to do any
14960         necessary target tests.  Also expect it to provide a vector type.
14961         Remove the type_in handling.
14963 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14965         * tree-vect-patterns.c (vect_pattern_detected): New function.
14966         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
14967         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
14968         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
14969         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
14970         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14971         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14972         (vect_recog_mask_conversion_pattern)
14973         (vect_try_gather_scatter_pattern): Likewise.
14975 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14977         * tree-vect-patterns.c (vect_get_internal_def): New function.
14978         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14979         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
14980         (search_type_for_mask_1): Use it.
14982 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14984         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14985         redundant WIDEN_SUM_EXPR handling.
14986         (vect_recog_sad_pattern): Likewise.
14988 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14990         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14991         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
14992         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
14993         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
14994         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
14996 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14998         * tree-vect-stmts.c (vectorizable_call): Make sure that we
14999         use the stmt_vec_info of the original bb statement for the
15000         new zero assignment, even if the call is part of a pattern.
15002 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15004         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15005         that the sequence is attached to the original statement rather
15006         than the pattern statement.
15007         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15008         PATTERN_DEF_SEQ from the original statement rather than
15009         the main pattern statement.
15010         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15011         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15012         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15014 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15016         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15017         definition statements before the early exit for statements that aren't
15018         live or relevant.
15019         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15020         split out from...
15021         (vect_transform_loop): ...here.  Process pattern definition
15022         statements without first checking whether the main pattern
15023         statement is live or relevant.
15025 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15027         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15028         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15030 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
15032         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15033         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15034         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15035         (expand_block_compare): Change select_block_compare_mode call.
15036         (expand_strncmp_align_check): Use new functions, fix comment.
15037         (emit_final_str_compare_gpr): New function.
15038         (expand_strn_compare): Refactor and clean up code.
15039         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15041 2018-06-19  Tony Reix  <tony.reix@atos.com>
15042             Damien Bergamini  <damien.bergamini@atos.com>
15043             David Edelsohn  <dje.gcc@gmail.com>
15045         * collect2.c (static_obj): New variable.
15046         (static_libs): New variable.
15047         (is_in_list): Uncomment declaration.
15048         (main): Track AIX libraries linked statically.
15049         (is_in_list): Uncomment definition.
15050         (scan_prog_file): Don't add AIX shared libraries initializer
15051         to constructor list if linking statically.
15053 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
15055         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15056         constant.
15057         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15059 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
15061         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15062         blocks.
15064 2018-06-19  Martin Liska  <mliska@suse.cz>
15066         * config/i386/i386.c (ix86_can_inline_p): Do not use
15067         ipa_fn_summaries::get_create.
15068         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15069         get.
15070         (devirtualization_time_bonus): Likewise.
15071         (ipcp_propagate_stage): Likewise.
15072         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15073         (edge_set_predicate): Likewise.
15074         (evaluate_conditions_for_known_args): Likewise.
15075         (evaluate_properties_for_edge): Likewise.
15076         (ipa_call_summary::reset): Tranform to ...
15077         (ipa_call_summary::~ipa_call_summary): ... this.
15078         (ipa_fn_summary::reset): Transform to ...
15079         (ipa_fn_summary::~ipa_fn_summary): ... this.
15080         (ipa_fn_summary_t::remove): Rename to ...
15081         (ipa_fn_summary_t::remove_callees): ... this.
15082         (ipa_fn_summary_t::duplicate): Use placement new
15083         instead of memory copy.
15084         (ipa_call_summary_t::duplicate): Likewise.
15085         (ipa_call_summary_t::remove): Remove.
15086         (dump_ipa_call_summary): Change get_create to get.
15087         (ipa_dump_fn_summary): Dump only when summary exists.
15088         (analyze_function_body): Use symbol_summary::get instead
15089         of get_create.
15090         (compute_fn_summary): Likewise.
15091         (estimate_edge_devirt_benefit): Likewise.
15092         (estimate_edge_size_and_time): Likewise.
15093         (inline_update_callee_summaries): Likewise.
15094         (remap_edge_change_prob): Likewise.
15095         (remap_edge_summaries): Likewise.
15096         (ipa_merge_fn_summary_after_inlining): Likewise.
15097         (write_ipa_call_summary): Likewise.
15098         (ipa_fn_summary_write): Likewise.
15099         (ipa_free_fn_summary): Likewise.
15100         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15101         (struct ipa_call_summary): Likewise.
15102         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15103         of get_create.
15104         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15105         (estimate_size_after_inlining): Likewise.
15106         (estimate_growth): Likewise.
15107         (growth_likely_positive): Likewise.
15108         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15109         (inline_call): Likewise.
15110         * ipa-inline.c (caller_growth_limits): Likewise.
15111         (can_inline_edge_p): Likewise.
15112         (can_inline_edge_by_limits_p): Likewise.
15113         (compute_uninlined_call_time): Likewise.
15114         (compute_inlined_call_time): Likewise.
15115         (want_inline_small_function_p): Likewise.
15116         (edge_badness): Likewise.
15117         (update_caller_keys): Likewise.
15118         (update_callee_keys): Likewise.
15119         (inline_small_functions): Likewise.
15120         (inline_to_all_callers_1): Likewise.
15121         (dump_overall_stats): Likewise.
15122         (early_inline_small_functions): Likewise.
15123         (early_inliner): Likewise.
15124         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15125         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15126         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15127         * ipa-split.c (execute_split_functions): Likewise.
15128         * symbol-summary.h: Likewise.
15129         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15131 2018-06-19  Richard Biener  <rguenther@suse.de>
15133         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15134         (vectorize_loops): ... here.  Fix dbgcnt handling.
15135         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15137 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15139         PR target/86197
15140         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15141         ieee128 argument takes up only one (vector) register, not two (floating
15142         point) registers.
15144 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15146         * gimplify.c (gimplify_init_constructor): Really never clear for an
15147         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15149 2018-06-19  Richard Biener  <rguenther@suse.de>
15151         PR tree-optimization/86179
15152         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15153         after failed recognition.
15155 2018-06-18  Martin Sebor  <msebor@redhat.com>
15157         PR middle-end/85602
15158         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15159         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15160         Handle integer subtraction.
15161         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15162         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15164 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15166         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15167         param from rtx to rtx_insn *.
15168         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15169         param.
15170         (frv_ifcvt_modify_insn): Likwise.
15171         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15172         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15173         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15174         as_a <rtx_insn *> cast to local "unprotected_region" once
15175         it's been established that it's not NULL or pc_rtx.
15176         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15177         param "sethi" from rtx to rtx_insn *.
15178         (nds32_group_float_insns): Likewise for param "insn".
15179         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15180         param.
15181         (vax_output_int_subtract): Likewise.
15182         * config/vax/vax.c (vax_output_int_add): Likewise for param
15183         "insn".
15184         (vax_output_int_subtract): Likewise.
15185         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15186         (emit_pattern_after): Likewise for param "after".
15187         (emit_insn_after): Likewise.
15188         (emit_jump_insn_after): Likewise.
15189         (emit_call_insn_after): Likewise.
15190         (emit_debug_insn_after): Likewise.
15191         (emit_pattern_before): Likewise for param "before".
15192         (emit_insn_before): Likewise.
15193         (emit_jump_insn_before): Likewise.
15194         * final.c (get_insn_template): Likewise for param "insn", removing
15195         a cast.
15196         * output.h (get_insn_template): Likewise for 2nd param.
15197         * rtl.h (emit_insn_before): Likewise.
15198         (emit_jump_insn_before): Likewise.
15199         (emit_debug_insn_before_noloc): Likewise.
15200         (emit_insn_after): Likewise.
15201         (emit_jump_insn_after): Likewise.
15202         (emit_call_insn_after): Likewise.
15203         (emit_debug_insn_after): Likewise.
15204         (set_insn_deleted): Likewise for param.
15206 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15208         PR target/85358
15209         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15210         floating point modes, so that IFmode is numerically greater than
15211         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15212         to declare the ordering.  This prevents IFmode from being
15213         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15214         machine.  Include rs6000-modes.h to share the fractional values
15215         between genmodes* and the rest of the compiler.
15216         (IFmode): Likewise.
15217         (KFmode): Likewise.
15218         (TFmode): Likewise.
15219         * config/rs6000/rs6000-modes.h: New file.
15220         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15221         meaning of rs6000_long_double_size so that 126..128 selects an
15222         appropriate 128-bit floating point type.
15223         (rs6000_option_override_internal): Likewise.
15224         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15225         (TARGET_LONG_DOUBLE_128): Change the meaning of
15226         rs6000_long_double_size so that 126..128 selects an appropriate
15227         128-bit floating point type.
15228         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15229         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15230         source and destination to match the standard usage.
15231         (truncifkf2): Likewise.
15232         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15233         ISA 2.07 to use an explicit clobber, instead of passing in a
15234         temporary.
15235         (copysign<mode>3_soft): Likewise.
15237 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15239         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15240         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15241         (vect_slp_analyze_instance_dependence): Likewise.
15242         (vect_enhance_data_refs_alignment): Likewise.
15243         (vect_analyze_data_refs_alignment): Likewise.
15244         (vect_slp_analyze_and_verify_instance_alignment
15245         (vect_analyze_data_ref_accesses): Likewise.
15246         (vect_prune_runtime_alias_test_list): Likewise.
15247         (vect_analyze_data_refs): Likewise.
15248         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15249         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15250         (vect_analyze_scalar_cycles_1): Likewise.
15251         (vect_get_loop_niters): Likewise.
15252         (vect_analyze_loop_form_1): Likewise.
15253         (vect_update_vf_for_slp): Likewise.
15254         (vect_analyze_loop_operations): Likewise.
15255         (vect_analyze_loop): Likewise.
15256         (vectorizable_induction): Likewise.
15257         (vect_transform_loop): Likewise.
15258         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15259         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15260         (vect_make_slp_decision): Likewise.
15261         (vect_detect_hybrid_slp): Likewise.
15262         (vect_slp_analyze_operations): Likewise.
15263         (vect_slp_bb): Likewise.
15264         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15265         (vectorizable_bswap): Likewise.
15266         (vectorizable_call): Likewise.
15267         (vectorizable_simd_clone_call): Likewise.
15268         (vectorizable_conversion): Likewise.
15269         (vectorizable_assignment): Likewise.
15270         (vectorizable_shift): Likewise.
15271         (vectorizable_operation): Likewise.
15272         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15274 2018-06-18  Martin Sebor  <msebor@redhat.com>
15276         PR tree-optimization/81384
15277         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15278         * builtins.c (expand_builtin_strnlen): New function.
15279         (expand_builtin): Call it.
15280         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15281         * builtins.def (BUILT_IN_STRNLEN): New.
15282         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15283         Warn for bounds in excess of maximum object size.
15284         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15285         single-value ranges.  Handle strnlen.
15286         (handle_builtin_strlen): Handle strnlen.
15287         (strlen_check_and_optimize_stmt): Same.
15288         * doc/extend.texi (Other Builtins): Document strnlen.
15290 2018-06-18  Maya Rashish  <coypu@sdf.org>
15292         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15293         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15294         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15296         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15297         here to ...
15298         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15300 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15302         * tree.c (escaped_string::escape): Replace cast to char * by
15303         const_cast<char *> (unescaped).
15305 2018-06-18  Nick Clifton  <nickc@redhat.com>
15307         PR 84195
15308         * tree.c (escaped_string): New class.  Converts an unescaped
15309         string into its escaped equivalent.
15310         (warn_deprecated_use): Use the new class to convert the
15311         deprecation message, if present.
15312         (test_escaped_strings): New self test.
15313         (test_c_tests): Add test_escaped_strings.
15314         * doc/extend.texi (deprecated): Add a note that the
15315         deprecation message is affected by the -fmessage-length
15316         option, and that control characters will be escaped.
15317         (#pragma GCC error): Document this pragma.
15318         (#pragma GCC warning): Likewise.
15319         * doc/invoke.texi (-fmessage-length): Document this option's
15320         effect on the #warning and #error preprocessor directives and
15321         the deprecated attribute.
15323 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15325         * tree.c (decl_value_expr_lookup): Revert latest change.
15326         (decl_value_expr_insert): Likewise.
15328 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15330         * gimplify.c (nonlocal_vlas): Delete.
15331         (nonlocal_vla_vars): Likewise.
15332         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15333         referenced VLAs.
15334         (gimplify_body): Do not create and destroy nonlocal_vlas.
15335         * tree-nested.c: Include diagnostic.h.
15336         (use_pointer_in_frame): Tweak.
15337         (lookup_field_for_decl): Add assertion and declare the transformation.
15338         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15339         internal error when the reference is in a wrong context.  Do not
15340         create a debug decl by default.
15341         (note_nonlocal_block_vlas): Delete.
15342         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15343         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15344         create a debug decl by default.
15345         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15346         call is in a wrong context.
15347         (fixup_vla_decls): New function.
15348         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15349         debug variables were created.
15350         * tree.c (decl_value_expr_lookup): Add checking assertion.
15351         (decl_value_expr_insert): Likewise.
15353 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15355         PR middle-end/82479
15356         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15357         * tree-scalar-evolution.c (interpret_expr): Likewise.
15358         (expression_expensive_p): Likewise.
15359         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15360         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15361         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15362         (ssa_defined_by_minus_one_stmt_p): New.
15364 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15366         PR middle-end/64946
15367         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15368         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15369         * dojump.c (do_jump): Likewise.
15370         * expr.c (expand_expr_real_2): Check operand type's sign.
15371         * fold-const.c (const_unop): Handle ABSU_EXPR.
15372         (fold_abs_const): Likewise.
15373         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15374         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15375         (strip_sign_op_1): Likesise.
15376         * match.pd: Add new pattern to generate ABSU_EXPR.
15377         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15378         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15379         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15380         * tree-inline.c (estimate_operator_cost): Likewise.
15381         * tree-pretty-print.c (dump_generic_node): Likewise.
15382         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15383         * tree.def (ABSU_EXPR): New.
15385 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15387         PR middle-end/86095
15388         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15389         documented as preserved for backward compatibility only.
15390         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15392         PR rtl-optimization/86108
15393         * bb-reorder.c (create_forwarder_block): Renamed to ...
15394         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15395         jump from new landing pad to the second part.
15396         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15397         Adjust callers.
15399 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15401         PR middle-end/85878
15402         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15403         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15404         Only call store_expr for halves if the mode is the same.
15406         PR middle-end/86123
15407         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15408         Fix up comment formatting.
15410 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15412         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15413         (typed_splay_tree::closure,
15414         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15415         (typed_splay_tree::typed_splay_tree,
15416         typed_splay_tree::operator =): Declared private.
15417         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15418         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15419         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15420         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15421         typed_splay_tree::splay_tree_splay,
15422         typed_splay_tree::splay_tree_foreach_helper,
15423         typed_splay_tree::splay_tree_insert,
15424         typed_splay_tree::splay_tree_remove,
15425         typed_splay_tree::splay_tree_lookup,
15426         typed_splay_tree::splay_tree_predecessor,
15427         typed_splay_tree::splay_tree_successor,
15428         typed_splay_tree::splay_tree_min,
15429         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15430         (typed_splay_tree::root, typed_splay_tree::comp,
15431         typed_splay_tree::delete_key,
15432         typed_splay_tree::delete_value): New data members.
15433         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15434         typed_splay_tree::remove.
15436 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15438         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15439         -mginv and -mno-ginv to the assembler.
15440         * config/mips/mips.opt (-mcrc): New option.
15441         (-mginv): Likewise.
15442         * doc/invoke.text (-mcrc): Document.
15443         (-mginv): Likewise.
15445 2018-06-15  Nick Clifton  <nickc@redhat.com>
15447         PR 84195
15448         * tree.c (escaped_string): New class.  Converts an unescaped
15449         string into its escaped equivalent.
15450         (warn_deprecated_use): Use the new class to convert the
15451         deprecation message, if present.
15452         (test_escaped_strings): New self test.
15453         (test_c_tests): Add test_escaped_strings.
15454         * doc/extend.texi (deprecated): Add a note that the
15455         deprecation message is affected by the -fmessage-length
15456         option, and that control characters will be escaped.
15457         (#pragma GCC error): Document this pragma.
15458         (#pragma GCC warning): Likewise.
15459         * doc/invoke.texi (-fmessage-length): Document this option's
15460         effect on the #warning and #error preprocessor directives and
15461         the deprecated attribute.
15463 2018-06-15  Richard Biener  <rguenther@suse.de>
15465         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15466         here, also noting vector size used.
15467         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15468         size used in MSG_OPTIMIZED_LOCATIONS dump.
15469         (pass_slp_vectorize::execute): Adjust.
15471 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15473         PR target/85968
15474         * config/arc/arc.c (arc_return_address_register): Fix
15475         if-condition.
15477 2018-06-15  Richard Biener  <rguenther@suse.de>
15479         PR middle-end/86159
15480         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15481         leave useless conversion stripping to force_gimple_operand_gsi.
15482         (gimplify_build2): Likewise.
15483         (gimplify_build1): Likewise.
15485 2018-06-15  Richard Biener  <rguenther@suse.de>
15487         PR middle-end/86076
15488         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15489         before adjusting their block.
15491 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15493         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15494         multilibs for *-*-rtems*.
15495         * config/riscv/t-rtems: New file.
15497 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15499         PR middle-end/86122
15500         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15501         unsigned_type_for returns NULL.
15503         PR target/85945
15504         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15505         subregs of multi-word pseudos unless the float mode has word size.
15507 2018-06-14  Richard Biener  <rguenther@suse.de>
15509         PR middle-end/86139
15510         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15511         duplicate and harmful type_hash_canon.
15512         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15514 2018-06-14  Richard Biener  <rguenther@suse.de>
15516         PR ipa/86124
15517         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15518         NULL cgraph_node.
15520 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15522         * config/rtems.h (STDINT_LONG32): Define.
15524 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15525             Prachi Godbole  <prachi.godbole@imgtec.com>
15527         * config/mips/mips-cpus.def: Define P6600.
15528         * config/mips/mips-tables.opt: Regenerate.
15529         * config/mips/mips.c (mips_ucbranch_type): New enum.
15530         (mips_rtx_cost_data): Add support for P6600.
15531         (mips_issue_rate): Likewise.
15532         (mips_multipass_dfa_lookahead): Likewise.
15533         (mips_avoid_hazard): Likewise.
15534         (mips_reorg_process_insns): Likewise.
15535         (mips_classify_branch_p6600): New function.
15536         * config/mips/mips.h (TUNE_P6600): New define.
15537         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15538         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15539         * config/mips/mips.md: Include p6600.md.
15540         (processor): Add p6600.
15541         * config/mips/p6600.md: New file.
15542         * doc/invoke.texi: Add p6600 to supported architectures.
15544 2018-06-13  Martin Sebor  <msebor@redhat.com>
15546         PR tree-optimization/86114
15547         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15548         of integer types.
15549         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15551 2018-06-13  Richard Biener  <rguenther@suse.de>
15553         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15554         Properly set vector type of the intermediate stmt.
15555         * tree-vect-stmts.c (vectorizable_operation): The destination
15556         var always has vectype_out type.
15558 2018-06-13  Jeff Law  <law@redhat.com>
15560         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15561         integer 0 for argument to print_rtl_with_bb.
15562         (rl78_reorg): Likewise.
15564 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15566         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15567         from rtx to rtx_insn *.
15568         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15569         "label".
15570         (add_sched_insns_for_speculation): Likewise for local "target",
15571         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15572         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15573         from rtx_insn ** to rtx_code_label **.
15574         (reorg_emit_nops): Likewise.
15575         (c6x_reorg): Likewise for local "call_labels".
15576         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15577         rtx to rtx_insn *.
15578         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
15579         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
15580         the loops over LABEL_REFS.
15581         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
15582         braf_label.
15583         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15584         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
15585         (split_branches): Strengthen local "olabel" from rtx to
15586         rtx_insn *, adding a safe_as_a cast.
15587         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
15588         to "rtx_insn *".
15589         (add_insn_after): Likewise for first two params.
15590         (add_insn_before): Likewise.
15591         (remove_insn): Likewise for param.
15592         (emit_pattern_before_noloc): Likewise for second and third params.
15593         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
15594         (emit_call_insn_before_noloc): Likewise.
15595         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
15596         to "rtx_insn *".
15597         (emit_barrier_before): Likewise.
15598         (emit_label_before): Strengthen "label" param from "rtx" to
15599         "rtx_code_label *".  Strengthen "before" param from "rtx" to
15600         "rtx_insn *".
15601         (emit_insn_after_1): Strengthen "after" param from "rtx" to
15602         "rtx_insn *".
15603         (emit_pattern_after_noloc): Likewise.
15604         (emit_insn_after_noloc): Likewise.
15605         (emit_jump_insn_after_noloc): Likewise.
15606         (emit_call_insn_after_noloc): Likewise.
15607         (emit_debug_insn_after_noloc): Likewise.
15608         (emit_barrier_after): Likewise.
15609         (emit_label_after): Likewise for both params.
15610         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
15611         "loc" param from "int" to "location_t".
15612         (emit_insn_after_setloc): Likewise.
15613         (emit_jump_insn_after_setloc): Likewise.
15614         (emit_call_insn_after_setloc): Likewise.
15615         (emit_debug_insn_after_setloc): Likewise.
15616         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
15617         "loc" param from "int" to "location_t".
15618         (emit_pattern_before): Convert NULL_RTX to NULL.
15619         (emit_insn_before_setloc): Convert "loc" param from "int" to
15620         "location_t".
15621         (emit_jump_insn_before_setloc): Likewise.
15622         (emit_call_insn_before_setloc): Likewise.
15623         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
15624         rtx_insn *.  Convert "loc" param from "int" to "location_t".
15625         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
15626         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
15627         Convert 3rd param from "int" to "location_t".
15628         (emit_barrier_before, emit_barrier_after, next_real_insn):
15629         Strengthen param from rtx to rtx_insn *.
15630         (emit_label_before): Strengthen 1st param from "rtx" to
15631         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
15632         "rtx_insn *".
15633         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
15634         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
15635         Strengthen 2nd param from "rtx" to "rtx_insn *".
15636         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
15637         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
15638         Likewise. Convert 3rd param from "int" to "location_t".
15639         (emit_label_after): Strengthen 1st param from "rtx" to
15640         "rtx_code_label *".
15641         (next_real_insn, remove_insn): Strengthen param from "rtx" to
15642         "rtx_insn *".
15643         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
15644         from "rtx" to "rtx_insn *".
15646 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
15648         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
15649         bodies streamed in with -Q.
15650         * dumpfile.c (dump_files): Add lto-stream-out dump file.
15651         * dumpfile.h (tree_dump_index): Add lto_stream_out.
15652         * gimple-streamer-out.c: Include gimple-pretty-print.h
15653         (output_bb): Dump stmts streamed.
15654         * lto-section-out.c: Include print-tree.h
15655         (lto_begin_section): Dump sections created.
15656         (lto_output_decl_index): Dump decl encoded.
15657         * lto-streamer-out.c: Include print-tree.h
15658         (create_output_block): Dump output block created.
15659         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
15660         (output_function): Dump function output.
15661         (output_constructor): Dump constructor streamed.
15662         (write_global_stream): Output indexes encoded.
15663         (produce_asm_for_decls): Dump streams encoded.
15664         * lto-streamer.c (streamer_dump_file): New global var.
15665         * lto-streamer.h (streamer_dump_file): Declare.
15666         * passes.c (ipa_write_summaries): Initialize streamer dump.
15667         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
15668         in.
15670 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
15672         PR target/86048
15673         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
15674         offsets for register save directives.  Emit a second batch of save
15675         directives, if need be, when the function accesses prior frames.
15677 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15679         * config/arc/fpu.md (fmasf4): Force operand to register.
15680         (fnmasf4): Likewise.
15682 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15684         * config/arc/arc-protos.h (arc_pad_return): Remove.
15685         * config/arc/arc.c (machine_function): Remove force_short_suffix
15686         and size_reason.
15687         (arc_print_operand): Adjust printing of '&'.
15688         (arc_verify_short): Remove conditional printing of short suffix.
15689         (arc_final_prescan_insn): Remove reference to size_reason.
15690         (pad_return): New function.
15691         (arc_reorg): Call pad_return.
15692         (arc_pad_return): Remove.
15693         (arc_init_machine_status): Remove reference to force_short_suffix.
15694         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
15695         (attr length): When attribute iscompact is true force to 2
15696         regardless; in the case of maybe check if we want to force the
15697         instruction to have 4 bytes length.
15698         (nopv): Change it to generate 4 byte long nop as well.
15699         (blockage): New pattern.
15700         (simple_return): Remove call to arc_pad_return.
15701         (p_return_i): Likewise.
15703 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15705         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
15707 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15709         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
15710         ARC cores.
15712 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15714         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
15715         for ARC700 and ARCv2.
15717 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
15719         PR target/86076
15720         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
15721         operands[2] instead of operands[1].
15724 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15726         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
15727         case, check whether the outer register overlaps an unallocatable
15728         register, not just whether it fits the required class.
15730 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15732         * poly-int.h (can_div_trunc_p): Add new overload in which all values
15733         are poly_ints.
15734         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
15735         (memrefs_conflict_p): Likewise.
15736         (init_alias_analysis): Likewise.
15737         * cfgexpand.c (expand_debug_expr): Likewise.
15738         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
15739         * cse.c (fold_rtx): Likewise.
15740         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
15741         * expr.c (emit_block_move_hints): Likewise.
15742         (clear_storage_hints, push_block, emit_push_insn): Likewise.
15743         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
15744         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
15745         (emit_group_store): Likewise.
15746         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
15747         to read the PRE/POST_MODIFY increment.
15748         * calls.c (store_one_arg): Use strip_offset.
15749         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
15750         poly_int_rtx_p.
15751         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
15752         by a VEC_SELECT.
15753         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
15754         (simplify_binary_operation_1): Extend CONST_INT handling to
15755         poly_int_rtx_p.
15756         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
15757         than a HOST_WIDE_INT.
15758         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
15759         poly_int64.
15760         (adjust_mems, add_stores): Update accodingly.
15761         (vt_canonicalize_addr): Track polynomial offsets.
15762         (emit_note_insn_var_location): Likewise.
15763         (vt_add_function_parameter): Likewise.
15764         (vt_initialize): Likewise.
15766 2018-06-12  Jeff Law  <law@redhat.com>
15768         * config.gcc (alpha*-*-freebsd*): Remove.
15769         * config/alpha/freebsd.h: Remove.
15771 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
15773         PR other/69968
15774         * spellcheck-tree.c (levenshtein_distance): Rename to...
15775         (get_edit_distance): ...this, and update for underlying renaming.
15776         * spellcheck-tree.h (levenshtein_distance): Rename to...
15777         (get_edit_distance): ...this.
15778         * spellcheck.c (levenshtein_distance): Rename to...
15779         (get_edit_distance): ...this.  Convert from Levenshtein distance
15780         to Damerau-Levenshtein distance by supporting transpositions of
15781         adjacent characters.  Rename "v1" to "v_next" and "v0" to
15782         "v_one_ago".
15783         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
15784         (selftest::test_edit_distance_unit_test_oneway): ...this, and
15785         update for underlying renaming.
15786         (selftest::levenshtein_distance_unit_test): Rename to...
15787         (selftest::test_get_edit_distance_unit): ...this, and update for
15788         underlying renaming.
15789         (selftest::test_find_closest_string): Add example from PR 69968
15790         where transposition helps
15791         (selftest::test_metric_conditions): Update for renaming.
15792         (selftest::test_metric_conditions): Likewise.
15793         (selftest::spellcheck_c_tests): Likewise.
15794         * spellcheck.h (levenshtein_distance): Rename both overloads to...
15795         (get_edit_distance): ...this.
15796         (best_match::consider): Update for renaming.
15798 2018-06-12  Martin Sebor  <msebor@redhat.com>
15800         PR tree-optimization/85259
15801         * builtins.c (compute_objsize): Handle constant offsets.
15802         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
15803         true iff a warning has been issued.
15804         * gimple.h (gimple_nonartificial_location): New function.
15805         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
15806         gimple_nonartificial_location and handle -Wno-system-headers.
15807         (handle_builtin_stxncpy): Same.
15809 2018-06-12  Martin Sebor  <msebor@redhat.com>
15811         PR c/85931
15812         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
15814 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
15816         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
15817         BUILTIN_VEC_XST entries for pointer to double and long long.
15819 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
15821         PR target/85990
15822         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15823         Update comments.
15824         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15825         Likewise.
15827 2018-06-12  Martin Liska  <mliska@suse.cz>
15829         * doc/options.texi: Document IntegerRange.
15831 2018-06-12  Martin Liska  <mliska@suse.cz>
15833         * config/i386/i386.opt: Make MPX-related options as Deprecated.
15834         * opt-functions.awk: Handle Deprecated flag.
15835         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
15836         and report error.
15837         (read_cmdline_option): Report warning for a deprecated option.
15838         * opts.h (struct cl_option): Add new field cl_deprecated.
15839         (CL_ERR_DEPRECATED): New.
15841 2018-06-12  Martin Liska  <mliska@suse.cz>
15843         * doc/options.texi: Document Deprecated option flag.
15845 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15847         * config/arc/arc-arch.h (arc_extras): New enum.
15848         (arc_cpu_t):Add field extra.
15849         (arc_cpu_types): Consider the extras.
15850         * config/arc/arc-cpus.def: Add extras info.
15851         * config/arc/arc-opts.h (processor_type): Consider extra field.
15852         * config/arc/arc.c (arc_override_options): Handle extra field.
15854 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15856         * config/arc/arc-arch.h: Update ARC_OPTX macro.
15857         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
15858         field.
15859         * config/arc/arc.c (arc_init): Update pic warning.
15860         (irq_range): Update irq range parsing warnings.
15861         (arc_override_options): Update various warning messages.
15862         (arc_handle_aux_attribute): Likewise.
15864 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
15866         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
15868 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15870         * doc/sourcebuild.texi: Document usage of line number 0 in verify
15871         compiler messages directives.
15873 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
15875         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
15876         * config/mips/mips-tables.opt: Regenerate.
15877         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
15878         mips64r6.
15879         * doc/invoke.texi: Document -march=i6500.
15881 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
15883         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
15884         (i6400_gpmul): Add cpu_unit.
15885         (i6400_gpdiv): Likewise.
15886         (i6400_msa_add_d): Update reservations.
15887         (i6400_msa_int_add) Likewise.
15888         (i6400_msa_short_logic3) Likewise.
15889         (i6400_msa_short_logic2) Likewise.
15890         (i6400_msa_short_logic) Likewise.
15891         (i6400_msa_move) Likewise.
15892         (i6400_msa_cmp) Likewise.
15893         (i6400_msa_short_float2) Likewise.
15894         (i6400_msa_div_d) Likewise.
15895         (i6400_msa_long_logic1) Likewise.
15896         (i6400_msa_long_logic2) Likewise.
15897         (i6400_msa_mult) Likewise.
15898         (i6400_msa_long_float2) Likewise.
15899         (i6400_msa_long_float4) Likewise.
15900         (i6400_msa_long_float5) Likewise.
15901         (i6400_msa_long_float8) Likewise.
15902         (i6400_fpu_fadd): Include frint type.
15903         (i6400_fpu_store): New define_insn_reservation.
15904         (i6400_fpu_load): Likewise.
15905         (i6400_fpu_move): Likewise.
15906         (i6400_fpu_fcmp): Likewise.
15907         (i6400_fpu_fmadd): Likewise.
15908         (i6400_int_mult): Include imul3nc type and update reservation.
15909         (i6400_int_div): Include idiv3 type and update reservation.
15910         (i6400_int_load): Update to check type not move_type.
15911         (i6400_int_store): Likewise.
15912         (i6400_int_prefetch): Set zero latency.
15914 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
15916         * gcc.c: Document new %@{...} sequence.
15917         (LINK_COMMAND_SPEC): Use it for the -L switches.
15918         (cpp_unique_options): Use it for the -I switches.
15919         (at_file_argbuf): New global variable.
15920         (in_at_file): Likewise.
15921         (alloc_args): Create at_file_argbuf.
15922         (clear_args): Truncate at_file_argbuf.
15923         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
15924         (open_at_file): New function.
15925         (close_at_file): Likewise.
15926         (create_at_file): Delete.
15927         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
15928         <'o'>: Likewise.
15929         <'@'>: New case.
15930         (validate_switches_from_spec): Deal with %@{...} sequence.
15931         (validate_switches): Likewise.
15932         (driver::finalize): Call clear_args.
15934 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
15936         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
15938 2018-06-11  Martin Sebor  <msebor@redhat.com>
15940         * doc/invoke.texi (-Wall): List -Wc++17-compat.
15941         (Wno-class-memaccess): Add @opindex.
15942         (Wno-templates, Wno-multiple-inheritance): Same.
15943         (Wno-virtual-inheritance, Wno-namespaces): Same.
15944         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
15945         (Wno-format-overflow, Wno-format-truncation): Same.
15946         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
15947         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
15948         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
15949         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
15950         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
15951         (Wno-misspelled-isr): Same.
15953 2018-06-11  Martin Sebor  <msebor@redhat.com>
15955         * PR tree-optimization/86083
15956         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
15958 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
15960         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
15962 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
15964         PR target/85755
15965         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
15966         on the correct operand.
15967         (*movdi_internal64): Ditto.
15969 2018-06-11  Martin Liska  <mliska@suse.cz>
15971         PR tree-optimization/86089
15972         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
15974 2018-06-11  Julia Koval  <julia.koval@intel.com>
15976         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
15977         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
15978         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
15980 2018-06-11  Olivier Hainque  <hainque@adacore.com>
15982         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
15983         for Ada with strict dwarf2.
15985 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
15987         PR target/85755
15988         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
15989         addresses.
15991 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
15993         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
15995 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
15997         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
15998         TARGET_ELF.
16000 2018-06-08  Martin Liska  <mliska@suse.cz>
16002         * tree-cfg.h (debug_function): Fix argument type to match
16003         implementation.
16005 2018-06-08  Martin Liska  <mliska@suse.cz>
16007         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16008         Remove usage of MPX-related (and removed) fields.
16009         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16011 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
16013         * cfg.c (debug): Use TDF_NONE rather than 0.
16014         * cfghooks.c (debug): Likewise.
16015         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16016         (struct dump_option_value_info): Convert to...
16017         (struct kv_pair): ...this template type.
16018         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16019         rather than 0.
16020         (optinfo_verbosity_options): Likewise.
16021         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16022         OPTGROUP_NONE.
16023         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16024         than int for "optgroup_flags" param.
16025         (dump_generic_expr_loc): Use dump_flags_t rather than int for
16026         "dump_kind" param.
16027         (dump_dec): Likewise.
16028         (dump_finish): Use TDF_NONE rather than 0.
16029         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16030         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
16031         than 0.  Update for change to option_ptr.
16032         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16033         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
16034         0.  Update for changes to optinfo_verbosity_options and
16035         optgroup_options.
16036         (opt_info_switch_p): Convert optgroup_flags from int to
16037         optgroup_flags_t.
16038         (dump_basic_block): Use dump_flags_t rather than int
16039         for "dump_kind" param.
16040         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16041         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16042         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16043         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16044         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16045         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16046         TDF_NONE): Convert from macros to...
16047         (enum dump_flag): ...this new enum.
16048         (dump_flags_t): Update to use enum.
16049         (operator|, operator&, operator~, operator|=, operator&=):
16050         Implement for dump_flags_t.
16051         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16052         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16053         Convert from macros to...
16054         (enum optgroup_flag): ...this new enum.
16055         (optgroup_flags_t): New typedef.
16056         (operator|, operator|=): Implement for optgroup_flags_t.
16057         (struct dump_file_info): Convert field "alt_flags" to
16058         dump_flags_t.  Convert field "optgroup_flags" to
16059         optgroup_flags_t.
16060         (dump_basic_block): Use dump_flags_t rather than int for param.
16061         (dump_generic_expr_loc): Likewise.
16062         (dump_dec): Likewise.
16063         (dump_register): Convert param "optgroup_flags" to
16064         optgroup_flags_t.
16065         (opt_info_enable_passes): Likewise.
16066         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16067         than 0.
16068         * gimple-pretty-print.c (debug): Likewise.
16069         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16070         (merged_store_group::apply_stores): Likewise.
16071         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16072         * gimple.c (verify_gimple_pp): Likewise.
16073         * graphite-poly.c (print_pbb_body): Likewise.
16074         * passes.c (pass_manager::register_one_dump_file): Convert
16075         local "optgroup_flags" to optgroup_flags_t.
16076         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16077         (debug): Likewise.
16078         (debug_body): Likewise.
16079         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16080         to optgroup_flags_t.
16081         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16082         than 0.
16083         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16084         (convert_mult_to_fma): Likewise.
16085         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16086         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16087         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16088         "dump_kind" to dump_flags_t.
16090 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16092         * config/rs6000/rs6000.c (min, max): Delete.
16094 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16096         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16097         -mabi=spe and -mabi=no-spe.
16099 2018-06-08  Martin Liska  <mliska@suse.cz>
16101         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16102         where we expect an existing summary.
16104 2018-06-08  Martin Liska  <mliska@suse.cz>
16106         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16107         * ipa-inline.h (estimate_edge_growth): Likewise.
16109 2018-06-08  Martin Liska  <mliska@suse.cz>
16111         * cgraph.c (function_version_hasher::hash): Use
16112         cgraph_node::get_uid ().
16113         (function_version_hasher::equal):
16114         * cgraph.h (cgraph_node::get_uid): New method.
16115         * ipa-inline.c (update_caller_keys): Use
16116         cgraph_node::get_uid ().
16117         (update_callee_keys): Likewise.
16118         * ipa-utils.c (searchc): Likewise.
16119         (ipa_reduced_postorder): Likewise.
16120         * lto-cgraph.c (input_node): Likewise.
16121         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16122         * symbol-summary.h (symtab_insertion): Likewise.
16123         (symtab_removal): Likewise.
16124         (symtab_duplication): Likewise.
16125         * tree-pretty-print.c (dump_function_header): Likewise.
16126         * tree-sra.c (convert_callers_for_node): Likewise.
16128 2018-06-08  Martin Liska  <mliska@suse.cz>
16130         * cgraph.c (symbol_table::create_edge): Always assign a new
16131         unique number.
16132         (symbol_table::free_edge): Do not recycle numbers.
16133         * cgraph.h (cgraph_edge::get): New method.
16134         * symbol-summary.h (symtab_removal): Use it.
16135         (symtab_duplication): Likewise.
16136         (call_summary::hashable_uid): Remove.
16138 2018-06-08  Martin Liska  <mliska@suse.cz>
16140         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16141         (initialize_growth_caches): Remove.
16142         (free_growth_caches): Likewise.
16143         (do_estimate_edge_time): Use edge_growth_cache.
16144         (do_estimate_edge_size): Likewise.
16145         (do_estimate_edge_hints): Likewise.
16146         * ipa-inline.c (reset_edge_caches): Likewise.
16147         (recursive_inlining): Likewise.
16148         (inline_small_functions): Likewise.
16149         * ipa-inline.h (initialize_growth_caches): Remove.
16150         (estimate_edge_size): Likewise.
16151         (estimate_edge_time): Likewise.
16152         (estimate_edge_hints): Likewise.
16153         (reset_edge_growth_cache): Likewise.
16154         * symbol-summary.h (call_summary::remove): New method.
16156 2018-06-08  Martin Liska  <mliska@suse.cz>
16158         * ipa-cp.c (class edge_clone_summary): New summary.
16159         (grow_edge_clone_vectors): Remove.
16160         (ipcp_edge_duplication_hook): Remove.
16161         (class edge_clone_summary_t): New call_summary class.
16162         (ipcp_edge_removal_hook): Remove.
16163         (edge_clone_summary_t::duplicate): New function.
16164         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16165         (create_specialized_node): Likewise.
16166         (ipcp_driver): Initialize edge_clone_summaries and do not
16167         register hooks.
16169 2018-06-08  Martin Liska  <mliska@suse.cz>
16171         * symbol-summary.h (get): New function.
16172         (call_summary::m_initialize_when_cloning): New class member.
16174 2018-06-08  Martin Liska  <mliska@suse.cz>
16176         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16177         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16178         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16179         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16180         (remove_cgraph_node_from_order): Use the removed_nodes set.
16181         (do_per_function_toporder): Likwise.
16182         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16183         instead of summary_uid.
16184         (symtab_removal): Likewise.
16185         (symtab_duplication): Likewise.
16187 2018-06-08  Martin Liska  <mliska@suse.cz>
16189         * ipa-cp.c (ipcp_store_bits_results): Use
16190         ipcp_transformation_sum.
16191         (ipcp_store_vr_results): Likewise.
16192         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16193         to ...
16194         (ipcp_transformation_initialize): ... this.
16195         (ipa_set_node_agg_value_chain):
16196         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16197         (write_ipcp_transformation_info): Likewise.
16198         (read_ipcp_transformation_info): Likewise.
16199         (ipcp_update_bits): Likewise.
16200         (ipcp_update_vr): Likewise.
16201         (ipcp_transform_function): Likewise.
16202         * ipa-prop.h: Rename ipcp_transformation_summary to
16203         ipcp_transformation.
16204         (class ipcp_transformation_t): New function summary.
16205         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16206         (ipa_get_agg_replacements_for_node): Likewise.
16208 2018-06-08  Martin Liska  <mliska@suse.cz>
16210         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16211         of struct.
16212         (class funct_state_summary_t): New function_summary class.
16213         (has_function_state): Remove.
16214         (get_function_state): Likewise.
16215         (set_function_state): Likewise.
16216         (add_new_function): Likewise.
16217         (funct_state_summary_t::insert): New function.
16218         (duplicate_node_data): Remove.
16219         (remove_node_data): Remove.
16220         (funct_state_summary_t::duplicate): New function.
16221         (register_hooks): Create new funct_state_summaries.
16222         (pure_const_generate_summary): Use it.
16223         (pure_const_write_summary): Likewise.
16224         (pure_const_read_summary): Likewise.
16225         (propagate_pure_const): Likewise.
16226         (propagate_nothrow): Likewise.
16227         (dump_malloc_lattice): Likewise.
16228         (propagate_malloc): Likewise.
16229         (execute): Do not register hooks, just remove summary
16230         instead.
16231         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16232         constructor.
16234 2018-06-08  Martin Liska  <mliska@suse.cz>
16236         * ipa-reference.c (remove_node_data): Remove.
16237         (duplicate_node_data): Likewise.
16238         (class ipa_ref_var_info_summary_t): New class.
16239         (class ipa_ref_opt_summary_t): Likewise.
16240         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16241         (get_reference_optimization_summary): Use
16242         ipa_ref_opt_sum_summaries.
16243         (set_reference_vars_info): Remove.
16244         (set_reference_optimization_summary): Likewise.
16245         (ipa_init): Create summaries.
16246         (init_function_info): Use function summary.
16247         (ipa_ref_opt_summary_t::duplicate): New function.
16248         (ipa_ref_opt_summary_t::remove): New function.
16249         (get_read_write_all_from_node): Fix GNU coding style.
16250         (propagate): Use function summary.
16251         (write_node_summary_p): Fix GNU coding style.
16252         (stream_out_bitmap): Likewise.
16253         (ipa_reference_read_optimization_summary): Use function summary.
16254         (ipa_reference_c_finalize): Do not release hooks.
16256 2018-06-08  Martin Liska  <mliska@suse.cz>
16258         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16259         (analyze_function_body): Extract multiple calls of get_create.
16260         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16261         * ipa-inline.c (recursive_inlining): Use ::get method.
16262         * ipa-inline.h (estimate_edge_growth): Likewise.
16264 2018-06-08  Martin Liska  <mliska@suse.cz>
16266         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16267         HSA_INVALID.
16268         (hsa_function_summary::hsa_function_summary): Use the new enum
16269         value.
16270         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16271         * hsa-gen.c (hsa_get_host_function): Likewise.
16272         (get_brig_function_name): Likewise.
16273         * ipa-hsa.c (process_hsa_functions): Likewise.
16274         (ipa_hsa_write_summary): Likewise.
16275         * symbol-summary.h (symtab_duplication): Use ::get function/
16276         (get): New function.
16278 2018-06-08  Martin Liska  <mliska@suse.cz>
16280         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16281         of get.
16282         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16283         (hsa_register_kernel): Likewise.
16284         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16285         * hsa-gen.c (hsa_get_host_function): Likewise.
16286         (get_brig_function_name): Likewise.
16287         (generate_hsa): Likewise.
16288         (pass_gen_hsail::execute): Likewise.
16289         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16290         (devirtualization_time_bonus): Likewise.
16291         (ipcp_propagate_stage): Likewise.
16292         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16293         (edge_set_predicate): Likewise.
16294         (evaluate_conditions_for_known_args): Likewise.
16295         (evaluate_properties_for_edge): Likewise.
16296         (ipa_fn_summary::reset): Likewise.
16297         (ipa_fn_summary_t::duplicate): Likewise.
16298         (dump_ipa_call_summary): Likewise.
16299         (ipa_dump_fn_summary): Likewise.
16300         (analyze_function_body): Likewise.
16301         (compute_fn_summary): Likewise.
16302         (estimate_edge_devirt_benefit): Likewise.
16303         (estimate_edge_size_and_time): Likewise.
16304         (estimate_calls_size_and_time): Likewise.
16305         (estimate_node_size_and_time): Likewise.
16306         (inline_update_callee_summaries): Likewise.
16307         (remap_edge_change_prob): Likewise.
16308         (remap_edge_summaries): Likewise.
16309         (ipa_merge_fn_summary_after_inlining): Likewise.
16310         (ipa_update_overall_fn_summary): Likewise.
16311         (read_ipa_call_summary): Likewise.
16312         (inline_read_section): Likewise.
16313         (write_ipa_call_summary): Likewise.
16314         (ipa_fn_summary_write): Likewise.
16315         (ipa_free_fn_summary): Likewise.
16316         * ipa-hsa.c (process_hsa_functions): Likewise.
16317         (ipa_hsa_write_summary): Likewise.
16318         (ipa_hsa_read_section): Likewise.
16319         * ipa-icf.c (sem_function::merge): Likewise.
16320         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16321         (do_estimate_edge_time): Likewise.
16322         (estimate_size_after_inlining): Likewise.
16323         (estimate_growth): Likewise.
16324         (growth_likely_positive): Likewise.
16325         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16326         (inline_call): Likewise.
16327         * ipa-inline.c (caller_growth_limits): Likewise.
16328         (can_inline_edge_p): Likewise.
16329         (can_inline_edge_by_limits_p): Likewise.
16330         (compute_uninlined_call_time): Likewise.
16331         (compute_inlined_call_time): Likewise.
16332         (want_inline_small_function_p): Likewise.
16333         (edge_badness): Likewise.
16334         (update_caller_keys): Likewise.
16335         (update_callee_keys): Likewise.
16336         (recursive_inlining): Likewise.
16337         (inline_small_functions): Likewise.
16338         (inline_to_all_callers_1): Likewise.
16339         (dump_overall_stats): Likewise.
16340         (early_inline_small_functions): Likewise.
16341         (early_inliner): Likewise.
16342         * ipa-inline.h (estimate_edge_growth): Likewise.
16343         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16344         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16345         * ipa-prop.h (IPA_NODE_REF): Likewise.
16346         (IPA_EDGE_REF): Likewise.
16347         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16348         (propagate_malloc): Likewise.
16349         * ipa-split.c (execute_split_functions): Likewise.
16350         * symbol-summary.h: Rename get to get_create.
16351         (get): Likewise.
16352         (get_create): Likewise.
16353         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16355 2018-06-08  Martin Liska  <mliska@suse.cz>
16357         * symbol-summary.h (release): Move definition out of class
16358         declaration.
16359         (symtab_removal): Likewise.
16360         (symtab_duplication): Likewise.
16362 2018-06-08  Martin Liska  <mliska@suse.cz>
16364         * symbol-summary.h (function_summary): Move constructor
16365         implementation out of class declaration.
16366         (release): Likewise.
16367         (symtab_insertion): Likewise.
16368         (symtab_removal): Likewise.
16369         (symtab_duplication): Likewise.
16370         (get): Likewise.
16372 2018-06-08  Martin Liska  <mliska@suse.cz>
16374         * Makefile.in: Remove support for MPX (macros, related functions,
16375         fields in cgraph_node, ...).
16376         * builtin-types.def (BT_BND): Likewise.
16377         (BT_FN_BND_CONST_PTR): Likewise.
16378         (BT_FN_CONST_PTR_BND): Likewise.
16379         (BT_FN_VOID_PTR_BND): Likewise.
16380         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16381         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16382         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16383         (expand_builtin_mempcpy_with_bounds): Likewise.
16384         (expand_builtin_memset_with_bounds): Likewise.
16385         (expand_builtin_memset_args): Likewise.
16386         (std_expand_builtin_va_start): Likewise.
16387         (expand_builtin): Likewise.
16388         (expand_builtin_with_bounds): Likewise.
16389         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16390         (DEF_LIB_BUILTIN_CHKP): Likewise.
16391         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16392         (DEF_CHKP_BUILTIN): Likewise.
16393         (BUILT_IN_MEMCPY): Likewise.
16394         (BUILT_IN_MEMMOVE): Likewise.
16395         (BUILT_IN_MEMPCPY): Likewise.
16396         (BUILT_IN_MEMSET): Likewise.
16397         (BUILT_IN_STPCPY): Likewise.
16398         (BUILT_IN_STRCAT): Likewise.
16399         (BUILT_IN_STRCHR): Likewise.
16400         (BUILT_IN_STRCPY): Likewise.
16401         (BUILT_IN_STRLEN): Likewise.
16402         (BUILT_IN_MEMCPY_CHK): Likewise.
16403         (BUILT_IN_MEMMOVE_CHK): Likewise.
16404         (BUILT_IN_MEMPCPY_CHK): Likewise.
16405         (BUILT_IN_MEMSET_CHK): Likewise.
16406         (BUILT_IN_STPCPY_CHK): Likewise.
16407         (BUILT_IN_STRCAT_CHK): Likewise.
16408         (BUILT_IN_STRCPY_CHK): Likewise.
16409         * calls.c (store_bounds): Likewise.
16410         (emit_call_1): Likewise.
16411         (special_function_p): Likewise.
16412         (maybe_warn_nonstring_arg): Likewise.
16413         (initialize_argument_information): Likewise.
16414         (finalize_must_preallocate): Likewise.
16415         (compute_argument_addresses): Likewise.
16416         (expand_call): Likewise.
16417         * cfgexpand.c (expand_call_stmt): Likewise.
16418         (expand_return): Likewise.
16419         (expand_gimple_stmt_1): Likewise.
16420         (pass_expand::execute): Likewise.
16421         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16422         (cgraph_node::remove): Likewise.
16423         (cgraph_node::dump): Likewise.
16424         (cgraph_node::verify_node): Likewise.
16425         * cgraph.h (chkp_function_instrumented_p): Likewise.
16426         (symtab_node::get_alias_target): Likewise.
16427         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16428         (cgraph_local_p): Likewise.
16429         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16430         (cgraph_edge::rebuild_references): Likewise.
16431         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16432         (walk_polymorphic_call_targets): Likewise.
16433         (cgraph_node::expand_thunk): Likewise.
16434         (symbol_table::output_weakrefs): Likewise.
16435         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16436         (ix86_handle_option): Likewise.
16437         * config/i386/constraints.md: Likewise.
16438         * config/i386/i386-builtin-types.def (BND): Likewise.
16439         (VOID): Likewise.
16440         (PVOID): Likewise.
16441         (ULONG): Likewise.
16442         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16443         (BDESC_FIRST): Likewise.
16444         (BDESC): Likewise.
16445         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16446         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16447         * config/i386/i386.c (enum reg_class): Likewise.
16448         (ix86_target_string): Likewise.
16449         (ix86_option_override_internal): Likewise.
16450         (ix86_conditional_register_usage): Likewise.
16451         (ix86_valid_target_attribute_inner_p): Likewise.
16452         (ix86_set_indirect_branch_type): Likewise.
16453         (ix86_set_current_function): Likewise.
16454         (ix86_function_arg_regno_p): Likewise.
16455         (init_cumulative_args): Likewise.
16456         (ix86_function_arg_advance): Likewise.
16457         (ix86_function_arg): Likewise.
16458         (ix86_pass_by_reference): Likewise.
16459         (ix86_function_value_regno_p): Likewise.
16460         (ix86_function_value_1): Likewise.
16461         (ix86_function_value_bounds): Likewise.
16462         (ix86_return_in_memory): Likewise.
16463         (ix86_setup_incoming_vararg_bounds): Likewise.
16464         (ix86_va_start): Likewise.
16465         (indirect_thunk_need_prefix): Likewise.
16466         (print_reg): Likewise.
16467         (ix86_print_operand): Likewise.
16468         (ix86_expand_call): Likewise.
16469         (ix86_output_function_return): Likewise.
16470         (reg_encoded_number): Likewise.
16471         (BDESC_VERIFYS): Likewise.
16472         (ix86_init_mpx_builtins): Likewise.
16473         (ix86_init_builtins): Likewise.
16474         (ix86_emit_cmove): Likewise.
16475         (ix86_emit_move_max): Likewise.
16476         (ix86_expand_builtin): Likewise.
16477         (ix86_builtin_mpx_function): Likewise.
16478         (ix86_get_arg_address_for_bt): Likewise.
16479         (ix86_load_bounds): Likewise.
16480         (ix86_store_bounds): Likewise.
16481         (ix86_load_returned_bounds): Likewise.
16482         (ix86_store_returned_bounds): Likewise.
16483         (ix86_class_likely_spilled_p): Likewise.
16484         (ix86_hard_regno_mode_ok): Likewise.
16485         (x86_order_regs_for_local_alloc): Likewise.
16486         (ix86_mitigate_rop): Likewise.
16487         (ix86_bnd_prefixed_insn_p): Likewise.
16488         (ix86_mpx_bound_mode): Likewise.
16489         (ix86_make_bounds_constant): Likewise.
16490         (ix86_initialize_bounds): Likewise.
16491         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16492         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16493         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16494         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16495         (TARGET_CHKP_BOUND_MODE): Likewise.
16496         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16497         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16498         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16499         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16500         * config/i386/i386.h (TARGET_MPX): Likewise.
16501         (TARGET_MPX_P): Likewise.
16502         (VALID_BND_REG_MODE): Likewise.
16503         (FIRST_BND_REG): Likewise.
16504         (LAST_BND_REG): Likewise.
16505         (enum reg_class): Likewise.
16506         (BND_REG_P): Likewise.
16507         (BND_REGNO_P): Likewise.
16508         (BNDmode): Likewise.
16509         (ADJUST_INSN_LENGTH): Likewise.
16510         * config/i386/i386.md: Likewise.
16511         * config/i386/i386.opt: Likewise.
16512         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16513         (defined): Likewise.
16514         (LINK_MPX): Likewise.
16515         (MPX_SPEC): Likewise.
16516         (LIBMPX_SPEC): Likewise.
16517         (LIBMPXWRAPPERS_SPEC): Likewise.
16518         (CHKP_SPEC): Likewise.
16519         * config/i386/predicates.md: Likewise.
16520         * dbxout.c (dbxout_type): Likewise.
16521         * doc/extend.texi: Likewise.
16522         * doc/invoke.texi: Likewise.
16523         * doc/md.texi: Likewise.
16524         * doc/tm.texi: Likewise.
16525         * doc/tm.texi.in: Likewise.
16526         * dwarf2out.c (is_base_type): Likewise.
16527         (gen_formal_types_die): Likewise.
16528         (gen_subprogram_die): Likewise.
16529         (gen_type_die_with_usage): Likewise.
16530         (gen_decl_die): Likewise.
16531         (dwarf2out_late_global_decl): Likewise.
16532         * expr.c (expand_assignment): Likewise.
16533         (emit_storent_insn): Likewise.
16534         (store_expr_with_bounds): Likewise.
16535         (store_expr): Likewise.
16536         (expand_expr_real_1): Likewise.
16537         * expr.h (store_expr_with_bounds): Likewise.
16538         * function.c (use_register_for_decl): Likewise.
16539         (struct bounds_parm_data): Likewise.
16540         (assign_parms_augmented_arg_list): Likewise.
16541         (assign_parm_find_entry_rtl): Likewise.
16542         (assign_parm_is_stack_parm): Likewise.
16543         (assign_parm_load_bounds): Likewise.
16544         (assign_bounds): Likewise.
16545         (assign_parms): Likewise.
16546         (expand_function_start): Likewise.
16547         * gcc.c (CHKP_SPEC): Likewise.
16548         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16549         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16550         (wrestrict_dom_walker::check_call): Likewise.
16551         * gimple.c (gimple_build_call_from_tree): Likewise.
16552         * gimple.h (enum gf_mask): Likewise.
16553         (gimple_call_with_bounds_p): Likewise.
16554         (gimple_call_set_with_bounds): Likewise.
16555         * gimplify.c (gimplify_init_constructor): Likewise.
16556         * ipa-cp.c (initialize_node_lattices): Likewise.
16557         (propagate_constants_across_call): Likewise.
16558         (find_more_scalar_values_for_callers_subset): Likewise.
16559         * ipa-hsa.c (process_hsa_functions): Likewise.
16560         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16561         * ipa-icf.c (sem_function::merge): Likewise.
16562         * ipa-inline.c (early_inliner): Likewise.
16563         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16564         (warn_function_cold): Likewise.
16565         (propagate_pure_const): Likewise.
16566         * ipa-ref.h (enum GTY): Likewise.
16567         * ipa-split.c (find_retbnd): Likewise.
16568         (consider_split): Likewise.
16569         (split_function): Likewise.
16570         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16571         * ipa.c (walk_polymorphic_call_targets): Likewise.
16572         (symbol_table::remove_unreachable_nodes): Likewise.
16573         (process_references): Likewise.
16574         (cgraph_build_static_cdtor_1): Likewise.
16575         * lto-cgraph.c (lto_output_node): Likewise.
16576         (output_refs): Likewise.
16577         (compute_ltrans_boundary): Likewise.
16578         (input_overwrite_node): Likewise.
16579         (input_node): Likewise.
16580         (input_cgraph_1): Likewise.
16581         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
16582         * passes.c (pass_manager::execute_early_local_passes): Likewise.
16583         (class pass_chkp_instrumentation_passes): Likewise.
16584         (make_pass_chkp_instrumentation_passes): Likewise.
16585         * passes.def: Likewise.
16586         * rtl.h (struct GTY): Likewise.
16587         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
16588         * stor-layout.c (layout_type): Likewise.
16589         * symtab.c: Likewise.
16590         * target.def: Likewise.
16591         * targhooks.c (default_chkp_bound_type): Likewise.
16592         (default_chkp_bound_mode): Likewise.
16593         (default_builtin_chkp_function): Likewise.
16594         (default_chkp_function_value_bounds): Likewise.
16595         (default_chkp_make_bounds_constant): Likewise.
16596         (default_chkp_initialize_bounds): Likewise.
16597         * targhooks.h (default_chkp_bound_type): Likewise.
16598         (default_chkp_bound_mode): Likewise.
16599         (default_builtin_chkp_function): Likewise.
16600         (default_chkp_function_value_bounds): Likewise.
16601         (default_chkp_make_bounds_constant): Likewise.
16602         (default_chkp_initialize_bounds): Likewise.
16603         * toplev.c (compile_file): Likewise.
16604         (process_options): Likewise.
16605         * tree-core.h (DEF_BUILTIN): Likewise.
16606         (DEF_BUILTIN_CHKP): Likewise.
16607         * tree-inline.c (declare_return_variable): Likewise.
16608         (remap_gimple_stmt): Likewise.
16609         (copy_bb): Likewise.
16610         (initialize_inlined_parameters): Likewise.
16611         (expand_call_inline): Likewise.
16612         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
16613         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
16614         (make_pass_ipa_chkp_produce_thunks): Likewise.
16615         (make_pass_chkp): Likewise.
16616         (make_pass_chkp_opt): Likewise.
16617         (make_pass_chkp_instrumentation_passes): Likewise.
16618         * tree-pretty-print.c (dump_generic_node): Likewise.
16619         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
16620         * tree-ssa-dce.c (propagate_necessity): Likewise.
16621         (eliminate_unnecessary_stmts): Likewise.
16622         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
16623         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
16624         * tree-ssa-sccvn.h: Likewise.
16625         * tree-ssa-strlen.c (get_string_length): Likewise.
16626         (valid_builtin_call): Likewise.
16627         (adjust_last_stmt): Likewise.
16628         (handle_builtin_strchr): Likewise.
16629         (handle_builtin_strcpy): Likewise.
16630         (handle_builtin_stxncpy): Likewise.
16631         (handle_builtin_memcpy): Likewise.
16632         (handle_builtin_strcat): Likewise.
16633         (strlen_check_and_optimize_stmt): Likewise.
16634         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
16635         * tree-streamer-in.c: Likewise.
16636         * tree-streamer.c (record_common_node): Likewise.
16637         * tree.c (tree_code_size): Likewise.
16638         (wide_int_to_tree_1): Likewise.
16639         (type_contains_placeholder_1): Likewise.
16640         (build_common_tree_nodes): Likewise.
16641         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
16642         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
16643         (POINTER_BOUNDS_P): Likewise.
16644         (BOUNDED_TYPE_P): Likewise.
16645         (BOUNDED_P): Likewise.
16646         (CALL_WITH_BOUNDS_P): Likewise.
16647         (pointer_bounds_type_node): Likewise.
16648         * value-prof.c (gimple_ic): Likewise.
16649         * var-tracking.c (vt_add_function_parameters): Likewise.
16650         * varasm.c (make_decl_rtl): Likewise.
16651         (assemble_start_function): Likewise.
16652         (output_constant): Likewise.
16653         (maybe_assemble_visibility): Likewise.
16654         * varpool.c (ctor_for_folding): Likewise.
16655         * chkp-builtins.def: Remove.
16656         * ipa-chkp.c: Remove.
16657         * ipa-chkp.h: Remove.
16658         * rtl-chkp.c: Remove.
16659         * rtl-chkp.h: Remove.
16660         * tree-chkp-opt.c: Remove.
16661         * tree-chkp.c: Remove.
16662         * tree-chkp.h: Remove.
16664 2018-06-07  Carl Love  <cel@us.ibm.com>
16666         * config/rs6000/vsx.md (vextract_fp_from_shorth,
16667         vextract_fp_from_shortl): Add BE support.
16669 2018-06-07  Paul Koning  <ni1d@arrl.net>
16671         * compare-elim.c (try_merge_compare): Don't merge compare if
16672         address contains a side effect.
16673         (try_eliminate_compare): Likewise.
16675 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
16677         * config.gcc: Support "tremont".
16678         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
16679         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16680         PROCESSOR_TREMONT.
16681         * config/i386/i386.c (m_TREMONT): Define.
16682         (processor_target_table): Add "tremont".
16683         (PTA_TREMONT): Define.
16684         (ix86_lea_outperforms): Add TARGET_TREMONT.
16685         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
16686         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
16687         and M_INTEL_GOLDMONT_PLUS.
16688         (fold_builtin_cpu): Add "tremont".
16689         (ix86_add_stmt_cost): Add TARGET_TREMONT.
16690         (ix86_option_override_internal): Add "tremont".
16691         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
16692         (processor_type): Add PROCESSOR_TREMONT.
16693         * config/i386/x86-tune.def: Add m_TREMONT.
16694         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
16696 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16698         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
16699         symbol defined for msp430i* devices to be lower case.
16701 2018-06-07  Richard Biener  <rguenther@suse.de>
16703         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
16704         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
16705         Properly wrap signed arithmetic if overflow wraps.
16707 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
16709         PR tree-optimization/69615
16710         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
16711         of a cast from a same precision integral SSA_NAME in a bb dominated
16712         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
16713         cast to utype if rhs2 has already a compatible type.
16715 2018-06-07  Richard Biener  <rguenther@suse.de>
16717         PR tree-optimization/85935
16718         * graphite-scop-detection.c (find_params_in_bb): Analyze
16719         condition operands with respect to the correct loop.  Assert
16720         the analysis doesn't fail.
16722 2018-06-04  Carl Love  <cel@us.ibm.com>
16724         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
16725         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
16726         as it is slightly cheaper.
16727         (first_match_or_eos_index_<mode>):
16728         Calculate index using natural element order.
16729         (first_match_index_<mode>):
16730         Calculate index using natural element order.
16731         (first_match_or_eos_index_<mode>):
16732         Calculate index using natural order.
16733         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
16734         for BE and LE modes.
16735         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
16736         P9V_BUILTIN_VCLZLSBB_V16QI.
16737         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
16738         specific.
16740 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16742         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
16743         indentation and line wrap for many prototypes.  Add missing
16744         @smallexample directives around block of prototypes for vec_xl and
16745         vec_xst.
16747 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
16749         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
16750         track if we pass or return IEEE 128-bit floating point.
16751         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
16752         C++ mangling that is compatible with GCC 8.1.
16753         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
16754         (init_cumulative_args): Note if we pass or return IEEE 128-bit
16755         floating point types.
16756         (rs6000_function_arg_advance_1): Likewise.
16757         (rs6000_mangle_type): Optionally generate mangled names that match
16758         what GCC 8.1 generated for IEEE 128-bit floating point types.
16759         (rs6000_globalize_decl_name): If we have an external function that
16760         passes or returns IEEE 128-bit types, generate a weak reference
16761         from the mangled name used in GCC 8.1 to the current mangled
16762         name.
16763         (rs6000_init_builtins): Make __ibm128 use the long double type if
16764         long double is IBM extended double.  Make __float128 use the long
16765         double type if long double is IEEE 128-bit.
16767         PR target/85657
16768         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
16769         macro for __ibm128 built-in functions.
16770         (PACK_IF): Add __ibm128 pack/unpack functions.
16771         (UNPACK_IF): Likewise.
16772         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
16773         enable long double built-in functions if long double is IEEE
16774         128-bit floating point.
16775         (rs6000_invalid_builtin): Update long double built-in function
16776         error message.
16777         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
16778         functions, adjust the built-in function to use the long double
16779         built-in function if __ibm128 and long double are the same type.
16780         * doc/extend.texi (PowerPC builtins): Update documention for
16781         __builtin_{,un}pack_longdouble.  Add documentation for
16782         __builtin_{,un}pack_ibm128.
16784 2018-06-06  Jim Wilson  <jimw@sifive.com>
16786         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
16787         (struct machine_function): New field interrupt_mode.
16788         (riscv_handle_type_attribute): New function.  Add forward declaration.
16789         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
16790         (riscv_expand_epilogue): Check interrupt_mode field.
16791         (riscv_set_current_function): Check interrupt attribute args and
16792         set interrupt_mode field.
16793         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
16794         (riscv_sret, riscv_uret): New.
16795         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
16796         new arguments to interrupt attribute.
16798 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
16800         PR target/63177
16801         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
16802         Don't handle -mcpu=power8 if -mpower9-vector is also used.
16804 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16806         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
16807         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
16808         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16809         several redundant entries.
16811 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
16813         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
16814         type from "rtx" to "rtx_insn *".
16815         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
16816         for local "call_insn", removing cast.
16817         (ix86_expand_call): Likewise, introducing a "call_insn" local.
16819 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
16821         PR tree-optimization/86066
16822         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
16823         for BIT_INSERT_EXPR stores.
16825 2018-06-06  Richard Biener  <rguenther@suse.de>
16827         PR tree-optimization/86062
16828         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16829         component refs ontop
16830         of to be offsetted base.
16832 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16834         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
16835         to be static and remove check on interrupt attribute name.
16837 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16839         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16840         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
16842 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
16844         PR target/79924
16845         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
16846         second argument.
16847         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
16848         Remove second argument, change how error is called.
16849         (aarch64_layout_arg): Remove second argument from
16850         aarch64_err_no_fpadvsimd call.
16851         (aarch64_init_cumulative_args): Ditto.
16852         (aarch64_gimplify_va_arg_expr): Ditto.
16853         * config/aarch64/aarch64.md (mov<mode>): Ditto.
16855 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
16857         * config/i386/i386.md (simple_return_indirect_internal): New expander.
16858         (*simple_return_indirect_internal<mode>): Rename from
16859         simple_return_indirect_internal.  Use W mode iterator.
16860         (rstorssp): New expander.
16861         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
16862         (clrssbsy): New expander.
16863         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
16865 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16867         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
16868         __typeof__.
16869         (cmse_check_pointed_object): Likewise.
16871 2018-06-05  Martin Liska  <mliska@suse.cz>
16873         PR gcov-profile/47618
16874         * doc/invoke.texi: Document how -fprofile-dir format
16875         is extended.
16877 2018-06-05  Richard Biener  <rguenther@suse.de>
16879         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
16880         removal pretend DOM info isn't available so we do not update
16881         it and only remove edges, not dominated blocks.  Actually free
16882         DOM info in case we removed something.  Remove unreachable blocks.
16883         (mfb_keep_latches): Work with either DOM info or marked backedges.
16884         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
16885         first.  Mark backedges if DOM info isn't available.
16886         (Re-)compute DOM info after cleanup_control_flow_pre.
16888 2018-06-05  Richard Biener  <rguenther@suse.de>
16890         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
16891         (locus_discrim_hasher::hash): Adjust.
16892         (locus_discrim_hasher::equal): Likewise.
16893         (next_discriminator_for_locus): Work on line directly.
16894         (same_line_p): Pass in expanded locus1 as well.
16895         (assign_discriminators): Avoid redundant location expansions.
16897 2018-06-05  Richard Biener  <rguenther@suse.de>
16899         PR tree-optimization/86046
16900         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
16901         if required after clearing TREE_ADDRESSABLE.
16903 2018-06-05  Richard Biener  <rguenther@suse.de>
16905         PR tree-optimization/86047
16906         * tree-ssa-loop.c (for_each_index): Glob handling of all
16907         decls and constants and really handle all of them.
16909 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16911         PR target/81497
16912         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
16913         qualifier_void_pointer and qualifier_const_void_pointer.
16914         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
16915         (arm_init_builtins): Handle the above.
16916         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
16917         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
16918         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
16919         void intrinsics.
16921 2018-06-05  Martin Liska  <mliska@suse.cz>
16923         * auto-profile.c (read_autofdo_file): Do not use
16924         gcov_ctr_summary struct.
16925         (afdo_callsite_hot_enough_for_early_inline): Likewise.
16926         * coverage.c (struct counts_entry): Likewise.
16927         (read_counts_file): Read just single summary entry.
16928         (get_coverage_counts): Use gcov_summary struct.
16929         * coverage.h (get_coverage_counts): Likewise.
16930         * gcov-dump.c (dump_working_sets): Likewise.
16931         (tag_summary): Dump just single summary.
16932         * gcov-io.c (gcov_write_summary): Write just histogram
16933         summary.
16934         (gcov_read_summary): Read just single summary.
16935         (compute_working_sets): Use gcov_summary struct.
16936         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
16937         of GCOV_COUNTERS_SUMMABLE.
16938         (GCOV_COUNTERS_SUMMABLE): Remove.
16939         (GCOV_FIRST_VALUE_COUNTER): Replace with
16940         GCOV_COUNTER_V_INTERVAL.
16941         (struct gcov_ctr_summary): Remove.
16942         (struct gcov_summary): Directly use fields of former
16943         gcov_ctr_summary.
16944         (compute_working_sets): Use gcov_summary struct.
16945         * gcov.c (read_count_file): Do not use ctrs fields.
16946         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
16947         struct.
16948         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
16949         struct.
16950         * profile.c: Likewise.
16951         * profile.h: Likewise.
16953 2018-06-05  Martin Liska  <mliska@suse.cz>
16955         PR gcov-profile/84846
16956         * gcov.c (output_lines): Print working directory only
16957         in intermediate format.
16959 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
16961         * config/s390/s390-builtin-types.def: Add void function type.
16962         * config/s390/s390-builtins.def: Use the function type for the
16963         tbeginc builtin.
16965 2018-06-04  Jim Wilson  <jimw@sifive.com>
16967         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
16968         to int.
16969         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
16970         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
16971         handle EH_RETURN_DATA_REGNO registers properly.
16972         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
16973         (riscv_expand_epilogue): Update comment.  Change argument name and
16974         type.  Update code to use new name and type.  Pass new args to
16975         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
16976         EXCEPTION_RETURN.
16977         * config/riscv/riscv.md (NORMAL_RETURN): New.
16978         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
16979         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16980         (eh_return): Call gen_eh_return_internal and emit barrier.
16981         (eh_return_internal): Call riscv_expand_epilogue.
16983 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
16985         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
16986         bit_insertion field and declare can_be_merged_into method.
16987         (merged_store_group::can_be_merged_into): New method.
16988         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
16989         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
16990         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
16992 2018-06-04  Richard Biener  <rguenther@suse.de>
16994         PR tree-optimization/85955
16995         * builtins.c (fold_builtin_sincos): Convert pointers to
16996         destination to appropriate type before dereferencing.
16998 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
17000         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17002 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
17004         * expr.c (expand_expr_real_1): Force the operand into memory if
17005         its TYPE_MODE is BLKmode and if there is no integer mode for
17006         the number of bits being extracted.
17008 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17010         PR target/85832
17011         PR target/86036
17012         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17013         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17015 2018-06-04  Richard Biener  <rguenther@suse.de>
17017         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17018         (cleanup_tree_cfg_noloop): ... single caller.  Do
17019         start_recording_case_labels later.
17021 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
17023         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17024         to _IMMINTRIN_H_INCLUDED.
17025         * config/i386/pconfigintrin.h: Ditto.
17026         * config/i386/waitpkgintrin.h: Ditto.
17027         * config/i386/immintrin.h: Add includes for sgxintrin.h,
17028         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17029         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17030         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17031         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17032         waitpkgintrin.h and cldemoteintrin.h.
17034 2018-06-04  Richard Biener  <rguenther@suse.de>
17036         PR tree-optimization/86038
17037         * tracer.c (find_best_successor): Check probability for
17038         being initialized, bail out if not.
17040 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
17042         PR target/86003
17043         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17044         of bits to ignore when comparing architectures.
17046 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17048         PR tree-optimization/69615
17049         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17050         maximum or minimum of the type, try to merge it also as if
17051         range1 is + [-, x - 1] or + [x + 1, -].
17053         PR c++/86025
17054         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17056 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
17058         PR tree-optimization/86034
17059         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17060         the unsigned bitfield type in a bit insertion sequence if it does not
17061         have a larger precision than the bitfield size.
17062         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17064 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
17066         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17068 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17070         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17071         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17072         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17073         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17075 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17077         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17078         Disable -fdelete-null-pointer-checks for ELF toolchain.
17080 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17081             Kito Cheng  <kito.cheng@gmail.com>
17083         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17084         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17085         (nds32*-*-*): Add glibc and uclibc conditions.
17086         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17087         (TARGET_EXCEPT_UNWIND_INFO): Define.
17088         * config/nds32/elf.h: New file.
17089         * config/nds32/linux.h: New file.
17090         * config/nds32/nds32-elf.opt: New file.
17091         * config/nds32/nds32-linux.opt: New file.
17092         * config/nds32/nds32-fp-as-gp.c
17093         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17094         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17095         TARGET_LINUX_ABI.
17096         (nds32_asm_file_end): Ditto.
17097         (nds32_print_operand): Ditto.
17098         (nds32_insert_attributes): Ditto.
17099         (nds32_init_libfuncs): New function.
17100         (TARGET_HAVE_TLS): Define.
17101         (TARGET_INIT_LIBFUNCS): Define.
17102         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17103         spec content.
17104         (TARGET_ELF): Apply different mcmodel setting.
17105         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17106         been migrated into elf.h and linux.h files.
17107         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17108         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17109         (mcmodel): The content has been migrated into nds32-elf.opt and
17110         nds32-linux.opt files.
17111         * config/nds32/t-elf: New file.
17112         * config/nds32/t-linux: New file.
17114 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17115             Shiva Chen  <shiva0217@gmail.com>
17117         * config/nds32/constants.md (unspec_volatile_element): Add
17118         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17119         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17120         optimization.
17121         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17122         (make_pass_nds32_fp_as_gp): Declare.
17123         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17124         optmization pass.
17125         (nds32_asm_function_end_prologue): Remove unused asm output.
17126         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17127         (nds32_asm_file_start): Output necessary fp_as_gp information.
17128         (nds32_option_override): Adjust register usage.
17129         (nds32_expand_prologue): Consider fp_as_gp situation.
17130         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17131         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17132         (epilogue): Ditto.
17133         (return): Ditto.
17134         (simple_return): Ditto.
17135         (omit_fp_begin): Output special directive for fp_as_gp.
17136         (omit_fp_end): Output special directive for fp_as_gp.
17137         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17138         mforbid-fp-as-gp): New options.
17140 2018-06-01  Mark Wielaard  <mark@klomp.org>
17142         * dwarf2out.c (dwarf2out_finish): Remove generation of
17143         DW_AT_loclists_base.
17145 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17147         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17148         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17149         (struct merged_store_group): Add bit_insertion field.
17150         (dump_char_array): Use standard hexadecimal format.
17151         (merged_store_group::merged_store_group): Set bit_insertion to false.
17152         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17153         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17154         also print the mask in the dump file.
17155         (pass_store_merging::gate): Minor tweak.
17156         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17157         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17158         stores with INTEGER_CST stores.
17159         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17160         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17161         and use it throughout.  Generate bit insertion sequences if need be.
17162         (pass_store_merging::process_store): Remove redundant condition.
17163         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17165 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17167         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17168         the 128-bit floating point types.  Fix function comment.
17170 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17172         * config/aarch64/aarch64-simd.md
17173         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17174         mnemonics.
17175         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17176         mnemonics.
17178 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17180         PR tree-optimization/85989
17181         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17182         variable.
17183         (backprop::intersect_uses): Check it when deciding whether this
17184         is a backedge reference.
17185         (backprop::process_block): Add each phi to m_visited_phis
17186         after visiting it, then clear it at the end.
17188 2018-06-01  Richard Biener  <rguenther@suse.de>
17190         * tree-vectorizer.h (vect_dr_stmt): New function.
17191         (vect_get_load_cost): Adjust.
17192         (vect_get_store_cost): Likewise.
17193         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17194         Use vect_dr_stmt instead of DR_SMTT.
17195         (vect_record_base_alignments): Likewise.
17196         (vect_calculate_target_alignment): Likewise.
17197         (vect_compute_data_ref_alignment): Likewise and make static.
17198         (vect_update_misalignment_for_peel): Likewise.
17199         (vect_verify_datarefs_alignment): Likewise.
17200         (vector_alignment_reachable_p): Likewise.
17201         (vect_get_data_access_cost): Likewise.  Pass down
17202         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17203         (vect_get_peeling_costs_all_drs): Likewise.
17204         (vect_peeling_hash_get_lowest_cost): Likewise.
17205         (vect_enhance_data_refs_alignment): Likewise.
17206         (vect_find_same_alignment_drs): Likewise.
17207         (vect_analyze_data_refs_alignment): Likewise.
17208         (vect_analyze_group_access_1): Likewise.
17209         (vect_analyze_group_access): Likewise.
17210         (vect_analyze_data_ref_access): Likewise.
17211         (vect_analyze_data_ref_accesses): Likewise.
17212         (vect_vfa_segment_size): Likewise.
17213         (vect_small_gap_p): Likewise.
17214         (vectorizable_with_step_bound_p): Likewise.
17215         (vect_prune_runtime_alias_test_list): Likewise.
17216         (vect_analyze_data_refs): Likewise.
17217         (vect_supportable_dr_alignment): Likewise.
17218         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17219         (vect_gen_prolog_loop_niters): Likewise.
17220         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17221         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17222         modify DR_STMT.
17223         (vect_recog_mask_conversion_pattern): Likewise.
17224         (vect_try_gather_scatter_pattern): Likewise.
17225         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17226         to vect_get_store_cost.
17227         (vect_get_store_cost): Get stmt_info instead of DR.
17228         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17229         (vect_get_load_cost): Get stmt_info instead of DR.
17231 2018-06-01  Richard Biener  <rguenther@suse.de>
17233         PR middle-end/86017
17234         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17235         vars wrapped in MEM_REFs.
17237 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17239         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17240         Fix subreg tests so that we only return a choice between
17241         GENERAL_REGS and FP_REGS if the original classes included both.
17243 2018-06-01  Richard Biener  <rguenther@suse.de>
17245         PR ipa/85960
17246         * tree-ssa-structalias.c (get_function_part_constraint):
17247         Handle NULL fi->decl.
17248         (find_func_aliases_for_call): Properly handle indirect
17249         fi from direct call.
17250         (find_func_clobbers): Likewise.
17251         (ipa_pta_execute): Likewise.
17252         (create_variable_info_for): For functions that are ifunc_resolver
17253         resolve to a varinfo that contains the result of the resolver call.
17254         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17255         aliases.
17257 2018-05-31  Michael Collison  <michael.collison@arm.com>
17259         * config/aarch64/aarch64.md:
17260         (*fix_to_zero_extenddfdi2): New pattern.
17261         * gcc.target/aarch64/fix_extend1.c: New testcase.
17263 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17265         PR middle-end/78809
17266         PR middle-end/83026
17267         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17268         and BUILT_IN_STRNCMP_EQ.
17269         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17270         BUILT_IN_STRNCMP_EQ.
17271         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17272         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17273         (gimple_fold_builtin): Likewise.
17274         * tree-ssa-strlen.c (compute_string_length): New function.
17275         (determine_min_obsize): New function.
17276         (handle_builtin_string_cmp): New function to handle calls to
17277         string compare functions.
17278         (strlen_optimize_stmt): Add handling to builtin string compare
17279         calls.
17280         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17281         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17282         * tree.c (build_common_builtin_nodes): Add new defines of
17283         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17285 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17287         PR target/85984
17288         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17289         functions with naked attribute.
17291 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17293         * config/i386/sse.md (avx_vec_concat<mode>):
17294         Substitute concat_tg_mode mode attribute with xtg_mode.
17295         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17296         (concat_tg_mode): Remove mode attribute.
17298 2018-05-31  Martin Sebor  <msebor@redhat.com>
17300         PR c/82063
17301         * calls.c (alloc_max_size): Correct a logic error/typo.
17302         Treat excessive arguments as infinite.  Warn for invalid arguments.
17303         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17305 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17307         PR target/85829
17308         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17309         and movx for Haswell.
17311 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17312             Cesar Philippidis  <cesar@codesourcery.com>
17314         PR middle-end/85879
17315         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17316         when emitting error on private/firstprivate reductions.
17317         * omp-low.c (lower_omp_target): Avoid reference-type processing
17318         on pointers for firstprivate clause.
17320 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17322         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17323         (st1x2): Likewise.
17324         (st1x3): Likewise.
17325         * config/aarch64/aarch64-simd.md
17326         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17327         (aarch64_ld1_x3_<mode>): Likewise
17328         (aarch64_st1x2<VALLDIF:mode>): Likewise
17329         (aarch64_st1_x2_<mode>): Likewise
17330         (aarch64_st1x3<VALLDIF:mode>): Likewise
17331         (aarch64_st1_x3_<mode>): Likewise
17332         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17333         (vld1_s8_x3): Likewise.
17334         (vld1_u16_x3): Likewise.
17335         (vld1_s16_x3): Likewise.
17336         (vld1_u32_x3): Likewise.
17337         (vld1_s32_x3): Likewise.
17338         (vld1_u64_x3): Likewise.
17339         (vld1_s64_x3): Likewise.
17340         (vld1_f16_x3): Likewise.
17341         (vld1_f32_x3): Likewise.
17342         (vld1_f64_x3): Likewise.
17343         (vld1_p8_x3): Likewise.
17344         (vld1_p16_x3): Likewise.
17345         (vld1_p64_x3): Likewise.
17346         (vld1q_u8_x3): Likewise.
17347         (vld1q_s8_x3): Likewise.
17348         (vld1q_u16_x3): Likewise.
17349         (vld1q_s16_x3): Likewise.
17350         (vld1q_u32_x3): Likewise.
17351         (vld1q_s32_x3): Likewise.
17352         (vld1q_u64_x3): Likewise.
17353         (vld1q_s64_x3): Likewise.
17354         (vld1q_f16_x3): Likewise.
17355         (vld1q_f32_x3): Likewise.
17356         (vld1q_f64_x3): Likewise.
17357         (vld1q_p8_x3): Likewise.
17358         (vld1q_p16_x3): Likewise.
17359         (vld1q_p64_x3): Likewise.
17360         (vst1_s64_x2): Likewise.
17361         (vst1_u64_x2): Likewise.
17362         (vst1_f64_x2): Likewise.
17363         (vst1_s8_x2): Likewise.
17364         (vst1_p8_x2): Likewise.
17365         (vst1_s16_x2): Likewise.
17366         (vst1_p16_x2): Likewise.
17367         (vst1_s32_x2): Likewise.
17368         (vst1_u8_x2): Likewise.
17369         (vst1_u16_x2): Likewise.
17370         (vst1_u32_x2): Likewise.
17371         (vst1_f16_x2): Likewise.
17372         (vst1_f32_x2): Likewise.
17373         (vst1_p64_x2): Likewise.
17374         (vst1q_s8_x2): Likewise.
17375         (vst1q_p8_x2): Likewise.
17376         (vst1q_s16_x2): Likewise.
17377         (vst1q_p16_x2): Likewise.
17378         (vst1q_s32_x2): Likewise.
17379         (vst1q_s64_x2): Likewise.
17380         (vst1q_u8_x2): Likewise.
17381         (vst1q_u16_x2): Likewise.
17382         (vst1q_u32_x2): Likewise.
17383         (vst1q_u64_x2): Likewise.
17384         (vst1q_f16_x2): Likewise.
17385         (vst1q_f32_x2): Likewise.
17386         (vst1q_f64_x2): Likewise.
17387         (vst1q_p64_x2): Likewise.
17388         (vst1_s64_x3): Likewise.
17389         (vst1_u64_x3): Likewise.
17390         (vst1_f64_x3): Likewise.
17391         (vst1_s8_x3): Likewise.
17392         (vst1_p8_x3): Likewise.
17393         (vst1_s16_x3): Likewise.
17394         (vst1_p16_x3): Likewise.
17395         (vst1_s32_x3): Likewise.
17396         (vst1_u8_x3): Likewise.
17397         (vst1_u16_x3): Likewise.
17398         (vst1_u32_x3): Likewise.
17399         (vst1_f16_x3): Likewise.
17400         (vst1_f32_x3): Likewise.
17401         (vst1_p64_x3): Likewise.
17402         (vst1q_s8_x3): Likewise.
17403         (vst1q_p8_x3): Likewise.
17404         (vst1q_s16_x3): Likewise.
17405         (vst1q_p16_x3): Likewise.
17406         (vst1q_s32_x3): Likewise.
17407         (vst1q_s64_x3): Likewise.
17408         (vst1q_u8_x3): Likewise.
17409         (vst1q_u16_x3): Likewise.
17410         (vst1q_u32_x3): Likewise.
17411         (vst1q_u64_x3): Likewise.
17412         (vst1q_f16_x3): Likewise.
17413         (vst1q_f32_x3): Likewise.
17414         (vst1q_f64_x3): Likewise.
17415         (vst1q_p64_x3): Likewise.
17417 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17419         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17420         user_label_prefix to name.
17422         * tree-core.h: Update comment about the format of NAME string
17423         passed to handler in attribute_spec.
17425         * config/msp430/msp430.md: Remove erroneous subreg expression from
17426         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17427         zero_extend{q,h}isi2.
17429 2018-05-30  Borislav Petkov  <bp@suse.de>
17431         * doc/extend.texi: Document some architecture specific
17432         constraints and sort entries.
17434 2018-05-30  Martin Sebor  <msebor@redhat.com>
17436         PR middle-end/85369
17437         * builtins.c (expand_builtin_stpcpy_1): New function.
17438         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17439         only if the former succeeds.
17441 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17443         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17444         in saphira.
17446 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17448         * doc/invoke.texi (-flinker-output): Document
17450 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17452         * passes.c (ipa_write_summaries): Only modify statements if body
17453         is in memory.
17454         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17455         incrementally linking.
17456         (ipa_passes): Likewise.
17457         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17458         pass down resolution info.
17459         * common.opt (flag_incremental_link): Update info.
17460         * gcc.c (plugin specs): Turn flinker-output=* to
17461         -plugin-opt=-linker-output-known
17462         * toplev.c (compile_file): Also cut compilation when doing incremental
17463         link.
17464         * flag-types. (enum lto_partition_model): Add
17465         LTO_LINKER_OUTPUT_NOLTOREL.
17466         (invoke.texi): Add -flinker-output docs.
17467         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17468         link same way as WPA; do not stream in dead initializers.
17470         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17471         darf2out_register_external_decl): Support incremental link.
17473 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17475         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17477 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17479         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17480         it down to simple_object_copy_lto_debug_sections.
17481         (run_gcc): Determine incremental LTO link time and configure
17482         lto1 into non-wpa mode, disable renaming of debug sections.
17484 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17486         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17487         descriptions of various incorrectly documented functions.
17489 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17491         Revert:
17492         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17493         address check not strict.
17495 2018-05-30  Richard Biener  <rguenther@suse.de>
17497         PR tree-optimization/85964
17498         * tracer.c (better_p): Drop initialized count check, we only
17499         call the function with initialized counts now.
17500         (find_best_successor): Do find a best edge if one
17501         has uninitialized count.
17502         (find_best_predecessor): Likewise.  Do BB frequency check only
17503         if count is initialized.
17505 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17507         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17508         (aarch64_ldrstr_offset_compare): New.
17509         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17510         load/store orderings.
17511         (aarch64_gen_adjusted_ldpstp): Likewise.
17513 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17515         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17516         Check for subset of GENERAL_REGS and FP_REGS.
17517         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17518         r=w alternative.
17520 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17522         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17523         and wi::to_poly_offset.  Add the current offset and then check
17524         whether the sum fits, rather than using an unchecked addition of
17525         a checked term.  Check for a shwi rather than a uhwi.
17526         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17527         (store_constructor): Use poly_int_tree_p.
17528         (expand_expr_real_1): Likewise.
17529         * function.c (assign_temp): Likewise.
17530         * fold-const.c (const_binop): Use poly_int_tree_p and
17531         wi::to_poly_offset.
17532         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17533         division.
17534         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17535         to_poly_offset for MEM offsets.
17536         * ipa-icf.c (sem_variable::equals): Likewise.
17537         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17538         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17539         wi::to_poly_offset for BIT_FIELD_REF offsets.
17540         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17541         wi::to_poly_offset.
17542         * var-tracking.c (emit_note_insn_var_location): Use
17543         tree_to_poly_uint64.
17545 2018-05-29  Jim Wilson  <jimw@sifive.com>
17547         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17549 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17551         PR target/85950
17552         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17553         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17554         sequence.
17555         (sse4_1_round<mode>2): Use nonimmediate_operand
17556         for operand 1 predicate.
17558 2018-05-29  Martin Sebor  <msebor@redhat.com>
17559             Richard Biener  <rguenther@suse.de>
17561         PR testsuite/85888
17562         * calls.c (get_size_range): Call determine_value_range instead
17563         of get_value_range..
17564         * tree-vrp.h (determine_value_range): Declared new function.
17565         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17567 2018-05-29  Richard Biener  <rguenther@suse.de>
17569         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17570         sure to use non-pattern stmts for get_earlier_stmt arguments.
17571         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17572         called on pattern stmts.
17573         (get_later_stmt): Likewise.
17575 2018-05-29  Martin Liska  <mliska@suse.cz>
17577         PR gcov-profile/85759
17578         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
17579         env variables.
17581 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
17583         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
17584         VEC_UNPACK_*_EXPR.
17585         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
17586         VEC_PACK_*_EXPR.
17588         PR target/85918
17589         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17590         VEC_PACK_FLOAT_EXPR): New tree codes.
17591         * tree-pretty-print.c (op_code_prio): Handle
17592         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17593         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17594         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17595         * tree-inline.c (estimate_operator_cost): Likewise.
17596         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
17597         * fold-const.c (const_binop): Likewise.
17598         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
17599         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17600         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17601         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
17602         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17603         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17604         * expr.c (expand_expr_real_2): Likewise.
17605         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
17606         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
17607         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
17608         optabs.
17609         * optabs.c (expand_widen_pattern_expr): For
17610         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
17611         sign from result type rather than operand's type.
17612         (expand_binop_directly): For vec_packu_float_optab and
17613         vec_packs_float_optab allow result type to be different from operand's
17614         type.
17615         * optabs-tree.c (optab_for_tree_code): Handle
17616         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17617         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
17618         * tree-vect-generic.c (expand_vector_operations_1):  Handle
17619         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17620         VEC_PACK_FLOAT_EXPR.
17621         * tree-vect-stmts.c (supportable_widening_operation): Handle
17622         FIX_TRUNC_EXPR.
17623         (supportable_narrowing_operation): Handle FLOAT_EXPR.
17624         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
17625         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
17626         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
17627         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
17628         mode attributes.
17629         (vec_pack<floatprefix>_float_<mode>): New expander.
17630         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
17631         attributes.
17632         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
17633         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
17634         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
17635         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
17636         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
17637         Document.
17638         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
17639         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
17640         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17641         VEC_PACK_FLOAT_EXPR): Document.
17643 2018-05-29  Richard Biener  <rguenther@suse.de>
17645         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
17646         member.
17647         (stmt_vec_info_vec): Make pointer.
17648         (init_stmt_vec_info_vec): Remove.
17649         (free_stmt_vec_info_vec): Likewise.
17650         (set_stmt_vec_info_vec): New function.
17651         (free_stmt_vec_infos): Likewise.
17652         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
17653         (set_vinfo_for_stmt): Likewise.
17654         (get_earlier_stmt): Likewise.
17655         (get_later_stmt): Likewise.
17656         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
17657         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
17658         (vec_info::~vec_info): Free stmt_vec_infos.
17659         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
17660         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
17661         (pass_slp_vectorize::execute): Likewise.
17662         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
17663         (free_stmt_vec_info_vec): Likewise.
17664         (set_stmt_vec_info_vec): New function.
17665         (free_stmt_vec_infos): Likewise.
17666         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
17667         the global stmt_vec_info_vec.
17668         * tree-parloops.c (gather_scalar_reductions): Use
17669         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
17670         vector.
17672 2018-05-29  Richard Biener  <rguenther@suse.de>
17674         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
17676 2018-05-29  Martin Liska  <mliska@suse.cz>
17677             David Malcolm  <dmalcolm@redhat.com>
17679         * vec.c (test_reverse): New.
17680         (vec_c_tests): Add new test.
17681         * vec.h (vl_ptr>::reverse): New function.
17683 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
17685         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
17687         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
17688         and later.
17690 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17692         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
17694 2018-05-28  Richard Biener  <rguenther@suse.de>
17696         PR tree-optimization/85933
17697         * tree-vect-data-refs.c (vect_record_base_alignments): Only
17698         look at stmts marked as vectorizable.
17700 2018-05-28  Richard Biener  <rguenther@suse.de>
17702         PR tree-optimization/85934
17703         * tree-vect-generic.c (expand_vector_operations_1): Hoist
17704         vector boolean check before scalar optimization.
17706 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
17708         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
17709         for armv5te.
17711 2018-05-28  Mark Wielaard  <mark@klomp.org>
17713         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
17714         if it is an expression containing a minus sign.
17716 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
17718         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
17720 2018-05-27  Paul Koning  <ni1d@arrl.net>
17722         * config/pdp11/pdp11.md (truncsihi2): Remove.
17724 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
17725             Chung-Ju Wu  <jasonwucj@gmail.com>
17727         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
17728         implementation.
17729         (unaligned_store_dw): Ditto.
17730         * config/nds32/nds32-memory-manipulation.c
17731         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17732         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
17733         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
17734         (emit_setmem_word_loop): Rename to ...
17735         (emit_setmem_doubleword_loop): ... this.
17736         (nds32_gen_dup_4_byte_to_word_value): New function.
17737         (nds32_gen_dup_8_byte_to_double_word_value): New function.
17738         (nds32_expand_setmem_loop): Refine implementation.
17739         (nds32_expand_setmem_loop_v3m): Ditto.
17740         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
17741         pattern.
17743 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17745         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
17747 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17749         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
17750         (nds32_init_machine_status): Initialize machine->attr_naked_p and
17751         machine->attr_no_prologue_p.
17752         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
17753         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
17754         (nds32_expand_epilogue): Consider attr_naked_p.
17755         (nds32_expand_epilogue_v3pop): Likewise.
17756         (nds32_can_use_return_insn): Likewise.
17757         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
17758         attr_no_prologue_p fields.
17759         * config/nds32/nds32.opt (mret-in-naked-func): New option.
17761 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
17763         PR target/85918
17764         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
17765         attributes.
17766         * config/i386/sse.md
17767         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
17768         Rename to ...
17769         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
17770         ... this.
17771         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
17772         Rename to ...
17773         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
17774         ... this.
17775         (*<floatsuffix>floatv2div2sf2): Rename to ...
17776         (*float<floatunssuffix>v2div2sf2): ... this.
17777         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
17778         (float<floatunssuffix>v2div2sf2_mask): ... this.
17779         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
17780         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
17781         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
17782         to ...
17783         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
17784         ... this.
17785         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17786         Rename to ...
17787         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17788         ... this.
17789         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17790         Rename to ...
17791         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17792         ... this.
17793         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
17794         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
17795         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
17796         gen_ufix_truncv8dfv8si2.
17797         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
17798         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
17799         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
17800         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
17801         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
17802         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
17803         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
17804         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
17806 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17808         PR target/85900
17809         PR target/85345
17810         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
17812 2018-05-25  Jim Wilson  <jimw@sifive.com>
17814         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
17815         * config/riscv/riscv.c (struct machine_function): Add
17816         interrupt_handler_p and attribute_checked_p fields.
17817         (riscv_attribute_table): Add interrupt.
17818         (riscv_interrupt_type_p): New.
17819         (riscv_save_reg_p): Save extra regs for interrupt handler.
17820         (riscv_use_save_libcall): Return false  for interrupt handler.
17821         (riscv_first_stack_step): Add forward declaration.
17822         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
17823         for interrupt handler with large frame.  Use it for saved reg list.
17824         (riscv_expand_prologue): Move flag_stack_usage_info support to
17825         eliminate duplication.
17826         (riscv_expand_epilogue): Generate mret for interrupt handler.
17827         (riscv_epilogue_uses): New.
17828         (riscv_can_use_return_insn): Return false for interrupt handler.
17829         (riscv_function_ok_for_sibcall): Likewise.
17830         (riscv_set_current_function): Add interrupt handler support.
17831         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
17832         * config/riscv/riscv.md (UNSPECV_MRET): New.
17833         (GP_REGNUM): New.
17834         (riscv_frflags, riscv_fsflags): Use tab after opcode.
17835         (riscv_mret): New.
17836         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
17838 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
17840         PR tree-optimization/85712
17841         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
17842         this candidate has already been replaced in-situ by a copy.
17844 2018-05-25  Jason Merrill  <jason@redhat.com>
17846         PR c++/80485 - inline function non-zero address.
17847         * symtab.c (nonzero_address): Check DECL_COMDAT.
17849 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
17851         PR target/83628
17852         * config/alpha/alpha.md (ashlsi3): New insn pattern.
17853         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
17854         extension of SImode operation.  Use const123_operand predicate.
17855         (*saddsi_1): Remove.
17856         (*saddl_se_1): Ditto.
17857         (*ssubsi_1): Ditto.
17858         (*ssubl_se_1): Ditto.
17859         * config/alpha/predicates.md (const123_operand): New predicate.
17860         * config/alpha/constraints.md (P): Use IN_RANGE.
17862 2018-05-25  Richard Biener  <rguenther@suse.de>
17864         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
17865         defaulted to true.
17866         (ref_maybe_used_by_stmt_p): Likewise.
17867         (stmt_may_clobber_ref_p): Likewise.
17868         (stmt_may_clobber_ref_p_1): Likewise.
17869         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
17870         and pass it along.
17871         (ref_maybe_used_by_stmt_p): Likewise.
17872         (stmt_may_clobber_ref_p): Likewise.
17873         (stmt_may_clobber_ref_p_1): Likewise.
17874         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
17875         the alias oracle to disambiguate DRs with stmts DR analysis
17876         couldn't handle.
17877         (vect_analyze_data_refs): Do not give up on not analyzable
17878         DRs for BB vectorization.  Remove code truncating the dataref
17879         vector.
17881 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
17883         PR target/85832
17884         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17885         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
17886         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
17888 2018-05-25  Richard Biener  <rguenther@suse.de>
17890         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
17891         function, combining stmt data ref gathering and fatal analysis
17892         parts.
17893         (vect_analyze_data_refs): Remove now redudnant code and simplify.
17894         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
17895         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
17896         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
17897         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
17899 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17901         PR tree-optimization/85720
17902         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
17903         SCC if all partitions are builtins.
17904         (version_loop_by_alias_check): New parameter.  Generate cancelable
17905         runtime alias check if all partitions are builtins.
17906         (distribute_loop): Update call to above function.
17908 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17910         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
17911         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
17912         (parm_default_def_partition_arg): Ditto.
17913         (set_parm_default_def_partition): Ditto.
17914         (get_parm_default_def_partitions): Ditto and make it static.
17915         (get_undefined_value_partitions): Ditto and make it static.
17916         (remove_ssa_form): Refactor call to init_var_map here.
17917         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
17918         computation for loop region.
17919         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
17920         (register_default_def): Delete.
17921         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
17922         (parm_default_def_partition_arg): Ditto.
17923         (set_parm_default_def_partition): Ditto.
17924         (get_parm_default_def_partitions): Ditto and make it static.
17925         (get_undefined_value_partitions): Ditto and make it static.
17926         (coalesce_with_default, coalesce_with_default): Update comment.
17927         (create_coalesce_list_for_region): New func factored out from
17928         create_outofssa_var_map.
17929         (populate_coalesce_list_for_outofssa): New func factored out from
17930         create_outofssa_var_map and coalesce_ssa_name.
17931         (create_outofssa_var_map): Delete.
17932         (coalesce_ssa_name): Refactor to support live range computation.
17933         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
17934         (get_parm_default_def_partitions): Delete.
17935         (get_undefined_value_partitions): Ditto.
17936         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
17937         computation for loop region.
17938         (new_tree_live_info, loe_visit_block): Ditto.
17939         (live_worklist, set_var_live_on_entry): Ditto.
17940         (calculate_live_on_exit, verify_live_on_entry): Ditto.
17941         * tree-ssa-live.h (struct _var_map): New fields.
17942         (init_var_map): Change decl.
17943         (region_contains_p): New.
17945 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17947         * tree-ssa-live.h (live_merge_and_clear): Delete.
17949 2018-05-25  Richard Biener  <rguenther@suse.de>
17951         PR c++/85912
17952         * tree-dump.c (dequeue_and_dump): Remove access to removed
17953         operand 2 of a SWITCH_EXPR.
17955 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17957         * doc/sourcebuild.texi (vect_double_cond_arith): Include
17958         multiplication and division.
17959         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
17960         (cond_udiv@var{m}, cond_umod@var{m}): Document.
17961         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
17962         (cond_udiv_optab, cond_umod_optab): New optabs.
17963         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
17964         (IFN_COND_RDIV): New internal functions.
17965         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
17966         TRUNC_MOD_EXPR and RDIV_EXPR.
17967         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
17968         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
17969         New unspecs.
17970         (SVE_INT_BINARY): Include mult.
17971         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
17972         (optab, sve_int_op): Handle mult.
17973         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
17974         UNSPEC_COND_DIV.
17975         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
17976         for SVE_INT_BINARY_SD.
17978 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17980         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
17981         (optab, sve_int_op): Handle div and udiv.
17982         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
17983         for SVE_INT_BINARY_SD.
17984         (*<optab><mode>3): New insn for the same.
17986 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17988         * tree-vect-patterns.c: Include predict.h.
17989         (vect_recog_divmod_pattern): Restrict check for division support
17990         to when optimizing for size.
17992 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17994         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
17995         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
17996         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
17997         (gimple_match_op::set_op): Likewise.
17998         (gimple_resimplify4): Declare.
17999         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18000         (expr::gen_transform): Likewise.
18001         (decision_tree::gen): Generate a simplification routine for 4 operands.
18002         * gimple-match-head.c (gimple_simplify): Add an overload for
18003         4 operands.  In the top-level function, handle up to 4 call
18004         arguments and call gimple_resimplify4.
18005         (gimple_resimplify4): New function.
18006         (build_call_internal): Pass a fourth operand.
18007         (maybe_push_to_seq): Likewise.
18008         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18009         Fold VEC_COND_EXPRs of an operation and a default value into
18010         an IFN_COND_* function if possible.
18011         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18012         New unspecs.
18013         (SVE_COND_FP_BINARY): Include them.
18014         (optab, sve_fp_op): Handle them.
18015         (SVE_INT_BINARY_REV): New code iterator.
18016         (SVE_COND_FP_BINARY_REV): New int iterator.
18017         (commutative): New int attribute.
18018         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18019         Declare.
18020         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18021         function.
18022         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18023         (*cond_<optab><mode>): New patterns for reversed operands.
18025 2018-05-25  Richard Biener  <rguenther@suse.de>
18027         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18028         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18029         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18030         (STMT_VINFO_GROUPED_ACCESS): Adjust.
18031         * tree-vect-data-refs.c (everywhere): Adjust users.
18032         * tree-vect-loop.c (everywhere): Likewise.
18033         * tree-vect-slp.c (everywhere): Likewise.
18034         * tree-vect-stmts.c (everywhere): Likewise.
18035         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18037 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18039         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18040         Rename to...
18041         (gcc_cv_as_section_exclude): ... this.
18042         Try Solaris as #exclude syntax.
18043         * configure: Regenerate.
18044         * config.in: Regenerate.
18045         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18046         SECTION_EXCLUDE.
18047         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18048         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18050         * varasm.c (default_elf_asm_named_section): Don't check if
18051         HAVE_GAS_SECTION_EXCLUDE is defined.
18053 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18055         * doc/md.texi: Update the documentation of the cond_* optabs
18056         to mention the new final operand.  Fix GET_MODE_NUNITS call.
18057         Describe the scalar case too.
18058         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18059         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18060         instead of 2.
18061         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18062         (get_conditional_internal_fn): Update comment.
18063         * tree-vect-loop.c (vectorizable_reduction): Pass the original
18064         accumulator value as a final argument to conditional functions.
18065         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18066         a define_expand and add an "else" operand.  Assert for now that
18067         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
18068         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18069         (*cond_<optab><mode>): New patterns.
18070         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18071         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18072         (UNSPEC_COND_EOR): Delete.
18073         (optab): Remove associated mappings.
18074         (SVE_INT_BINARY): New code iterator.
18075         (sve_int_op): Remove int attribute and add "minus" to the code
18076         attribute.
18077         (SVE_COND_INT_OP): Delete.
18078         (SVE_COND_FP_OP): Rename to...
18079         (SVE_COND_FP_BINARY): ...this.
18081 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18083         * optabs.c (can_reuse_operands_p): New function.
18084         (maybe_legitimize_operands): Try to reuse the results for
18085         earlier operands.
18087 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18089         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18090         Add {q} suffix to insn mnemonic.
18092 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18094         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18095         (msp430_warn_func_return): New.
18097 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18099         * fold-const.c (tree_nonzero_bits): New function.
18100         * fold-const.h (tree_nonzero_bits): Likewise.
18101         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18102         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18104 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18106         PR target/85900
18107         PR target/85345
18108         * varasm.c (assemble_alias): Check ifunc_resolver only on
18109         FUNCTION_DECL.
18111 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18113         PR target/85903
18114         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18115         when memory input operand is handled.
18117 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18119         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18120         global.
18121         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18123 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18125         * match.pd: Delay FMA folds until after vectorization.
18127 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18129         PR target/83009
18130         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18131         address check not strict.
18133 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18135         * gimple-match.h (gimple_match_op): New class.
18136         (mprts_hook): Replace parameters with a gimple_match_op *.
18137         (maybe_build_generic_op): Likewise.
18138         (gimple_simplified_result_is_gimple_val): Replace parameters with
18139         a const gimple_match_op *.
18140         (gimple_simplify): Replace code_helper * and tree * parameters with
18141         a gimple_match_op * parameter.
18142         (gimple_resimplify1): Replace code_helper *, tree and tree *
18143         parameters with a gimple_match_op * parameter.
18144         (gimple_resimplify2): Likewise.
18145         (gimple_resimplify3): Likewise.
18146         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18147         parameters with a gimple_match_op * parameter.
18148         * gimple-match-head.c (gimple_simplify): Change prototypes of
18149         auto-generated functions to take a gimple_match_op * instead of
18150         separate code_helper * and tree * parameters.  Make the same
18151         change in the top-level overload and update calls to the
18152         gimple_resimplify routines.  Update calls to the auto-generated
18153         functions and to maybe_push_res_to_seq in the publicly-facing
18154         operation-specific gimple_simplify overloads.
18155         (gimple_match_op::MAX_NUM_OPS): Define.
18156         (gimple_resimplify1): Replace rcode and ops with a single res_op
18157         parameter.  Update call to gimple_simplify.
18158         (gimple_resimplify2): Likewise.
18159         (gimple_resimplify3): Likewise.
18160         (mprts_hook): Replace parameters with a gimple_match_op *.
18161         (maybe_build_generic_op): Likewise.
18162         (build_call_internal): Replace type, nargs and ops with
18163         a gimple_match_op *.
18164         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18165         with a single gimple_match_op *.  Update calls to mprts_hook,
18166         build_call_internal and gimple_simplified_result_is_gimple_val.
18167         Factor out code that is common to the tree_code and combined_fn cases.
18168         * genmatch.c (expr::gen_transform): Replace tem_code and
18169         tem_ops with a gimple_match_op called tem_op.  Update calls
18170         to the gimple_resimplify functions and maybe_push_res_to_seq.
18171         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18172         res_ops.  Update call to the gimple_resimplify functions.
18173         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18174         (decision_tree::gen): Make the functions take a gimple_match_op *
18175         called res_op instead of separate res_code and res_ops parameters.
18176         Update call accordingly.
18177         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18178         and ops with a single res_op parameter.  Update calls to
18179         maybe_build_generic_op and maybe_push_res_to_seq.
18180         (fold_stmt_1): Update calls to gimple_simplify and
18181         replace_stmt_with_simplification.
18182         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18183         and gimple_simplified_result_is_gimple_val.
18184         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18185         gimple_simplify.
18186         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18187         with a gimple_match_op *.
18188         (vn_nary_build_or_lookup): Likewise.  Update call to
18189         vn_nary_build_or_lookup_1.
18190         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18191         gimple_match_op *.  Update calls to the gimple_resimplify routines
18192         and to gimple_simplified_result_is_gimple_val.
18193         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18194         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18195         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18196         (visit_nary_op): Likewise.
18197         (visit_reference_op_load): Likewise.
18199 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18201         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18202         modifier for printing the step amount.
18204 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18206         PR target/78849
18207         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18208         types.
18210 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18212         * doc/sourcebuild.texi (Endianness): New subsubsection.
18214 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18216         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18217         <prefetch_dynamic_strides>: New const bool field.
18218         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18219         prefetch_dynamic_strides.
18220         (exynosm1_prefetch_tune): Likewise.
18221         (thunderxt88_prefetch_tune): Likewise.
18222         (thunderx_prefetch_tune): Likewise.
18223         (thunderx2t99_prefetch_tune): Likewise.
18224         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18225         false.
18226         (aarch64_override_options_internal): Update to set
18227         PARAM_PREFETCH_DYNAMIC_STRIDES.
18228         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18229         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18230         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18231         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18232         prefetch-dynamic-strides setting.
18234 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18236         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18237         <minimum_stride>: New const int field.
18238         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18239         minimum_stride field defaulting to -1.
18240         (exynosm1_prefetch_tune): Likewise.
18241         (thunderxt88_prefetch_tune): Likewise.
18242         (thunderx_prefetch_tune): Likewise.
18243         (thunderx2t99_prefetch_tune): Likewise.
18244         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18245         <default_opt_level>: Set to 3.
18246         (aarch64_override_options_internal): Update to set
18247         PARAM_PREFETCH_MINIMUM_STRIDE.
18248         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18249         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18250         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18251         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18252         stride is constant and is below the minimum stride threshold.
18254 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18256         * config/arm/arm-cpus.in (mode26): Delete.
18257         (armv4): Delete mode26 reference.
18258         * config/arm/arm.c (arm_configure_build_target): Delete use of
18259         isa_bit_mode26.
18261 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18263         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18264         New insn pattern.
18265         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18266         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18267         for non-SSE modes.
18268         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18269         (floatunsdidf2): Ditto.
18271 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18273         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18274         (fixuns_trunc<mode>si2_avx512f): Ditto.
18275         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18276         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18277         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18279 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18281         PR rtl-optimization/79985
18282         * df-scan.c (df_insn_refs_collect): Remove special case for
18283         global registers and asm statements.
18285 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18287         * extend.texi (Global Register Variables): Rewrite the bullet list.
18288         Note that the register is available for allocation. Note that access
18289         via inline asm must use constraints. Add note about async-signal
18290         handlers. Remove paragraph about automagic register selection.
18292 2018-05-23  Richard Biener  <rguenther@suse.de>
18294         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18295         of fixed offset from memset VN.
18297 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18299         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18300         first_interp field.
18301         (alloc_cand_and_find_basis): Initialize first_interp field.
18302         (slsr_process_mul): Modify first_interp field.
18303         (slsr_process_add): Likewise.
18304         (slsr_process_cast): Modify first_interp field for each new
18305         interpretation.
18306         (slsr_process_copy): Likewise.
18307         (dump_candidate): Dump first_interp field.
18308         (replace_mult_candidate): Process all interpretations, not just
18309         subsequent ones.
18310         (replace_rhs_if_not_dup): Likewise.
18311         (replace_one_candidate): Likewise.
18313 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18315         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18316         Add new boolean.
18317         (aarch64_needs_frame_chain): New function.
18318         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18320 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18322         PR target/84882
18323         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18324         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18325         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18326         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18327         as true for strict-align.
18328         (aarch64_can_inline_p): Perform checks even when callee has no
18329         attributes to check for strict alignment.
18330         * doc/extend.texi (AArch64 Function Attributes): Document
18331         no-strict-align.
18332         * doc/invoke.texi: (AArch64 Options): Likewise.
18334 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18336         PR tree-optimization/85853
18337         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18338         the handling of the root of the node to...
18339         (vect_slp_analyze_node_operations_1): ...this new function,
18340         and run the whole thing with the child nodes' def types
18341         set according to their SLP node's def type.
18343 2018-05-23  Richard Biener  <rguenther@suse.de>
18345         PR middle-end/85874
18346         * tree-data-ref.c (create_runtime_alias_checks): Defer
18347         and ignore overflow warnings.
18349 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18351         PR tree-optimization/85822
18352         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18353         constants.
18355 2018-05-23  Richard Biener  <rguenther@suse.de>
18357         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18358         memset constants via native_interpret_expr.
18360 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18362         PR target/85345
18363         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18364         attribute.
18365         (cgraph_node::create_alias): Likewise.
18366         (cgraph_node::get_availability): Check ifunc_resolver instead
18367         of looking up ifunc attribute.
18368         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18369         * varasm.c (do_assemble_alias): Likewise.
18370         (assemble_alias): Likewise.
18371         (default_binds_local_p_3): Likewise.
18372         * cgraph.h (cgraph_node): Add ifunc_resolver.
18373         (cgraph_node::only_called_directly_or_aliased_p): Return false
18374         for IFUNC resolver.
18375         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18376         attribute.
18377         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18378         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18379         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18380         instead of looking up ifunc attribute.
18382 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18384         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18386 2018-05-22  Martin Sebor  <msebor@redhat.com>
18388         PR middle-end/85359
18389         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18390         only when expasion succeeds.
18391         (expand_builtin_strcmp): Same.
18392         (expand_builtin_strncmp): Same.
18394 2018-05-22  Martin Sebor  <msebor@redhat.com>
18396         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18398 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18399             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18401         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18402         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18403         code.
18404         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18405         Add check for legitimate_address.
18406         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18407         (aarch64_swap_ldrstr_operands): New.
18408         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18409         Define prototype.
18411 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18412             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18414         * config/aarch64/aarch64.md: New patterns to generate stp
18415         and ldp.
18416         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18417         single words and double words.
18418         (load_pair_sw, load_pair_dw): Likewise.
18419         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18420         Delete.
18421         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18422         Delete.
18423         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18424         for different mode ldpstp and add peephole for merged zero stores.
18425         Likewise for loads.
18426         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18427         Add size check.
18428         (aarch64_gen_store_pair): Rename calls to match new patterns.
18429         (aarch64_gen_load_pair): Rename calls to match new patterns.
18430         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18431         (load_pair<DREG:mode><DREG2:mode>): ... This.
18432         (store_pair<mode>): Rename to...
18433         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18434         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18435         New mode iterators.
18436         (V_INT_EQUIV): Handle SImode.
18437         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18438         New predicate.
18440 2018-05-22  Martin Sebor  <msebor@redhat.com>
18442         PR c/85623
18443         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18444         or ajust the presumed bound on an operation to avoid unnecessary
18445         warnings.
18447 2018-05-22  Martin Sebor  <msebor@redhat.com>
18449         PR tree-optimization/85826
18450         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18451         assuming that a DECL necesarily has a constant size.
18453 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18455         PR middle-end/85862
18456         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18458 2018-05-22  Richard Biener  <rguenther@suse.de>
18460         PR tree-optimization/85834
18461         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18462         non-constant and non-zero memset arguments.
18464 2018-05-22  Martin Liska  <mliska@suse.cz>
18466         PR ipa/85607
18467         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18469 2018-05-22  Richard Biener  <rguenther@suse.de>
18471         PR tree-optimization/85863
18472         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18473         comparisons when vectype is specified.
18474         (vectorizable_condition): Do not specify vectype for
18475         vect_is_simple_cond when SLP vectorizing.
18477 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18479         PR target/85657
18480         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18481         define __ibm128 as long double.
18482         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18483         as a distinct type when IEEE 128-bit support is enabled.
18484         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18485         128-bit types to use the correct functions.
18486         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18487         convert between 128-bit floating point types that have different
18488         modes but the same representation, instead of using gen_lowpart to
18489         makean alias.
18490         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18491         KFmode.
18492         (IFKF_reg): New attributes to give the register constraints for
18493         IFmode and KFmode.
18494         (extend<mode>tf2_internal): New insns to mark an explicit
18495         conversion between 128-bit floating point types that have a
18496         different mode but share the same representation.
18498 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18500         PR tree-optimization/85814
18501         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18502         a null return from get_strinfo when unsharing the next
18503         strinfo in the chain.
18505 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18507         PR gcc/84923
18508         * varasm.c (weak_finish): Clean up weak_decls.
18510 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18512         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18513         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18514         UNSPEC_UADALP values.
18515         * config/aarch64/iterators.md (ABAL): New int iterator.
18516         (ABDL2): Likewise.
18517         (ADALP): Likewise.
18518         (sur): Add mappings for the above.
18519         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18520         New define_insn.
18521         (aarch64_<sur>abal<mode>_4): Likewise.
18522         (aarch64_<sur>adalp<mode>_3): Likewise.
18523         (<sur>sadv16qi): New define_expand.
18525 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18527         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18528         (*movdf_internal): Ditto.
18529         (*rcpsf2_sse): Ditto.
18530         (*rsqrtsf2_sse): Ditto.
18531         (*sqrt<mode>2_sse): Ditto.
18533 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18535         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18536         eor3q<mode>4.
18537         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18538         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18539         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18540         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18541         vbcaxq_s64): New.
18542         * config/aarch64/arm_neon.h: Likewise.
18543         * config/aarch64/iterators.md (VQ_I): New.
18545 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18547         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18548         arc*-*-linux*.
18549         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18550         MULTILIB_DIRNAMES
18552 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18554         * config/nds32/constraints.md (S): New constraint.
18555         * config/nds32/nds32.md (call_internal): Use constraint S.
18556         (call_value_internal): Likewise.
18557         (sibcall_internal): Likewise.
18558         (sibcall_value_internal): Likewise.
18560 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18561             Chung-Ju Wu  <jasonwucj@gmail.com>
18563         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18564         into consideration.
18566 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18567             Chung-Ju Wu  <jasonwucj@gmail.com>
18569         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18570         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18571         (nds32_rtx_costs_impl): Simplify.
18572         (nds32_address_cost_impl): Simplify.
18573         (nds32_init_rtx_costs): New function.
18574         (nds32_rtx_costs_speed_prefer): Likewise.
18575         (nds32_rtx_costs_size_prefer): Likewise.
18576         (nds32_address_cost_speed_prefer): Likewise.
18577         (nds32_address_cost_speed_fwprop): Likewise.
18578         (nds32_address_cost_size_prefer): Likewise.
18579         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
18580         * config/nds32/nds32.c (nds32_option_override): Use
18581         nds32_init_rtx_costs function.
18583 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18585         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
18586         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
18587         (TARGET_PIPELINE_N8): Likewise.
18588         (TARGET_PIPELINE_N10): Likewise.
18589         (TARGET_PIPELINE_N13): Likewise.
18590         (TARGET_PIPELINE_GRAYWOLF): Likewise.
18592 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
18594         * config/nds32/nds32-fpu.md: Update copyright year.
18596 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18598         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18600 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18602         * config/nds32/nds32.c
18603         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
18604         * config/nds32/nds32.opt (minline-asm-r15): New option.
18606 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18608         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
18609         MASK_HW_ABS.
18610         * config/nds32/nds32.md (abssi2): New pattern.
18612 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
18614         * config/i386/i386.md (rex64namesuffix): New mode attribute.
18615         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
18616         Merge insn pattern from sse_cvtsi2ss<round_name> and
18617         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
18618         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
18619         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
18620         using SWI48 mode iterator.
18621         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
18622         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
18623         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
18624         pattern from sse_cvttss2si<round_saeonly_name>
18625         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
18626         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
18627         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
18628         using SWI48 mode iterator.
18629         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
18630         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
18631         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
18632         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
18633         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
18634         using SWI48 mode iterator.
18635         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
18636         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
18637         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
18638         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
18639         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
18640         SWI48 mode iterator.
18641         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
18642         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
18643         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
18644         pattern from sse_cvttsd2si<round_saeonly_name>
18645         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
18647 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18649         * config/nds32/nds32-md-auxiliary.c
18650         (nds32_valid_smw_lwm_base_p): Refine.
18651         (nds32_output_smw_single_word): Refine.
18652         (nds32_output_smw_double_word): New.
18653         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
18655 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18657         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
18658         (nds32_output_stack_pop): Refine.
18659         (nds32_expand_unaligned_load): Refine.
18660         (nds32_expand_unaligned_store): Refine.
18662 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18663             Chung-Ju Wu  <jasonwucj@gmail.com>
18665         * config/nds32/constants.md: Add TP_REGNUM constant.
18666         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
18667         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
18668         UNSPEC_ADD32.
18669         * config/nds32/nds32-doubleword.md: Consider flag_pic.
18670         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
18671         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
18672         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
18673         and PIC code generation.
18674         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
18675         code generation.
18676         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
18677         optimization.
18678         * config/nds32/nds32.md: Support TLS and PIC.
18679         * config/nds32/nds32.c: Support TLS and PIC.
18680         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
18681         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
18682         predicate.
18684 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18686         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
18687         mode with E_ prefix.
18689 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18690             Chung-Ju Wu  <jasonwucj@gmail.com>
18692         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
18693         * config/nds32/nds32-md-auxiliary.c
18694         (symbolic_reference_mentioned_p): New.
18695         (nds32_legitimize_ict_address): New.
18696         (nds32_expand_ict_move): New.
18697         (nds32_indirect_call_referenced_p): New.
18698         (nds32_symbol_binds_local_p): Delete.
18699         (nds32_long_call_p): Modify.
18700         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18701         * config/nds32/nds32-protos.h
18702         (symbolic_reference_mentioned_p): Declare.
18703         (nds32_legitimize_ict_address): Declare.
18704         (nds32_expand_ict_move): Declare.
18705         (nds32_indirect_call_referenced_p): Declare.
18706         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
18707         (nds32_relax_group): Use nds32_ict_const_p as condition.
18708         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
18709         (nds32_asm_file_start): Output ict_model directive in asm code.
18710         (nds32_legitimate_address_p): Consider indirect call.
18711         (nds32_print_operand): Consider indirect call.
18712         (nds32_print_operand_address): Consider indirect call.
18713         (nds32_insert_attributes): Handle "indirect_call" attribute.
18714         (TARGET_LEGITIMATE_ADDRESS_P): Define.
18715         (TARGET_LEGITIMATE_CONSTANT_P): Define.
18716         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
18717         (TARGET_DELEGITIMIZE_ADDRESS): Define.
18718         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
18719         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
18720         (TARGET_ICT_MODEL_SMALL): Define.
18721         (TARGET_ICT_MODEL_LARGE): Define.
18722         * config/nds32/nds32.md (movsi): Consider ict model.
18723         (call, call_value): Consider ict model.
18724         (sibcall, sibcall_value): Consider ict model.
18725         * config/nds32/nds32.opt (mict-model): New option.
18726         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
18727         model.
18729 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
18730             Monk Chiang  <sh.chiang04@gmail.com>
18731             Jim Wilson <jimw@sifive.com>
18733         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
18734         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
18735         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
18736         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
18737         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
18738         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
18739         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
18740         compute save_libcall_adjustment properly.
18741         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
18742         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
18743         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
18744         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
18745         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
18746         (ABI_SPEC): Handle mabi=ilp32e.
18747         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
18748         (RVE): Add RVE mask.
18749         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
18750         <-march>: Add rv32e as an example.
18752 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
18754         PR c++/82899
18755         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
18756         (intra_create_variable_infos): Handle C++ constructors.
18758 2018-05-18  Martin Liska  <mliska@suse.cz>
18760         * passes.def: Remove a redundant pass.
18762 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
18764         PR bootstrap/85838
18765         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
18767 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18769         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
18770         (ARMv4): Update.
18771         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
18772         (ARMv6m): Update.
18773         (armv2, armv2a, armv3, armv3m): Delete architectures.
18774         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
18775         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
18776         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
18777         Delete cpus.
18778         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
18779         (*mulsidi3adddi): Likewise.
18780         (mulsidi3): Likewise.
18781         (*mulsidi3_nov6): Likewise.
18782         (umulsidi3): Likewise.
18783         (umulsidi3_nov6): Likewise.
18784         (umaddsidi4): Likewise.
18785         (*umulsidi3adddi): Likewise.
18786         (smulsi3_highpart): Likewise.
18787         (*smulsi3_highpart_nov6): Likewise.
18788         (umulsi3_highpart): Likewise.
18789         (*umulsi3_highpart_nov6): Likewise.
18790         * config/arm/arm.h (arm_arch3m): Delete.
18791         * config/arm/arm.c (arm_arch3m): Delete.
18792         (arm_option_override_internal): Update armv3-related comment.
18793         (arm_configure_build_target): Delete use of isa_bit_mode32.
18794         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
18795         (arm_rtx_costs_internal): Delete check of arm_arch3m.
18796         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
18797         (mulsa3): Likewise.
18798         (mulusa3): Likewise.
18799         * config/arm/arm-protos.h (arm_arch3m): Delete.
18800         * config/arm/arm-tables.opt: Regenerate.
18801         * config/arm/arm-tune.md: Likewise.
18802         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
18803         deleted architectures.
18805 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18807         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
18808         (armv5t, armv5te): New features.
18809         (ARMv5, ARMv5e): Delete fgroups.
18810         (ARMv5t, ARMv5te): Adjust for above changes.
18811         (ARMv6m): Likewise.
18812         (armv5, armv5e): Delete arches.
18813         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
18814         arm_arch5.
18815         (*call_reg_arm): Likewise.
18816         (*call_value_reg_armv5): Likewise.
18817         (*call_value_reg_arm): Likewise.
18818         (*call_symbol): Likewise.
18819         (*call_value_symbol): Likewise.
18820         (*sibcall_insn): Likewise.
18821         (*sibcall_value_insn): Likewise.
18822         (clzsi2): Likewise.
18823         (prefetch): Likewise.
18824         (define_split and define_peephole2 dependent on arm_arch5):
18825         Likewise.
18826         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
18827         arm_arch5e.
18828         (TARGET_ARM_QBIT): Likewise.
18829         (TARGET_DSP_MULTIPLY): Likewise.
18830         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
18831         (arm_arch5, arm_arch5e): Delete.
18832         (arm_arch5t, arm_arch5te): Declare.
18833         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
18834         (arm_arch5t): Declare.
18835         (arm_option_reconfigure_globals): Update for the above.
18836         (arm_options_perform_arch_sanity_checks): Update comment, replace
18837         use of arm_arch5 with arm_arch5t.
18838         (use_return_insn): Likewise.
18839         (arm_emit_call_insn): Likewise.
18840         (output_return_instruction): Likewise.
18841         (arm_final_prescan_insn): Likewise.
18842         (arm_coproc_builtin_available): Likewise.
18843         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
18844         arm_arch5e with arm_arch5t and arm_arch5te.
18845         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
18846         (arm_arch5t, arm_arch5te): Declare.
18847         * config/arm/arm-tables.opt: Regenerate.
18848         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
18849         * config/arm/t-multilib: Likewise.
18850         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
18851         instead of arm_arch5.
18852         (*call_reg_thumb1): Likewise.
18853         (*call_value_reg_thumb1_v5): Likewise.
18854         (*call_value_reg_thumb1): Likewise.
18855         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
18856         unreachable path.
18857         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
18859 2018-05-18  Martin Liska  <mliska@suse.cz>
18861         PR gcov-profile/84846
18862         * doc/gcov.texi: Document -t option of gcov tool.
18864 2018-05-18  Martin Liska  <mliska@suse.cz>
18866         PR gcov-profile/84846
18867         * gcov.c (print_usage): Add new -t option.
18868         (process_args): Handle the option.
18869         (generate_results): Use stdout as output when requested by
18870         the option.
18872 2018-05-18  Martin Liska  <mliska@suse.cz>
18874         PR gcov-profile/84846
18875         * coverage.c (coverage_init): Write PWD to .gcno file.
18876         * doc/gcov.texi: Document how working directory is printed.
18877         * gcov-dump.c (dump_gcov_file): Print PWD.
18878         * gcov.c (output_intermediate_file): Likewise.
18879         (read_graph_file): Read PWD string.
18880         (output_lines): Print PWD.
18882 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18884         PR middle-end/85817
18885         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
18886         for retval and return false if all args to phi are zero.
18888 2018-05-18  Richard Biener  <rguenther@suse.de>
18890         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
18891         method.
18892         (evrp_dom_walker::before_dom_children): Call it.
18894 2018-05-18  Richard Biener  <rguenther@suse.de>
18896         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
18897         results when processing array refs with variable index.
18899 2018-05-18  Toon Moene  <toon@moene.org>
18901         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
18902         directly after that of -floop-interchange. Indicate that both
18903         options are enabled by default when specifying -O3.
18905 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18907         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
18908         iterator.  Delete separate integer-mode vec_set<mode> expander.
18909         (aarch64_simd_vec_setv2di): Delete.
18910         (vec_setv2di): Delete.
18911         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
18912         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
18913         the "w, r" alternative.
18915 2018-05-18  Martin Liska  <mliska@suse.cz>
18917         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
18918         * tree-pass.h (make_pass_lower_switch_O0): New function.
18919         * tree-switch-conversion.c (node_has_low_bound): Remove.
18920         (node_has_high_bound): Likewise.
18921         (node_is_bounded): Likewise.
18922         (class pass_lower_switch): Make it a template type and create
18923         two instances.
18924         (pass_lower_switch::execute): Add template argument.
18925         (make_pass_lower_switch): New function.
18926         (make_pass_lower_switch_O0): New function.
18927         (do_jump_if_equal): Remove.
18928         (emit_case_nodes): Simplify to just handle all 3 cases and leave
18929         all the hard work to tree optimization passes.
18931 2018-05-18  Martin Liska  <mliska@suse.cz>
18933         * dbgcnt.c (limit_low): Renamed from limit.
18934         (limit_high): New variable.
18935         (dbg_cnt_is_enabled): Check for upper limit.
18936         (dbg_cnt): Adjust dumping.
18937         (dbg_cnt_set_limit_by_index): Add new argument for high
18938         value.
18939         (dbg_cnt_set_limit_by_name): Likewise.
18940         (dbg_cnt_process_single_pair): Parse new format.
18941         (dbg_cnt_process_opt): Use strtok.
18942         (dbg_cnt_list_all_counters): Remove 'value' and add
18943         'limit_high'.
18944         * doc/invoke.texi: Document changes.
18946 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
18948         * doc/sourcebuild.texi (scalar_all_fma): Document.
18949         * tree.def (FMA_EXPR): Delete.
18950         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
18951         * internal-fn.c (ternary_direct): New macro.
18952         (expand_ternary_optab_fn): Likewise.
18953         (direct_ternary_optab_supported_p): Likewise.
18954         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
18955         * builtins.c (fold_builtin_fma): Delete.
18956         (fold_builtin_3): Don't call it.
18957         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
18958         * expr.c (expand_expr_real_2): Likewise.
18959         * fold-const.c (operand_equal_p): Likewise.
18960         (fold_ternary_loc): Likewise.
18961         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
18962         * gimple.c (DEFTREECODE): Likewise.
18963         * gimplify.c (gimplify_expr): Likewise.
18964         * optabs-tree.c (optab_for_tree_code): Likewise.
18965         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18966         * tree-eh.c (operation_could_trap_p): Likewise.
18967         (stmt_could_throw_1_p): Likewise.
18968         * tree-inline.c (estimate_operator_cost): Likewise.
18969         * tree-pretty-print.c (dump_generic_node): Likewise.
18970         (op_code_prio): Likewise.
18971         * tree-ssa-loop-im.c (stmt_cost): Likewise.
18972         * tree-ssa-operands.c (get_expr_operands): Likewise.
18973         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
18974         * fold-const-call.h (fold_fma): Delete.
18975         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
18976         CFN_FNMA and CFN_FNMS.
18977         (fold_fma): Delete.
18978         * genmatch.c (combined_fn): New enum.
18979         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
18980         (commutative_op): New function.
18981         (commutate): Use it.  Handle more than 2 operands.
18982         (dt_operand::gen_gimple_expr): Use commutative_op.
18983         (parser::parse_expr): Allow :c to be used with non-binary
18984         operators if the commutative operand is known.
18985         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
18986         CFN_FMS, CFN_FNMA and CFN_FNMS.
18987         (backprop::process_assign_use): Remove FMA_EXPR handling.
18988         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
18989         (gen_hsa_fma): New function.
18990         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
18991         IFN_FNMA and IFN_FNMS.
18992         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
18993         * gimple-fold.h (follow_all_ssa_edges): Declare.
18994         * gimple-fold.c (follow_all_ssa_edges): New function.
18995         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
18996         gimple_build interface and use follow_all_ssa_edges to fold the result.
18997         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
18998         instead of checking for optabs directly.
18999         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19000         rather than FMA_EXPRs.
19001         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19002         call to IFN_FMA instead of an FMA_EXPR.
19004 2018-05-17  Jim Wilson  <jimw@sifive.com>
19006         * expr.c (do_tablejump): When converting index to Pmode, if we have a
19007         sign extended promoted subreg, and the range does not have the sign bit
19008         set, then do a sign extend.
19010         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19011         test, check for sign extended subreg and/or constant operands, and
19012         do a sign extend in that case.
19014 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
19016         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19017         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19018         Add untyped.
19019         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19020         Change logics_shift_reg to logics_shift_imm.
19021         (thunderx2t99_fp_loadpair_basic): Delete.
19022         (thunderx2t99_fp_storepair_basic): Delete.
19023         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19024         (thunderx2t99_asimd_polynomial): Delete.
19025         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19026         and neon_fp_mul_d_scalar_q.
19027         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19028         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19029         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19030         (thunderx2t99_asimd_lut): Add missing tbl types.
19031         (thunderx2t99_asimd_ext): Delete.
19032         (thunderx2t99_asimd_load1_1_mult): Delete.
19033         (thunderx2t99_asimd_load1_2_mult): Delete.
19034         (thunderx2t99_asimd_load1_ldp): New.
19035         (thunderx2t99_asimd_load1): New.
19036         (thunderx2t99_asimd_load2): Add missing *load2* types.
19037         (thunderx2t99_asimd_load3): New.
19038         (thunderx2t99_asimd_load4): New.
19039         (thunderx2t99_asimd_store1_1_mult): Delete.
19040         (thunderx2t99_asimd_store1_2_mult): Delete.
19041         (thunderx2t99_asimd_store2_mult): Delete.
19042         (thunderx2t99_asimd_store2_onelane): Delete.
19043         (thunderx2t99_asimd_store_stp): New.
19044         (thunderx2t99_asimd_store1): New.
19045         (thunderx2t99_asimd_store2): New.
19046         (thunderx2t99_asimd_store3): New.
19047         (thunderx2t99_asimd_store4): New.
19049 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
19051         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19052         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
19054 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
19055             Segher Boessenkool  <segher@kernel.crashing.org>
19057         PR target/85698
19058         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19059         operand.
19061 2018-05-17  Richard Biener  <rguenther@suse.de>
19063         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19064         for pruning loop and prune defs feeding only already visited PHIs.
19066 2018-05-17  Richard Biener  <rguenther@suse.de>
19068         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19070 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
19071             Richard Biener  <rguenther@suse.de>
19073         PR tree-optimization/85793
19074         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19075         for VMAT_ELEMENTWISE.
19077 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19079         * internal-fn.h (lookup_internal_fn): Declare
19080         * internal-fn.c (lookup_internal_fn): New function.
19081         * gimple.c (gimple_build_call_from_tree): Handle calls to
19082         internal functions.
19083         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19084         internal function names.
19085         * tree-pretty-print.c (dump_generic_node): Likewise.
19086         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19088 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19090         * gimple-fold.h (gimple_build): Make the function forms take
19091         combined_fn rather than built_in_function.
19092         (gimple_simplify): Likewise.
19093         * gimple-match-head.c (gimple_simplify): Likewise.
19094         * gimple-fold.c (gimple_build): Likewise.
19095         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19096         rather than gimple_build_call_internal.
19097         (get_initial_defs_for_reduction): Likewise.
19098         (vect_create_epilog_for_reduction): Likewise.
19099         (vectorizable_live_operation): Likewise.
19101 2018-05-17  Martin Liska  <mliska@suse.cz>
19103         * gimple-ssa-sprintf.c (format_directive): Do not use
19104         space in between 'G_' and '('.
19106 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19108         PR target/85323
19109         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19110         even if the mask is not all ones.
19112         PR target/85323
19113         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19114         vector.
19115         (ix86_gimple_fold_builtin): Likewise.
19117         PR target/85323
19118         * config/i386/i386.c: Include tree-vector-builder.h.
19119         (ix86_vector_shift_count): New function.
19120         (ix86_fold_builtin): Fold shift builtins by scalar count.
19121         (ix86_gimple_fold_builtin): Likewise.
19123         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19124         _mm512_setzero): New intrinsics.
19126 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19127             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19129         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19130         code generation for cases where splatting a value is not useful.
19131         * simplify-rtx.c (simplify_ternary_operation): Simplify
19132         vec_merge across a vec_duplicate and a paradoxical subreg forming
19133         a vector mode to a vec_concat.
19135 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19137         * config.gcc: Support "goldmont-plus".
19138         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19139         "goldmont-plus".
19140         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19141         PROCESSOR_GOLDMONT_PLUS.
19142         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19143         (processor_target_table): Add "goldmont-plus".
19144         (PTA_GOLDMONT_PLUS): Define.
19145         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19146         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19147         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19148         (fold_builtin_cpu): Add "goldmont-plus".
19149         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19150         (ix86_option_override_internal): Add "goldmont-plus".
19151         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19152         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19153         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19154         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19156 2018-05-17  Richard Biener  <rguenther@suse.de>
19158         PR tree-optimization/85757
19159         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19160         remove defs that only feed that PHI from further processing.
19162 2018-05-16  Jim Wilson  <jimw@sifive.com>
19164         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19165         asterisk to name.
19166         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19168 2018-05-16  Mark Wielaard  <mark@klomp.org>
19170         * dwarf2out.c (count_index_strings): New function.
19171         (output_indirect_strings): Call count_index_strings and generate
19172         header for dwarf_version >= 5.
19174 2018-05-16  Mark Wielaard  <mark@klomp.org>
19176         * dwarf2out.c (dwarf_FORM): New function.
19177         (set_indirect_string): Use dwarf_FORM.
19178         (reset_indirect_string): Likewise.
19179         (size_of_die): Likewise.
19180         (value_format): Likewise.
19181         (output_die): Likewise.
19182         (add_skeleton_AT_string): Likewise.
19183         (output_macinfo_op): Likewise.
19184         (index_string): Likewise.
19185         (output_index_string_offset): Likewise.
19186         (output_index_string): Likewise.
19187         (count_index_strings): Likewise.
19189 2018-05-16  Carl Love  <cel@us.ibm.com>
19191         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19192         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19194 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19196         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19197         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19199 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19201         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19202         (fnma<mode>4): Likewise.
19203         (fms<mode>4): Likewise.
19204         (fnms<mode>4): Likewise.
19205         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19206         (aarch64_fnma<mode>4): Likewise.
19207         (aarch64_fms<mode>4): Likewise.
19208         (aarch64_fnms<mode>4): Likewise.
19209         (aarch64_fnmadd<mode>4): Likewise.
19211 2018-05-16  Jason Merrill  <jason@redhat.com>
19213         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19215 2018-05-16  Richard Biener  <rguenther@suse.de>
19217         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19218         (dump_stmt_cost): Declare.
19219         (add_stmt_cost): Dump cost we add.
19220         (add_stmt_costs): New function.
19221         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19222         No longer exported.
19223         (vect_analyze_stmt): Adjust prototype.
19224         (vectorizable_condition): Likewise.
19225         (vectorizable_live_operation): Likewise.
19226         (vectorizable_reduction): Likewise.
19227         (vectorizable_induction): Likewise.
19228         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19229         cost vector to pass to vectorizable_ and record afterwards.
19230         (vect_model_reduction_cost): Take cost vector argument and adjust.
19231         (vect_model_induction_cost): Likewise.
19232         (vectorizable_reduction): Likewise.
19233         (vectorizable_induction): Likewise.
19234         (vectorizable_live_operation): Likewise.
19235         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19236         SLP_TREE_NUMBER_OF_VEC_STMTS.
19237         (vect_analyze_slp_cost_1): Remove.
19238         (vect_analyze_slp_cost): Likewise.
19239         (vect_slp_analyze_node_operations): Take visited args and
19240         a target cost vector.  Avoid processing already visited stmt sets.
19241         (vect_slp_analyze_operations): Use a local cost vector to gather
19242         costs and register those of non-discarded instances.
19243         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19244         (vect_schedule_slp_instance): Remove copying of
19245         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19246         zero.
19247         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19248         adding cost.  Record cost entry location.
19249         (vect_prologue_cost_for_slp_op): Function to compute cost of
19250         a constant or invariant generated for SLP vect in the prologue,
19251         split out from vect_analyze_slp_cost_1.
19252         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19253         (vect_model_promotion_demotion_cost): Likewise.
19254         (vect_model_store_cost): Likewise, make static.
19255         (vect_model_load_cost): Likewise.
19256         (vectorizable_bswap): Add cost vector arg and adjust.
19257         (vectorizable_call): Likewise.
19258         (vectorizable_simd_clone_call): Likewise.
19259         (vectorizable_conversion): Likewise.
19260         (vectorizable_assignment): Likewise.
19261         (vectorizable_shift): Likewise.
19262         (vectorizable_operation): Likewise.
19263         (vectorizable_store): Likewise.
19264         (vectorizable_load): Likewise.
19265         (vectorizable_condition): Likewise.
19266         (vectorizable_comparison): Likewise.
19267         (can_vectorize_live_stmts): Likewise.
19268         (vect_analyze_stmt): Likewise.
19269         (vect_transform_stmt): Adjust calls to vectorizable_*.
19270         * tree-vectorizer.c: Include gimple-pretty-print.h.
19271         (dump_stmt_cost): New function.
19273 2018-05-16  Richard Biener  <rguenther@suse.de>
19275         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19276         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19277         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19278         (check_name): New callback.
19279         (dse_classify_store): Track cycles via a visited bitmap of PHI
19280         defs and simplify handling of in-loop and across loop dead stores
19281         and properly fail for loop-variant refs.  Handle byte-tracking with
19282         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19283         limiting the walk.
19285 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19287         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19288         (vect_get_mask_type_for_stmt): Likewise.
19289         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19290         split out from...
19291         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19292         to determine the statement's vector type and the vector type that
19293         should be used for calculating nunits.  Deal with cases in which
19294         the type has to be deferred.
19295         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19296         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19297         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19298         (vect_determine_vf_for_stmt): New functions, split out from...
19299         (vect_determine_vectorization_factor): ...here.
19300         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19301         (vect_get_mask_type_for_stmt): New functions, split out from
19302         vect_determine_vectorization_factor.
19304 2018-05-16  Richard Biener  <rguenther@suse.de>
19306         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19307         verify the [VEC_]COND_EXPR embedded comparison.
19309 2018-05-15  Martin Sebor  <msebor@redhat.com>
19311         PR tree-optimization/85753
19312         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19313         RECORD_TYPE in addition to ARRAY_TYPE.
19315 2018-05-15  Martin Sebor  <msebor@redhat.com>
19317         PR middle-end/85643
19318         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19320 2018-05-15  Richard Biener  <rguenther@suse.de>
19322         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19323         add by_clobber_p one.  Change algorithm to collect all defs
19324         representing uses we need to walk and try reducing them to
19325         a single one before failing.
19326         (dse_dom_walker::dse_optimize_stmt): Adjust.
19328 2018-05-13  Mark Wielaard  <mark@klomp.org>
19330         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19331         (size_of_loc_descr): Likewise.
19332         (output_loc_operands): Likewise.
19333         (output_loc_operands_raw): Likewise.
19334         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19335         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19336         (hash_loc_operands): Likewise.
19337         (compare_loc_operands): Likewise.
19339 2018-05-14  Mark Wielaard  <mark@klomp.org>
19341         * dwarf2out.c (count_index_addrs): New function.
19342         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19344 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19346         PR tree-optimization/83648
19347         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19348         return value as malloc candidate.
19350 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19352         PR ipa/85734
19353         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19354         param as true in call to suggest_attribute.
19356 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19358         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19359         -mreadonly-in-sdata.
19361 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19363         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19364         New pattern.
19365         (aarch64_crypto_aesd_fused): Likewise.
19367 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19369         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19370         (movsi_aarch64): Likewise.
19371         (load_pairsi): Likewise.
19372         (load_pairdi): Likewise.
19373         (store_pairsi): Likewise.
19374         (store_pairdi): Likewise.
19375         (load_pairsf): Likewise.
19376         (load_pairdf): Likewise.
19377         (store_pairsf): Likewise.
19378         (store_pairdf): Likewise.
19379         (zero_extend): Likewise.
19380         (trunc): Swap alternatives.
19381         (fcvt_target): Add '?' to prefer w over r.
19383 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19385         PR target/85756
19386         * config/i386/i386.md: Disallow non-commutative arithmetics in
19387         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19388         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19389         in the peephole2 before it.
19391 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19393         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19394         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19395         (ix86_handle_option): Handle -mcldemote.
19396         * config.gcc: New header.
19397         * config/i386/cldemoteintrin.h: New file.
19398         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19399         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19400         -mcldemote.
19401         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19402         OPTION_MASK_ISA_CLDEMOTE.
19403         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19404         (ix86_valid_target_attribute_inner_p): Ditto.
19405         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19406         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19407         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19408         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19409         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19410         (cldemote): New.
19411         * config/i386/i386.opt: Add -mcldemote.
19412         * config/i386/x86intrin.h: New header.
19413         * doc/invoke.texi: Add -mcldemote.
19415 2018-05-14  Richard Biener  <rguenther@suse.de>
19417         * doc/match-and-simplify.texi: Adjust :s documentation.
19419 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19421         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19422         intended memcpy size.
19423         (REORDER_45): Likewise.
19425 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19427         * sort.cc: New file.
19428         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19429         * vec.c (qsort_chk): Use gcc_qsort.
19430         * Makefile.in (OBJS-libcommon): Add sort.o.
19431         (build/sort.o): New target.  Use it...
19432         (BUILD_RTL): ... here, and...
19433         (build/gencfn-macros): ... here, and...
19434         (build/genmatch): ... here.
19436 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19437             Chung-Ju Wu  <jasonwucj@gmail.com>
19439         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19440         * config/nds32/nds32-graywolf.md: New file.
19441         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19442         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19443         pipeline.
19444         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19445         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19446         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19447         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19448         * config/nds32/pipelines.md: Include n15 settings.
19450 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19451             Chung-Ju Wu  <jasonwucj@gmail.com>
19453         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19454         * config/nds32/nds32-n13.md: New file.
19455         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19456         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19457         pipeline.
19458         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19459         * config/nds32/nds32.md (pipeline_model): Add n13.
19460         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19461         * config/nds32/pipelines.md: Include n13 settings.
19463 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19464             Chung-Ju Wu  <jasonwucj@gmail.com>
19466         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19467         * config/nds32/nds32-n10.md: New file.
19468         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19469         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19470         pipeline.
19471         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19472         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19473         * config/nds32/nds32.md (pipeline_model): Add n10.
19474         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19475         * config/nds32/pipelines.md: Include n10 settings.
19477 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19478             Kito Cheng  <kito.cheng@gmail.com>
19479             Chung-Ju Wu  <jasonwucj@gmail.com>
19481         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19482         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19483         Add enum values for DSP extension instructions.
19484         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19485         New constraints.
19486         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19487         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19488         New code iterators.
19489         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19490         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19491         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19492         * config/nds32/nds32-intrinsic.md: Likewise.
19493         * config/nds32/nds32_intrinsic.h: Likewise.
19494         * config/nds32/nds32-md-auxiliary.c: Likewise.
19495         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19496         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19497         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19498         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19499         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19500         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19501         TYPE_DMAC in switch statement.
19502         * config/nds32/nds32.c: New checking and implementation for DSP
19503         extension instructions.
19504         * config/nds32/nds32.h: Likewise.
19505         * config/nds32/nds32.md: Likewise.
19506         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19507         * config/nds32/predicates.md: Implement new predicates for DSP
19508         extension.
19510 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19512         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19513         Reformat alternatives and attributes so it is easier to identify
19514         which constraints/attributes go with which instruction.
19515         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19516         (mov<mode>_softfloat32, FMOVE64): Likewise.
19517         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19518         (mov<mode>_softfloat64, FMOVE64): Likewise.
19520 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19522         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19523         subsection.
19524         (Basic PowerPC Built-in Functions): The new name of the
19525         subsection previously known as "PowerPC Built-in Functions".
19526         (Basic PowerPC Built-in Functions Available on all Configurations):
19527         New subsubsection.
19528         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19529         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19530         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19531         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19533 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19535         PR ipa/85655
19536         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19537         single const.
19539 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19541         PR target/85733
19542         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19544 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19546         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19547         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19548         (ix86_handle_option): Handle -mwaitpkg.
19549         * config.gcc: New header.
19550         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19551         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19552         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19553         function type.
19554         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19555         OPTION_MASK_ISA_WAITPKG.
19556         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19557         (ix86_option_override_internal): Add PTA_WAITPKG.
19558         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19559         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19560         IX86_BUILTIN_TPAUSE.
19561         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19562         __builtin_ia32_umwait and __builtin_ia32_tpause.
19563         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19564         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19565         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19566         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19567         UNSPECV_TPAUSE): New.
19568         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19569         * config/i386/i386.opt: Add -mwaitpkg.
19570         * config/i386/waitpkgintrin.h: New file.
19571         * config/i386/x86intrin.h: New header.
19572         * doc/invoke.texi: Add -mwaitpkg.
19574 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19576         PR target/85606
19577         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
19578         equivalent.
19579         (cortex-m0): Use armv6s-m isa.
19580         (cortex-m0plus): Likewise.
19581         (cortex-m1): Likewise.
19582         (cortex-m0.small-multiply): Likewise.
19583         (cortex-m0plus.small-multiply): Likewise.
19584         (cortex-m1.small-multiply): Likewise.
19586 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
19587             Jakub Jelinek  <jakub@redhat.com>
19589         PR tree-optimization/85692
19590         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
19591         source permute as well.
19593 2018-05-11  Martin Liska  <mliska@suse.cz>
19595         PR sanitizer/85556
19596         * doc/extend.texi: Document LLVM style format for no_sanitize
19597         attribute.
19599 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
19601         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
19602         mode_supports_vsx_dform_quad to mode_supports_dq_form.
19603         (mode_supports_vsx_dform_quad): Likewise.
19604         (mode_supports_vmx_dform): Move these functions to be next to the
19605         other mode_supports functions.
19606         (mode_supports_dq_form): Likewise.
19607         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
19608         mode_supports_dq_form.
19609         (reg_offset_addressing_ok_p): Likewise.
19610         (offsettable_ok_by_alignment): Likewise.
19611         (rs6000_legitimate_offset_address_p): Likewise.
19612         (legitimate_lo_sum_address_p): Likewise.
19613         (rs6000_legitimize_address): Likewise.
19614         (rs6000_legitimize_reload_address): Likewise.
19615         (rs6000_secondary_reload_inner): Likewise.
19616         (rs6000_preferred_reload_class): Likewise.
19617         (rs6000_output_move_128bit): Likewise.
19619 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19621         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
19622         Generate SImode target register for null target.
19623         <case IX86_BUILTIN_XGETBV>: Ditto.
19624         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
19625         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
19627 2018-05-10  Carl Love  <cel@us.ibm.com>
19629         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19630         dcbtt and dcbtstt if operands[2] is 0.
19632 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19634         PR target/85693
19635         * config/i386/sse.md (usadv64qi): New expander.
19637 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
19639         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
19640         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
19641         -maltivec=be support.
19642         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
19643         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
19644         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
19645         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
19646         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
19647         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
19648         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
19649         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
19650         altivec_vsumsws): Adjust.
19651         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
19652         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
19653         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
19654         support.
19655         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
19656         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
19657         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
19658         (altivec_lve<VI_char>x): Delete expand.
19659         (*altivec_lve<VI_char>x_internal): Rename to...
19660         (altivec_lve<VI_char>x): ... this.
19661         (altivec_lvxl_<mode>): Delete expand.
19662         (*altivec_lvxl_<mode>_internal): Rename to ...
19663         (altivec_lvxl_<mode>): ... this.
19664         (altivec_stvxl_<mode>): Delete expand.
19665         (*altivec_stvxl_<mode>_internal): Rename to ...
19666         (altivec_stvxl_<mode>): ... this.
19667         (altivec_stve<VI_char>x): Delete expand.
19668         (*altivec_stve<VI_char>x_internal): Rename to ...
19669         (altivec_stve<VI_char>x): ... this.
19670         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
19671         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
19672         reduc_plus_scal_<mode>): Adjust.
19673         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
19674         comment.
19675         (rs6000_cpu_cpp_builtins): Adjust.
19676         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
19677         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
19678         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
19679         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
19680         -maltivec=be support.
19681         (rs6000_split_vec_extract_var): Adjust.
19682         (rs6000_split_v4si_init): Adjust.
19683         (swap_selector_for_mode): Delete.
19684         (altivec_expand_lvx_be, altivec_expand_stvx_be,
19685         altivec_expand_stvex_be): Delete.
19686         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
19687         -maltivec=be support.
19688         (rs6000_gimple_fold_builtin): Ditto.
19689         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
19690         Adjust.
19691         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
19692         (TARGET_DIRECT_MOVE_64BIT): Adjust.
19693         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
19694         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
19695         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
19696         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
19697         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
19698         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
19699         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
19700         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
19701         anonymous split): Adjust.
19702         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
19703         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
19705 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
19707         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
19708         when --with-gxx-include-dir is also specified.
19709         * configure: Regenerate.
19711 2018-05-09  Jim Wilson  <jimw@sifive.com>
19713         PR target/84797
19714         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
19715         * config/riscv/t-withmultilib: New.
19716         * config/riscv/withmultilib.h: New.
19717         * doc/install.texi: Document RISC-V --with-multilib-list support.
19719 2018-05-09  Richard Biener  <rguenther@suse.de>
19721         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19722         vector.
19723         (vect_bb_vectorization_profitable_p): Adjust.  Compute
19724         actual scalar cost using the cost vector and the add_stmt_cost
19725         machinery.
19727 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19729         PR rtl-optimization/85645
19730         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
19731         in the REG_CFA_REGISTER note for LR, don't leave it empty.
19733 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19735         PR rtl-optimization/85645
19736         * shrink-wrap.c (spread_components): Return a boolean saying if
19737         anything was changed.
19738         (try_shrink_wrapping_separate): Iterate spread_components until
19739         nothing changes anymore.
19741 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19743         PR rtl-optimization/85645
19744         * regrename.c (build_def_use): Also kill the chains that include the
19745         destination of a REG_CFA_REGISTER note.
19747 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19749         PR rtl-optimization/85645
19750         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
19751         insn that has a REG_CFA_REGISTER note.
19753 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
19755         * cfgexpand.c (expand_clobber): New function.
19756         (expand_gimple_stmt_1): Use it.
19757         * tree-vect-stmts.c (vect_clobber_variable): New function,
19758         split out from...
19759         (vectorizable_simd_clone_call): ...here.
19760         (vectorizable_store): Emit a clobber either side of an
19761         IFN_STORE_LANES sequence.
19762         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
19764 2018-05-09  Tom de Vries  <tom@codesourcery.com>
19766         PR target/85626
19767         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
19768         (define_insn "trap_if_false"): Add exit after trap.
19770 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
19772         PR rtl-optimization/85638
19773         * bb-reorder.c: Include common/common-target.h.
19774         (create_forwarder_block): New function extracted from...
19775         (fix_up_crossing_landing_pad): ...here.  Rename into...
19776         (dw2_fix_up_crossing_landing_pad): ...this.
19777         (sjlj_fix_up_crossing_landing_pad): New function.
19778         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
19779         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
19780         from both partitions and exit the loop after one iteration.
19782 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19784         Revert:
19785         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19786         subsection.
19787         (Basic PowerPC Built-in Functions): The new name of the
19788         subsection previously known as "PowerPC Built-in Functions".
19789         (Basic PowerPC Built-in Functions Available on all Configurations):
19790         New subsubsection.
19791         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19792         subsubsection.
19793         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19794         subsubsection.
19795         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19796         subsubsection.
19797         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19798         subsubsection.
19800 2018-05-08  Jim Wilson  <jimw@sifive.com>
19802         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
19803         (LD_EMUL_SUFFIX): New.
19804         (LINK_SPEC): Use it.
19806 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19808         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19809         subsection.
19810         (Basic PowerPC Built-in Functions): The new name of the
19811         subsection previously known as "PowerPC Built-in Functions".
19812         (Basic PowerPC Built-in Functions Available on all Configurations):
19813         New subsubsection.
19814         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19815         subsubsection.
19816         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19817         subsubsection.
19818         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19819         subsubsection.
19820         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19821         subsubsection.
19823 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19825         PR target/85683
19826         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
19827         after cmpelim optimization.
19829 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
19831         * config.gcc: Support "goldmont".
19832         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
19833         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19834         PROCESSOR_GOLDMONT.
19835         * config/i386/i386.c (m_GOLDMONT): Define.
19836         (processor_target_table): Add "goldmont".
19837         (PTA_GOLDMONT): Define.
19838         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
19839         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
19840         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
19841         (fold_builtin_cpu): Add "goldmont".
19842         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
19843         (ix86_option_override_internal): Add "goldmont".
19844         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
19845         (processor_type): Add PROCESSOR_GOLDMONT.
19846         * config/i386/i386.md: Add CPU "glm".
19847         * config/i386/glm.md: New file.
19848         * config/i386/x86-tune.def: Add m_GOLDMONT.
19849         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
19851 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19853         PR target/85572
19854         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
19855         E_V4DImode.
19856         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
19857         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
19858         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
19860         PR target/85317
19861         * config/i386/i386.c (ix86_fold_builtin): Handle
19862         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
19864         PR target/85480
19865         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
19866         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
19868 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
19870         PR target/85658
19871         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
19872         (check_arch): Likewise.
19873         (check_fpu): Return the result rather than printing it.
19874         (end arch): Fix operator precedence.
19875         (end cpu): Likewise.
19876         (END): Print the result from check_fpu.
19878 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19879             Alan Hayward  <alan.hayward@arm.com>
19880             David Sherwood  <david.sherwood@arm.com>
19882         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
19883         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
19884         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
19885         (*fcmuo<mode>_and): New patterns.
19887 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19889         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
19890         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
19891         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
19892         (cmp_op, sve_imm_con): New code attributes.
19893         (SVE_COND_INT_CMP, imm_con): Delete.
19894         (cmp_op): Remove above unspecs from int attribute.
19895         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
19896         to...
19897         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
19898         comparison-specific unspecs.
19899         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
19900         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
19901         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
19902         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
19903         (*vec_fcm<cmp_op><mode>): Rename to...
19904         (*fcm<cmp_op><mode>): ...this and adjust likewise.
19905         (*vec_fcmuo<mode>): Rename to...
19906         (*fcmuo<mode>): ...this and adjust likewise.
19907         (*pred_fcm<cmp_op><mode>): New pattern.
19908         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
19909         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
19910         functions.
19911         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
19912         and UNORDERED.
19913         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
19914         (aarch64_emit_sve_predicated_cond): New function.
19915         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
19916         (aarch64_emit_unspec_cond_or): Replace with...
19917         (aarch64_emit_sve_or_conds): ...this new function.  Use
19918         aarch64_emit_sve_ptrue_op for the individual comparisons and
19919         aarch64_emit_binop to OR them together.
19920         (aarch64_emit_inverted_unspec_cond): Replace with...
19921         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
19922         aarch64_emit_sve_ptrue_op for the comparison and
19923         aarch64_emit_unop to invert the result.
19924         (aarch64_expand_sve_vec_cmp_float): Update after the above
19925         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
19927 2018-05-07  Nathan Sidwell  <nathan@acm.org>
19929         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
19930         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
19931         (Backwards Compatibility): Likewise.
19933 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19935         PR bootstrap/85681
19936         Revert:
19937         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19939         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19940         <prefetch_dynamic_strides>: New const bool field.
19941         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19942         prefetch_dynamic_strides.
19943         (exynosm1_prefetch_tune): Likewise.
19944         (thunderxt88_prefetch_tune): Likewise.
19945         (thunderx_prefetch_tune): Likewise.
19946         (thunderx2t99_prefetch_tune): Likewise.
19947         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19948         to false.
19949         (aarch64_override_options_internal): Update to set
19950         PARAM_PREFETCH_DYNAMIC_STRIDES.
19951         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19952         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19953         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19954         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19955         prefetch-dynamic-strides setting.
19957         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19959         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19960         <minimum_stride>: New const int field.
19961         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19962         minimum_stride field.
19963         (exynosm1_prefetch_tune): Likewise.
19964         (thunderxt88_prefetch_tune): Likewise.
19965         (thunderx_prefetch_tune): Likewise.
19966         (thunderx2t99_prefetch_tune): Likewise.
19967         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19968         (aarch64_override_options_internal): Update to set
19969         PARAM_PREFETCH_MINIMUM_STRIDE.
19970         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19971         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19972         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19973         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19974         stride is constant and is below the minimum stride threshold.
19976 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19978         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
19979         to 512.
19981 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19983         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19984         <prefetch_dynamic_strides>: New const bool field.
19985         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19986         prefetch_dynamic_strides.
19987         (exynosm1_prefetch_tune): Likewise.
19988         (thunderxt88_prefetch_tune): Likewise.
19989         (thunderx_prefetch_tune): Likewise.
19990         (thunderx2t99_prefetch_tune): Likewise.
19991         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19992         to false.
19993         (aarch64_override_options_internal): Update to set
19994         PARAM_PREFETCH_DYNAMIC_STRIDES.
19995         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19996         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19997         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19998         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19999         prefetch-dynamic-strides setting.
20001 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20003         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20004         <minimum_stride>: New const int field.
20005         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20006         minimum_stride field.
20007         (exynosm1_prefetch_tune): Likewise.
20008         (thunderxt88_prefetch_tune): Likewise.
20009         (thunderx_prefetch_tune): Likewise.
20010         (thunderx2t99_prefetch_tune): Likewise.
20011         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20012         (aarch64_override_options_internal): Update to set
20013         PARAM_PREFETCH_MINIMUM_STRIDE.
20014         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20015         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20016         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20017         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20018         stride is constant and is below the minimum stride threshold.
20020 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
20022         PR c++/85659
20023         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20024         the type is addressable.  Don't force op into register if it has
20025         BLKmode.
20027 2018-05-05  Roland McGrath  <mcgrathr@google.com>
20029         PR other/77609
20030         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20031         any section for which we don't know a specific type it should have,
20032         regardless of name.  Previously this was done only for the exact
20033         names ".init_array", ".fini_array", and ".preinit_array".
20034         (default_elf_asm_named_section): Add comment about
20035         relationship with default_section_type_flags and SECTION_NOTYPE.
20036         (get_section): Don't consider it a type conflict if one side has
20037         SECTION_NOTYPE and the other doesn't, as long as neither has the
20038         SECTION_BSS et al used in the default_section_type_flags logic.
20040 2018-05-05  Tom de Vries  <tom@codesourcery.com>
20042         PR target/85653
20043         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20044         (workaround_barsyncs): New function.
20045         (nvptx_reorg): Use workaround_barsyncs.
20046         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20047         (define_expand "nvptx_membar_cta"): New define_expand.
20048         (define_insn "*nvptx_membar_cta"): New insn.
20050 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
20052         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20053         To improve optimization opportunities.
20054         * builtin-types.def: The new needed builtin types for the above.
20056 2018-05-04  Richard Biener  <rguenther@suse.de>
20058         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20059         * gimple-ssa-store-merging.c
20060         (imm_store_chain_info::output_merged_store): Remove redundant create,
20061         release split_store vector contents on failure.
20062         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20063         scalar stmt vector on cache hit.
20065 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
20067         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20068         Xilinx FP support.
20069         * config.gcc (powerpc-xilinx-eabi*): Remove.
20070         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20071         support.
20072         (fusion_addis_mem_combo_load): Ditto.
20073         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20074         FP support.
20075         (rs6000_cpu_cpp_builtins): Ditto.
20076         * config/rs6000/rs6000-linux.c
20077         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20078         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20079         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20080         support.
20081         (rs6000_setup_reg_addr_masks): Ditto.
20082         (rs6000_init_hard_regno_mode_ok): Ditto.
20083         (rs6000_option_override_internal): Ditto.
20084         (legitimate_lo_sum_address_p): Ditto.
20085         (rs6000_legitimize_address): Ditto.
20086         (rs6000_legitimize_reload_address): Ditto.
20087         (rs6000_legitimate_address_p): Ditto.
20088         (abi_v4_pass_in_fpr): Ditto.
20089         (setup_incoming_varargs): Ditto.
20090         (rs6000_gimplify_va_arg): Ditto.
20091         (rs6000_split_multireg_move): Ditto.
20092         (rs6000_savres_strategy): Ditto.
20093         (rs6000_emit_prologue_components): Ditto.
20094         (rs6000_emit_epilogue_components): Ditto.
20095         (rs6000_emit_prologue): Ditto.
20096         (rs6000_emit_epilogue): Ditto.
20097         (rs6000_elf_file_end): Ditto.
20098         (rs6000_function_value): Ditto.
20099         (rs6000_libcall_value): Ditto.
20100         * config/rs6000/rs6000.h: Ditto.
20101         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20102         (TARGET_MINMAX): ... this.  New.
20103         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20104         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20105         (*movsi_internal1_single): Delete.
20106         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20107         mfpu=, mxilinx-fpu): Delete.
20108         * config/rs6000/singlefp.h: Delete.
20109         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20110         * config/rs6000/t-rs6000: Ditto.
20111         * config/rs6000/t-xilinx: Delete.
20112         * config/rs6000/titan.md: Adjust for fp_type removal.
20113         * config/rs6000/vsx.md: Remove Xilinx FP support.
20114         (VStype_simple): Delete.
20115         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20116         * config/rs6000/xfpu.h: Delete.
20117         * config/rs6000/xfpu.md: Delete.
20118         * config/rs6000/xilinx.h: Delete.
20119         * config/rs6000/xilinx.opt: Delete.
20120         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20121         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20123 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20125         PR libgomp/85639
20126         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20127         if ignore == 0.
20129 2018-05-04  Richard Biener  <rguenther@suse.de>
20131         PR middle-end/85627
20132         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20133         (expand_complex_div_wide): Likewise.
20134         (expand_complex_operations_1): Likewise.
20135         (expand_complex_libcall): Preserve EH info of the original stmt.
20136         (tree_lower_complex): Handle removed blocks.
20137         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20138         on complex multiplication and division libcall builtins.
20140 2018-05-04  Richard Biener  <rguenther@suse.de>
20142         PR middle-end/85574
20143         * fold-const.c (negate_expr_p): Restrict negation of operand
20144         zero of a division to when we know that can happen without
20145         overflow.
20146         (fold_negate_expr_1): Likewise.
20148 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20150         PR libstdc++/85466
20151         * real.h (real_nextafter): Declare.
20152         * real.c (real_nextafter): New function.
20153         * fold-const-call.c (fold_const_nextafter): New function.
20154         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20155         CASE_CFN_NEXTTOWARD.
20156         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20157         even when arg1_mode is different from arg0_mode.
20159 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20161         * doc/extend.texi (Deprecated Features): Remove
20162         -ffriend-injection.
20163         (Backwards Compatibility): Likewise.
20164         * doc/invoke.texi (C++ Language Options): Likewise.
20165         (C++ Dialect Options): Likewise.
20167 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20169         PR target/85530
20170         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20171         _mm512_mask_mullox_epi64): New intrinsics.
20173 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20175         PR testsuite/85106
20176         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20177         dump files): Add offload-tree.
20179 2018-05-03  Richard Biener  <rguenther@suse.de>
20181         PR tree-optimization/85615
20182         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20183         to loops not nested in BBs loop father to avoid creating multi-entry
20184         loops.
20186 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20188         PR tree-optimization/70291
20189         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20190         arguments.  Change return type to tree.  Emit libcall as a new
20191         statement rather than replacing existing one when inplace_p is true.
20192         (expand_complex_multiplication_components): New function.
20193         (expand_complex_multiplication): Expand floating-point complex
20194         multiplication using the above.
20195         (expand_complex_division): Rename inner_type parameter to type.
20196         Update expand_complex_libcall call-site.
20197         (expand_complex_operations_1): Update expand_complex_multiplication
20198         and expand_complex_division call-sites.
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): In condition require that
20206         the highest significant bit of the shift count mask is clear.  In
20207         check whether and[sq]i3 is needed verify that all significant bits
20208         of the shift count other than the highest are set.
20210 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20212         PR libgomp/82428
20213         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20214         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20215         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20216         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20217         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20218         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20219         __builtin_goacc_parlevel_size.
20221 2018-05-02  Richard Biener  <rguenther@suse.de>
20223         PR tree-optimization/85597
20224         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20225         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20226         directly.
20228 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20230         PR testsuite/85106
20231         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20232         dump files): Add ltrans-tree.
20234 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20236         PR testsuite/85106
20237         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20238         dump files): Add wpa-ipa.
20240 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20242         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20243         powerpc*-*-linux*paired* target.
20244         * config/rs6000/750cl.h: Delete.
20245         * config/rs6000/paired.h: Delete.
20246         * config/rs6000/paired.md: Delete.
20247         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20248         float support.
20249         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20250         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20251         comment.  Remove paired float support.
20252         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20253         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20254         VECTOR_PAIRED.
20255         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20256         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20257         declarations.
20258         * config/rs6000/rs6000.c: Remove paired float support.
20259         (paired_expand_vector_init, paired_expand_vector_move,
20260         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20261         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20262         paired_expand_builtin, paired_expand_predicate_builtin,
20263         paired_init_builtins): Delete.
20264         * config/rs6000/rs6000.h: Remove paired float support.
20265         * config/rs6000/rs6000.md: Remove paired float support.
20266         (move_from_CR_ov_bit): Delete.
20267         * config/rs6000/rs6000.opt (mpaired): Delete.
20268         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20269         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20271 2018-05-02  Richard Biener  <rguenther@suse.de>
20273         PR middle-end/85567
20274         * gimplify.c (gimplify_save_expr): When in SSA form allow
20275         SAVE_EXPRs to compute to SSA vars.
20277 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20279         PR target/85582
20280         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20281         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20282         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20283         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20285 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20287         PR tree-optimization/85586
20288         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20289         exit early for statements in the same group if the accesses are
20290         not strided.
20292 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20294         PR lto/85451
20295         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20296         error message.
20298 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20300         PR tree-optimization/85143
20301         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20303 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20305         PR lto/85451
20306         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20307         not found" error message.
20309 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20311         PR other/83786
20312         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20313         * vec.c (test_ordered_remove_if): New function.
20314         (vec_c_tests): Call test_ordered_remove_if.
20315         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20316         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20317         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20318         VEC_ORDERED_REMOVE_IF.
20320 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20322         PR tree-optimization/82665
20323         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20324         pointer subtraction where arguments come from a memchr call.
20326 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20328         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20329         --push-state --as-needed and --pop-state instead of --as-needed and
20330         --no-as-needed if ld supports it.
20331         * configure: Regenerated.
20333         PR web/85578
20334         * doc/install.texi2html: Replace _002d with - and _002a with * in
20335         generated html files using sed.
20337 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20339         PR c++/85523
20340         * gcc-rich-location.c (blank_line_before_p): New function.
20341         (use_new_line): New function.
20342         (gcc_rich_location::add_fixit_insert_formatted): New function.
20343         * gcc-rich-location.h
20344         (gcc_rich_location::add_fixit_insert_formatted): New function.
20346 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20348         * selftest.c (assert_streq): Rename "expected" and "actual" to
20349         "val1" and "val2".  Extend NULL-handling to cover both inputs
20350         symmetrically, while still requiring both to be non-NULL for a pass.
20351         * selftest.h (assert_streq): Rename "expected" and "actual" to
20352         "val1" and "val2".
20353         (ASSERT_EQ): Likewise.
20354         (ASSERT_EQ_AT): Likewise.
20355         (ASSERT_KNOWN_EQ): Likewise.
20356         (ASSERT_KNOWN_EQ_AT): Likewise.
20357         (ASSERT_NE): Likewise.
20358         (ASSERT_MAYBE_NE): Likewise.
20359         (ASSERT_MAYBE_NE_AT): Likewise.
20360         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20361         the assertion to pass.
20362         (ASSERT_STREQ_AT): Likewise.
20364 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20366         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20367         interaction with -pie.
20369 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20371         * selftest.h: Fix alphabetization of per-source-file selftest
20372         declarations.
20374 2018-04-30  Jason Merrill  <jason@redhat.com>
20376         PR c++/61982 - dead stores to destroyed objects.
20377         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20378         of clobber.
20380 2018-04-30  Jason Merrill  <jason@redhat.com>
20382         * tree.c (build_clobber): New.
20383         * tree.h: Declare it.
20384         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20386 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20388         * diagnostic-show-locus.c (layout::layout): Update for
20389         location_get_source_line returning a char_span.
20390         (struct char_span): Move to input.h.
20391         (struct correction): Update for fields in char_span becoming
20392         private.
20393         (struct source_line): Update for location_get_source_line
20394         returning a char_span.
20395         (layout::print_line): Likewise.
20396         * edit-context.c (edited_file::print_content): Likewise.
20397         (edited_file::print_diff_hunk): Likewise.
20398         (edited_file::print_run_of_changed_lines): Likewise.
20399         (edited_file::get_num_lines): Likewise.
20400         (edited_line::edited_line): Likewise.
20401         * final.c (asm_show_source): Likewise.
20402         * input.c (location_get_source_line): Convert return type
20403         from const char * to char_span, losing the final "line_len"
20404         param.
20405         (dump_location_info): Update for the above.
20406         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20407         when handling the literal within the line.
20408         (test_reading_source_line): Update for location_get_source_line
20409         returning a char_span.
20410         * input.h (class char_span): Move here from
20411         diagnostic-show-locus.c, converting from a struct to a class.
20412         Make data members private.
20413         (char_span::operator bool): New.
20414         (char_span::length): New.
20415         (char_span::get_buffer): New.
20416         (char_span::operator[]): New.
20417         (char_span::subspan): Make const.
20418         (char_span::xstrdup): New.
20419         (location_get_source_line): Convert return type from const char *
20420         to char_span, losing the final "line_size" param.
20422 2018-04-30  Jan Hubicka  <jh@suse.cz>
20424         * lto-wrapper.c (ltrans_priorities): New static var.
20425         (cmp_priority): New.
20426         (run_gcc): Read priorities and if doing parallel build order
20427         the Makefile by them.
20429 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20431         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20433 2018-04-30  Richard Biener  <rguenther@suse.de>
20435         * tree-cfg.c (verify_address): Remove base argument, add
20436         flag whether to check TREE_ADDRESSABLE and do that.
20437         (verify_expr): Remove.
20438         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20439         (verify_gimple_assign_single): Likewise.
20440         (verify_gimple_switch): Likewise.
20441         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20442         piece from verify_expr.
20443         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20445 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20447         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20449 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20451         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20452         (small_data_pattern): Likewise.
20453         (arc_rewrite_small_data): Likewise.
20454         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20455         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20456         (get_symbol_alignment): New function.
20457         (legitimate_small_data_address_p): Likewise.
20458         (legitimate_scaled_address): Update, call
20459         legitimate_small_data_address_p.
20460         (output_sdata): New static variable.
20461         (arc_print_operand): Update how we handle small data operands.
20462         (arc_print_operand_address): Likewise.
20463         (arc_legitimate_address_p): Update, use
20464         legitimate_small_data_address_p.
20465         (arc_rewrite_small_data_p): Remove.
20466         (arc_rewrite_small_data_1): Likewise.
20467         (arc_rewrite_small_data): Likewise.
20468         (small_data_pattern): Likewise.
20469         (compact_sda_memory_operand): Update to use
20470         legitimate_small_data_address_p and get_symbol_alignment.
20471         (prepare_move_operands): Don't rewite sdata pattern.
20472         (prepare_extend_operands): Remove.
20473         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20474         pattern.
20475         (zero_extendqisi2): Likewise.
20476         (zero_extendhisi2): Likewise.
20477         (extendqihi2): Likewise.
20478         (extendqisi2): Likewise.
20479         (extendhisi2): Likewise.
20480         (addsi3): Likewise.
20481         (subsi3): Likewise.
20482         (andsi3): Likewise.
20483         * config/arc/constraints.md (Usd): Change it to memory constraint.
20485 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20487         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20488         as source of std instructions.
20489         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20490         allow 6-bit constants as source for store instructions.
20491         (movdi_insn): Update instruction pattern to allow 6-bit constants
20492         as source for store instructions.
20494 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20496         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20498 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20499             Sandra Loosemore <sandra@codesourcery.com>
20501         * dumpfile.c (dump_open): Allow '-' for stdout.
20502         * doc/invoke.texi (Developer Options): Document dump filename
20503         determination early.  Document stdin/stdout selection.
20505 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20507         Microblaze Target: PIC data text relative
20509         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20510         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20511         Add declaration.
20512         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20513         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20514         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20515         New addressing mode for data-text relative position indepenedent code.
20516         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20517         'ADDRESS_SYMBOLIC_TXT_REL'.
20518         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20519         (microblaze_legitimate_pic_operand): Exclude function calls from
20520         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20521         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20522         addresses cases.
20523         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20524         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20525         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20526         for 'address + offset'.
20527         (microblaze_expand_prologue): Add new function prologue call for
20528         'r20' assignation.
20529         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20530         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20531         table in case of TARGET_PIC_DATA_TEXT_REL.
20532         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20533         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20534         Add new macros 'UNSPEC_TEXT',
20535         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20536         + exclude function calls from 'UNSPEC_PLT' in case of data text
20537         relative mode.
20538         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20539         new target hook for generating address diff vector tables in case of
20540         flag_pic.
20541         * doc/tm.texi : Regenerate.
20542         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20543         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20544         of addr diff vector generation.
20545         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20546         target hook definition.
20547         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20548         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20549         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20550         option for data text relative.
20552 2018-04-30  Richard Biener  <rguenther@suse.de>
20554         * tree-chrec.h (evolution_function_is_constant_p): Remove
20555         redundant check.
20556         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20558 2018-04-30  Richard Biener  <rguenther@suse.de>
20560         PR bootstrap/85571
20561         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20563 2018-04-30  Richard Biener  <rguenther@suse.de>
20565         PR tree-optimization/28364
20566         PR tree-optimization/85275
20567         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20568         copying first exit test.
20570 2018-04-28  Mark Wielaard  <mark@klomp.org>
20572         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20573         dwarf_version >= 5.
20574         (dwarf_AT): Handle DW_AT_addr_base.
20575         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20577 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
20579         PR target/84431
20580         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
20581         (*ashl<dwi>3_doubleword_mask_1): Ditto.
20582         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
20583         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
20585 2018-04-28  Richard Biener  <rguenther@suse.de>
20587         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
20588         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
20589         to reflect use.  Only add interesting stmts.
20591 2018-04-27  Martin Jambor  <mjambor@suse.cz>
20593         PR ipa/85549
20594         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
20595         the jump function allows for passing through aggregate values.
20597 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
20599         * input.h (in_system_header_at): Convert from macro to inline
20600         function.
20601         (from_macro_expansion_at): Likewise.
20602         (from_macro_definition_at): Likewise.
20604 2018-04-27  Jeff Law  <law@redhat.com>
20606         * config.gcc: Mark tile* targets as deprecated/obsolete.
20608 2018-04-27  Richard Biener  <rguenther@suse.de>
20610         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
20611         fix for ILP32.
20613 2018-04-27  Richard Biener  <rguenther@suse.de>
20615         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
20617 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
20619         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
20620         with Yd constraint. Set "preferred_for_speed" attribute from
20621         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
20622         with Yd constraint.
20623         (*movdi_internal): Ditto.
20624         (movti_interunit splitters): Remove
20625         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
20626         (movdi_interunit splitters): Ditto.
20627         * config/i386/constraints.md (Ye): Remove.
20628         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
20630 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20632         PR target/85512
20633         * config/aarch64/constraints.md (Usg): Limit to 31.
20634         (Usj): Limit to 63.
20636 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
20638         PR tree-optimization/85529
20639         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
20640         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
20641         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
20642         zero extension or masking of the MSB bit.
20643         (optimize_range_tests): Add FIRST_BB argument, pass it through
20644         to optimize_range_tests_var_bound.
20645         (maybe_optimize_range_tests, reassociate_bb): Adjust
20646         optimize_range_tests callers.
20648 2018-04-26  Richard Biener  <rguenther@suse.de>
20649             Jakub Jelinek  <jakub@redhat.com>
20651         * cgraph.h (symbol_table): Just declare debug method here.
20652         * symtab.c (symbol_table::debug): Define.
20654 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
20656         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
20658 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
20660         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
20661         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
20662         (*movdi_internal): Substitute Yi and Yj constraint with x
20663         and Ym and Yn constraint with y constraint.  Update "isa"
20664         attribute and set "preferred_for_speed" attribute from
20665         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
20666         (*movsi_internal): Ditto.
20667         (*movdf_internal): Ditto.
20668         (*movsf_internal): Ditto.
20669         (*zero_extendsidi2): Ditto.
20670         * config/i386/sse.md (vec_set<mode>_0): Ditto.
20671         (sse2_loadld): Ditto.
20672         (*vec_extract<ssevecmodelower>_0): Ditto.
20673         (*vec_extractv4si_0_zext_sse4): Ditto.
20674         (vec_concatv2di): Ditto.
20675         (*vec_dup<mode>): Ditto.
20676         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
20677         * config/i386/constraints.md (Yi): Remove.
20678         (Yj): Remove.
20679         (Ym): Remove.
20680         (Yn): Remove.
20682 2018-04-26  Nathan Sidwell  <nathan@acm.org>
20684         * dumpfile.c (dump_open): New.
20685         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
20686         (dump_finish): Detect stdio/stderr by value not name.
20688 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
20690         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
20692 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20694         PR target/84952
20695         * config/nvptx/nvptx.c (verify_neutering_jumps)
20696         (verify_neutering_labels): New function
20697         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
20699 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20701         PR target/84025
20702         * config/nvptx/nvptx.c (needs_neutering_p): New function.
20703         (nvptx_single): Use needs_neutering_p to skip over insns that do not
20704         need neutering.
20706 2018-04-26  Richard Biener <rguenther@suse.de>
20707             Tom de Vries  <tom@codesourcery.com>
20709         PR lto/85422
20710         * lto-streamer-out.c (output_function): Fixup loops if required to match
20711         discovery done in the reader.
20713 2018-04-26  Richard Biener  <rguenther@suse.de>
20715         PR tree-optimization/85116
20716         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
20717         have a loop exit from the single latch predecessor.  Remove
20718         case of header with just condition.
20719         (ch_base::copy_headers): Exclude infinite loops from any
20720         processing.
20721         (pass_ch::execute): Record exits.
20723 2018-04-26  Richard Biener  <rguenther@suse.de>
20725         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
20726         prologue cost vector and pass it to vect_get_load_cost.
20727         (vect_get_peeling_costs_all_drs): Likewise.
20728         (vect_peeling_hash_get_lowest_cost): Likewise.
20729         (vect_enhance_data_refs_alignment): Likewise.
20731 2018-04-26  Richard Biener  <rguenther@suse.de>
20733         PR middle-end/85450
20734         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
20735         checking of integer<->pointer conversions.
20736         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
20737         sign-/zero-extending pointer types.
20738         (expand_omp_for_static_chunk): Likewise.
20740 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
20741             Jean Lee  <xiaoyur347@gmail.com>
20743         * config/mips/mips.c (mips_asan_shadow_offset): New function.
20744         (TARGET_ASAN_SHADOW_OFFSET): Define.
20745         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
20746         true for -fsanitize=address.
20748 2018-04-25  Mark Wielaard  <mark@klomp.org>
20750         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
20751         shorter ones.
20753 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20755         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
20756         than "alu", remove explicit "memory" and "imm_disp" attributes.
20757         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
20759         PR middle-end/85414
20760         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
20761         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
20762         gen_lowpart_no_emit.
20764 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
20766         PR target/85473
20767         * config/i386/i386.c (ix86_expand_builtin): Change memory
20768         operand to XI, extend p0 to Pmode.
20769         * config/i386/i386.md: Change unspec volatile and operand
20770         1 mode to XI, change operand 0 mode to P.
20772 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20774         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
20775         GET_MODE_MASK before any checking.
20776         (nds32_can_use_bset_p): Likewise.
20777         (nds32_can_use_btgl_p): Likewise.
20779 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20781         * config/nds32/nds32-doubleword.md: New define_split pattern for
20782         illegal register number.
20784 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20786         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
20788 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20790         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
20792 2018-04-25  Richard Biener  <rguenther@suse.de>
20794         * lto-streamer.h (LTO_major_version): Bump to 8.
20796 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20798         * BASE-VER: Set to 9.0.0.
20800 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
20802         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
20803         in __abskf2 and __powikf2.
20805 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20807         PR target/85512
20808         * config/aarch64/constraints.md (Usg, Usj): New constraints.
20809         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
20810         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
20811         Use the above on operand 2.  Reindent.
20812         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
20814 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20816         PR target/85485
20817         * common/config/i386/i386-common.c (ix86_handle_option): Don't
20818         handle OPT_mcet.
20819         * config/i386/i386.opt (mcet): Removed.
20820         * doc/install.texi: Remove -mcet documentation.
20821         * doc/invoke.texi: Likewise.
20823 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20825         PR target/85485
20826         * doc/install.texi: Remove -mcet from bootstrap-cet.
20828 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20830         PR target/85511
20831         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
20832         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
20833         if TARGET_64BIT.
20835         PR target/85503
20836         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
20837         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
20838         containing a CONST_VECTOR.
20840 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
20842         * doc/install.texi: Update newlib dependency for nvptx.
20844 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20846         PR target/85508
20847         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
20848         instead of INTVAL when shifting x left.
20850 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
20852         PR tree-optimization/85478
20853         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
20854         vect_grouped_store_supported for single element vectors.
20856 2018-04-24  Richard Biener  <rguenther@suse.de>
20858         PR target/85491
20859         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
20860         load cost increase to the case of non-constant step.
20862 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20864         PR target/84828
20865         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
20866         destination if any_malformed_asm.
20868 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
20870         PR middle-end/85496
20871         * expr.c (store_field): In the bitfield case, if the value comes from
20872         a function call and is returned in registers by means of a PARALLEL,
20873         do not change the mode of the temporary unless BLKmode and VOIDmode.
20875 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
20877         PR rtl-optimization/85423
20878         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
20879         dependencies to debug insns when the previous insn is non-debug.
20881 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20883         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
20884         enums into a single definition.
20885         (fls): Fix predicates and printing.
20886         (seti): Likewise.
20888 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20890         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
20891         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
20892         and short u6 immediate.
20893         (check_if_valid_sleep_operand): Remove.
20894         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
20896 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20898         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
20899         flag_always_save_lp condition.
20900         * config/nds32/nds32.opt (malways-save-lp): New option.
20902 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20904         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
20905         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
20906         * config/nds32/nds32.h
20907         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
20908         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
20910 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20912         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
20913         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
20915 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20916             Chung-Ju Wu  <jasonwucj@gmail.com>
20918         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
20919         Declare.
20920         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
20921         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
20923 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20925         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
20927 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20929         * config/nds32/nds32-protos.h (nds32_data_alignment,
20930         nds32_local_alignment): Declare.
20931         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
20932         nds32_local_alignment): New functions.
20933         (TARGET_CONSTANT_ALIGNMENT): Define.
20934         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
20936 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20938         * config/nds32/nds32.c
20939         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
20940         (TARGET_MODES_TIEABLE_P): Likewise.
20942 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20944         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
20945         level Ofast and Og.
20947 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
20948             Chung-Ju Wu  <jasonwucj@gmail.com>
20950         * config/nds32/constants.md (unspec_volatile_element): Add enum values
20951         for unaligned access.
20952         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20953         unaligned access.
20954         * config/nds32/nds32-intrinsic.md: Likewise.
20955         * config/nds32/nds32_intrinsic.h: Likewise.
20956         * config/nds32/nds32.h (nds32_builtins): Likewise.
20957         * config/nds32/nds32.opt (munaligned-access): New option.
20958         * config/nds32/nds32.c (nds32_asm_file_start): Display
20959         flag_unaligned_access status.
20961 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20963         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
20964         -mno-relax is present.
20965         * config/riscv/linux.h (LINK_SPEC): Ditto.
20967 2018-04-20  Martin Sebor  <msebor@redhat.com>
20969         PR c/85365
20970         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
20971         for null pointers.
20972         (gimple_fold_builtin_stxcpy_chk): Same.
20973         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
20975 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
20977         PR target/85456
20978         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
20979         __powikf2 when long double is IEEE 128-bit.
20981 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20983         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
20984         step to make sure stack always aligned.
20986 2018-04-20  Carl Love  <cel@us.ibm.com>
20988         PR target/83402
20989         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
20990         size check for arg0.
20992 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
20993             Tom de Vries  <tom@codesourcery.com>
20995         PR target/85445
20996         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
20997         Emit insns for calls too.
20998         (nvptx_find_par): Always look for worker-level predecessor insn.
20999         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
21000         calls.
21001         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21002         (nvptx_process_pars): Propagate frames for calls.
21004 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
21006         PR target/85469
21007         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21008         Removed.
21009         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21010         (ix86_handle_option): Don't handle OPT_mibt.
21011         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21012         __SHSTK__.
21013         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21014         has_ibt and ibt.
21015         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21016         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21017         (ix86_target_macros): Define __CET__ with flag_cf_protection
21018         for -fcf-protection.
21019         * config/i386/i386.c (isa2_opts): Remove -mibt.
21020         * config/i386/i386.h (TARGET_IBT): Removed.
21021         (TARGET_IBT_P): Likewise.
21022         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21023         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21024         * config/i386/i386.opt (mcet): Update help message.
21025         (mshstk): Likewise.
21026         (mibt): Removed.
21027         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
21028         -mcet as an alias for -mshstk.
21030 2018-04-20  Richard Biener <rguenther@suse.de>
21032         PR middle-end/85475
21033         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21034         complexity by forcing a single use of the multiply operand.
21036 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21038         ipa/85449
21039         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21040         recursion dependency to only apply to non-clones.
21042 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21044         ipa/85447
21045         * ipa-cp.c (create_specialized_node): Check that clones of
21046         self-recursive edges exist during IPA-CP.
21048 2018-04-19  Toon Moene  <toon@moene.org>
21050         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21051         by -O3.
21053 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21055         PR tree-optimization/85467
21056         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21057         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
21058         VECTOR_CST element to type.
21060 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21062         PR target/85397
21063         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21064         * config/i386/i386.md (builtin_setjmp_setup): Removed.
21065         (builtin_longjmp): Likewise.
21066         (save_stack_nonlocal): New pattern.
21067         (restore_stack_nonlocal): Likewise.
21069 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21071         PR target/85404
21072         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21073         Replace ASM_OUTPUT_LABEL with fprintf.
21075 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21077         PR target/85417
21078         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21079         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21080         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21081         define __IBT__ and __SHSTK__ for -fcf-protection.
21082         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21083         TARGET_IBT.
21084         (ix86_trampoline_init): Likewise.
21085         (x86_output_mi_thunk): Likewise.
21086         (ix86_notrack_prefixed_insn_p): Likewise.
21087         (ix86_option_override_internal): Don't disallow -fcf-protection.
21088         * config/i386/i386.md (rdssp<mode>): Also enable for
21089         -fcf-protection.
21090         (incssp<mode>): Likewise.
21091         (nop_endbr): Likewise.
21092         * config/i386/i386.opt (mcet): Change help message to built-in
21093         functions only.
21094         (mibt): Likewise.
21095         (mshstk): Likewise.
21096         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21097         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21098         enable CET built-in functions.
21100 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21102         * common/config/i386/i386-common.c
21103         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21104         OPTION_MASK_ISA_MOVDIRI_UNSET,
21105         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21106         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21107         * config.gcc (movdirintrin.h): New header.
21108         * config/i386/cpuid.h (bit_MOVDIRI,
21109         bit_MOVDIR64B): New bits.
21110         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21111         and -mmvodir64b.
21112         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21113         (VOID, PVOID, PCVOID)): New function types.
21114         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21115         __builtin_ia32_directstoreu_u64,
21116         __builtin_ia32_movdir64b): New builtins.
21117         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21118         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21119         and -mmovdiri.
21120         (ix86_valid_target_attribute_inner_p): Ditto.
21121         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21122         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21123         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21124         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21125         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21126         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21127         (movdiri<mode>, movdir64b_<mode>): New.
21128         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21129         * config/i386/immintrin.h: Include movdirintrin.h.
21130         * config/i386/movdirintrin.h: New file.
21131         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21133 2018-04-19  Richard Biener  <rguenther@suse.de>
21135         PR middle-end/85455
21136         * cfg.c (clear_bb_flags): When loop state says we have
21137         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21139 2018-04-19  Richard Biener  <rguenther@suse.de>
21141         PR tree-optimization/84737
21142         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21143         copying restrict info.
21144         (vect_setup_realignment): Use it.
21145         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21146         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21147         the first DR to all generated stores.
21148         (vectorizable_load): Likewise for loads.
21150 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21152         PR tree-optimization/85446
21153         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21154         the integral and pointer types to have the same precision.
21156         * doc/install.texi: Document --disable-cet being the default and
21157         --enable-cet=auto.
21159 2018-04-18  Martin Liska  <mliska@suse.cz>
21161         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21162         style.
21164 2018-04-18  Martin Liska  <mliska@suse.cz>
21166         Revert
21167         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21169         PR ipa/83983
21170         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21171         arguments if they are comparable.
21173 2018-04-18  Martin Liska  <mliska@suse.cz>
21175         Revert
21176         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21178         PR lto/84805
21179         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21180         incomplete types.
21182 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21184         PR target/85388
21185         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21186         ENDBR after calling __morestack.
21188 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21190         PR jit/85384
21191         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21192         by using gcc_base_ver to generate a gcc_driver_version, and use
21193         it when generating GCC_DRIVER_NAME.
21194         * configure: Regenerate.
21196 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21198         PR target/81084
21199         * config.gcc: Obsolete powerpc*-*-*spe*.
21201 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21203         PR debug/84637
21204         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21205         (stabstr_D): Change type of unum from unsigned int to
21206         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21207         type.
21209 2018-04-17  Jim Wilson  <jimw@sifive.com>
21211         PR 84856
21212         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21213         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21214         Set arg_pointer_offset after using pretend_args_size.
21216 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21218         PR rtl-optimization/85431
21219         * dse.c (record_store): Ignore zero width stores.
21221         PR sanitizer/85230
21222         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21223         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21224         __builtin_stack_restore rather than after it.
21225         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21226         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21227         argument instead of virtual_dynamic_stack_rtx.
21229 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21231         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21232         New prototype.
21233         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21234         Add note to error message to explain internal mapping of overloaded
21235         built-in function name to non-overloaded built-in function name.
21236         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21237         function.
21239 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21241         PR target/85424
21242         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21243         where the inputs overlap with the output.
21245 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21247         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21248         (=v, v) alternative and explicit "memory" attribute.
21249         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21250         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21251         attributes.
21252         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21253         "sselog1" type instead of "sselog".
21254         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21255         "sselog".  Remove explicit "memory" attribute.
21256         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21257         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21258         attributes.
21259         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21260         "sselog1" type instead of "sselog".  Remove explicit "memory"
21261         attribute.
21262         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21263         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21264         attribute.
21265         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21266         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21267         attributes.
21268         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21269         "sselog1" type instead of "sselog".  Remove explicit "memory"
21270         attribute.
21271         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21272         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21273         attribute.
21275         PR target/85430
21276         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21278         PR middle-end/85414
21279         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21280         on a SUBREG.
21282 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21284         PR ipa/85421
21285         * ipa-cp.c (create_specialized_node): Call
21286         expand_all_artificial_thunks if necessary.
21288 2018-04-17  Martin Liska  <mliska@suse.cz>
21290         PR lto/85405
21291         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21292         in message, remote space in between '_G' and '('.
21294 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21296         PR target/85281
21297         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21298         avx512f_vmcmp<mode>3<round_saeonly_name>,
21299         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21300         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21301         avx512f_rndscale<mode><round_saeonly_name>,
21302         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21303         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21304         Use %<iptr>2 instead of %2 for -masm=intel.
21305         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21306         avx512f_vcvttss2usi<round_saeonly_name>,
21307         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21308         -masm=intel.
21309         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21310         avx512f_vcvttsd2usi<round_saeonly_name>,
21311         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21312         Use %q1 instead of %1 for -masm=intel.
21313         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21314         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21315         of %3 for -masm=intel.
21316         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21317         -masm=intel.
21318         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21319         -masm=intel.
21320         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21321         -masm=intel.
21322         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21323         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21324         %g1.
21325         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21326         -masm=intel.
21327         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21328         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21329         %g1 and one with %0 and %1.
21330         (avx512er_vmrcp28<mode><round_saeonly_name>,
21331         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21332         %1 for -masm=intel.
21333         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21334         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21335         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21336         of %0 and %{%4%} for -masm=intel.
21337         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21338         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21339         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21340         order of %0 and %{%5%}%{z%} for -masm=intel.
21342 2018-04-17  Jan Hubicka  <jh@suse.cz>
21344         PR lto/85405
21345         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21347 2018-04-17  Martin Liska  <mliska@suse.cz>
21349         PR ipa/85329
21350         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21351         for target_clone error message.  Make default implementation
21352         clone to be a local declaration.
21353         (separate_attrs): Add new argument and check for an empty
21354         string.
21355         (expand_target_clones): Handle it.
21356         (ipa_target_clone): Make redirection just for target_clones
21357         functions.
21359 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21360             Tom de Vries  <tom@codesourcery.com>
21362         PR middle-end/84955
21363         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21364         tiled basic blocks without omp continue statements.
21366 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21368         PR target/83660
21369         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21370         vec_extract expression as having side effects to make sure it gets
21371         a cleanup point.
21373 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21375         PR target/85403
21376         * config/i386/i386.c (get_builtin_code_for_version): Check
21377         error_mark_node.
21379 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21381         PR target/84331
21382         * config.gcc: Support "skylake".
21383         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21384         PROCESSOR_SKYLAKE.
21385         * config/i386/i386.c (m_SKYLAKE): Define.
21386         (processor_target_table): Add "skylake".
21387         (ix86_option_override_internal): Add "skylake".
21388         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21389         PROCESSOR_CANNONLAKE.
21390         (get_builtin_code_for_version): Fix priority for
21391         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21392         PROCESSOR_SKYLAKE-AVX512.
21393         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21394         (processor_type): Add PROCESSOR_SKYLAKE.
21396 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21397             Jason Merrill  <jason@redhat.com>
21399         PR c++/85112
21400         * convert.c (convert_to_integer_1): Use direct recursion for
21401         enumeral types and types with a precision less than the number
21402         of bits in their mode.
21404 2018-04-16  Julia Koval  <julia.koval@intel.com>
21406         PR target/84413
21407         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21408         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21410 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21412         PR target/85293
21413         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21414         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21415         and -mno-direct-move.
21417 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21419         PR target/83402
21420         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21421         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21422         Ensure negative shifts result in {0}.
21424 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21426         PR rtl-optimization/79916
21427         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21428         regs (if any) to define how to gnerate SD moves when LRA is in
21429         progress.
21431 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21433         PR rtl-optimization/85393
21434         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21435         * except.c (expand_dw2_landing_pad_for_region): Make static.
21436         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21437         a label and unconditional jump to old_bb, rather than
21438         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21439         basic block.
21441         PR rtl-optimization/85376
21442         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21443         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21444         instead of a specific value.
21446 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21447             Bin Cheng  <bin.cheng@arm.com>
21449         PR tree-optimization/82965
21450         PR tree-optimization/83991
21451         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21452         by_profile_only parameter.
21453         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21454         information if the loop was predicted to iterate too many times.
21455         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21457 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21459         PR lto/71991
21460         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21461         always inline.
21463 2018-04-13  Martin Liska  <mliska@suse.cz>
21464             Jakub Jelinek  <jakub@redhat.com>
21466         PR middle-end/81657
21467         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21468         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21469         * builtins.c (expand_builtin_memory_copy_args): Use
21470         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21471         handle dest_addr == pc_rtx.
21473 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21475         PR target/85291
21476         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21477         asked to not generate direct moves.
21478         (fix_trunc<mode>si2_stfiwx): Similar.
21479         (fix_trunc<mode>si2_internal): Similar.
21481 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21483         PR debug/83157
21484         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21485         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21486         lookup if dest in some wider mode is known to be const0_rtx and
21487         if so, record permanent equivalence for it to be ZERO_EXTEND of
21488         the narrower mode destination.
21490 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21492         * lto-streamer-out.c (output_function): Revert 259346.
21493         * omp-expand.c (expand_oacc_for): Likewise.
21495 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21497         PR rtl-optimization/85354
21498         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21499         * sel-sched.c (sel_global_init): ... here.
21501 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21503         PR target/85238
21504         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21505         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21506         mode for PE-COFF targets.
21507         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21508         (i386_pe_asm_lto_end): Likewise.
21509         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21510         (TARGET_ASM_LTO_END): Likewise.
21511         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21512         (i386_pe_asm_lto_start): New function.
21513         (i386_pe_asm_lto_end): Likewise.
21515 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21516             Richard Biener  <rguenther@suse.de>
21518         PR middle-end/84955
21519         * lto-streamer-out.c (output_function): Fix CFG loop state before
21520         streaming out.
21521         * omp-expand.c (expand_oacc_for): Handle calls to internal
21522         functions like regular functions.
21524 2018-04-12  Richard Biener  <rguenther@suse.de>
21526         PR lto/85371
21527         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21528         for the early LTO debug to properly generate references to it
21529         during DIE emission.  Do not re-use that for the skeleton for
21530         split-dwarf.
21531         (dwarf2out_early_finish): Likewise.
21533 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21535         PR target/85328
21536         * config/i386/sse.md
21537         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21538         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21539         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21540         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21541         and output is a reg, avoid creating invalid lowpart subreg, but
21542         instead split into a 512-bit move.  Don't split if not AVX512VL,
21543         input is xmm16+ reg and output is a mem.
21544         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21545         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21546         xmm16+ reg and output is a mem.
21548 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21550         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21551         also for flag_dwarf2_cfi_asm.
21553 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21555         PR rtl-optimization/85342
21556         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21557         a bool scalar var inside of the loop instead.  Don't try to update
21558         recog_data.operand after failed apply_change_group.
21560 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21562         PR target/85296
21563         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21564         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21565         array with flexible array member as array without given dimension.
21566         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21567         argument for undefined param to true.
21569 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21571         PR target/85321
21572         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21573         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21574         from PowerPC section.
21575         * config/rs6000/sysv4.opt (mcall-): Improve help text.
21576         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21577         help text that is too long.
21578         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
21579         help text that is too long.
21580         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
21581         help text that is too long.
21583 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
21585         * config/alpha/alpha.md (stack_probe_internal): Rename
21586         from "probe_stack".  Update all callers.
21588 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21590         PR rtl-optimization/84566
21591         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
21592         sched_macro_fuse_insns.
21594 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21596         PR target/84301
21597         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
21598         (compute_block_dependences): ... from here.
21600 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21602         PR tree-optimization/85331
21603         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
21604         from int to HOST_WIDE_INT.
21606 2018-04-11  Martin Jambor  <mjambor@suse.cz>
21608         PR ipa/84149
21609         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
21610         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
21611         not the same as the source val.
21612         (cgraph_edge_brings_value_p): New parameter.
21613         (gather_edges_for_value): Pass destination value to
21614         cgraph_edge_brings_value_p.
21615         (perhaps_add_new_callers): Likewise.
21616         (get_info_about_necessary_edges): Likewise and exclude values brought
21617         only by self-recursive edges.
21618         (create_specialized_node): Redirect only clones of self-calling edges.
21619         (+self_recursive_pass_through_p): New function.
21620         (find_more_scalar_values_for_callers_subset): Use it.
21621         (find_aggregate_values_for_callers_subset): Likewise.
21622         (known_aggs_to_agg_replacement_list): Removed.
21623         (decide_whether_version_node): Re-calculate known constants for all
21624         remaining context clones.
21626 2018-04-11  Richard Biener  <rguenther@suse.de>
21628         PR lto/85339
21629         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
21630         from early DWARF output.
21631         (dwarf2out_early_finish): Output line info unconditionally into
21632         early DWARF and add reference to it.
21634 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21636         PR target/85281
21637         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
21638         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
21639         other than V2DFmode using iptr mode attribute.
21640         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
21642 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21644         PR rtl-optimization/84659
21645         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
21647 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21649         PR debug/85302
21650         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
21651         SIZEP is NULL.
21652         (output_loc_list): Pass address of a dummy size variable even in the
21653         locview handling loop.
21654         (index_location_lists): Add comment on why skip_loc_list_entry can't
21655         call size_of_locs.
21657 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21659         PR target/85261
21660         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
21661         into register.
21663 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
21665         PR target/85321
21666         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21667         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
21668         and -mstring-compare-inline-limit.
21670 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21672         PR target/85287
21673         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
21674         for stack clash protection in a register whenever we need it to be in
21675         a register.
21677 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21679         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
21680         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
21682 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21684         PR target/85321
21685         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
21686         the help text.
21687         (mlong-double-): Ditto.
21688         * config/rs6000/sysv4.opt (msdata=): Ditto.
21689         (mtls-size=): Ditto.
21691 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21693         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
21694         erroneous entries for
21695         "vector int vec_ldl (int, long int *)", and
21696         "vector unsigned int vec_ldl (int, unsigned long int *)".
21697         Add comments and entries for
21698         "vector bool char vec_ldl (int, bool char *)",
21699         "vector bool short vec_ldl (int, bool short *)",
21700         "vector bool int vec_ldl (int, bool int *)",
21701         "vector bool long long vec_ldl (int, bool long long *)",
21702         "vector pixel vec_ldl (int, pixel *)",
21703         "vector long long vec_ldl (int, long long *)",
21704         "vector unsigned long long vec_ldl (int, unsigned long long *)".
21705         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
21706         type tree bool_long_long_type_node and correct definition of
21707         bool_V2DI_type_node to make reference to this new type tree.
21708         (rs6000_mangle_type): Replace erroneous reference to
21709         bool_long_type_node with bool_long_long_type_node.
21710         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
21711         comments to emphasize sign distinctions for char and int types and
21712         replace RS6000_BTI_bool_long constant with
21713         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
21714         use of RS6000_BTI_pixel.
21715         (bool_long_type_node): Remove this macro definition.
21716         (bool_long_long_type_node): New macro definition
21718 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21720         PR rtl-optimization/85300
21721         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
21722         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
21723         simplify_unary_operation fails.
21725 2018-04-10  Martin Liska  <mliska@suse.cz>
21727         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
21728         cgraph_edge and ipa_ref.
21730 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21732         PR target/85177
21733         PR target/85255
21734         * config/i386/sse.md
21735         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
21736         computation of the VEC_MERGE selector from mask.
21737         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
21738         Fix decoding of the VEC_MERGE selector into mask.
21740 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21742         PR tree-optimization/85286
21743         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
21745 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21747         * final.c (final_1): Set insn_last_address as well as
21748         insn_current_address.
21750 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21752         PR target/85173
21753         * explow.c (emit_stack_probe): Call validize_mem on memory location
21754         before passing it to gen_probe_stack.  Create address operand and
21755         legitimize it for the probe_stack_address case.
21757 2018-04-09  Jan Hubicka  <jh@suse.cz>
21759         PR lto/85078
21760         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
21761         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
21762         * tree.c (free_lang_data_in_type): Fix handling of binfos;
21763         walk basetypes.
21764         (free_lang_data): Rebuild type inheritance graph.
21766 2018-04-09  Martin Sebor  <msebor@redhat.com>
21768         * invoke.texi (-finline-small-functions): Mention other optimization
21769         options.
21770         (-findirect-inlining, -fpartial-inlining): Same.
21771         (-finline-functions-called-once): Same.
21772         (-freorder-blocks-and-partition): Same.
21774 2018-04-09  Jan Hubicka  <jh@suse.cz>
21776         PR rtl/84058
21777         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
21778         jumps; choose last target that matches the criteria (i.e.
21779         no partition changes for non-crossing jumps).
21780         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
21781         support for redirecting crossing jumps to non-crossing.
21783 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
21785         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
21786         also for naked functions.
21788 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
21790         * config/arc/arc.md (add_shift): New pattern.
21791         (add_shift2): Likewise.
21792         (sub_shift): Likewise.
21793         (sub_shift_cmp0_noout): Likewise.
21794         (compare_si_ashiftsi): Likewise.
21795         (xbfu_cmp0_noout): New combine pattern.
21796         (xbfu_cmp0"): Likewise.
21797         (movsi_set_cc_insn): Place the predicable variant first.
21798         (commutative_binary_cmp0_noout): Remove clobber.
21799         (commutative_binary_cmp0): New pattern.
21800         (noncommutative_binary_cmp0): Likewise.
21801         (noncommutative_binary_cmp0_noout): Likewise.
21802         (noncommutative_binary_comparison_result_used): Removed.
21803         (rsub_cmp0): New pattern.
21804         (rsub_cmp0_noout): Likewise.
21805         (extzvsi): Changed, keep only meaningful variants.
21806         (SQH, SEZ): New iterators.
21807         (SQH_postfix): New mode attribute.
21808         (SEZ_prefix): New code attribute.
21809         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
21810         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
21811         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
21812         of numerical value.
21813         (noncommutative_operator): Check the availability of barrel
21814         shifter option.
21816 2018-04-09  Richard Biener  <rguenther@suse.de>
21818         PR tree-optimization/85284
21819         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
21820         Only use the niter constraining form of simple_iv when the exit
21821         is always executed.
21823 2018-04-09  Tom de Vries  <tom@codesourcery.com>
21825         PR target/84041
21826         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21827         (define_expand "*memory_barrier"): New define_expand.
21828         (define_insn "memory_barrier"): New insn.
21830 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21832         PR rtl-optimization/80463
21833         PR rtl-optimization/83972
21834         PR rtl-optimization/83480
21836         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
21837         correct producer for the insn.
21838         (tidy_control_flow): Fixup seqnos in case of debug insns.
21840 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21842         PR rtl-optimization/83913
21844         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
21845         different sched-times when merging exprs.
21847 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21849         PR rtl-optimization/83962
21851         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
21852         tidy_fallthru_edge and tidy_control_flow.
21854 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21856         PR rtl-optimization/83530
21858         * sel-sched.c (force_next_insn): New global variable.
21859         (remove_insn_for_debug): When force_next_insn is true, also leave only
21860         next insn in the ready list.
21861         (sel_sched_region): When the region wasn't scheduled, make another pass
21862         over it with force_next_insn set to 1.
21864 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21866         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
21867         into tm_file.
21868         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21869         for interrupt control.
21870         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21871         functions for interrupt control.
21872         * config/nds32/nds32-intrinsic.md: Likewise.
21873         * config/nds32/nds32_intrinsic.h: Likewise.
21874         * config/nds32/nds32.h (nds32_builtins): Likewise.
21876 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21878         * config/nds32/nds32.c (nds32_init_machine_status,
21879         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
21880         strict_aligned_p field.
21881         (nds32_expand_to_rtl_hook): New function.
21882         (TARGET_EXPAND_TO_RTL_HOOK): Define.
21883         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
21885 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21886             Chung-Ju Wu  <jasonwucj@gmail.com>
21888         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
21889         * config/nds32/nds32-n7.md: New file.
21890         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
21891         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
21892         pipeline.
21893         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
21894         * config/nds32/nds32.md (pipeline_model): Add n7.
21895         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21896         * config/nds32/pipelines.md: Include n7 settings.
21898 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21899             Chung-Ju Wu  <jasonwucj@gmail.com>
21901         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
21902         * config/nds32/nds32-e8.md: New file.
21903         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
21904         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
21905         pipeline.
21906         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
21907         * config/nds32/nds32.md (pipeline_model): Add e8.
21908         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
21909         * config/nds32/pipelines.md: Include e8 settings.
21911 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21912             Chung-Ju Wu  <jasonwucj@gmail.com>
21914         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
21915         * config/nds32/nds32-n8.md: New file.
21916         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
21917         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
21918         pipeline.
21919         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
21920         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
21921         * config/nds32/nds32.md (pipeline_model): Add n8.
21922         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
21923         * config/nds32/pipelines.md: Include n8 settings.
21925 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21926             Chung-Ju Wu  <jasonwucj@gmail.com>
21928         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
21929         * config/nds32/nds32-n9-2r1w.md: New file.
21930         * config/nds32/nds32-n9-3r2w.md: New file.
21931         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
21932         nds32_register_ports): New or modify for cpu n9.
21933         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
21934         pipeline.
21935         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
21936         * config/nds32/nds32-utils.c: New file.
21937         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
21938         TARGET_MUL_SLOW): Define.
21939         * config/nds32/nds32.md (pipeline_model): New attribute.
21940         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
21941         New options that support cpu n9.
21942         * config/nds32/pipelines.md: Include n9 settings.
21943         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
21945 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21947         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
21948         information if necessary.
21949         (output_cond_branch_compare_zero): Likewise.
21950         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
21951         (nds32_target_alignment): Refine for alignment.
21952         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
21953         (FUNCTION_BOUNDARY): Modify.
21954         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
21955         align case.
21956         * config/nds32/nds32.opt (malways-align, malign-functions): New.
21958 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21960         * config/nds32/constants.md (unspec_volatile_element): Add values for
21961         TLB operation and data prefetch.
21962         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21963         functions for TLB operation and data prefetch.
21964         * config/nds32/nds32-intrinsic.md: Likewise.
21965         * config/nds32/nds32_intrinsic.h: Likewise.
21966         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
21967         (nds32_print_operand): Likewise.
21968         * config/nds32/nds32.h (nds32_builtins): Likewise.
21970 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21971         Andrew Pinski <pinsika@gcc.gnu.org>
21973         PR middle-end/82976
21974         * match.pd: Use constant_boolean_node of correct type instead of
21975         boolean_true_node or boolean_false_node for simplifying
21976         pointer comparisons to zero.
21978 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21980         PR tree-optimization/80021
21981         * tree.c (verify_type_variant): Make error call in verify_variant_match
21982         translatable and remove final full stop.
21984 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21986         * config/nds32/constants.md (unspec_volatile_element): Add
21987         UNSPEC_VOLATILE_EH_RETURN.
21988         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
21989         nds32_output_stack_pop): Support dwarf exception handling process.
21990         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
21991         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
21992         exception handling process.
21993         (nds32_compute_stack_frame): Likewise.
21994         (nds32_return_addr_rtx): Likewise.
21995         (nds32_initial_elimination_offset): Likewise.
21996         (nds32_expand_prologue): Likewise.
21997         (nds32_expand_epilogue): Likewise.
21998         (nds32_dynamic_chain_address): New function.
21999         * config/nds32/nds32.h (machine_function): Add fields for dwarf
22000         exception handling.
22001         (DYNAMIC_CHAIN_ADDRESS): Define.
22002         (EH_RETURN_DATA_REGNO): Define.
22003         (EH_RETURN_STACKADJ_RTX): Define.
22004         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22005         patterns for dwarf exception handling.
22007 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22009         * config/nds32/nds32.h: Clean up obsolete macros.
22011 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22013         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22014         Add enum values for particular instructions.
22015         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22016         particular intrinsic functions.
22017         * config/nds32/nds32-intrinsic.md: Likewise.
22018         * config/nds32/nds32_intrinsic.h: Likewise.
22019         * config/nds32/nds32.h (nds32_builtins): Likewise.
22020         * config/nds32/nds32.md (type): Add pbsad and pbsada.
22021         (btst, ave): New patterns for particular instructions.
22023 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22025         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22026         Add enum values for atomic load/store and memory sync.
22027         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22028         and memory sync.
22029         * config/nds32/nds32-intrinsic.md: Likewise.
22030         * config/nds32/nds32_intrinsic.h: Likewise.
22031         * config/nds32/nds32.h (nds32_builtins): Likewise.
22033 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22035         PR tree-optimization/85257
22036         * fold-const.c (native_encode_vector): If not all elts could fit
22037         and off is -1, return 0 rather than offset.
22038         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22039         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22040         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
22041         adjust buffer in native_interpret_expr call.
22043 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22045         * config/nds32/constants.md (unspec_volatile_element): Add cache
22046         control enum values.
22047         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22048         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22049         * config/nds32/nds32.c (nds32_cctl_names): New.
22050         (nds32_print_operand): Handle cache control register names.
22051         * config/nds32/nds32.h (nds32_builtins): New enum values.
22052         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22053         macros.
22054         * config/nds32/nds32.md (type): Add mmu.
22055         * config/nds32/pipelines.md (simple_insn): Add mmu.
22057 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22059         * config/nds32/nds32.md (type): Remove call.
22060         * config/nds32/pipelines.md (simple_insn): Likewise.
22062 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22064         * config/nds32/constants.md (unspec_volatile_element): Add
22065         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22066         UNSPEC_VOLATILE_FMFCFG.
22067         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22068         description for fmfcfg and fmfcsr.
22069         (bdesc_1arg): Add fmtcsr.
22070         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22071         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22072         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22073         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22074         unspec_fmfcfg): New patterns.
22075         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22076         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22077         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22078         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22079         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22080         __nds32__fmfcfg): Define.
22082 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22084         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22085         intrinsic register names.
22086         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22087         intrinsic register enum values and macros.
22089 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22091         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22092         for load/store addressing form.
22093         (nds32_print_operand_address): Likewise.
22095 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22097         PR target/85196
22098         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22099         based on LABEL_REF.  Remove useless assertion.
22100         (pic_address_needs_scratch): Fix formatting.
22101         (sparc_legitimize_pic_address): Minor tweaks.
22102         (sparc_delegitimize_address): Adjust assertion accordingly.
22103         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22104         into symbolic_operand.
22105         (movsi_high_pic_label_ref): Likewise.
22106         (movsi_lo_sum_pic_label_ref): Likewise.
22107         (movdi_pic_label_ref): Likewise.
22108         (movdi_high_pic_label_ref): Likewise.
22109         (movdi_lo_sum_pic_label_ref): Likewise.
22111 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22113         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22114         custom LIB_SPEC setup.
22116 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22117             Kito Cheng  <kito.cheng@gmail.com>
22119         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22120         * config/riscv/freebsd.h: New.
22122 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22124         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22125         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22126         file.
22128 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22129             Kito Cheng  <kito.cheng@gmail.com>
22131         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22132         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22133         * config/nds32/nds32-protos.h (nds32_output_call,
22134         nds32_output_return): Declare.
22135         * config/nds32/nds32.md: Refine all the call and return patterns.
22137 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22139         PR debug/85252
22140         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22141         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22143         PR rtl-optimization/84872
22144         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22145         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22146         EDGE_CROSSING edge.
22148 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22150         * expr.c (copy_blkmode_to_reg): Revert 254862.
22151         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22153 2018-04-06  Richard Biener  <rguenther@suse.de>
22155         PR middle-end/85244
22156         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22157         after seeing a component reference with an adjacent field.  Treat
22158         refs to arrays at struct end of external decls similar to
22159         refs to unconstrained commons.
22161 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22163         PR sanitizer/85213
22164         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22165         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22166         recursive calls.
22167         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22168         save_p here.
22170 2018-04-06  Richard Biener  <rguenther@suse.de>
22172         PR middle-end/85180
22173         * alias.c (find_base_term): New wrapper around find_base_term
22174         unwinding CSELIB_VAL_PTR changes.
22175         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22176         recursion.
22178 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22180         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22181         instructions.
22182         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22183         constant definitions.
22184         ("nop"): lr 0,0 -> nopr r0
22185         ("nop_lr0", "nop_lr1"): New insn definitions.
22187 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22189         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22190         NDS32_V3PUSH_AVAILABLE_P macro.
22192 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22193             Chung-Ju Wu  <jasonwucj@gmail.com>
22195         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22196         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22197         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22198         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22199         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22200         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22201         * config/nds32/constraints.md: New constraints and checking for hard
22202         float configuration.
22203         * config/nds32/iterators.md: New mode iterator and attribute for hard
22204         float configuration.
22205         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22206         patterns.
22207         * config/nds32/nds32-fpu.md: New file.
22208         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22209         deal with hard float code generation.
22210         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22211         ARCH_V3S.
22212         (abi_type, float_reg_number): New enum type.
22213         * config/nds32/nds32-predicates.c: New predicates for hard float.
22214         * config/nds32/nds32-protos.h: Declare functions for hard float.
22215         * config/nds32/nds32.c: Implementation for hard float configuration.
22216         * config/nds32/nds32.h: Definitions for hard float configuration.
22217         * config/nds32/nds32.md: Include hard float machine description and
22218         modify patterns for hard float configuration.
22219         * config/nds32/nds32.opt: New options for hard float configuration.
22220         * config/nds32/predicates.md: New predicates for hard float
22221         configuration.
22223 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22225         * common/config/nds32/nds32-common.c
22226         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22228 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22230         PR middle-end/85195
22231         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22232         CONSTRUCTOR_ELT (ctor, ...)->value.
22234 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22236         PR target/85193
22237         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22239 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22241         PR target/85204
22242         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22243         cond jump.
22245 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22246             Kito Cheng  <kito.cheng@gmail.com>
22248         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22249         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22250         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22251         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22253 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22254             Kito Cheng  <kito.cheng@gmail.com>
22256         * config/nds32/constraints.md (Ufe): New memory constraint.
22257         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22258         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22259         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22260         operands.
22261         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22262         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22264 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22266         * config/nds32/nds32.md: Use optimize_size in the condition for
22267         alu-shift instructions.
22269 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22271         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22273 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22275         * config/nds32/nds32.md (negsi2): Refine pattern.
22277 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22278             Chung-Ju Wu  <jasonwucj@gmail.com>
22280         * config/nds32/iterators.md (shift_rotate): New code iterator.
22281         (shift): New code attribute.
22282         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22283         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22284         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22285         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22286         bit-wise operations.
22287         (andsi3, *andsi3): Ditto.
22288         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22289         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22290         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22291         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22292         nds32_ior_operand, nds32_xor_operand): New predicates.
22294 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22296         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22297         (addsi3, subsi3): ... this.
22299 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22301         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22303 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22305         * config/nds32/nds32.md: Adjust indention.
22307 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22309         * config/nds32/nds32.md (feature): New attribute.
22311 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22313         * config/nds32/nds32.md (subtype): New attribute.
22315 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22317         PR target/85203
22318         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22319         expansion to perform a bitwise AND of the argument followed by a
22320         boolean negation of the result.
22322 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22324         PR rtl-optimization/84878
22325         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22326         the basic block.  Assert the use reference is not artificial and that
22327         it has an associated insn.
22329 2018-04-04  Michael Matz  <matz@suse.de>
22331         * builtins.c (compute_objsize): Pass correct operand
22332         to array_at_struct_end_p.
22334 2018-04-04  Richard Biener  <rguenther@suse.de>
22336         PR lto/85176
22337         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22338         from contexts for DINFO_LEVEL_TERSE and below.
22340 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22342         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22343         resiter_operand condition.
22344         * config/nds32/nds32.md (*move<mode>): Ditto.
22346 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22347             Monk Chiang  <sh.chiang04@gmail.com>
22349         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22351 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22353         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22355 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22356             Kito Cheng  <kito.cheng@gmail.com>
22358         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22359         nds32_cond_code_str, output_cond_branch,
22360         output_cond_branch_compare_zero, nds32_expand_cbranch,
22361         nds32_expand_cstore, nds32_expand_movcc,
22362         nds32_output_cbranchsi4_equality_zero,
22363         nds32_output_cbranchsi4_equality_reg,
22364         nds32_output_cbranchsi4_equality_reg_or_const_int,
22365         nds32_output_cbranchsi4_greater_less_zero: New functions.
22366         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22367         nds32_expand_cstore, nds32_expand_movcc,
22368         nds32_output_cbranchsi4_equality_zero,
22369         nds32_output_cbranchsi4_equality_reg,
22370         nds32_output_cbranchsi4_equality_reg_or_const_int,
22371         nds32_output_cbranchsi4_greater_less_zero): Declare.
22372         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22373         nds32_rimm11s_operand): New predicates.
22374         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22375         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22376         patterns.
22378 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22380         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22381         * config/nds32/nds32.md: Ditto.
22382         * config/nds32/pipelines.md: Ditto.
22384 2018-04-04  Richard Biener  <rguenther@suse.de>
22386         PR tree-optimization/85168
22387         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22388         propagating abnormals.
22390 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22392         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22394 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22395             Kito Cheng  <kito.cheng@gmail.com>
22397         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22398         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22399         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22400         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22401         * config/nds32/nds32.md (sibcall_internal): New.
22402         (sibcall_register): Remove.
22403         (sibcall_immediate): Remove.
22404         (sibcall_value_internal): New.
22405         (sibcall_value_register): Remove.
22406         (sibcall_value_immediate): Remove.
22407         * config/nds32/predicates.md (nds32_general_register_operand): New.
22408         (nds32_call_address_operand): New.
22410 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22412         PR rtl-optimization/85167
22413         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22414         bb_defs if *split_p, instead preinitialize it to NULL.
22416         PR tree-optimization/85156
22417         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22418         evaluating the argument multiple times.
22420 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22422         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22423         than vector.
22424         (_mm_cvtpd_ps): Likewise.
22425         (_mm_cvttpd_epi32): Likewise.
22426         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22427         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22428         vector, pixel, and bool following altivec.h include.
22430 2018-04-03  Martin Sebor  <msebor@redhat.com>
22432         * doc/extend.texi (Common Function Attributes): Clarify.
22433         (const attribute): Likewise.
22434         (pure attribute): Likewise.
22436 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22438         PR target/85169
22439         * config/i386/i386.c (ix86_expand_vector_set): Use
22440         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22442 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22444         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22445         instructions when changing rounding bits to preserve precision bits
22446         in the x87 control word.
22448 2018-04-03  Martin Liska  <mliska@suse.cz>
22450         PR tree-optimization/82491
22451         * rtl.h (strip_offset_and_add): Replace += suboffset with
22452         poly_uint64 () + suboffset.
22454 2018-03-29  Martin Liska  <mliska@suse.cz>
22455             Martin Jambor  <mjambor@suse.cz>
22457         PR ipa/84947
22458         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22459         param_type is not an integral or pointer type.
22461 2018-04-03  Richard Biener  <rguenther@suse.de>
22463         * sese.h (recompute_all_dominators): Remove.
22465 2018-04-02  Martin Sebor  <msebor@redhat.com>
22467         * doc/invoke.texi (-Wrestrict): Fix typos.
22469 2018-04-02  Jim Wilson  <jimw@sifive.com>
22471         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22472         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22473         (<optab>di3, <optab>si3_extend): Likewise.
22474         (<optab>si3_mask, <optab>si3_mask_1): New.
22475         (<optab>di3_mask, <optab>di3_mask_1): New.
22476         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22477         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22478         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22480 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22482         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22483         example.
22485 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22487         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22488         (nds32_canonicalize_comparison): New function.
22490 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22491             Kito Cheng  <kito.cheng@gmail.com>
22492             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22494         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22495         * config/nds32/constants.md (unspec_volatile_element): Add
22496         UNSPEC_VOLATILE_RELAX_GROUP.
22497         * config/nds32/nds32-relax-opt.c: New file.
22498         * config/nds32/nds32-predicates.c
22499         (nds32_symbol_load_store_p): New function.
22500         * config/nds32/nds32-protos.h
22501         (nds32_symbol_load_store_p): Declare function.
22502         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22503         * config/nds32/nds32.c
22504         (nds32_register_pass): New function to register pass.
22505         (nds32_register_passes): New function to register passes.
22506         * config/nds32/nds32.md (relax_group): New pattern.
22507         * config/nds32/nds32.opt (mrelax-hint): New option.
22508         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22510 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22512         * config/nds32/t-nds32: Modify files dependency.
22514 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22516         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22517         (PROFILE_HOOK): Define its implementation.
22519 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22521         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22522         type and 32-bit size.
22524 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22526         PR middle-end/85090
22527         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22528         (V_128_256): New mode iterator.
22529         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22530         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22531         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22532         of V.
22533         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22534         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22536 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22538         PR target/83315
22539         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22540         NaN inputs correctly.
22542 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22544         PR target/80546
22545         * config/rs6000/vsx.md (??r): New mode attribute.
22546         (*vsx_mov<mode>_64bit): Use it.
22547         (*vsx_mov<mode>_32bit): Likewise.
22549 2018-03-30  Martin Sebor  <msebor@redhat.com>
22551         PR tree-optimization/84818
22552         * builtins.c (check_access): Use warning_n.
22554 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22556         PR target/83822
22557         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22558         condition.
22559         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22560         condition.
22562 2018-03-30  Julia Koval  <julia.koval@intel.com>
22564         PR target/84413
22565         * x86-tune.def (movx, partial_reg_dependency): Enable for
22566         m_SKYLAKE_AVX512.
22568 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22570         PR inline-asm/84985
22571         * lra-constraints.c (process_alt_operands): Move setting
22572         this_alternative_matches below.
22574 2018-03-29  Martin Liska  <mliska@suse.cz>
22576         PR lto/84995.
22577         * doc/invoke.texi: Document how LTO works with debug info.
22578         Describe auto-load support of binutils.  Mention 'x86-64'
22579         as valid option value of -march option.
22581 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
22583         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
22585         PR c/85094
22586         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
22587         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
22588         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
22589         checking.
22591 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22593         PR target/84912
22594         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
22595         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
22596         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
22597         for RS6000_BTM_POWERPC64.
22598         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
22599         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
22600         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
22601         definition.
22602         (DIVDE): Use it.
22603         (DIVDEU): Likewise.
22605 2018-03-28  Carl Love  <cel@us.ibm.com>
22607         Revert
22608         2017-09-27  Carl Love  <cel@us.ibm.com>
22610         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
22611         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
22612         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
22613         fctiw instruction.
22615 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22617         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
22618         instead of __vector bool.
22619         (_mm_max_pu8): Likewise.
22620         (_mm_min_pi16): Likewise.
22622 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22624         PR target/84912
22625         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
22626         (DIVWEUO): Likewise.
22627         (DIVDEO): Likewise.
22628         (DIVDEUO): Likewise.
22629         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
22630         DIVWEUO and DIVDEUO.
22631         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
22632         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
22633         (div_extend): Likewise.
22634         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
22635         builtin function.
22636         (__builtin_divweuo): Likewise.
22637         (__builtin_divdeo): Likewise.
22638         (__builtin_divdeuo): Likewise.
22640 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22642         PR target/85095
22643         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
22644         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
22646         PR tree-optimization/82004
22647         * gimple-match-head.c (optimize_pow_to_exp): New function.
22648         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
22649         Don't fold to exp if optimize_pow_to_exp is false.
22651 2018-03-28  Martin Liska  <mliska@suse.cz>
22653         PR other/84819
22654         * calls.c (initialize_argument_information): Fix trailing space.
22655         * common.opt: Fix typo and provide better explanation for
22656         -fsanitize-coverage option.
22657         * config/i386/i386.opt: Fix typo.
22659 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22660             Martin Liska  <mliska@suse.cz>
22662         PR sanitizer/85081
22663         * gimplify.c (asan_poison_variable): Don't do the check for
22664         gimplify_omp_ctxp here.
22665         (gimplify_decl_expr): Do it here.
22666         (gimplify_target_expr): Likewise.
22668 2018-03-28  Martin Liska  <mliska@suse.cz>
22670         PR target/84988
22671         * config/i386/i386.c (ix86_function_arg_advance): Do not call
22672         chkp_type_bounds_count if MPX is not enabled.
22674 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
22676         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
22678 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
22680         PR target/84914
22681         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
22682         function to create the function decl for complex long double
22683         multiply and divide for -mabi=ieeelongdouble.
22684         (init_float128_ieee): Call it.
22686 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
22688         PR target/85044
22689         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
22690         -fcf-protection=branch -mibt.
22691         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
22693 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22695         PR target/81863
22696         * config/arm/arm.c (arm_valid_symbolic_address): Handle
22697         arm_word_relocations.
22699 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
22701         PR target/85056
22702         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
22703         extern array declarations.
22705 2018-03-27  Richard Biener  <rguenther@suse.de>
22707         PR middle-end/84067
22708         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
22709         explicit single_use checks.
22711 2018-03-27  Richard Biener  <rguenther@suse.de>
22713         PR tree-optimization/85082
22714         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
22715         Valueize the VUSE.
22717 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22719         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
22720         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
22721         Turn on fasynchronous-unwind-tables and funwind-tables.
22723 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
22725         PR target/85073
22726         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
22727         (*bmi_blsr_<mode>_ccz): Ditto.
22729 2018-03-26  Tom de Vries  <tom@codesourcery.com>
22731         PR tree-optimization/85063
22732         * omp-general.c (offloading_function_p): New function.  Factor out
22733         of ...
22734         * omp-offload.c (pass_omp_target_link::gate): ... here.
22735         * omp-general.h (offloading_function_p): Declare.
22736         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
22737         with attribute omp declare target for offloading functions.
22739 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
22741         PR tree-optimization/84005
22742         * tree-data-ref.h (get_base_for_alignment): Declare.
22743         * tree-data-ref.c (get_base_for_alignment_1): New function.
22744         (get_base_for_alignment): Likewise.
22745         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
22746         get_base_for_alignment to find a suitable base object, instead
22747         of always using drb->base_address.
22749 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22751         PR inline-asm/85022
22752         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
22753         known size by default.
22755 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
22757         PR inline-asm/85030
22758         * lra-constraints.c (process_alt_operands): Don't match BLKmode
22759         and non BLKmode operands.
22761 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22763         PR target/85026
22764         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
22765         Clean up attributes.
22767 2018-03-23  Richard Biener  <rguenther@suse.de>
22769         PR debug/85020
22770         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
22771         we are going to emit early debug for LTO.
22773 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22775         PR inline-asm/85034
22776         * function.c (match_asm_constraints_1): Don't optimize if input
22777         doesn't satisfy general_operand predicate for output's mode.
22779         PR inline-asm/85022
22780         * alias.c (write_dependence_p): Don't require for x_canonicalized
22781         non-VOIDmode if x has VOIDmode.
22783         PR sanitizer/85029
22784         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
22785         just don't try to optimize it rather than assert it never happens.
22787 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22789         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
22790         macro expansions for definition of ST_INTERNAL_<mode> and
22791         LD_INTERNAL_<mode> builtins.
22792         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
22793         Remove prototype.
22794         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
22795         function.
22796         (altivec_expand_st_builtin): Likewise.
22797         (altivec_expand_builtin): Remove calls to deleted functions.
22798         (rs6000_address_for_altivec): Delete this function.
22799         * config/rs6000/vector.md: Remove expands for
22800         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
22802 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
22804         PR target/84826
22805         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
22806         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
22807         re-computing once computed.
22808         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
22809         (arm_init_machine_status): Initialize
22810         machine->static_chain_stack_bytes.
22812 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22814         PR target/84760
22815         * doc/extend.texi: Add four new prototypes for vec_ld.
22816         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
22817         definitions for more logical presentation.
22818         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
22819         entries for V1TI variants of __builtin_altivec_ld builtin.
22820         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
22821         handling of V1TI variant of LVX icode pattern.
22822         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
22823         (rs6000_gimple_fold_builtin): Likewise.
22824         (altivec_init_builtins): Add code to define
22825         __builtin_altivec_lvx_v1ti function.
22827 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22829         PR inline-asm/84941
22830         * function.c (match_asm_constraints_1): Don't do the optimization
22831         if input isn't a REG, SUBREG, MEM or constant.
22833 2018-03-22  Tom de Vries  <tom@codesourcery.com>
22835         PR tree-optimization/84956
22836         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
22837         bb_has_abnormal_pred.
22839 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22841         PR sanitizer/85018
22842         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
22843         DECL_INITIAL (decl) to decl at the end.
22844         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
22845         adjust the comment.
22847 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
22849         * doc/extend.texi (__builtin_tgmath): Document when complex
22850         integer types are treated as _Complex _Float64.
22852 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22854         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
22856 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22858         PR tree-optimization/84960
22859         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
22860         if it is ENTRY block, move them into single succ of ENTRY in that case.
22862 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
22864         PR tree-optimization/84811
22865         * poly-int.h (poly_span_traits): Remove the T3 parameter and
22866         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
22867         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
22868         (known_subrange_p): Update accordingly.  Cast each value involved
22869         in the size comparison, rather than casting the result of the
22870         subtraction.
22872 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22874         PR tree-optimization/84982
22875         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
22876         by flipping the least significant bit rather than all bits from
22877         bitpos to bitpos + bitsize - 1.
22879 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22881         * doc/extend.texi (Deprecated Features): Remove mention of
22882         long-deleted deprecations.
22884 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22886         PR jit/84288
22887         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
22888         * configure: Regenerate.
22890 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22892         PR tree-optimization/83126
22893         * tree-parloops.c (num_phis): New function.
22894         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
22896 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22898         * doc/extend.texi (Deprecated Features): Update deprecated flags,
22899         mention anon-struct/union members and trailing attributes.
22901 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
22903         PR tree-optimization/84969
22904         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
22905         builtin memset partitions if they set different rhs values.
22907 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22909         PR rtl-optimization/84989
22910         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
22911         VEC_DUPLICATE with scalar result mode.
22913 2018-03-21  Martin Liska  <mliska@suse.cz>
22915         PR ipa/84963
22916         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
22917         not intended return statement.
22919 2018-03-21  Martin Liska  <mliska@suse.cz>
22921         PR target/84988
22922         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
22923         (chkp_find_bound_slots_1): Limit number of iterations.
22925 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
22927         PR target/84838
22928         * Minor grammar fixes for x86 options.
22930 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22932         PR debug/84875
22933         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
22934         holding REG_CFA_RESTORE notes, instead turn them into a USE.
22936 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
22938         PR target/83789
22939         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
22940         (altivec_lvx_<mode>_1op): Likewise.
22941         (altivec_stvx_<mode>_2op): Likewise.
22942         (altivec_stvx_<mode>_1op): Likewise.
22943         (altivec_lvx_<VM2:mode>): New define_expand.
22944         (altivec_stvx_<VM2:mode>): Likewise.
22945         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
22946         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22947         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
22948         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22949         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
22950         (rs6000_gen_lvx): Likewise.
22951         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
22952         (altivec_expand_stv_builtin): Likewise.
22953         (altivec_expand_builtin): Likewise.
22954         * config/rs6000/vector.md: Likewise.
22956 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22958         PR target/82518
22959         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
22960         BYTES_BIG_ENDIAN.
22962 2018-03-20  Richard Biener  <rguenther@suse.de>
22964         PR target/84986
22965         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
22966         sign-conversions as zero, fall back to standard scalar_stmt
22967         cost for the rest.
22969 2018-03-20  Martin Liska  <mliska@suse.cz>
22971         PR ipa/84825
22972         * predict.c (rebuild_frequencies): Handle case when we have
22973         PROFILE_ABSENT, but flag_guess_branch_prob is false.
22975 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22977         PR target/84990
22978         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
22979         flag_section_anchors.
22980         * varasm.c (use_blocks_for_decl_p): Remove hack for
22981         dw2_force_const_mem.
22983         PR target/84845
22984         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
22985         to ...
22986         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
22987         be created, use lowpart_subreg of operands[0] rather than operands[0]
22988         itself.
22989         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
22990         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
22991         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
22992         and n constraint instead of aarch64_shift_imm_di and Usd.
22993         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
22994         (*aarch64_<optab>_reg_minus<mode>3): ... this.
22996 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
22998         PR target/82989
22999         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23000         to favor GPR over NEON registers.
23001         (<shift>di3_neon): Likewise.
23003 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23005         PR target/84952
23006         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23007         (nvptx_process_pars): Emit bar.sync asap and alap.
23009 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23011         PR target/84954
23012         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23013         seen_label if seen_label is already set.
23015 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23017         PR target/84945
23018         * config/i386/i386.c (fold_builtin_cpu): For features above 31
23019         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23020         Use 1U instead of 1.  Formatting fixes.
23022         PR c/84953
23023         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23024         instead of TREE_TYPE (s1) for the return value.
23026 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
23028         PR tree-optimization/84946
23029         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23030         bitsize + bitsize in poly_uint64 rather than poly_int64.
23032         PR sanitizer/78651
23033         * dwarf2asm.c: Include fold-const.c.
23034         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23035         of decl rather than decl itself.
23037         PR rtl-optimization/84643
23038         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23040 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
23042         PR sanitizer/78651
23043         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23044         calling assemble_variable.
23046 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
23048         PR target/81647
23049         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23050         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23052 2018-03-19  Jim Wilson  <jimw@sifive.com>
23054         PR bootstrap/84856
23055         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23056         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23057         (riscv_first_stack_step): Likewise.
23058         (riscv_option_override): Use STACK_BOUNDARY instead of
23059         MIN_STACK_BOUNDARY.
23060         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23061         MIN_STACK_BOUNDARY.
23062         (BIGGEST_ALIGNMENT): Set to 128.
23063         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23064         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23065         STACK_BOUNDARY.
23067 2018-03-19  Richard Biener  <rguenther@suse.de>
23069         PR tree-optimization/84933
23070         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23071         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23073 2018-03-19  Richard Biener  <rguenther@suse.de>
23075         PR tree-optimization/84859
23076         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23077         (cond_if_else_store_replacement): Perform sinking operation on
23078         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23079         Generalize what a BB with a single eligible store is.
23081 2018-03-19  Richard Biener  <rguenther@suse.de>
23083         PR tree-optimization/84929
23084         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23085         chrec_is_positive against non-chrec arg.
23087 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23089         PR target/84711
23090         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23092 2018-03-18  Martin Liska  <mliska@suse.cz>
23094         PR rtl-optimization/84635
23095         * regrename.c (build_def_use): Use matches_mode only when
23096         matches >= 0.
23098 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23100         PR tree-optimization/84913
23101         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23102         vectorize chains of COND_EXPRs.
23104 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23106         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23108 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23110         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23112 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23114         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23116 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23117             Kito Cheng  <kito.cheng@gmail.com>
23119         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23120         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23121         (nds32_adjust_reg_alloc_order): New function.
23122         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23124 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23126         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23127         nds32_print_operand, nds32_print_operand_address): Use
23128         HOST_WIDE_INT_PRINT_DEC instead.
23130 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23132         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23134 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23136         PR target/84902
23137         * config/i386/i386.c (initial_ix86_tune_features,
23138         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23139         unsigned long long.
23140         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23141         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23142         rather than 1u << ix86_tune.  Formatting fix.
23143         (ix86_option_override_internal): Change ix86_arch_mask from
23144         unsigned int to unsigned HOST_WIDE_INT, initialize to
23145         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23146         (ix86_function_specific_restore): Likewise.
23148 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23150         PR target/84899
23151         * postreload.c (reload_combine_recognize_pattern): Perform
23152         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23153         truncate_int_for_mode the result for the destination's mode.
23155         PR c/84909
23156         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23157         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23158         Likewise.
23159         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23160         Likewise.
23162 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23164         PR target/84876
23165         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23166         regno_allocno_class_array and sorted_pseudos.
23167         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23168         insns where regno is used.
23170 2018-03-16  Martin Liska  <mliska@suse.cz>
23172         PR ipa/84833
23173         * multiple_target.c (create_dispatcher_calls): Redirect
23174         reference in the symbol table.
23176 2018-03-16  Martin Liska  <mliska@suse.cz>
23178         PR ipa/84722
23179         * multiple_target.c (create_dispatcher_calls): Redirect also
23180         an alias.
23182 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23184         PR c++/79937
23185         PR c++/82410
23186         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23187         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23188         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23190 2018-03-16  Julia Koval  <julia.koval@intel.com>
23192         * doc/invoke.texi (Skylake Server): Add CLWB.
23193         Cannonlake): Remove CLWB.
23195 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23197         PR tree-optimization/84841
23198         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23199         1 << 3.
23200         (FLOAT_ONE_CONST_TYPE): Define.
23201         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23202         (sort_by_operand_rank): Put entries with higher constant_type last
23203         rather than first to match comments.
23205 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23207         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23208         split predicate.
23210 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23212         PR c++/79085
23213         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23214         check and use address of target always.
23216 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23218         PR target/84574
23219         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23220         (indirect_thunk_bnd_needed): Likewise.
23221         (indirect_thunks_used): Likewise.
23222         (indirect_thunks_bnd_used): Likewise.
23223         (indirect_return_needed): New.
23224         (indirect_return_bnd_needed): Likewise.
23225         (output_indirect_thunk_function): Add a bool argument for
23226         function return.
23227         (output_indirect_thunk_function): Don't generate alias for
23228         function return thunk.
23229         (ix86_code_end): Call output_indirect_thunk_function to generate
23230         function return thunks.
23231         (ix86_output_function_return): Set indirect_return_bnd_needed
23232         and indirect_return_needed instead of indirect_thunk_bnd_needed
23233         and indirect_thunk_needed.
23235 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23237         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23238         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23239         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23241 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23242             Paul Hua <paul.hua.gm@gmail.com>
23244         PR c/84852
23245         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23247 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23249         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23250         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23251         resp. SFmode cases.
23253 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23255         PR target/84711
23256         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23257         instead of GET_MODE_SIZE when comparing Units.
23259 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23261         PR target/68256
23262         * varasm.c (hash_section): Return an unchangeble hash value
23263         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23264         Return !aarch64_can_use_per_function_literal_pools_p ().
23266 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23268         PR target/84860
23269         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23270         rather than address of cmode as last argument to prepare_cmp_insn.
23272 2018-03-15  Julia Koval  <julia.koval@intel.com>
23274         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23275         F_AVX512VNNI, F_AVX512BITALG): New.
23277 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23279         PR target/83451
23280         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23281         insn for floating-point loads and stores.
23283 2018-03-14  Carl Love  <cel@us.ibm.com>
23285         * config/rs6000/rs6000-c.c: Add macro definitions for
23286         ALTIVEC_BUILTIN_VEC_PERMXOR.
23287         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23288         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23289         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23290         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23291         UNSPEC_VPERMXOR.
23292         * config/doc/extend.texi: Add prototypes for vec_permxor.
23294 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23296         PR c/84852
23297         * diagnostic-show-locus.c (class layout_point): Convert m_line
23298         from int to linenum_type.
23299         (line_span::comparator): Use linenum "compare" function when
23300         comparing line numbers.
23301         (test_line_span): New function.
23302         (layout_range::contains_point): Convert param "row" from int to
23303         linenum_type.
23304         (layout_range::intersects_line_p): Likewise.
23305         (layout::will_show_line_p): Likewise.
23306         (layout::print_source_line): Likewise.
23307         (layout::should_print_annotation_line_p): Likewise.
23308         (layout::print_annotation_line): Likewise.
23309         (layout::print_leading_fixits): Likewise.
23310         (layout::annotation_line_showed_range_p): Likewise.
23311         (struct line_corrections): Likewise for field m_row.
23312         (line_corrections::line_corrections): Likewise for param "row".
23313         (layout::print_trailing_fixits): Likewise.
23314         (layout::get_state_at_point): Likewise.
23315         (layout::get_x_bound_for_row): Likewise.
23316         (layout::print_line): Likewise.
23317         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23318         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23319         * input.c (selftest::test_linenum_comparisons): New function.
23320         (selftest::input_c_tests): Call it.
23321         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23322         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23323         * selftest.h (ASSERT_GT): New macro.
23324         (ASSERT_GT_AT): New macro.
23325         (ASSERT_LT): New macro.
23326         (ASSERT_LT_AT): New macro.
23328 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23330         PR rtl-optimization/84780
23331         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23333 2018-03-14  Martin Liska  <mliska@suse.cz>
23335         * tree.c (record_node_allocation_statistics): Use
23336         get_stats_node_kind.
23337         (get_stats_node_kind): New function extracted from
23338         record_node_allocation_statistics.
23339         (free_node): Use get_stats_node_kind.
23341 2018-03-14  Richard Biener  <rguenther@suse.de>
23343         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23344         that the value-set of ANTIC_IN doesn't grow.
23346         Revert
23347         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23348         member.
23349         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23350         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23352 2018-03-14  Julia Koval  <julia.koval@intel.com>
23354         * config.gcc (icelake-client, icelake-server): New.
23355         (icelake): Remove.
23356         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23357         (initial_ix86_arch_features): Ditto.
23358         (PTA_SKYLAKE): Add SGX.
23359         (PTA_ICELAKE): Remove.
23360         (PTA_ICELAKE_CLIENT): New.
23361         (PTA_ICELAKE_SERVER): New.
23362         (ix86_option_override_internal): Split up icelake on icelake client and
23363         icelake server.
23364         (get_builtin_code_for_version): Ditto.
23365         (fold_builtin_cpu): Ditto.
23366         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23367         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23368         * config/i386/i386.h (processor_type): Ditto.
23369         * doc/invoke.texi: Ditto.
23371 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23373         PR sanitizer/83392
23374         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23375         INTEGER_CST offset, add it together with bitpos / 8 and
23376         sign extend based on POINTER_SIZE.
23378         PR target/84844
23379         Revert
23380         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23382         PR target/78090
23383         * config/i386/constraints.md (Yc): New register constraint.
23384         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23385         Use Yc constraint for alternative 2 of operand 0.  Remove
23386         preferred_for_speed attribute.
23388 2018-03-14  Richard Biener  <rguenther@suse.de>
23390         PR tree-optimization/84830
23391         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23392         with the old one to avoid oscillations.
23394 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23396         PR target/83712
23397         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23398         pseudos.
23399         (assign_by_spills): Return a flag of reload assignment failure.
23400         Do not process the reload assignment failures.  Do not spill other
23401         reload pseudos if they has the same reg class.  Update n if
23402         necessary.
23403         (lra_assign): Add a return arg.  Set up from the result of
23404         assign_by_spills call.
23405         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23406         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23407         usage_insns if it is not NULL.
23408         (spill_hard_reg_in_range): New function.
23409         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23410         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23411         function prototypes.
23412         (lra_assign): Change prototype.
23413         * lra.c (lra): Add code to deal with fails by splitting hard reg
23414         live ranges.
23416 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23418         * config/riscv/riscv.opt (mrelax): New option.
23419         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23420         "norelax" when riscv_mrelax is disabled.
23421         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23423 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23425         PR target/84743
23426         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23427         reassociation for int modes.
23429 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23431         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23432         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23433         for big-endian.
23434         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23435         * config/aarch64/aarch64-sve.md
23436         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23437         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23438         (*extend<mode><Vwide>2): Rename to...
23439         (aarch64_sve_extend<mode><Vwide>2): ...this.
23440         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23441         renaming the old pattern to...
23442         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23443         unsigned packs.
23444         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23445         define_expand, renaming the old pattern to...
23446         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23447         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23448         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23449         account when deciding which SVE instruction the optab should use.
23450         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23452 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23454         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23455         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23456         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23457         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23458         (tlsdesc_small_<mode>): Turn a define_expand and use
23459         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23460         (tlsdesc_small_advsimd_<mode>): ...this.
23461         (tlsdesc_small_sve_<mode>): New pattern.
23463 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23465         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23466         (UNSPEC_UMUL_HIGHPART): New constants.
23467         (MUL_HIGHPART): New int iteraor.
23468         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23469         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23470         define_expand.
23471         (*<su>mul<mode>3_highpart): New define_insn.
23473 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23475         PR lto/84805
23476         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23477         incomplete types.
23479 2018-03-13  Martin Liska  <mliska@suse.cz>
23481         PR ipa/84658.
23482         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23483         vector.
23484         (sem_item_optimizer::~sem_item_optimizer): Release it.
23485         (sem_item_optimizer::merge_classes): Register variable aliases.
23486         (sem_item_optimizer::fixup_pt_set): New function.
23487         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23488         * ipa-icf.h: Declare new variables and functions.
23490 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23492         PR middle-end/84834
23493         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23494         integer_pow2p@2 and test integer_pow2p in condition.
23495         (A < 0 ? C : 0): Similarly for @1.
23497         PR middle-end/84831
23498         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23499         characters starting at p contain '\0' character, don't look beyond
23500         that.
23502         PR target/84827
23503         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23504         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23506         PR target/84828
23507         * reg-stack.c (change_stack): Change update_end var from int to
23508         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23509         also call set_block_for_insn on the newly added insns and rescan.
23511         PR target/84786
23512         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23513         on the last operand.
23515         PR c++/84704
23516         * tree.c (stabilize_reference_1): Return save_expr (e) for
23517         STATEMENT_LIST even if it doesn't have side-effects.
23519 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23521         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23523 2018-03-12  Renlin Li  <renlin.li@arm.com>
23525         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23526         aarch64_output_scalar_simd_mov_immediate.
23528 2018-03-12  Martin Sebor  <msebor@redhat.com>
23530         PR tree-optimization/83456
23531         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23532         for perfectly overlapping calls to memcpy.
23533         (gimple_fold_builtin_memory_chk): Same.
23534         (gimple_fold_builtin_strcpy): Handle no-warning.
23535         (gimple_fold_builtin_stxcpy_chk): Same.
23536         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23538 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23540         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23541         parameter.  Use it for SFmode.
23542         (rs6000_function_arg_advance_1): Adjust.
23543         (rs6000_function_arg): Adjust.
23544         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23546 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23548         PR rtl-optimization/84169
23549         PR rtl-optimization/84780
23550         * combine.c (can_combine_p): Check for a 2-insn combination whether
23551         the destination register is used between the two insns, too.
23553 2018-03-12  Richard Biener  <rguenther@suse.de>
23555         PR tree-optimization/84803
23556         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23557         for refs DR analysis didn't process.
23559 2018-03-12  Richard Biener  <rguenther@suse.de>
23561         PR tree-optimization/84777
23562         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23563         force-vectorize loops ignore whether we are optimizing for size.
23565 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23567         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23568         (TARGET_MD_ASM_ADJUST): Define.
23570 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23571             Kito Cheng  <kito.cheng@gmail.com>
23572             Chung-Ju Wu  <jasonwucj@gmail.com>
23574         * config/nds32/nds32.c (nds32_compute_stack_frame,
23575         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23576         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23577         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
23578         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
23579         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
23580         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
23581         * config/nds32/nds32.md (prologue, epilogue): Use macro
23582         NDS32_V3PUSH_AVAILABLE_P to do checking.
23584 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
23586         PR debug/58150
23587         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
23588         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
23589         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
23590         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
23591         addition of most attributes on !orig_type_die or the attribute not
23592         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
23594 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23595             Chung-Ju Wu  <jasonwucj@gmail.com>
23597         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
23598         __NDS32_VH__ macro.
23599         * config/nds32/nds32.opt (mvh): New option.
23601 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23602             Chung-Ju Wu  <jasonwucj@gmail.com>
23604         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
23605         function.
23606         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
23607         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
23608         definition.
23610 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23611             Chung-Ju Wu  <jasonwucj@gmail.com>
23613         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
23614         function.
23615         * config/nds32/nds32-multiple.md (strlensi): New pattern.
23616         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
23618 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
23619             Kito Cheng  <kito.cheng@gmail.com>
23620             Chung-Ju Wu  <jasonwucj@gmail.com>
23622         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
23623         UNSPEC_FFMISM and UNSPEC_FLMISM.
23624         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
23625         for ffb, ffmism and flmism.
23626         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
23627         (unspec_ffmism): Ditto.
23628         (unspec_flmism): Ditto.
23629         (nds32_expand_builtin_impl): Check if string extension is available.
23630         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
23631         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
23633 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
23635         Reverting patch:
23636         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23638         PR target/83712
23639         * lra-assigns.c (assign_by_spills): Return a flag of reload
23640         assignment failure.  Do not process the reload assignment
23641         failures.  Do not spill other reload pseudos if they has the same
23642         reg class.
23643         (lra_assign): Add a return arg.  Set up from the result of
23644         assign_by_spills call.
23645         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23646         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23647         usage_insns if it is not NULL.
23648         (spill_hard_reg_in_range): New function.
23649         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23650         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23651         function prototypes.
23652         (lra_assign): Change prototype.
23653         * lra.c (lra): Add code to deal with fails by splitting hard reg
23654         live ranges.
23656 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
23658         PR target/84807
23659         * config/i386/i386.opt: Replace Enforcment with Enforcement.
23661 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
23663         PR debug/84620
23664         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
23665         (dw_val_node): Add val_symbolic_view.
23666         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
23667         (symview_upper_bound): New.
23668         (new_line_info_table): Initialize symviews_since_reset.
23669         (dwarf2out_source_line): Count symviews_since_reset and set
23670         symview_upper_bound.
23671         (dw_val_equal_p): Handle symview.
23672         (add_AT_symview): New.
23673         (print_dw_val): Handle symview.
23674         (attr_checksum, attr_checksum_ordered): Likewise.
23675         (same_dw_val_p, size_of_die): Likewise.
23676         (value_format, output_die): Likewise.
23677         (add_high_low_attributes): Use add_AT_symview for entry_view.
23678         (dwarf2out_finish): Reset symview_upper_bound, clear
23679         zero_view_p.
23681 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
23683         PR target/83969
23684         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
23685         Add strict argument and use it.
23686         (rs6000_split_multireg_move): Update for new strict argument.
23687         (mem_operand_gpr): Disallow all non-offsettable addresses.
23688         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
23690 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
23692         PR target/84772
23693         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
23694         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
23695         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
23697         PR c++/84767
23698         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
23699         decl, use remap_type if we want to use the type.
23701 2018-03-09  Martin Sebor  <msebor@redhat.com>
23703         PR tree-optimization/84526
23704         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23705         Remove dead code.
23706         (builtin_access::generic_overlap): Be prepared to handle non-array
23707         base objects.
23709 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
23711         PR rtl-optimization/84682
23712         * lra-constraints.c (process_address_1): Check is_address flag
23713         for address constraints.
23714         (process_alt_operands): Likewise.
23715         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
23716         preprocess_constraints.
23717         * recog.h (preprocess_constraints): Add oploc parameter.
23718         Adjust callers.
23719         * recog.c (preprocess_constraints): Test address_operand for
23720         CT_ADDRESS constraints.
23722 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23724         PR target/83712
23725         * lra-assigns.c (assign_by_spills): Return a flag of reload
23726         assignment failure.  Do not process the reload assignment
23727         failures.  Do not spill other reload pseudos if they has the same
23728         reg class.
23729         (lra_assign): Add a return arg.  Set up from the result of
23730         assign_by_spills call.
23731         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23732         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23733         usage_insns if it is not NULL.
23734         (spill_hard_reg_in_range): New function.
23735         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23736         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23737         function prototypes.
23738         (lra_assign): Change prototype.
23739         * lra.c (lra): Add code to deal with fails by splitting hard reg
23740         live ranges.
23742 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23744         PR target/83193
23745         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
23746         Accept complain bool parameter.  Only emit errors if it is true.
23747         (arm_parse_cpu_option_name): Likewise.
23748         (arm_target_thumb_only): Adjust callers of the above.
23749         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
23750         prototype to take a default true bool parameter.
23751         (arm_parse_arch_option_name): Likewise.
23753 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
23754             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
23756         PR jit/64089
23757         PR jit/84288
23758         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
23759         * configure: Regenerate.
23760         * configure.ac ("linker --version-script option"): New.
23761         ("linker soname option"): New.
23763 2018-03-09  Richard Biener  <rguenther@suse.de>
23765         PR tree-optimization/84775
23766         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
23767         immediate uses of predicate stmts and mark them modified.
23769         Revert
23770         PR tree-optimization/84178
23771         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23772         to caller.
23773         (version_loop_for_if_conversion): Delay update_ssa call.
23774         (tree_if_conversion): Delay update_ssa until after predicate
23775         insertion.
23777 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
23779         PR target/84763
23780         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
23781         when the function accesses prior frames.
23783 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23785         PR debug/84456
23786         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
23787         gen_llsym, otherwise call maybe_gen_llsym.
23789         PR inline-asm/84742
23790         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
23791         has ',' character inside of it.
23793 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23795         PR target/84748
23796         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
23797         as clobbering CC_REGNUM.
23799 2018-03-08  Richard Biener  <rguenther@suse.de>
23801         PR middle-end/84552
23802         * tree-scalar-evolution.c: Include tree-into-ssa.h.
23803         (follow_copies_to_constant): Do not follow SSA names registered
23804         for update.
23806 2018-03-08  Richard Biener  <rguenther@suse.de>
23808         PR tree-optimization/84178
23809         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23810         to caller.
23811         (version_loop_for_if_conversion): Delay update_ssa call.
23812         (tree_if_conversion): Delay update_ssa until after predicate
23813         insertion.
23815 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
23817         PR tree-optimization/84178
23818         * tree-if-conv.c (release_bb_predicate): Remove the
23819         the assertion that the stmts have NULL use_ops.
23820         Discard the statements, asserting that they haven't
23821         yet been added to a BB.
23823 2018-03-08  Richard Biener  <rguenther@suse.de>
23825         PR tree-optimization/84746
23826         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
23827         (phi_translate): Pass in destination ANTIC_OUT set.
23828         (phi_translate_1): Likewise.  For a simplified result lookup
23829         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
23830         (phi_translate_set): Adjust.
23831         (do_pre_regular_insertion): Likewise.
23832         (do_pre_partial_partial_insertion): Likewise.
23834 2018-03-08  Martin Liska  <mliska@suse.cz>
23836         PR gcov-profile/84735
23837         * doc/gcov.texi: Document usage of profile files.
23838         * gcov-io.h: Document changes in the format.
23840 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
23842         PR debug/84404
23843         PR debug/84408
23844         * dwarf2out.c (struct dw_line_info_table): Update comments for
23845         view == -1.
23846         (FORCE_RESET_NEXT_VIEW): New.
23847         (FORCE_RESETTING_VIEW_P): New.
23848         (RESETTING_VIEW_P): Check for -1 too.
23849         (ZERO_VIEW_P): Likewise.
23850         (new_line_info_table): Force-reset next view.
23851         (dwarf2out_begin_function): Likewise.
23852         (dwarf2out_source_line): Simplify zero_view_p initialization.
23853         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
23854         view directly.  Omit view when omitting .loc at line 0.
23856 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23858         PR tree-optimization/84740
23859         * tree-switch-conversion.c (process_switch): Call build_constructors
23860         only if info.phi_count is non-zero.
23862         PR tree-optimization/84739
23863         * tree-tailcall.c (find_tail_calls): Check call arguments against
23864         DECL_ARGUMENTS (current_function_decl) rather than
23865         DECL_ARGUMENTS (func) when checking for tail recursion.
23867 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23869         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
23870         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
23871         Volker Reichelt's entry and add entries for people that perform
23872         GCC fuzzy testing and report numerous bugs.
23874 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
23876         PR target/82411
23877         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
23878         readonly data in sdata, if that is disabled.
23879         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
23880         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
23881         -mreadonly-in-sdata option.
23883 2018-03-07  Martin Sebor  <msebor@redhat.com>
23885         PR tree-optimization/84468
23886         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
23887         basic block when looking for nul assignment.
23889 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
23891         PR target/84277
23892         * except.h (output_function_exception_table): Adjust prototype.
23893         * except.c (output_function_exception_table): Remove FNNAME parameter
23894         and add SECTION parameter.  Ouput one part of the table at a time.
23895         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
23896         the first part of the exception table and emit unwind directives.
23897         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
23898         (i386_pe_seh_cold_init): Likewise.
23899         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
23900         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
23901         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
23902         (ix86_output_call_insn): Emit a nop in one more case for SEH.
23903         * config/i386/winnt.c: Include except.h.
23904         (struct seh_frame_state): Add reg_offset, after_prologue and
23905         in_cold_section fields.
23906         (i386_pe_seh_end_prologue): Set seh->after_prologue.
23907         (i386_pe_seh_cold_init): New function.
23908         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
23909         to seh->in_cold_section.
23910         (seh_emit_push): Record the offset of the push.
23911         (seh_emit_save): Record the offet of the save.
23912         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
23913         Test seh->after_prologue to disregard the epilogue.
23914         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
23915         (i386_pe_end_cold_function): New function.
23917 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23919         PR fortran/84565
23920         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
23921         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
23923         PR c++/84704
23924         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
23925         on tmp_var.
23926         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
23927         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
23929         PR middle-end/84723
23930         * multiple_target.c: Include tree-inline.h and intl.h.
23931         (expand_target_clones): Diagnose and fail if node->definition and
23932         !tree_versionable_function_p (node->decl).
23934 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
23936         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
23937         sprint_ul.
23938         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
23939         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
23940         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
23942 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23944         PR target/84710
23945         * combine.c (try_combine): Use reg_or_subregno instead of handling
23946         just paradoxical SUBREGs and REGs.
23948 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
23950         * config/arc/arc.c (arc_finalize_pic): Remove function.
23951         (arc_must_save_register): We use single base PIC register, remove
23952         checks to save/restore the PIC register.
23953         (arc_expand_prologue): Likewise.
23954         * config/arc/arc-protos.h (arc_set_default_type_attributes):
23955         Remove.
23956         (arc_verify_short): Likewise.
23957         (arc_attr_type): Likewise.
23958         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
23959         (walk_stores): Likewise.
23960         (arc_address_cost): Make it static.
23961         (arc_verify_short): Likewise.
23962         (branch_dest): Likewise.
23963         (arc_attr_type): Likewise.
23964         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
23965         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
23966         (arc_final_prescan_insn): Remove inserting the nops due to
23967         hardware hazards.  It is done in reorg step.
23968         (insn_length_variant_t): Remove.
23969         (insn_length_parameters_t): Likewise.
23970         (arc_insn_length_parameters): Likewise.
23971         (arc_get_insn_variants): Likewise.
23972         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
23974 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23976         PR inline-asm/84683
23977         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
23978         assertion failure.
23980         PR tree-optimization/84687
23981         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
23982         on new_node->decl.
23983         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
23985 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23987         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
23988         Rename to ppc_speculation_barrier.
23989         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
23990         __builtin_ppc_speculation_barrier.
23992 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23994         PR target/84700
23995         * combine.c (combine_simplify_rtx): Don't try to simplify if
23996         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
23997         are equal to x.
23999 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
24001         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24002         to 32 bytes when compiling for POWER9.
24004 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24006         PR target/84564
24007         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24008         regparm >= 3 with no arg reg available also for calls with
24009         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
24011         PR target/84524
24012         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24013         orig,vex.
24014         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
24016 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
24018         PR target/84264
24019         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24021 2018-03-05  Richard Biener  <rguenther@suse.de>
24023         PR tree-optimization/84486
24024         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24025         When inserting a __builtin_assume_aligned call set the LHS
24026         SSA name alignment info accordingly.
24028 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
24030         PR tree-optimization/84114
24031         * config/aarch64/aarch64.c (aarch64_reassociation_width)
24032         Avoid reassociation of FLOAT_MODE addition.
24034 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
24036         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24037         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24038         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24039         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24040         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24041         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24042         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24043         and -mwbnoinvd.
24044         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24045         __builtin_ia32_wbinvd): New builtins.
24046         (SPECIAL_ARGS2): New.
24047         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24048         (SPECIAL_ARGS2): New.
24049         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24050         (ix86_valid_target_attribute_inner_p): Ditto.
24051         (ix86_init_mmx_sse_builtins): Add special_args2.
24052         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24053         TARGET_WBNOINVD_P): New.
24054         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24055         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24056         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24057         * config/i386/immintrin.h (_wbinvd): New intrinsic.
24058         * config/i386/pconfigintrin.h: New file.
24059         * config/i386/wbnoinvdintrin.h: Ditto.
24060         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24061         wbnoinvdintrin.h.
24062         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24064 2018-03-05  Richard Biener  <rguenther@suse.de>
24066         PR tree-optimization/84670
24067         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24068         member.
24069         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24070         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24071         (compute_antic_aux): Only assert the number of values in ANTIC_IN
24072         doesn't grow if all successors (recursively) were visited at least
24073         once.
24075 2018-03-05  Richard Biener  <rguenther@suse.de>
24077         PR tree-optimization/84650
24078         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24079         if executed in the loop pipeline.
24081 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24083         * doc/configfiles.texi (Configuration Files): Move info about
24084         conditionalizing $target-protos.h to...
24085         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24086         differs from $target-protos.h.
24088 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24089             Chung-Ju Wu  <jasonwucj@gmail.com>
24091         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24092         * config/nds32/nds32-multiple.md (setmemsi): Define.
24093         * config/nds32/nds32-memory-manipulation.c
24094         (nds32_gen_dup_4_byte_to_word_value): New.
24095         (emit_setmem_word_loop): New.
24096         (emit_setmem_byte_loop): New.
24097         (nds32_expand_setmem_loop): New.
24098         (nds32_expand_setmem_loop_v3m): New.
24099         (nds32_expand_setmem_unroll): New.
24100         (nds32_expand_setmem): New.
24102 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24103             Chung-Ju Wu  <jasonwucj@gmail.com>
24105         * config/nds32/nds32-memory-manipulation.c
24106         (nds32_emit_load_store): New.
24107         (nds32_emit_post_inc_load_store): New.
24108         (nds32_emit_mem_move): New.
24109         (nds32_emit_mem_move_block): New.
24110         (nds32_expand_movmemsi_loop_unknown_size): New.
24111         (nds32_expand_movmemsi_loop_known_size): New.
24112         (nds32_expand_movmemsi_loop): New.
24113         (nds32_expand_movmemsi_unroll): New.
24114         (nds32_expand_movmemqi): Rename ...
24115         (nds32_expand_movmemsi): ... to this.
24116         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24117         (movmemsi): ... to this.
24118         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24119         (nds32_expand_movmemsi): ... to this.
24121 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24122             Monk Chiang  <sh.chiang04@gmail.com>
24123             Chung-Ju Wu  <jasonwucj@gmail.com>
24125         * config/nds32/nds32-protos.h
24126         (nds32_expand_load_multiple): New arguments.
24127         (nds32_expand_store_multiple): Ditto.
24128         (nds32_valid_multiple_load_store): Rename ...
24129         (nds32_valid_multiple_load_store_p): ... to this.
24130         * config/nds32/nds32-memory-manipulation.c
24131         (nds32_expand_load_multiple): Refine implementation.
24132         (nds32_expand_store_multiple): Ditto.
24133         * config/nds32/nds32-multiple.md
24134         (load_multiple): Update nds32_expand_load_multiple interface.
24135         (store_multiple): Update nds32_expand_store_multiple interface.
24136         * config/nds32/nds32-predicates.c
24137         (nds32_valid_multiple_load_store): Rename ...
24138         (nds32_valid_multiple_load_store_p): ... to this and refine
24139         implementation.
24140         * config/nds32/predicates.md
24141         (nds32_load_multiple_and_update_address_operation): New predicate.
24142         (nds32_store_multiple_and_update_address_operation): New predicate.
24144 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24145             Chung-Ju Wu  <jasonwucj@gmail.com>
24147         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24148         (combo): New attribute.
24149         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24151 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24153         * config/nds32/nds32.opt: Change -mcmodel= default value.
24155 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24156             Monk Chiang  <sh.chiang04@gmail.com>
24157             Chung-Ju Wu  <jasonwucj@gmail.com>
24159         * config/nds32/constants.md (unspec_element): New enum.
24160         * config/nds32/constraints.md (Umw): New constraint.
24161         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24162         * config/nds32/nds32-intrinsic.md: Likewise.
24163         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24164         (nds32_valid_smw_lwm_base_p): New.
24165         (nds32_output_smw_single_word): New.
24166         (nds32_output_lmw_single_word): New.
24167         (nds32_expand_unaligned_load): New.
24168         (nds32_expand_unaligned_store): New.
24169         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24170         (nds32_output_smw_single_word): Declare.
24171         (nds32_output_lmw_single_word): Declare.
24172         (nds32_expand_unaligned_load): Declare.
24173         (nds32_expand_unaligned_store): Declare.
24174         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24175         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24176         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24177         NDS32_BUILTIN_UASTORE_DW.
24178         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24179         predicate.
24181 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24182             Kito Cheng  <kito.cheng@gmail.com>
24183             Chung-Ju Wu  <jasonwucj@gmail.com>
24185         * config/nds32/nds32-intrinsic.c
24186         (nds32_expand_builtin_null_ftype_reg): Delete.
24187         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24188         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24189         (nds32_read_argument): New.
24190         (nds32_legitimize_target): Ditto.
24191         (nds32_legitimize_argument): Ditto.
24192         (nds32_check_constant_argument): Ditto.
24193         (nds32_expand_unop_builtin): Ditto.
24194         (nds32_expand_unopimm_builtin): Ditto.
24195         (nds32_expand_binop_builtin): Ditto.
24196         (nds32_builtin_decl_impl): Ditto.
24197         (builtin_description): Ditto.
24198         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24199         (nds32_init_builtins_impl): Ditto.
24200         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24201         (nds32_builtin_decl): New.
24202         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24203         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24205 2018-03-02  Jeff Law  <law@redhat.com>
24207         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24208         (redundant_insn, fill_simple_delay_slots): Likewise.
24209         (fill_slots_from_thread): Likewise.
24210         * resource.c (mark_referenced_resources): Likewise.
24211         (mark_set_resources, find_dead_or_set_registers): Likewise.
24213 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24215         * substring-locations.h (format_warning_va): Formatting fix for
24216         ATTRIBUTE_GCC_DIAG.
24217         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24218         argument.
24219         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24220         * substring-locations.c: Include intl.h.
24221         (format_warning_va): Turned into small wrapper around
24222         format_warning_n_va, renamed to ...
24223         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24224         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24225         use ngettext.
24226         (format_warning_at_substring_n): New function.
24227         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24228         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24229         format_warning_at_substring with just a shorter name instead of
24230         const function pointer.
24231         (fmtwarn_n): New function.
24232         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24233         appropriate, get rid of all the fmtstr temporaries, move conditionals
24234         with G_() wrapped string literals directly into fmtwarn arguments,
24235         cast dir.len to (int), formatting fixes.
24237 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24239         * doc/invoke.texi: Remove "Cilk Plus" references.
24241 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24242             Richard Biener  <rguenther@suse.de>
24244         PR ipa/84628
24245         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24246         for error or warning attributes if CALL_FROM_THUNK_P is set.
24247         Formatting fixes.
24249 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24251         PR target/56540
24252         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24253         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24255         PR target/56540
24256         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24257         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24259         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24260         instead of -1U in last predictors element's probability member.
24262 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24264         PR ipa/83983
24265         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24266         arguments if they are comparable.
24268 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24270         PR tree-optimization/84634
24271         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24272         masks and masked_loop_p with a single loop_masks, making sure it's
24273         null for bb vectorization.
24275 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24277         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24278         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24279         loop->force_vectorize to check whether there is no alias.
24281 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24283         PR target/84614
24284         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24285         prototypes.
24286         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24287         comments.
24288         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24289         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24290         instead of a loop around prev_real_insn.
24291         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24292         prev_real_insn.
24294         PR inline-asm/84625
24295         * config/i386/i386.c (ix86_print_operand): Use conditional
24296         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24297         zero vector.
24299 2018-03-02  Richard Biener  <rguenther@suse.de>
24301         PR tree-optimization/84427
24302         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24303         (bitmap_set_subtract_values): Rewrite to handle multiple
24304         exprs per value.
24305         (clean): Likewise.
24306         (prune_clobbered_mems): Likewise.
24307         (phi_translate): Take edge instead of pred/phiblock.
24308         (phi_translate_1): Likewise.
24309         (phi_translate_set): Likewise.  Insert all translated
24310         exprs for a value into the set, keeping possibly multiple
24311         expressions per value.
24312         (compute_antic_aux): Adjust for phi_translate changes.
24313         When intersecting union the expressions and prune those
24314         not in the final value set, keeping possibly multiple
24315         expressions per value.  Do not use value-insertion
24316         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24317         all expressions.  Add verification that the value-sets
24318         only shrink during iteration.
24319         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24320         (do_pre_regular_insertion): Likewise.
24321         (do_pre_partial_partial_insertion): Likewise.
24323 2018-03-02  Richard Biener  <rguenther@suse.de>
24325         PR target/82005
24326         * config/darwin.c (saved_debug_info_level): New static global.
24327         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24328         (darwin_asm_lto_end): Restore debug info generation settings.
24330 2018-03-01  Martin Liska  <mliska@suse.cz>
24332         PR sanitizer/82484
24333         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24334         volatile arguments.
24336 2018-03-01  Richard Biener  <rguenther@suse.de>
24338         PR debug/84645
24339         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24340         type annotation with LTO when debug was disabled at compile-time.
24342 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24344         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24345         XINT with INTVAL.
24346         (mips_final_postscan_insn): Likewise.
24348 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24350         PR rtl-optimization/84528
24351         * alias.c (init_alias_target): Add commentary.
24352         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24353         a unique base value if the frame pointer is not eliminated
24354         to the stack pointer.
24356 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24358         PR rtl-optimization/83327
24359         * lra-int.h (hard_regs_spilled_into): Declare.
24360         * lra.c (hard_regs_spilled_into): Define.
24361         (init_reg_info): Init hard_regs_spilled_into.
24362         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24363         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24364         (process_bb_lives): Handle hard_regs_spilled_into.
24365         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24366         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24368 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24370         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24371         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24372         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24373         * config/rs6000/aix72.h: New file.
24375 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24377         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24378         instead of warning_at with conditional singular and plural messages
24379         where possible.
24381         PR target/52991
24382         * stor-layout.c (update_alignment_for_field): For
24383         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24384         && !DECL_PACKED (field), do the alignment update, just use
24385         only desired_align instead of MAX (type_align, desired_align)
24386         as the alignment.
24387         (place_field): Don't do known_align < desired_align handling
24388         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24389         is non-NULL, instead do it after rli->prev_field handling and
24390         only if not within a bitfield word.  For DECL_PACKED (field)
24391         use type_align of BITS_PER_UNIT.
24393 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24395         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24396         superfluous parentheses and trailing spaces.
24398 2018-02-28  Richard Biener  <rguenther@suse.de>
24400         PR tree-optimization/84584
24401         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24402         SCoPs with fake exit edge.
24404 2018-02-28  Martin Liska  <mliska@suse.cz>
24406         PR testsuite/84597
24407         * timevar.c (timer::print): Fix format to properly print 100%
24408         values.
24410 2018-02-28  Richard Biener  <rguenther@suse.de>
24412         PR middle-end/84607
24413         * genmatch.c (capture_info::walk_match): Do not mark
24414         captured expressions without operands as expr_p given
24415         they act more like predicates and should be subject to
24416         "lost tail" side-effect preserving.
24418 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24420         PR rtl-optimization/81611
24421         * auto-inc-dec.c (attempt_change): Move dead note from
24422         mem_insn if it's the next use of regno
24423         (find_address): Take address use of reg holding
24424         non-incremented value.  Add parm to limit search to the named
24425         reg only.
24426         (merge_in_block): Attempt to use a mem insn that is the next
24427         use of the original regno.
24429 2018-02-27  Martin Sebor  <msebor@redhat.com>
24431         PR c++/83871
24432         * doc/invoke.texi (-Wmissing-attributes): New option.
24433         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24435 2018-02-27  Martin Sebor  <msebor@redhat.com>
24437         PR translation/84207
24438         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24439         n argument to unsigned HOST_WIDE_INT.
24440         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24441         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24442         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24443         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24445 2018-02-27  Richard Biener  <rguenther@suse.de>
24447         PR tree-optimization/84512
24448         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24449         Do not use the estimate returned from record_stmt_cost for
24450         the scalar iteration cost but sum properly using add_stmt_cost.
24452 2018-02-27  Richard Biener  <rguenther@suse.de>
24454         PR tree-optimization/84466
24455         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24456         Adjust last change to less strictly validate use operands.
24458 2018-02-27  Martin Liska  <mliska@suse.cz>
24460         PR gcov-profile/84548
24461         * gcov.c (process_file): Allow partial overlap and consider it
24462         also as group functions.
24463         (output_lines): Properly calculate range of lines for a group.
24465 2018-02-27  Martin Liska  <mliska@suse.cz>
24467         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24468         'ggc' suffixes.  Change first column width.
24469         (timer::print): Fix formatting of the column.
24471 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24473         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24474         preserve inline entry blocks for the sake of debug inline
24475         entry point markers alone.
24476         (remove_unused_locals): Suggest in comments a better place to
24477         force the preservation of inline entry blocks that are
24478         otherwise unused, but do not preserve them.
24480 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24482         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24484 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24486         PR target/84039
24487         * config/i386/constraints.md (Bs): Replace
24488         ix86_indirect_branch_register with
24489         TARGET_INDIRECT_BRANCH_REGISTER.
24490         (Bw): Likewise.
24491         * config/i386/i386.md (indirect_jump): Likewise.
24492         (tablejump): Likewise.
24493         (*sibcall_memory): Likewise.
24494         (*sibcall_value_memory): Likewise.
24495         Peepholes of indirect call and jump via memory: Likewise.
24496         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24497         (*sibcall_value_GOT_32): Likewise.
24498         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24499         (GOT_memory_operand): Likewise.
24500         (call_insn_operand): Likewise.
24501         (sibcall_insn_operand): Likewise.
24502         (GOT32_symbol_operand): Likewise.
24503         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24505 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24507         PR rtl-optimization/83496
24508         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24509         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24510         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24511         redundant insn, if any.
24512         (relax_delay_slots): Likewise.
24513         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24515 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24517         PR tree-optimization/83965
24518         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24519         that grouped statements are part of a reduction chain.  Return
24520         true if the statement is not marked as a reduction itself but
24521         is part of a group.
24522         (vect_recog_dot_prod_pattern): Don't check whether the statement
24523         is part of a group here.
24524         (vect_recog_sad_pattern): Likewise.
24525         (vect_recog_widen_sum_pattern): Likewise.
24527 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24529         PR debug/84545
24530         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24532 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24534         PR target/84530
24535         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24536         the bool argument.
24537         (ix86_output_indirect_function_return): New prototype.
24538         (ix86_split_simple_return_pop_internal): Likewise.
24539         * config/i386/i386.c (indirect_return_via_cx): New.
24540         (indirect_return_via_cx_bnd): Likewise.
24541         (indirect_thunk_name): Handle return va CX_REG.
24542         (output_indirect_thunk_function): Create alias for
24543         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24544         (ix86_output_indirect_jmp): Remove the bool argument.
24545         (ix86_output_indirect_function_return): New function.
24546         (ix86_split_simple_return_pop_internal): Likewise.
24547         * config/i386/i386.md (*indirect_jump): Don't pass false
24548         to ix86_output_indirect_jmp.
24549         (*tablejump_1): Likewise.
24550         (simple_return_pop_internal): Change it to define_insn_and_split.
24551         Call ix86_split_simple_return_pop_internal to split it for
24552         -mfunction-return=.
24553         (simple_return_indirect_internal): Call
24554         ix86_output_indirect_function_return instead of
24555         ix86_output_indirect_jmp.
24557 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24559         PR bootstrap/84405
24560         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24561         memset and value initialization afterwards.
24563 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24565         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24567 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24569         PR target/84521
24570         * common/config/aarch64/aarch64-common.c
24571         (aarch_option_optimization_table[]): Switch
24572         off fomit-frame-pointer
24574 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24575             Chung-Ju Wu  <jasonwucj@gmail.com>
24577         * config/nds32/nds32-multiple.md (load_multiple): Disallow
24578         volatile memory.
24579         (store_multiple): Ditto.
24581 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24583         * config.gcc: Add --with-cpu support for nds32 target.
24584         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
24585         * config/nds32/nds32.opt: Add -mcpu= option.
24587 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
24589         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
24590         isel=yes): Warn for these deprecated options.
24592 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
24594         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
24595         ISA_2_5_MASKS_EMBEDDED.
24597 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
24599         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
24600         p->max as pointers rather than using iterative_hash_expr.
24602 2018-02-23  Carl Love  <cel@us.ibm.com>
24604         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
24605         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
24606         BU_P8V_OVERLOAD_2.
24607         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
24608         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
24609         P8V_BUILTIN_VEC_VUNSIGNED2.
24611 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
24613         PR target/81572
24614         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
24615         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
24616         LRA_UNKNOWN_ALT.
24617         * lra-constraints.c (curr_insn_transform): Set up
24618         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
24619         LRA_UNKNOWN_ALT.
24620         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
24621         * lra-eliminations.c (spill_pseudos): Ditto.
24622         (process_insn_for_elimination): Ditto.
24623         * lra-lives.c (reg_early_clobber_p): Use the new macros.
24624         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
24625         LRA_NON_CLOBBERED_ALT.
24627 2018-02-22  Martin Sebor  <msebor@redhat.com>
24629         PR tree-optimization/84480
24630         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
24631         to maybe_diag_stxncpy_trunc.  Call it.
24632         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
24633         from gimple_fold_builtin_strcpy.  Print inlining stack.
24634         (handle_builtin_stxncpy): Print inlining stack.
24635         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
24637 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
24639         PR target/84176
24640         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
24641         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
24642         and -fcheck-pointer-bounds are used together.
24643         (indirect_thunk_prefix): New enum.
24644         (indirect_thunk_need_prefix): New function.
24645         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
24646         "_nt" instead of "_bnd" for NOTRACK prefix.
24647         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
24648         (output_indirect_thunk_function): Likewise.
24649         (): Likewise.
24650         (ix86_code_end): Update output_indirect_thunk_function calls.
24651         (ix86_output_indirect_branch_via_reg): Replace
24652         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
24653         (ix86_output_indirect_branch_via_push): Likewise.
24654         (ix86_output_function_return): Likewise.
24655         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
24656         incompatible with -fcf-protection=branch and
24657         -fcheck-pointer-bounds.
24659 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24661         PR target/83335
24662         * config/aarch64/aarch64.c (aarch64_print_address_internal):
24663         Change gcc_assert call to output_operand_lossage.
24665 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24667         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
24669 2018-02-22  DJ Delorie  <dj@redhat.com>
24670             Sebastian Perta  <sebastian.perta@renesas.com>
24671             Oleg Endo  <olegendo@gcc.gnu.org>
24673         * config/rx/rx.c (rx_rtx_costs): New function.
24674         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
24676 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24678         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
24680 2018-02-22  Martin Liska  <mliska@suse.cz>
24682         PR driver/83193
24683         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
24684         Add "native" as a possible value.
24686 2018-02-22  Martin Liska  <mliska@suse.cz>
24688         PR driver/83193
24689         * config/i386/i386.c (ix86_option_override_internal):
24690         Add "native" as a possible value for -march and -mtune.
24692 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
24694         PR target/84502
24695         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
24696         to all type variants.
24698         PR tree-optimization/84503
24699         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
24700         width as info->bitpos + info->bitsize - start.
24701         (merged_store_group::merge_overlapping): Simplify width computation.
24702         (check_no_overlap): New function.
24703         (imm_store_chain_info::try_coalesce_bswap): Compute expected
24704         start + width and last_order of the group, fail if check_no_overlap
24705         fails.
24706         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
24707         to group if check_no_overlap fails.
24709 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24711         * config/rs6000/altivec.md: Delete contraint arguments to
24712         define_expand, define_split, and define_peephole2, and in
24713         define_insn_and_split if always unused.
24714         * config/rs6000/darwin.md: Ditto.
24715         * config/rs6000/dfp.md: Ditto.
24716         * config/rs6000/rs6000.md: Ditto.
24717         * config/rs6000/sync.md: Ditto.
24718         * config/rs6000/vector.md: Ditto.
24719         * config/rs6000/vsx.md: Ditto.
24721 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24723         * config/rs6000/altivec.md: Write output control strings as braced
24724         blocks instead of double-quoted strings.
24725         * config/rs6000/darwin.md: Ditto.
24726         * config/rs6000/rs6000.md: Ditto.
24727         * config/rs6000/vector.md: Ditto.
24728         * config/rs6000/vsx.md: Ditto.
24730 2018-02-21  Jason Merrill  <jason@redhat.com>
24732         PR c++/84314 - ICE with templates and fastcall attribute.
24733         * attribs.c (build_type_attribute_qual_variant): Remove assert.
24735 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24737         * ipa-cp.c (determine_versionability): Fix comment typos.
24739 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24741         PR c/84229
24742         * ipa-cp.c (determine_versionability): Do not version functions caling
24743         va_arg_pack.
24745 2018-02-21  Martin Liska  <mliska@suse.cz>
24747         PR driver/83193
24748         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
24749         Add "native" as a possible value.
24750         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
24751         the macro when native cpu detection is available.
24753 2018-02-21  Martin Liska  <mliska@suse.cz>
24755         PR driver/83193
24756         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
24757         Add "native" as a possible value.
24758         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
24759         when native cpu detection is available.
24761 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
24762             Martin Sebor  <msebor@redhat.com>
24764         PR tree-optimization/84478
24765         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
24766         false.
24767         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
24768         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
24769         support which is conservatively correct, for 2 only stay conservative
24770         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
24771         argument to the 2 argument get_range_strlen, adjust 6 arg
24772         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
24773         false.
24774         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
24775         (gimple_fold_builtin_strlen): Pass true as last argument to
24776         get_range_strlen.
24778 2018-02-20  Martin Sebor  <msebor@redhat.com>
24780         PR middle-end/84095
24781         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
24782         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
24783         (builtin_memref::builtin_memref): Factor out parts into
24784         set_base_and_offset and call it.
24786 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24788         PR middle-end/84406
24789         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
24790         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
24791         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
24792         search at the associated MODE_INT.
24794 2018-02-20  Jeff Law  <law@redhat.com>
24796         PR middle-end/82123
24797         PR tree-optimization/81592
24798         PR middle-end/79257
24799         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
24800         for range data rather than using global data.
24801         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
24802         range data rather than using global data.
24803         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
24804         pass it to children as needed.
24805         (struct directive::fmtresult): Similarly.
24806         (struct directive::set_width): Similarly.
24807         (struct directive::set_precision): Similarly.
24808         (format_integer, format_directive, parse_directive): Similarly.
24809         (format_none): Accept unnamed vr_values parameter.
24810         (format_percent, format_floating, format_character): Similarly.
24811         (format_string, format_plain): Similarly.
24812         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
24813         the EVRP range analyzer for range data rather than using global data.
24814         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
24815         gimple-ssa-evrp-analyze.h
24816         (class sprintf_dom_walker): Add after_dom_children member function.
24817         Add evrp_range_analyzer member.
24818         (sprintf_dom_walker::before_dom_children): Call into the EVRP
24819         range analyzer as needed.
24820         (sprintf_dom_walker::after_dom_children): New member function.
24821         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
24822         if not optimizing.
24823         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
24824         (evrp_range_analyzer::pop_to_marker): Likewise.
24826 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24828         PR tree-optimization/84419
24829         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
24830         with the required type if its current type is compatible but
24831         different.
24833 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
24835         PR middle-end/82004
24836         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
24837         after vectorization.
24839 2018-02-20  Martin Liska  <mliska@suse.cz>
24841         PR driver/83193
24842         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
24843         possible values if we don't have a hint.
24845 2018-02-20  Martin Liska  <mliska@suse.cz>
24847         PR c/84310
24848         PR target/79747
24849         * final.c (shorten_branches): Build align_tab array with one
24850         more element.
24851         * opts.c (finish_options): Add alignment option limit check.
24852         (MAX_CODE_ALIGN): Likewise.
24853         (MAX_CODE_ALIGN_VALUE): Likewise.
24854         * doc/invoke.texi: Document maximum allowed option value for
24855         all -falign-* options.
24857 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
24859         PR target/84146
24860         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
24861         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
24862         * var-tracking.c (emit_note_insn_var_location): Remove all references
24863         to NOTE_INSN_CALL_ARG_LOCATION.
24864         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
24865         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
24866         Use copy_rtx_if_shared.
24867         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
24868         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
24869         (dwarf2out_var_location): Remove handling of
24870         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
24871         on call_insn.
24872         * final.c (final_scan_insn): Remove all references to
24873         NOTE_INSN_CALL_ARG_LOCATION.
24874         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
24875         before dumping final insns.
24876         * except.c (emit_note_eh_region_end): Remove all references to
24877         NOTE_INSN_CALL_ARG_LOCATION.
24878         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
24879         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
24880         * config/arc/arc.c (hwloop_optimize): Likewise.
24881         * config/arm/arm.c (create_fix_barrier): Likewise.
24882         * config/s390/s390.c (s390_chunkify_start): Likewise.
24883         * config/sh/sh.c (find_barrier): Likewise.
24884         * config/i386/i386.c (rest_of_insert_endbranch,
24885         ix86_seh_fixup_eh_fallthru): Likewise.
24886         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
24887         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
24888         * config/frv/frv.c (frv_function_prologue): Likewise.
24889         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
24890         reg note.
24891         (note_outside_basic_block_p): Remove all references to
24892         NOTE_INSN_CALL_ARG_LOCATION.
24893         * gengtype.c (adjust_field_rtx_def): Likewise.
24894         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
24895         Likewise.
24896         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
24897         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
24899         PR c++/84444
24900         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
24901         is ADDR_EXPR.
24903         PR tree-optimization/84452
24904         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
24905         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
24906         is NULL.
24908 2018-02-19  Martin Liska  <mliska@suse.cz>
24910         PR sanitizer/82183
24911         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
24913 2018-02-19  Martin Liska  <mliska@suse.cz>
24914             Richard Sandiford  <richard.sandiford@linaro.org>
24916         PR tree-optimization/82491
24917         * gimple-fold.c (get_base_constructor): Make earlier bail out
24918         to prevent ubsan.
24920 2018-02-19  Carl Love  <cel@us.ibm.com>
24922         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
24923         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
24924         BU_P8V_OVERLOAD_1.
24925         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
24926         P8V_BUILTIN_VEC_NEG.
24928 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
24930         * config/rl78/rl78.md (movdf): New define expand.
24932 2018-02-19  Martin Liska  <mliska@suse.cz>
24934         PR other/80589
24935         * doc/invoke.texi: Fix typo.
24936         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
24938 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
24940         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
24941         handle rs6000_single_float and rs6000_double_float specially for
24942         e500 family CPUs.
24944 2018-02-16  Jeff Law  <law@redhat.com>
24946         * config/rx/rx.c (add_pop_cfi_notes): New function.;
24947         (pop_regs): Use it.
24949 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24951         PR ipa/84425
24952         * ipa-inline.c (inline_small_functions): Fix a typo.
24954 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24956         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
24958 2018-02-16  Carl Love  <cel@us.ibm.com>
24960         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
24961         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
24962         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
24963         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
24964         expansion to P8V_BUILTIN_VEC_FLOAT2.
24966 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
24968         PR rtl-optimization/70023
24969         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
24970         src_regno into account.
24972 2018-02-16  Carl Love  <cel@us.ibm.com>
24974         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
24975         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
24976         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
24977         * config/rs6000/rs6000.c: Remove case statements for
24978         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
24979         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
24980         and P9V_BUILTIN_VEC_VINSERT4B.
24981         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
24982         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
24983         * config/rs6000/vsx.md:
24984         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
24985         vec_insert4b.
24987 2018-02-16  Carl Love  <cel@us.ibm.com>
24989         * config/rs6000/altivec.h: Add builtin names vec_extract4b
24990         vec_insert4b.
24991         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
24992         definitions.
24993         * config/rs6000/rs6000-c.c: Add the definitions for
24994         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
24995         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
24996         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
24997         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
24998         definition for insert4b and define insn *insert3b_internal.
24999         * doc/extend.texi: Add documentation for vec_extract4b.
25001 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25003         * doc/extend.texi (Backwards Compatibility): Mention friend
25004         injection.  Note for-scope is deprecated.
25005         * doc/invoke.texi (-ffriend-injection): Deprecate.
25007 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
25009         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25010         that moved to I2, also allow destinations that are a paradoxical
25011         subreg (instead of a normal reg).
25013 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
25015         PR target/83831
25016         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25017         to QImode.
25019 2018-02-16  Richard Biener  <rguenther@suse.de>
25021         PR tree-optimization/84037
25022         PR tree-optimization/84016
25023         PR target/82862
25024         * config/i386/i386.c (ix86_builtin_vectorization_cost):
25025         Adjust vec_construct for the fact we need additional higher latency
25026         128bit inserts for AVX256 and AVX512 vector builds.
25027         (ix86_add_stmt_cost): Scale vector construction cost for
25028         elementwise loads.
25030 2018-02-16  Richard Biener  <rguenther@suse.de>
25032         PR tree-optimization/84417
25033         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25034         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25035         (non_rewritable_lvalue_p): Likewise, use poly-ints.
25037 2018-02-16  Martin Liska  <mliska@suse.cz>
25039         PR sanitizer/84307
25040         * internal-fn.def (ASAN_CHECK): Set proper flags.
25041         (ASAN_MARK): Likewise.
25043 2018-02-16  Julia Koval  <julia.koval@intel.com>
25045         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25046         from PTA_CANNONLAKE.
25048 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25050         PR target/84272
25051         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25052         Use ++iter rather than iter++ for std::list iterators.
25053         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
25054         defer deleting them until all nodes in the forest are processed.  Do
25055         free even leaf nodes.  Change to_process into auto_vec.
25057         PR bootstrap/84405
25058         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25059         * vec.h (vec_default_construct): Use memset instead of placement new
25060         if BROKEN_VALUE_INITIALIZATION is defined.
25061         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25062         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25063         is defined.
25065         PR rtl-optimization/83723
25066         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25067         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
25068         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
25069         recursive calls.
25070         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25071         callers.
25072         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25074 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
25076         PR rtl-optimization/81443
25077         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25078         from inner REGs to paradoxical SUBREGs.
25080 2018-02-16  Richard Biener  <rguenther@suse.de>
25082         PR tree-optimization/84399
25083         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25084         For operands we can analyze at their definition make sure we can
25085         analyze them at each use as well.
25087 2018-02-16  Richard Biener  <rguenther@suse.de>
25089         PR tree-optimization/84190
25090         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25091         volatile accesses if the decl isn't volatile.
25093 2018-02-15  Jason Merrill  <jason@redhat.com>
25095         PR c++/84314 - ICE with templates and fastcall attribute.
25096         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25097         TYPE_CANONICAL on an existing type.
25099 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25101         PR tree-optimization/84383
25102         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25103         dstoff nor call operand_equal_p if dstbase is NULL.
25105         PR tree-optimization/84334
25106         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25107         also a CONSTANT_CLASS_P, punt.
25109 2018-02-14  Jim Wilson  <jimw@sifive.com>
25111         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25112         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25113         to where locals are set.  Add TARGET_RVC support.
25114         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25116 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25118         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25120 2018-02-14  Martin Sebor  <msebor@redhat.com>
25122         PR tree-optimization/83698
25123         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25124         arrays constrain the offset range to their bounds.
25125         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25126         (builtin_access::overlap): Avoid setting the size of overlap if it's
25127         already been set.
25128         (maybe_diag_overlap): Also consider arrays when deciding what values
25129         of offsets to include in diagnostics.
25131 2018-02-14  Martin Sebor  <msebor@redhat.com>
25133         PR c/84108
25134         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25135         that correspond to the kind of a declaration.
25137 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25139         PR target/83984
25140         * config/pa/pa.md: Load address of PIC label using the linkage table
25141         if the label is nonlocal.
25143 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25145         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25146         warning message if user requests -maltivec=be.
25147         * doc/invoke.texi: Document deprecation of -maltivec=be.
25149 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25151         PR target/84220
25152         * config/rs6000/rs6000-c.c: Update definitions for
25153         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25154         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25156 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25158         PR target/84239
25159         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25160         add _get_ssp intrinsics. Remove argument from
25161         __builtin_ia32_rdssp[d|q].
25162         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25163         * config/i386/i386-builtin.def: Remove argument from
25164         __builtin_ia32_rdssp[d|q].
25165         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25166         ix86_expand_special_args_builtin for _rdssp[d|q].
25167         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25168         Clear register before usage.
25169         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25170         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25172 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25174         PR tree-optimization/84357
25175         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25176         operand 1 of an ARRAY_REF too.
25178 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25180         PR target/83831
25181         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25182         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25183         declarations.
25184         (set_of_reg): New struct.
25185         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25186         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25187         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25188         functions.
25189         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25190         Split into bitclr, bitset, bitinvert patterns if appropriate.
25191         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25192         use rx_fuse_in_memory_bitop.
25193         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25194         to named insn, correct maximum insn length.
25196 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25198         PR target/79242
25199         * machmode.def: Define a complex mode for PARTIAL_INT.
25200         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25201         MODE_PARTIAL_INT.
25202         * doc/rtl.texi: Document CSPImode.
25203         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25204         handling.
25205         (msp430_hard_regno_nregs_with_padding): Likewise.
25207 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25209         PR target/84279
25210         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25212 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25214         PR rtl-optimization/84169
25215         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25216         we generated a parallel as new i3 and we split that to new i2 and i3
25217         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25218         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25219         those to i2, not i1.  Partially rewrite this scan code.
25221 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25223         PR c/82210
25224         * stor-layout.c (place_field): For variable length fields, adjust
25225         offset_align afterwards not just based on the field's alignment,
25226         but also on the size.
25228         PR middle-end/84309
25229         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25230         of exps and logs in the use_exp2 case.
25232 2018-02-13  Jeff Law  <law@redhat.com>
25234         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25235         entry for "vector".
25237         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25238         ARGS as unused.
25240 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25242         PR debug/84342
25243         PR debug/84319
25244         * common.opt (gas-loc-support, gas-locview-support): New.
25245         (ginline-points, ginternal-reset-location-views): New.
25246         * doc/invoke.texi: Document them.  Use @itemx where intended.
25247         (gvariable-location-views): Adjust.
25248         * target.def (reset_location_view): New.
25249         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25250         (TARGET_RESET_LOCATION_VIEW): New.
25251         * doc/tm.texi: Rebuilt.
25252         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25253         (dwarf2out_default_as_locview_support): New.
25254         (output_asm_line_debug_info): Use option variables.
25255         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25256         (output_loc_list): Likewise.
25257         (add_high_low_attributes): Check option variables.
25258         Don't output entry view attribute in strict mode.
25259         (gen_inlined_subroutine_die): Check option variables.
25260         (dwarf2out_inline_entry): Likewise.
25261         (init_sections_and_labels): Likewise.
25262         (dwarf2out_early_finish): Likewise.
25263         (maybe_reset_location_view): New, from...
25264         (dwarf2out_var_location): ... here.  Call it.
25265         * debug.h (dwarf2out_default_as_loc_support): Declare.
25266         (dwarf2out_default_as_locview_support): Declare.
25267         * hooks.c (hook_int_rtx_insn_0): New.
25268         * hooks.h (hook_int_rtx_insn_0): Declare.
25269         * toplev.c (process_options): Take -gas-loc-support and
25270         -gas-locview-support from dwarf2out.  Enable
25271         -gvariable-location-views by default only with locview
25272         assembler support.  Enable -ginternal-reset-location-views by
25273         default only if the target defines the corresponding hook.
25274         Enable -ginline-points by default if location views are
25275         enabled; force it disabled if statement frontiers are
25276         disabled.
25277         * tree-inline.c (expand_call_inline): Check option variables.
25278         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25280 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25282         PR tree-optimization/84321
25283         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25284         handling.  Also check whether the anti-range contains any values
25285         that satisfy the mask; switch to a VR_RANGE if not.
25287 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25289         PR sanitizer/84340
25290         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25292 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25294         PR c++/83990
25295         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25296         of call statements, also set location of a load to a temporary.
25298 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25300         * config/rl78/rl78.c (add_vector_labels): New function.
25301         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25302         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25303         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25304         which checks that no arguments are passed.
25305         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25306         * doc/extend.texi: Documentation for the new attribute.
25308 2018-02-13  Andreas Schwab  <schwab@suse.de>
25310         * config/riscv/linux.h (CPP_SPEC): Define.
25312 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25314         PR target/84335
25315         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25316         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25317         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25318         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25319         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25320         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25321         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25322         temporarily for AES and PCLMUL builtins.
25324         PR tree-optimization/84339
25325         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25326         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25327         Formatting fixes.
25329         PR middle-end/84309
25330         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25331         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25332         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25333         inline function.
25334         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25335         inline function.
25336         * omp-simd-clone.h: New file.
25337         * omp-simd-clone.c: Include omp-simd-clone.h.
25338         (expand_simd_clones): No longer static.
25339         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25340         cgraph.h and omp-simd-clone.h.
25341         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25342         (vect_recog_widen_shift_pattern): Formatting fix.
25343         (vect_pattern_recog_1): Don't check optab for calls.
25345         PR target/84336
25346         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25347         operands[2] into a REG before using gen_lowpart on it.
25349 2018-02-12  Jeff Law  <law@redhat.com>
25351         PR target/83760
25352         * config/sh/sh.c (find_barrier): Consider a sibling call
25353         a barrier as well.
25355         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25356         successfully back substituting a reg.
25358 2018-02-12  Richard Biener  <rguenther@suse.de>
25360         PR tree-optimization/84037
25361         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25362         parameter, move visited init to caller.
25363         (vect_slp_analyze_operations): Separate cost from validity
25364         check, initialize visited once for all instances.
25365         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25366         for all instances.
25367         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25368         out an assert.
25369         (vect_model_promotion_demotion_cost): Likewise.
25370         (vectorizable_bswap): Guard cost modeling with !slp_node
25371         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25372         SLP stmts.
25373         (vectorizable_call): Likewise.
25374         (vectorizable_conversion): Likewise.
25375         (vectorizable_assignment): Likewise.
25376         (vectorizable_shift): Likewise.
25377         (vectorizable_operation): Likewise.
25378         (vectorizable_store): Likewise.
25379         (vectorizable_load): Likewise.
25380         (vectorizable_condition): Likewise.
25381         (vectorizable_comparison): Likewise.
25383 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25385         PR sanitizer/84307
25386         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25387         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25388         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25390 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25392         PR middle-end/83665
25393         * params.def (inline-min-speedup): Increase from 8 to 15.
25394         (max-inline-insns-auto): Decrease from 40 to 30.
25395         * ipa-split.c (consider_split): Add some buffer for function to
25396         be considered inlining candidate.
25397         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25398         default values.
25400 2018-02-12  Richard Biener  <rguenther@suse.de>
25402         PR tree-optimization/84037
25403         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25404         matched stmts if we cannot swap the non-matched ones.
25406 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25408         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25409         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25410         _mm_maskz_scalef_round_ss): New intrinsics.
25411         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25412         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25413         __builtin_ia32_scalefss_round): Remove.
25414         (__builtin_ia32_scalefsd_mask_round,
25415         __builtin_ia32_scalefss_mask_round): New intrinsics.
25416         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25417         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25418         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25419         "<round_constraint>")): Changed to ...
25420         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25421         "<round_scalar_constraint>")): ... this.
25422         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25423         %0, %1, %2<round_op3>}"): Changed to ...
25424         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25425         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25426         %2<round_scalar_mask_op3>}"): ... this.
25427         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25429 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25431         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25432         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25433         (_mm_maskz_sqrt_round_ss): New intrinsics.
25434         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25435         (__builtin_ia32_sqrtsd_mask_round)
25436         (__builtin_ia32_sqrtss_mask_round): New builtins.
25437         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25438         (__builtin_ia32_sqrtss_round): Remove.
25439         (__builtin_ia32_sqrtsd_mask_round)
25440         (__builtin_ia32_sqrtss_mask_round): New builtins.
25441         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25442         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25443         ((match_operand:VF_128 1 "vector_operand"
25444         "xBm,<round_constraint>")): Changed to ...
25445         ((match_operand:VF_128 1 "vector_operand"
25446         "xBm,<round_scalar_constraint>")): ... this.
25447         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25448         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25449         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25450         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25451         %<iptr>1<round_scalar_mask_op3>}): ... this.
25452         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25453         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25455 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25457         PR target/84266
25458         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25459         Cast vec_cmpeq result to correct type.
25460         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25461         Cast vec_cmpgt result to correct type.
25463 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25465         * final.c (final_scan_insn_1): Renamed from...
25466         (final_scan_insn): ... this.  New wrapper, to recover
25467         seen from the outermost call in recursive ones.
25468         * config/sparc/sparc.c (output_return): Drop seen from call.
25469         (output_sibcall): Likewise.
25470         * config/visium/visium.c (output_branch): Likewise.
25472 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25474         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25475         function label.
25477 2018-02-10  Alan Modra  <amodra@gmail.com>
25479         PR target/84300
25480         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25481         Specify LR as an input.
25483 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25485         PR sanitizer/83987
25486         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25487         remove_member_access_dummy_vars): New functions.
25488         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25489         lower_omp_1, execute_lower_omp): Use them.
25491         PR rtl-optimization/84308
25492         * shrink-wrap.c (spread_components): Release todo vector.
25494 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25496         PR rtl-optimization/57193
25497         * ira-color.c (struct allocno_color_data): Add member
25498         conflict_allocno_hard_prefs.
25499         (update_conflict_allocno_hard_prefs): New.
25500         (bucket_allocno_compare_func): Add a preference based on
25501         conflict_allocno_hard_prefs.
25502         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25503         (color_allocnos): Remove a dead code.  Initiate
25504         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25506 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25508         PR target/84226
25509         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25510         constraint from =wa to wa.  Avoid a subreg on the output operand,
25511         instead use a pseudo and subreg it in a move.
25512         (p9_xxbrd_<mode>): Changed to ...
25513         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25514         (p9_xxbrd_v2df): New expander.
25515         (p9_xxbrw_<mode>): Changed to ...
25516         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25517         (p9_xxbrw_v4sf): New expander.
25519 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25521         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25523 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25525         PR target/83926
25526         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25527         multiply in 32-bit mode.
25528         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25529         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25530         mode.
25532 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25534         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25535         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25536         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25537         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25539 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25541         PR lto/84213
25542         * dwarf2out.c (is_trivial_indirect_ref): New function.
25543         (dwarf2out_late_global_decl): Do not generate a location
25544         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25545         and that are not defined in the current unit.
25547 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25549         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25550         instead of a libcall for UNORDERED.
25552 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25554         PR target/82641
25555         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25556         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25558 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25560         PR target/PR84295
25561         * config/s390/s390.c (s390_set_current_function): Invoke
25562         s390_indirect_branch_settings also if fndecl didn't change.
25564 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25566         * config/rs6000/rs6000.md (blockage): Set length to zero.
25568 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25570         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25572 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25574         PR sanitizer/84285
25575         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25576         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25577         -static-lib*san.
25579         PR debug/84252
25580         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
25581         PARALLEL incoming that failed vt_get_decl_and_offset check.
25583         PR middle-end/84237
25584         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
25585         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
25586         TREE_READONLY bit.
25587         (get_variable_section): For decls in named .bss* sections pass true as
25588         second argument to bss_initializer_p.
25590 2018-02-09  Marek Polacek  <polacek@redhat.com>
25591             Jakub Jelinek  <jakub@redhat.com>
25593         PR c++/83659
25594         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
25595         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
25596         Sync some changes from cxx_fold_indirect_ref.
25598 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25600         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
25601         markers.
25602         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
25603         (BLOCK_INLINE_ENTRY_LABEL): New.
25604         (dwarf2out_var_location): Disregard inline entry markers.
25605         (inline_entry_data): New struct.
25606         (inline_entry_data_hasher): New hashtable type.
25607         (inline_entry_data_hasher::hash): New.
25608         (inline_entry_data_hasher::equal): New.
25609         (inline_entry_data_table): New variable.
25610         (add_high_low_attributes): Add DW_AT_entry_pc and
25611         DW_AT_GNU_entry_view attributes if a pending entry is found
25612         in inline_entry_data_table.  Add old entry_pc attribute only
25613         if debug nonbinding markers are disabled.
25614         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
25615         markers are enabled.
25616         (block_within_block_p, dwarf2out_inline_entry): New.
25617         (dwarf2out_finish): Check that no entries remained in
25618         inline_entry_data_table.
25619         * final.c (reemit_insn_block_notes): Handle inline entry notes.
25620         (final_scan_insn, notice_source_line): Likewise.
25621         (rest_of_clean_state): Skip inline entry markers.
25622         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
25623         markers.
25624         * gimple.c (gimple_build_debug_inline_entry): New.
25625         * gimple.h (enum gimple_debug_subcode): Add
25626         GIMPLE_DEBUG_INLINE_ENTRY.
25627         (gimple_build_debug_inline_entry): Declare.
25628         (gimple_debug_inline_entry_p): New.
25629         (gimple_debug_nonbind_marker_p): Adjust.
25630         * insn-notes.def (INLINE_ENTRY): New.
25631         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
25632         inline entry marker notes.
25633         (print_insn): Likewise.
25634         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
25635         (INSN_DEBUG_MARKER_KIND): Likewise.
25636         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
25637         * tree-inline.c (expand_call_inline): Build and insert
25638         debug_inline_entry stmt.
25639         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
25640         inline entry blocks early, if nonbind markers are enabled.
25641         (dump_scope_block): Dump fragment info.
25642         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
25643         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
25644         (gimple_build_debug_inline_entry): New.
25645         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
25646         Enable/disable inline entry points too.
25647         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
25648         (DEBUG_INSN): Describe inline entry markers.
25650         * common.opt (gvariable-location-views): New.
25651         (gvariable-location-views=incompat5): New.
25652         * config.in: Rebuilt.
25653         * configure: Rebuilt.
25654         * configure.ac: Test assembler for view support.
25655         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
25656         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
25657         * dwarf2out.c (var_loc_view): New typedef.
25658         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
25659         (dwarf2out_locviews_in_attribute): New.
25660         (dwarf2out_locviews_in_loclist): New.
25661         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
25662         (enum dw_line_info_opcode): Add LI_adv_address.
25663         (struct dw_line_info_table): Add view.
25664         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
25665         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
25666         (zero_view_p): New variable.
25667         (ZERO_VIEW_P): New macro.
25668         (output_asm_line_debug_info): New.
25669         (struct var_loc_node): Add view.
25670         (add_AT_view_list, AT_loc_list): New.
25671         (add_var_loc_to_decl): Add view param.  Test it against last.
25672         (new_loc_list): Add view params.  Record them.
25673         (AT_loc_list_ptr): Handle loc and view lists.
25674         (view_list_to_loc_list_val_node): New.
25675         (print_dw_val): Handle dw_val_class_view_list.
25676         (size_of_die): Likewise.
25677         (value_format): Likewise.
25678         (loc_list_has_views): New.
25679         (gen_llsym): Set vl_symbol too.
25680         (maybe_gen_llsym, skip_loc_list_entry): New.
25681         (dwarf2out_maybe_output_loclist_view_pair): New.
25682         (output_loc_list): Output view list or entries too.
25683         (output_view_list_offset): New.
25684         (output_die): Handle dw_val_class_view_list.
25685         (output_dwarf_version): New.
25686         (output_compilation_unit_header): Use it.
25687         (output_skeleton_debug_sections): Likewise.
25688         (output_rnglists, output_line_info): Likewise.
25689         (output_pubnames, output_aranges): Update version comments.
25690         (output_one_line_info_table): Output view numbers in asm comments.
25691         (dw_loc_list): Determine current endview, pass it to new_loc_list.
25692         Call maybe_gen_llsym.
25693         (loc_list_from_tree_1): Adjust.
25694         (add_AT_location_description): Create view list attribute if
25695         needed, check it's absent otherwise.
25696         (convert_cfa_to_fb_loc_list): Adjust.
25697         (maybe_emit_file): Call output_asm_line_debug_info for test.
25698         (dwarf2out_var_location): Reset views as needed.  Precompute
25699         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
25700         attribute.  Set view.
25701         (new_line_info_table): Reset next view.
25702         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
25703         (dwarf2out_source_line): Likewise.  Output view resets and labels to
25704         the assembler, or select appropriate line info opcodes.
25705         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
25706         (optimize_string_length): Catch it.  Adjust.
25707         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
25708         dw_val_class_view_list, and remove it if no longer needed.
25709         (hash_loc_list): Hash view numbers.
25710         (loc_list_hasher::equal): Compare them.
25711         (optimize_location_lists): Check whether a view list symbol is
25712         needed, and whether the locview attribute is present, and
25713         whether they match.  Remove the locview attribute if no longer
25714         needed.
25715         (index_location_lists): Call skip_loc_list_entry for test.
25716         (dwarf2out_finish): Call output_asm_line_debug_info for test.
25717         Use output_dwarf_version.
25718         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
25719         (struct dw_val_node): Add val_view_list.
25720         * final.c (SEEN_NEXT_VIEW): New.
25721         (set_next_view_needed): New.
25722         (clear_next_view_needed): New.
25723         (maybe_output_next_view): New.
25724         (final_start_function): Rename to...
25725         (final_start_function_1): ... this.  Take pointer to FIRST,
25726         add SEEN parameter.  Emit param bindings in the initial view.
25727         (final_start_function): Reintroduce SEEN-less interface.
25728         (final): Rename to...
25729         (final_1): ... this.  Take SEEN parameter.  Output final pending
25730         next view at the end.
25731         (final): Reintroduce seen-less interface.
25732         (final_scan_insn): Output pending next view before switching
25733         sections or ending a block.  Mark the next view as needed when
25734         outputting variable locations.  Notify debug backend of section
25735         changes, and of location view changes.
25736         (rest_of_handle_final): Adjust.
25737         * toplev.c (process_options): Autodetect value for debug variable
25738         location views option.  Warn on incompat5 without -gdwarf-5.
25739         * doc/invoke.texi (gvariable-location-views): New.
25740         (gvariable-location-views=incompat5): New.
25741         (gno-variable-location-views): New.
25743 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
25745         PR tree-optimization/84136
25746         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
25747         that the result of find_edge is non-NULL.
25749 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25751         PR target/83008
25752         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
25753         storing integer register in SImode.  Fix cost of 256 and 512
25754         byte aligned SSE register store.
25756 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25758         * config/i386/i386.c (ix86_multiplication_cost): Fix
25759         multiplication cost for TARGET_AVX512DQ.
25761 2018-02-08  Marek Polacek  <polacek@redhat.com>
25763         PR tree-optimization/84238
25764         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
25765         get_range_strlen.
25767 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25769         PR tree-optimization/84265
25770         * tree-vect-stmts.c (vectorizable_store): Don't treat
25771         VMAT_CONTIGUOUS accesses as grouped.
25772         (vectorizable_load): Likewise.
25774 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25776         PR tree-optimization/81635
25777         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
25778         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
25779         (test_round_for_mask): New functions.
25780         (wide_int_cc_tests): Call test_round_for_mask.
25781         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
25782         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
25783         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
25784         range returned by get_range_info.
25786 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25788         PR ipa/81360
25789         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
25790         * symtab.c: Include builtins.h
25791         (symtab_node::output_to_lto_symbol_table_p): Move here
25792         from lto-streamer-out.c:output_symbol_p.
25793         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
25794         (output_symbol_p): Move all logic to symtab.c
25795         (produce_symtab): Update.
25797 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25799         * config/s390/s390-opts.h (enum indirect_branch): Define.
25800         * config/s390/s390-protos.h (s390_return_addr_from_memory)
25801         (s390_indirect_branch_via_thunk)
25802         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
25803         (enum s390_indirect_branch_type): Define.
25804         * config/s390/s390.c (struct s390_frame_layout, struct
25805         machine_function): Remove.
25806         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
25807         (indirect_branch_table_label_no, indirect_branch_table_name):
25808         Define variables.
25809         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
25810         (enum s390_indirect_branch_option): Define.
25811         (s390_return_addr_from_memory): New function.
25812         (s390_handle_string_attribute): New function.
25813         (s390_attribute_table): Add new attribute handler.
25814         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
25815         (s390_indirect_branch_via_thunk): New function.
25816         (s390_indirect_branch_via_inline_thunk): New function.
25817         (s390_function_ok_for_sibcall): When jumping via thunk disallow
25818         sibling call optimization for non z10 compiles.
25819         (s390_emit_call): Force indirect branch target to be a single
25820         register.  Add r1 clobber for non-z10 compiles.
25821         (s390_emit_epilogue): Emit return jump via return_use expander.
25822         (s390_reorg): Handle JUMP_INSNs as execute targets.
25823         (s390_option_override_internal): Perform validity checks for the
25824         new command line options.
25825         (s390_indirect_branch_attrvalue): New function.
25826         (s390_indirect_branch_settings): New function.
25827         (s390_set_current_function): Invoke s390_indirect_branch_settings.
25828         (s390_output_indirect_thunk_function):  New function.
25829         (s390_code_end): Implement target hook.
25830         (s390_case_values_threshold): Implement target hook.
25831         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
25832         macros.
25833         * config/s390/s390.h (struct s390_frame_layout)
25834         (struct machine_function): Move here from s390.c.
25835         (TARGET_INDIRECT_BRANCH_NOBP_RET)
25836         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
25837         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
25838         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
25839         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
25840         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
25841         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
25842         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
25843         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
25844         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
25845         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
25846         (mnemonic attribute): Add values which aren't recognized
25847         automatically.
25848         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
25849         pattern for branch conversion.  Fix mnemonic attribute.
25850         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
25851         indirect branch via thunk if requested.
25852         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
25853         ("*indirect_jump"): Disable for branch conversion using out of
25854         line thunks.
25855         ("indirect_jump_via_thunk<mode>_z10")
25856         ("indirect_jump_via_thunk<mode>")
25857         ("indirect_jump_via_inlinethunk<mode>_z10")
25858         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
25859         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
25860         ("casesi_jump_via_inlinethunk<mode>_z10")
25861         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
25862         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
25863         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
25864         ("*indirect2_jump"): Disable for branch conversion.
25865         ("casesi_jump"): Turn into expander and expand patterns for branch
25866         conversion.
25867         ("return_use"): New expander.
25868         ("*return"): Emit return via thunk and rename it to ...
25869         ("*return<mode>"): ... this one.
25870         * config/s390/s390.opt: Add new options and and enum for the
25871         option values.
25873 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25875         * lra-constraints.c (match_reload): Unconditionally use
25876         gen_lowpart_SUBREG, rather than selecting between that
25877         and equivalent gen_rtx_SUBREG code.
25879 2018-02-08  Richard Biener  <rguenther@suse.de>
25881         PR tree-optimization/84233
25882         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
25883         changed flag instead of boguously re-using phi_inserted.
25885 2018-02-08  Martin Jambor  <mjambor@suse.cz>
25887         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
25888         static local variables.
25890 2018-02-08  Richard Biener  <rguenther@suse.de>
25892         PR tree-optimization/84278
25893         * tree-vect-stmts.c (vectorizable_store): When looking for
25894         smaller vector types to perform grouped strided loads/stores
25895         make sure the mode is supported by the target.
25896         (vectorizable_load): Likewise.
25898 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25900         * config/aarch64/aarch64.c (aarch64_components_for_bb):
25901         Increase LDP/STP opportunities by adding adjacent callee-saves.
25903 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25905         PR rtl-optimization/84068
25906         PR rtl-optimization/83459
25907         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
25909 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
25911         PR tree-optimization/84224
25912         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
25913         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
25914         non-zero arguments.
25916 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
25918         PR target/84113
25919         * config/rs6000/altivec.md (*restore_world): Remove LR use.
25920         * config/rs6000/predicates.md (restore_world_operation): Adjust op
25921         count, remove one USE.
25923 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25925         * doc/install.texi (Configuration): Document the
25926         --with-long-double-format={ibm,ieee} PowerPC configuration
25927         options.
25929         PR target/84154
25930         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
25931         Convert from define_expand to be define_insn_and_split.  Rework
25932         float/double/_Float128 conversions to QI/HI/SImode to work with
25933         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
25934         conversions to QI/HImode types did a store and then a load to
25935         truncate the value.  For conversions to VSX registers, don't split
25936         the insn, instead emit the code directly.  Use the code iterator
25937         any_fix to combine signed and unsigned conversions.
25938         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
25939         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
25940         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25941         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
25942         (fix_<mode>di2_hw): Likewise.
25943         (fixuns_<mode>di2_hw): Likewise.
25944         (fix_<mode>si2_hw): Likewise.
25945         (fixuns_<mode>si2_hw): Likewise.
25946         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
25947         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25948         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
25949         fix<uns>_trunc<SFDF:mode>si2_p8.
25950         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
25951         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
25952         (fix<uns>_<mode>_mem): Likewise.
25953         (fctiw<u>z_<mode>_mem): Likewise.
25954         (fix<uns>_<mode>_mem): Likewise.
25955         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
25956         the register allocator from doing a direct move to the GPRs to do
25957         a store, and instead use the ISA 3.0 store byte/half-word from
25958         vector register instruction.  For IEEE 128-bit floating point,
25959         also optimize stores of 32-bit ints.
25960         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
25962 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
25964         * genextract.c (push_pathstr_operand): New function to support
25965         [a-zA-Z].
25966         (walk_rtx): Call push_pathstr_operand.
25967         (print_path): Support [a-zA-Z].
25969 2018-02-07  Richard Biener  <rguenther@suse.de>
25971         PR tree-optimization/84037
25972         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
25973         (cse_and_gimplify_to_preheader): Declare.
25974         (vect_get_place_in_interleaving_chain): Likewise.
25975         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25976         ivexpr_map.
25977         (_loop_vec_info::~_loop_vec_info): Delete it.
25978         (cse_and_gimplify_to_preheader): New function.
25979         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
25980         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
25981         (vectorizable_load): Likewise.  For grouped stores always base
25982         the IV on the first element.
25983         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
25984         condition before gimplifying.
25986 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25988         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
25989         *DIV_EXPR and *MOD_EXPR.
25991 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
25993         PR target/84248
25994         * config/i386/i386.c (ix86_option_override_internal): Mask out
25995         the CF_SET bit when checking -fcf-protection.
25997 2018-02-07  Tom de Vries  <tom@codesourcery.com>
25999         PR libgomp/84217
26000         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26001         enough.
26003 2018-02-07  Richard Biener  <rguenther@suse.de>
26005         PR tree-optimization/84204
26006         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26007         this place.
26009         PR tree-optimization/84205
26010         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26011         special-case isl_ast_op_zdiv_r.
26013         PR tree-optimization/84223
26014         * graphite-scop-detection.c (gather_bbs::before_dom_children):
26015         Only add conditions from within the region.
26016         (gather_bbs::after_dom_children): Adjust.
26018 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
26020         PR target/84209
26021         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26022         * config/avr/avr.md: Only post-reload split REG-REG moves if
26023         either register is GENERAL_REG_P.
26025 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26027         PR tree-optimization/84235
26028         * tree-ssa-scopedtables.c
26029         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26030         if the subtraction is performed in floating point type where NaNs are
26031         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26032         build 1.  Formatting fix.
26034 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
26036         PR target/84146
26037         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26038         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26039         and skip it regardless of bb boundaries.  Use CALL_P macro,
26040         don't test INSN_P (insn) together with CALL_P or JUMP_P check
26041         unnecessarily, formatting fix.
26043 2018-02-06  Michael Collison  <michael.collison@arm.com>
26045         * config/arm/thumb2.md:
26046         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26047         (*thumb_mov_notscc): Ditto.
26049 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
26051         PR target/84154
26052         * config/rs6000/rs6000.md (su code attribute): Use "u" for
26053         unsigned_fix, not "s".
26055 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26057         * configure.ac (gcc_fn_eh_frame_ro): New function.
26058         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26059         correct .eh_frame permissions.
26060         * configure: Regenerate.
26062 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
26064         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26065         irrelevant options.
26067 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26069         * config/rs6000/rs6000.c (rs6000_option_override_internal):
26070         Display warning message for -mno-speculate-indirect-jumps.
26072 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
26074         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26075         Undocumented.
26076         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26078 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26080         PR tree-optimization/84225
26081         * tree-eh.c (find_trapping_overflow): Only call
26082         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26084 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26086         PR target/84145
26087         * config/i386/i386.c: Reimplement the check of possible options
26088         -mibt/-mshstk conbination. Change error messages.
26089         * doc/invoke.texi: Fix a typo: remove extra '='.
26091 2018-02-06  Marek Polacek  <polacek@redhat.com>
26093         PR tree-optimization/84228
26094         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26096 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26098         PR target/82641
26099         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26100         emitted arch directives.
26101         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26102         __ARM_FEATURE_COPROC before changing architectures.
26104 2018-02-06  Richard Biener  <rguenther@suse.de>
26106         * config/i386/i386.c (print_reg): Fix typo.
26107         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26109 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26111         * configure: Regenerate.
26113 2018-02-05  Martin Sebor  <msebor@redhat.com>
26115         PR tree-optimization/83369
26116         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26117         inlining context.
26119 2018-02-05  Martin Liska  <mliska@suse.cz>
26121         * doc/invoke.texi: Cherry-pick upstream r323995.
26123 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26125         * ira.c (ira_init_register_move_cost): Adjust comment.
26127 2018-02-05  Martin Liska  <mliska@suse.cz>
26129         PR gcov-profile/84137
26130         * doc/gcov.texi: Fix typo in documentation.
26132 2018-02-05  Martin Liska  <mliska@suse.cz>
26134         PR gcov-profile/83879
26135         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26136         using dlopen functionality.
26138 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26140         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26141         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26142         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26143         _mm_maskz_range_round_ss): New intrinsics.
26144         (__builtin_ia32_rangesd128_round)
26145         (__builtin_ia32_rangess128_round): Remove.
26146         (__builtin_ia32_rangesd128_mask_round,
26147         __builtin_ia32_rangess128_mask_round): New builtins.
26148         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26149         __builtin_ia32_rangess128_round): Remove.
26150         (__builtin_ia32_rangesd128_mask_round,
26151         __builtin_ia32_rangess128_mask_round): New builtins.
26152         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26153         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26154         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26155         "<round_saeonly_constraint>")): Changed to ...
26156         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26157         "<round_saeonly_scalar_constraint>")): ... this.
26158         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26159         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26160         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26161         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26162         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26164 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26166         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26167         options.
26168         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26169         Remove all values except native, 8540 and 8548.
26171 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26173         * config/i386/i386.c (ix86_output_function_return): Pass
26174         INVALID_REGNUM, instead of -1, as invalid register number to
26175         indirect_thunk_name and output_indirect_thunk.
26177 2018-02-02  Julia Koval  <julia.koval@intel.com>
26179         * config.gcc: Add -march=icelake.
26180         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26181         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26182         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26183         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26184         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26185         (processor_target_table): Add icelake.
26186         (ix86_option_override_internal): Handle new PTAs.
26187         (get_builtin_code_for_version): Handle icelake.
26188         (M_INTEL_COREI7_ICELAKE): New.
26189         (fold_builtin_cpu): Handle icelake.
26190         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26191         * doc/invoke.texi: Add -march=icelake.
26193 2018-02-02  Julia Koval  <julia.koval@intel.com>
26195         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26196         to wide_int_bitmask.
26197         * wide-int-bitmask.h: New.
26199 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26201         PR target/84066
26202         * config/i386/i386.md: Replace Pmode with word_mode in
26203         builtin_setjmp_setup and builtin_longjmp to support x32.
26205 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26207         PR target/56010
26208         PR target/83743
26209         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26210         #include "opts.h".
26211         (rs6000_supported_cpu_names): New static variable.
26212         (linux_cpu_translation_table): Likewise.
26213         (elf_platform) <cpu>: Define new static variable and use it.
26214         Translate kernel AT_PLATFORM name to canonical name if needed.
26215         Error if platform name is unknown.
26217 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26219         PR target/84089
26220         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26222 2018-02-01  Jeff Law  <law@redhat.com>
26224         PR target/84128
26225         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26226         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26227         the scratch if RELEASE_VIA_POP is false.
26228         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26229         If we have to save a temporary register, decrement SIZE appropriately.
26230         Pass new arguments to release_scratch_register_on_entry.
26231         (ix86_adjust_stack_and_probe): Likewise.
26232         (ix86_emit_probe_stack_range): Pass new arguments to
26233         release_scratch_register_on_entry.
26235 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26237         PR rtl-optimization/84157
26238         * combine.c (change_zero_ext): Use REG_P predicate in
26239         front of HARD_REGISTER_P predicate.
26241 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26243         * config/avr/avr.c (avr_option_override): Move disabling of
26244         -fdelete-null-pointer-checks to...
26245         * common/config/avr/avr-common.c (avr_option_optimization_table):
26246         ...here.
26248 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26250         PR tree-optimization/81635
26251         * tree-data-ref.c (split_constant_offset_1): For types that
26252         wrap on overflow, try to use range info to prove that wrapping
26253         cannot occur.
26255 2018-02-01  Renlin Li  <renlin.li@arm.com>
26257         PR target/83370
26258         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26259         TAILCALL_ADDR_REGS.
26260         (aarch64_register_move_cost): Likewise.
26261         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26262         TAILCALL_ADDR_REGS.
26263         (REG_CLASS_NAMES): Likewise.
26264         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26265         TAILCALL_ADDR_REGS. Remove IP registers.
26266         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26268 2018-02-01  Richard Biener  <rguenther@suse.de>
26270         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26271         for specifying RPO order and allow NULL for that.
26272         * domwalk.c (dom_walker::dom_walker): Likewise.
26273         (dom_walker::walk): Handle NULL RPO order.
26274         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26275         in RPO order.
26276         (rewrite_update_dom_walker): Likewise.
26277         (mark_def_dom_walker): Likewise.
26279 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26281         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26282         (aarch64_maybe_expand_sve_subreg_move): Declare.
26283         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26284         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26285         predicate.
26286         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26287         that are semantically a reverse operation.
26288         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26289         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26290         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26291         functions.
26292         (aarch64_can_change_mode_class): For big-endian, forbid changes
26293         between two SVE modes if they have different element sizes.
26295 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26297         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26298         the TImode handling for big-endian targets.
26300 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26302         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26303         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26304         not just bytes.
26305         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26306         Remove BSWAP handing for big-endian targets and use the form of
26307         LD1RQ appropariate for the mode.
26309 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26311         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26312         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26313         duplicated element.
26315 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26317         PR tearget/83845
26318         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26319         check for operands that need to go through aarch64_sve_reload_be.
26321 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26323         PR tree-optimization/81661
26324         PR tree-optimization/84117
26325         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26326         * tree-eh.c: Include gimplify.h.
26327         (find_trapping_overflow, replace_trapping_overflow,
26328         rewrite_to_non_trapping_overflow): New functions.
26329         * tree-vect-loop.c: Include tree-eh.h.
26330         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26331         * tree-data-ref.c: Include tree-eh.h.
26332         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26334 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26336         PR rtl-optimization/84123
26337         * combine.c (change_zero_ext): Check if hard register satisfies
26338         can_change_dest_mode before calling gen_lowpart_SUBREG.
26340 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26342         PR target/82444
26343         * ira.c (ira_init_register_move_cost): Remove assert.
26345 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26347         PR rtl-optimization/84071
26348         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26349         * doc/tm.texi: Regenerate.
26351 2018-01-31  Richard Biener  <rguenther@suse.de>
26353         PR tree-optimization/84132
26354         * tree-data-ref.c (analyze_miv_subscript): Properly
26355         check whether evolution_function_is_affine_multivariate_p
26356         before calling gcd_of_steps_may_divide_p.
26358 2018-01-31  Julia Koval  <julia.koval@intel.com>
26360         PR target/83618
26361         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26362         * config/i386/i386.md (rdpid_rex64) New.
26363         (rdpid): Make 32bit only.
26365 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26367         PR lto/84105
26368         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26369         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26371 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26373         Revert
26374         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26376         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26378 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26380         PR rtl-optimization/84071
26381         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26382         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26384 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26386         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26387         (arc_attribute_table): Add 'aux' attribute.
26388         (arc_in_small_data_p): Consider aux like variables.
26389         (arc_is_aux_reg_p): New function.
26390         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26391         (arc_get_aux_arg): New function.
26392         (prepare_move_operands): Handle aux-register access.
26393         (arc_handle_aux_attribute): New function.
26394         * doc/extend.texi (ARC Variable attributes): Add subsection.
26396 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26398         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26399         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26400         (arc_attribute_table): Add 'uncached' attribute.
26401         (arc_print_operand): Print '.di' flag for uncached memory
26402         accesses.
26403         (arc_in_small_data_p): Do not consider for small data the uncached
26404         types.
26405         (arc_is_uncached_mem_p): New function.
26406         * config/arc/predicates.md (compact_store_memory_operand): Check
26407         for uncached memory accesses.
26408         (nonvol_nonimm_operand): Likewise.
26409         * doc/extend.texi (ARC Type Attribute): New subsection.
26411 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26413         PR c/84100
26414         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26415         falign-loops=): Add Optimization flag.
26417 2018-01-30  Jeff Law  <law@redhat.com>
26419         PR target/84064
26420         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26421         INT_REGISTERS_SAVED.  Check it prior to calling
26422         get_scratch_register_on_entry.
26423         (ix86_adjust_stack_and_probe): Similarly.
26424         (ix86_emit_probe_stack_range): Similarly.
26425         (ix86_expand_prologue): Corresponding changes.
26427 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26429         PR target/40411
26430         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26431         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26433 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26435         PR target/84112
26436         * lra-constraints.c (curr_insn_transform): Process AND in the
26437         address.
26439 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26441         PR rtl-optimization/83986
26442         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26443         dependence against last_pending_memory_flush in addition to
26444         pending_jump_insns.
26446 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26448         PR tree-optimization/81611
26449         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26450         copies.
26452 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26454         PR target/83758
26455         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26456         a reg rtx.
26458 2018-01-30  Richard Biener  <rguenther@suse.de>
26459             Jakub Jelinek  <jakub@redhat.com>
26461         PR tree-optimization/84111
26462         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26463         inner loops added during recursion, as they don't have up-to-date
26464         SSA form.
26466 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26468         PR ipa/81360
26469         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26470         (can_inline_edge_by_limits_p): ... here.
26471         (can_early_inline_edge_p, check_callers,
26472         update_caller_keys, update_callee_keys, recursive_inlining,
26473         add_new_edges_to_heap, speculation_useful_p,
26474         inline_small_functions,
26475         inline_small_functions, flatten_function,
26476         inline_to_all_callers_1): Update.
26478 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26480         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26481         zeros correctly.
26483 2018-01-30  Richard Biener  <rguenther@suse.de>
26485         PR tree-optimization/83008
26486         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26487         invariant and constant vector uses in stmts when they need
26488         more than one stmt.
26490 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26492         PR bootstrap/84017
26493         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26494         * configure: Regenerate.
26496 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26498         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26499         pattern.
26500         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26501         Use gen_rtx_REG rather than gen_lowpart.
26503 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26505         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26506         rather than 0 when creating partial subregs.
26508 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26510         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26511         of usage.
26513 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26515         PR target/81550
26516         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26517         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26518         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26519         flags.  This restores the settings used before the 2017-07-24.
26520         Turning off pre increment/decrement/modify allows IVOPTS to
26521         optimize DF/SF loops where the index is an int.
26523 2018-01-29  Richard Biener  <rguenther@suse.de>
26524             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26526         PR bootstrap/80867
26527         * tree-vect-stmts.c (vectorizable_call): Don't call
26528         targetm.vectorize_builtin_md_vectorized_function if callee is
26529         NULL.
26531 2018-01-22  Carl Love  <cel@us.ibm.com>
26533         * doc/extend.tex: Fix typo in second arg in
26534         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26536 2018-01-29  Richard Biener  <rguenther@suse.de>
26538         PR tree-optimization/84086
26539         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26540         (flush_ssaname_freelist): When SSA names were released reset
26541         the SCEV hash table.
26543 2018-01-29  Richard Biener  <rguenther@suse.de>
26545         PR tree-optimization/84057
26546         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26547         removed paths when removing edges.
26549 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26551         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26552         -mfunction-return=@var{choice}.
26554 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26556         PR diagnostic/84034
26557         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26558         Handle CR like TAB.
26559         (layout::print_source_line): Likewise.
26560         (test_get_line_width_without_trailing_whitespace): Add test cases.
26562 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26564         PR middle-end/84040
26565         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26566         debug insns.
26568 2018-01-26  Jim Wilson  <jimw@sifive.com>
26570         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26572         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26573         specified.
26575 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26577         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
26578         and CMP + SUB-immediate -> SUBS.
26580 2018-01-26  Martin Sebor  <msebor@redhat.com>
26582         PR tree-optimization/83896
26583         * tree-ssa-strlen.c (get_string_len): Rename...
26584         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
26585         Avoid assuming length is constant.
26586         (handle_char_store): Use HOST_WIDE_INT for string length.
26588 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
26590         PR target/81763
26591         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
26592         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
26594 2018-01-26  Richard Biener  <rguenther@suse.de>
26596         PR rtl-optimization/84003
26597         * dse.c (record_store): Only record redundant stores when
26598         the earlier store aliases at least all accesses the later one does.
26600 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
26602         PR rtl-optimization/83985
26603         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
26604         REG_CFA_RESTORE insns.
26605         (delete_unmarked_insns): Don't ignore separate shrink wrapping
26606         REG_CFA_RESTORE insns here.
26608         PR c/83989
26609         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
26610         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
26612 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26614         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
26615         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
26616         (arc_init): Likewise.
26617         (arc_override_options): Likewise.
26618         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
26619         value.
26620         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
26621         support.
26622         * config/arc/arc.h (TARGET_DBNZ): Define.
26623         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
26624         properly set the tune attribute.
26625         (dbnz): Use TARGET_DBNZ guard.
26626         * config/arc/arc.opt (mtune): Add core3 option.
26628 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26630         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
26631         recognize new pic like addresses.
26632         (arc_delegitimize_address): Clean up.
26634 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26636         * config/arc/arc-arches.def: Option mrf16 valid for all
26637         architectures.
26638         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
26639         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
26640         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
26641         * config/arc/arc-tables.opt: Regenerate.
26642         * config/arc/arc.c (arc_conditional_register_usage): Handle
26643         reduced register file case.
26644         (arc_file_start): Set must have build attributes.
26645         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
26646         mrf16 option value.
26647         * config/arc/arc.opt (mrf16): Add new option.
26648         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
26649         * config/arc/genmultilib.awk: Handle new mrf16 option.
26650         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
26651         * config/arc/t-multilib: Regenerate.
26652         * doc/invoke.texi (ARC Options): Document mrf16 option.
26654 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26656         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
26657         * config/arc/arc.c (arc_handle_secure_attribute): New function.
26658         (arc_attribute_table): Add 'secure_call' attribute.
26659         (arc_print_operand): Print secure call operand.
26660         (arc_function_ok_for_sibcall): Don't optimize tail calls when
26661         secure.
26662         (arc_is_secure_call_p): New function.  * config/arc/arc.md
26663         (call_i): Add support for sjli instruction.
26664         (call_value_i): Likewise.
26665         * config/arc/constraints.md (Csc): New constraint.
26667 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26668             John Eric Martin  <John.Martin@emmicro-us.com>
26670         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
26671         * config/arc/arc.c (_arc_jli_section): New struct.
26672         (arc_jli_section): New type.
26673         (rc_jli_sections): New static variable.
26674         (arc_handle_jli_attribute): New function.
26675         (arc_attribute_table): Add jli_always and jli_fixed attribute.
26676         (arc_file_end): New function.
26677         (TARGET_ASM_FILE_END): Define.
26678         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
26679         (arc_add_jli_section): New function.
26680         (jli_call_scan): Likewise.
26681         (arc_reorg): Call jli_call_scan.
26682         (arc_output_addsi): Remove 'S' from printing asm operand.
26683         (arc_is_jli_call_p): New function.
26684         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
26685         operand.
26686         (movhi_insn): Likewise.
26687         (movsi_insn): Likewise.
26688         (movsi_set_cc_insn): Likewise.
26689         (loadqi_update): Likewise.
26690         (load_zeroextendqisi_update): Likewise.
26691         (load_signextendqisi_update): Likewise.
26692         (loadhi_update): Likewise.
26693         (load_zeroextendhisi_update): Likewise.
26694         (load_signextendhisi_update): Likewise.
26695         (loadsi_update): Likewise.
26696         (loadsf_update): Likewise.
26697         (movsicc_insn): Likewise.
26698         (bset_insn): Likewise.
26699         (bxor_insn): Likewise.
26700         (bclr_insn): Likewise.
26701         (bmsk_insn): Likewise.
26702         (bicsi3_insn): Likewise.
26703         (cmpsi_cc_c_insn): Likewise.
26704         (movsi_ne): Likewise.
26705         (movsi_cond_exec): Likewise.
26706         (clrsbsi2): Likewise.
26707         (norm_f): Likewise.
26708         (normw): Likewise.
26709         (swap): Likewise.
26710         (divaw): Likewise.
26711         (flag): Likewise.
26712         (sr): Likewise.
26713         (kflag): Likewise.
26714         (ffs): Likewise.
26715         (ffs_f): Likewise.
26716         (fls): Likewise.
26717         (call_i): Remove 'S' asm letter, add jli instruction.
26718         (call_value_i): Likewise.
26719         * config/arc/arc.op (mjli-always): New option.
26720         * config/arc/constraints.md (Cji): New constraint.
26721         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
26722         operand.
26723         (subsf3_fpx): Likewise.
26724         (mulsf3_fpx): Likewise.
26725         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
26726         asm operand.
26727         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
26728         function attrbutes.
26729         * doc/invoke.texi (ARC): Document mjli-always option.
26731 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
26733         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
26734         avoid addition with 0 and use incw and decw where possible.
26736 2018-01-26  Richard Biener  <rguenther@suse.de>
26738         PR tree-optimization/81082
26739         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
26740         association if it requires casting to unsigned.
26741         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
26742         from fold_plusminus_mult_expr to catch important cases late when
26743         range info is available.
26745 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26747         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
26748         * configure.ac (hidden_linkonce): New test.
26749         * configure: Regenerate.
26750         * config.in: Regenerate.
26752 2018-01-26  Julia Koval  <julia.koval@intel.com>
26754         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
26755         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
26756         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
26757         _mm_mask_bitshuffle_epi64_mask): Fix type.
26758         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
26759         USI_FTYPE_V4DI_V4DI_USI): Remove.
26760         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
26761         __builtin_ia32_vpshufbitqmb256_mask,
26762         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
26763         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
26764         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
26766 2018-01-26  Alan Modra  <amodra@gmail.com>
26768         PR target/84033
26769         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
26770         UNSPEC_VBPERMQ.  Sort other unspecs.
26772 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
26774         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
26776 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
26778         PR middle-end/83055
26779         * predict.c (drop_profile): Do not push/pop cfun; update also
26780         node->count.
26781         (handle_missing_profiles): Fix logic looking for zero profiles.
26783 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
26785         PR middle-end/83977
26786         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
26787         on functions with #pragma omp declare simd or functions with simd
26788         attribute.
26789         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
26790         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
26791         Remove trailing \n from warning_at calls.
26793 2018-01-25  Tom de Vries  <tom@codesourcery.com>
26795         PR target/84028
26796         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26797         for neutered workers.
26799 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
26801         PR target/68467
26802         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
26803         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
26805 2018-01-24  Jeff Law  <law@redhat.com>
26807         PR target/83994
26808         * i386.c (get_probe_interval): Move to earlier point.
26809         (ix86_compute_frame_layout): If -fstack-clash-protection and
26810         the frame is larger than the probe interval, then use pushes
26811         to save registers rather than reg->mem moves.
26812         (ix86_expand_prologue): Remove conditional for int_registers_saved
26813         assertion.
26815 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
26817         PR target/84014
26818         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
26819         min/max for never referenced object.
26821 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
26823         PR middle-end/83977
26824         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
26825         here.
26826         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
26827         attributes from DECL_ATTRIBUTES (decl) without affecting
26828         DECL_ATTRIBUTES (current_function_decl).
26829         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
26830         functions with non-NULL DECL_ABSTRACT_ORIGIN.
26832 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
26834         PR tree-optimization/83979
26835         * fold-const.c (fold_comparison): Use constant_boolean_node
26836         instead of boolean_{true,false}_node.
26838 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
26840         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
26841         with zero counts.
26843 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26845         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26846         Simplify the clause that sets the length attribute.
26847         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26848         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
26849         clause that sets the length attribute.
26850         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26852 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26854         PR target/83589
26855         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
26856         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
26857         Add strict parameter.
26858         (prevent_branch_around_nothing): Insert dummy insn between branch to
26859         label and label with no ptx insn inbetween.
26860         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
26862 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26864         PR target/81352
26865         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26866         for neutered threads in warp.
26867         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
26869 2018-01-24  Richard Biener  <rguenther@suse.de>
26871         PR tree-optimization/83176
26872         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
26873         operands.
26875 2018-01-24  Richard Biener  <rguenther@suse.de>
26877         PR tree-optimization/82819
26878         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
26879         code generating pluses that are no-ops in the target precision.
26881 2018-01-24  Richard Biener  <rguenther@suse.de>
26883         PR middle-end/84000
26884         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
26886 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26888         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
26889         to merge probabilities.
26890         * predict.c (probably_never_executed): Also mark as cold functions
26891         with global 0 profile and guessed local profile.
26892         * profile-count.c (profile_probability::combine_with_count): New
26893         member function.
26894         * profile-count.h (profile_probability::operator*,
26895         profile_probability::operator*=, profile_probability::operator/,
26896         profile_probability::operator/=): Reduce precision to adjusted
26897         and set value to guessed on contradictory divisions.
26898         (profile_probability::combine_with_freq): Remove.
26899         (profile_probability::combine_wiht_count): Declare.
26900         (profile_count::force_nonzero):: Set to adjusted.
26901         (profile_count::probability_in):: Set quality to adjusted.
26902         * tree-ssa-tail-merge.c (replace_block_by): Use
26903         combine_with_count.
26905 2018-01-23  Andrew Waterman  <andrew@sifive.com>
26906             Jim Wilson  <jimw@sifive.com>
26908         * config/riscv/riscv.c (riscv_stack_boundary): New.
26909         (riscv_option_override): Set riscv_stack_boundary.  Handle
26910         riscv_preferred_stack_boundary_arg.
26911         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
26912         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
26913         (STACK_BOUNDARY): Set to riscv_stack_boundary.
26914         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
26915         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
26916         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
26918 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
26920         PR target/83905
26921         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
26922         of struct ix86_frame.
26923         (ix86_expand_epilogue): Likewise.  Add a local variable for
26924         the reg_save_offset field in struct ix86_frame.
26926 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
26928         PR tree-optimization/82604
26929         * tree-loop-distribution.c (enum partition_kind): New enum item
26930         PKIND_PARTIAL_MEMSET.
26931         (partition_builtin_p): Support above new enum item.
26932         (generate_code_for_partition): Ditto.
26933         (compute_access_range): Differentiate cases that equality can be
26934         proven at all loops, the innermost loops or no loops.
26935         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
26936         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
26937         (finalize_partitions, distribute_loop): Don't fuse partition of
26938         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
26939         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
26940         parloop is enabled.
26942 2018-01-23  Martin Liska  <mliska@suse.cz>
26944         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
26945         order to ignore the predictor.
26946         (PRED_POLYMORPHIC_CALL): Likewise.
26947         (PRED_RECURSIVE_CALL): Likewise.
26949 2018-01-23  Martin Liska  <mliska@suse.cz>
26951         * tree-profile.c (tree_profiling): Print function header to
26952         aware reader which function we are working on.
26953         * value-prof.c (gimple_find_values_to_profile): Do not print
26954         not interesting value histograms.
26956 2018-01-23  Martin Liska  <mliska@suse.cz>
26958         * profile-count.h (enum profile_quality): Add
26959         profile_uninitialized as the first value. Do not number values
26960         as they are zero based.
26961         (profile_count::verify): Update sanity check.
26962         (profile_probability::verify): Likewise.
26964 2018-01-23  Nathan Sidwell  <nathan@acm.org>
26966         * doc/invoke.texi (ffor-scope): Deprecate.
26968 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26970         PR tree-optimization/83510
26971         * domwalk.c (set_all_edges_as_executable): New function.
26972         (dom_walker::dom_walker): Convert bool param
26973         "skip_unreachable_blocks" to enum reachability.  Move setup of
26974         edge flags to set_all_edges_as_executable and only do it when
26975         reachability is REACHABLE_BLOCKS.
26976         * domwalk.h (enum dom_walker::reachability): New enum.
26977         (dom_walker::dom_walker): Convert bool param
26978         "skip_unreachable_blocks" to enum reachability.
26979         (set_all_edges_as_executable): New decl.
26980         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
26981         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
26982         "reachability".
26983         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
26984         but converting true to REACHABLE_BLOCKS.
26985         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
26986         * tree-vrp.c
26987         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
26988         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
26989         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
26990         REACHABLE_BLOCKS.
26991         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
26992         if check_all_array_refs will be called.
26994 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26996         * tree.c (selftest::test_location_wrappers): Add more test
26997         coverage.
26999 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27001         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27002         (selftest::test_bit_in_range): Likewise.
27004 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27006         PR testsuite/83888
27007         * doc/sourcebuild.texi (vect_float): Say that the selector
27008         only describes the situation when -funsafe-math-optimizations is on.
27009         (vect_float_strict): Document.
27011 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27013         PR tree-optimization/83965
27014         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27015         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27016         instead of checking only for a reduction.
27017         (vect_recog_widen_sum_pattern): Likewise.
27019 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27021         * predict.c (probably_never_executed): Only use precise profile info.
27022         (compute_function_frequency): Skip after inlining hack since we now
27023         have quality checking.
27025 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27027         * profile-count.h (profile_probability::very_unlikely,
27028         profile_probability::unlikely, profile_probability::even): Set
27029         precision to guessed.
27031 2018-01-23  Richard Biener  <rguenther@suse.de>
27033         PR tree-optimization/83963
27034         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27035         Properly terminate dominator walk when crossing the exit edge not
27036         when visiting its source block.
27038 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
27040         PR c++/83918
27041         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27042         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27044 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
27046         PR tree-optimization/83957
27047         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
27048         semicolon after for body surrounded by braces.
27050         PR tree-optimization/83081
27051         * profile-count.h (profile_probability::split): New method.
27052         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27053         Use profile_probability::split.
27054         (do_compare_rtx_and_jump): Fix adjustment of probabilities
27055         when splitting a single conditional jump into 2.
27057 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
27059         PR tree-optimization/69452
27060         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27061         decl.
27063 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27065         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27066         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27067         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27069 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27071         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27072         declaration.
27073         * config/rl78/rl78.md (movdi): New define_expand.
27074         * config/rl78/rl78.c (rl78_split_movdi): New function.
27076 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27078         PR target/83862
27079         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27080         no longer used.
27081         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27082         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27083         128-bit to produce an UNSPEC move to get the double word with the
27084         signbit and then a shift directly to do signbit.
27085         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27086         implementation with a new version that just does either a direct
27087         move or a regular move.  Move memory interface to separate insns.
27088         Move insns so they are next to the expander.
27089         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27090         with signbit move.  Split big and little endian case.
27091         (signbit<mode>2_dm_mem_le): Likewise.
27092         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27093         (signbit<mode>2_dm2): Likewise.
27095 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27097         * config/rl78/rl78.md (anddi3): New define_expand.
27099 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27101         * config/rl78/rl78.md (umindi3): New define_expand.
27103 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27105         * config/rl78/rl78.md (smindi3): New define_expand.
27107 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27109         * config/rl78/rl78.md (smaxdi3): New define_expand.
27111 2018-01-22  Carl Love  <cel@us.ibm.com>
27113         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27114         LVX_V1TI): Add macro expansion.
27115         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27116         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27117         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27118         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27119         Change check to determine if the instruction is a byte reversing
27120         entry.  Fix typo in comment.
27121         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27122         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27123         Add def_builtin calls for new builtins.
27124         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27125         Add define_insn expansion.
27127 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27129         * config/rl78/rl78.md (umaxdi3): New define_expand.
27131 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27133         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27134         for non-QImode registers.
27136 2018-01-22  Richard Biener  <rguenther@suse.de>
27138         PR tree-optimization/83963
27139         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27140         including the loop exit block.
27141         (scop_detection::merge_sese): Likewise.
27142         (scop_detection::add_scop): Do it here instead.
27144 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27146         * doc/sourcebuild.texi (arm_softfloat): Document.
27148 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27150         PR gcc/77734
27151         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27152         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27153         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27155 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27156             David Edelsohn  <dje.gcc@gmail.com>
27158         PR target/83946
27159         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27160         Change "crset eq" to "crset 2".
27161         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27162         (*call_indirect_aix<mode>_nospec): Likewise.
27163         (*call_value_indirect_aix<mode>_nospec): Likewise.
27164         (*call_indirect_elfv2<mode>_nospec): Likewise.
27165         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27166         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27167         change assembly output from . to $.
27168         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27169         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27170         (*tablejump<mode>_internal1_nospec): Likewise.
27172 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27174         PR target/80870
27175         * config/sh/sh_optimize_sett_clrt.cc:
27176         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27178 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27180         PR tree-optimization/83940
27181         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27182         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27183         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27184         use it to pass back the definition type.
27185         (vect_check_store_rhs): Likewise rhs_dt_out.
27186         (vect_build_gather_load_calls): Add a mask_dt argument and use
27187         it instead of a call to vect_is_simple_use.
27188         (vectorizable_store): Update calls to vect_check_load_store_mask
27189         and vect_check_store_rhs.  Use the dt returned by the latter instead
27190         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27191         instead of calls to vect_is_simple_use.  Pass the scalar rather
27192         than the vector operand to vect_is_simple_use when handling
27193         second and subsequent copies of an rhs value.
27194         (vectorizable_load): Update calls to vect_check_load_store_mask
27195         and vect_build_gather_load_calls.  Use the cached mask_dt and
27196         gs_info.offset_dt instead of calls to vect_is_simple_use.
27198 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27200         PR middle-end/83945
27201         * tree-emutls.c: Include gimplify.h.
27202         (lower_emutls_2): New function.
27203         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27204         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27205         it before further processing.
27207         PR target/83930
27208         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27209         UINTVAL (trueop1) instead of INTVAL (op1).
27211 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27213         PR debug/81570
27214         PR debug/83728
27215         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27216         INCOMING_FRAME_SP_OFFSET if not defined.
27217         (scan_trace): Add ENTRY argument.  If true and
27218         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27219         emit a note to adjust the CFA offset.
27220         (create_cfi_notes): Adjust scan_trace callers.
27221         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27222         INCOMING_FRAME_SP_OFFSET in the CIE.
27223         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27224         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27225         Likewise.
27226         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27227         * doc/tm.texi: Regenerated.
27229 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27231         PR rtl-optimization/83147
27232         * lra-constraints.c (remove_inheritance_pseudos): Use
27233         lra_substitute_pseudo_within_insn.
27235 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27236             Cesar Philippidis  <cesar@codesourcery.com>
27238         PR target/83920
27239         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27241 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27243         PR target/83790
27244         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27245         spaces for function labels.
27247 2018-01-19  Martin Liska  <mliska@suse.cz>
27249         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27250         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27251         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27252         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27253         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27254         (PRED_CONST_RETURN): Change from 69 to 65.
27255         (PRED_NULL_RETURN): Change from 91 to 71.
27256         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27257         (PRED_LOOP_GUARD): Change from 66 to 73.
27259 2018-01-19  Martin Liska  <mliska@suse.cz>
27261         * predict.c (predict_insn_def): Add new assert.
27262         (struct branch_predictor): Change type to signed integer.
27263         (test_prediction_value_range): Amend test to cover
27264         PROB_UNINITIALIZED.
27265         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27266         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27267         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27268         (PRED_LOOP_IV_COMPARE): Likewise.
27269         * predict.h (PROB_UNINITIALIZED): Define new constant.
27271 2018-01-19  Martin Liska  <mliska@suse.cz>
27273         * predict.c (dump_prediction): Add new format for
27274         analyze_brprob.py script which is enabled with -details
27275         suboption.
27276         * profile-count.h (precise_p): New function.
27278 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27280         PR tree-optimization/83922
27281         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27282         there are no statements that need masking.
27283         (vect_active_double_reduction_p): New function.
27284         (vect_analyze_loop_operations): Use it when handling phis that
27285         are not in the loop header.
27287 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27289         PR tree-optimization/83914
27290         * tree-vect-loop.c (vectorizable_induction): Don't convert
27291         init_expr or apply the peeling adjustment for inductions
27292         that are nested within the vectorized loop.
27294 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27296         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27297         instead of NEG.
27299 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27301         PR sanitizer/81715
27302         PR testsuite/83882
27303         * function.h (gimplify_parameters): Add gimple_seq * argument.
27304         * function.c: Include gimple.h and options.h.
27305         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27306         for the added local temporaries if needed.
27307         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27308         if there are any parameter cleanups, wrap whole body into a
27309         try/finally with the cleanups.
27311 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27313         PR target/82964
27314         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27315         Use GET_MODE_CLASS for scalar floating point.
27317 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27319         PR ipa/82256
27320         patch by PaX Team
27321         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27322         Fix call of call_cgraph_insertion_hooks.
27324 2018-01-18  Martin Sebor  <msebor@redhat.com>
27326         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27328 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27330         PR ipa/83619
27331         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27332         frequencies.
27334 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27336         PR other/70268
27337         * common.opt: (-ffile-prefix-map): New option.
27338         * opts.c (common_handle_option): Defer it.
27339         * opts-global.c (handle_common_deferred_options): Handle it.
27340         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27341         * file-prefix-map.h: New file.
27342         (remap_debug_filename, add_debug_prefix_map): ...here.
27343         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27344         * final.c (debug_prefix_map, add_debug_prefix_map
27345         remap_debug_filename): Move to...
27346         * file-prefix-map.c: New file.
27347         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27348         generalize, get rid of alloca(), use strrchr() instead of strchr().
27349         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27350         Implement in terms of add_prefix_map().
27351         (remap_macro_filename, remap_debug_filename): Implement in term of
27352         remap_filename().
27353         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27354         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27355         * dbxout.c: Include file-prefix-map.h.
27356         * varasm.c: Likewise.
27357         * vmsdbgout.c: Likewise.
27358         * xcoffout.c: Likewise.
27359         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27360         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27361         * doc/invoke.texi (-ffile-prefix-map): Document.
27362         (-fdebug-prefix-map): Update description.
27364 2018-01-18  Martin Liska  <mliska@suse.cz>
27366         * config/i386/i386.c (indirect_thunk_name): Document that also
27367         lfence is emitted.
27368         (output_indirect_thunk): Document why both instructions
27369         (pause and lfence) are generated.
27371 2018-01-18  Richard Biener  <rguenther@suse.de>
27373         PR tree-optimization/83887
27374         * graphite-scop-detection.c
27375         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27376         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27377         (scop_detection::merge_sese): Re-implement with a flood-fill
27378         algorithm that properly finds a SESE region if it exists.
27380 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27382         PR c/61240
27383         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27384         pointer_diff optimizations use view_convert instead of convert.
27386 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27388         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27389         Generate different code for -mno-speculate-indirect-jumps.
27390         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27391         (*call_indirect_aix<mode>): Disable for
27392         -mno-speculate-indirect-jumps.
27393         (*call_indirect_aix<mode>_nospec): New define_insn.
27394         (*call_value_indirect_aix<mode>): Disable for
27395         -mno-speculate-indirect-jumps.
27396         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27397         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27398         -mno-speculate-indirect-jumps.
27399         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27401 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27403         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27404         long double type, set the flags for noting the default long double
27405         type, even if we don't pass or return a long double type.
27407 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27409         PR ipa/83051
27410         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27411         failure.
27413 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27415         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27416         support for merge[hl].
27417         (fold_mergehl_helper): New helper function.
27418         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27419         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27420         (altivec_vmrglw_direct): Add xxmrglw insn.
27422 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27424         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27425         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27427 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27429         PR lto/83121
27430         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27431         call the lto_location_cache before reading the
27432         DECL_SOURCE_LOCATION of the types.
27434 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27435             Richard Sandiford  <richard.sandiford@linaro.org>
27437         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27438         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27439         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27440         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27441         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27442         Add declaration.
27443         * config/aarch64/constraints.md (aarch64_movti_operand):
27444         Limit immediates.
27445         * config/aarch64/predicates.md (Uti): Add new constraint.
27447 2018-01-17  Carl Love  <cel@us.ibm.com>
27449         * config/rs6000/vsx.md (define_expand xl_len_r,
27450         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27451         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27452         lxvll.
27453         (define_expand, define_insn): Move the shift left from  the
27454         define_insn to the define_expand for lxvl and stxvl instructions.
27455         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27456         and XL_LEN_R definitions to PURE.
27458 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27460         * config/i386/i386.c (indirect_thunk_name): Declare regno
27461         as unsigned int.  Compare regno with INVALID_REGNUM.
27462         (output_indirect_thunk): Ditto.
27463         (output_indirect_thunk_function): Ditto.
27464         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27465         in the call to output_indirect_thunk_function.
27467 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27469         PR middle-end/83884
27470         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27471         rather than the size of inner_type to determine the stack slot size
27472         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27474 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27476         PR target/83546
27477         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27478         to PTA_SILVERMONT.
27480 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27482         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27483         endian Linux systems to optionally enable multilibs for selecting
27484         the long double type if the user configured an explicit type.
27485         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27486         have no long double multilibs if not defined.
27487         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27488         warn if the user used -mabi={ieee,ibm}longdouble and we built
27489         multilibs for long double.
27490         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27491         appropriate multilib option.
27492         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27493         multilib options.
27494         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27495         for building long double multilibs.
27496         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27498 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27500         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27501         copies.
27503         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27504         64 bits.
27505         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27506         128 bits.
27508         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27509         variables.
27511         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27512         return value.
27514 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27516         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27517         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27519 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27521         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27522         different rtl trees depending on TARGET_64BIT.
27523         (rs6000_gen_lvx): Likewise.
27525 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27527         * config/visium/visium.md (nop): Tweak comment.
27528         (hazard_nop): Likewise.
27530 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27532         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27533         -mspeculate-indirect-jumps.
27534         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27535         for -mno-speculate-indirect-jumps.
27536         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27537         (*call_value_indirect_elfv2<mode>): Disable for
27538         -mno-speculate-indirect-jumps.
27539         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27540         (indirect_jump): Emit different RTL for
27541         -mno-speculate-indirect-jumps.
27542         (*indirect_jump<mode>): Disable for
27543         -mno-speculate-indirect-jumps.
27544         (*indirect_jump<mode>_nospec): New define_insn.
27545         (tablejump): Emit different RTL for
27546         -mno-speculate-indirect-jumps.
27547         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27548         (tablejumpsi_nospec): New define_expand.
27549         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27550         (tablejumpdi_nospec): New define_expand.
27551         (*tablejump<mode>_internal1): Disable for
27552         -mno-speculate-indirect-jumps.
27553         (*tablejump<mode>_internal1_nospec): New define_insn.
27554         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27555         option.
27557 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27559         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27560         callers updated.
27562 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27563             Richard Biener  <rguenth@suse.de>
27565         PR libgomp/83590
27566         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27567         return early, inline manually is_gimple_sizepos.  Make sure if we
27568         call gimplify_expr we don't end up with a gimple constant.
27569         * tree.c (variably_modified_type_p): Don't return true for
27570         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27571         * gimplify.h (is_gimple_sizepos): Remove.
27573 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27575         PR tree-optimization/83857
27576         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27577         vectorizable_live_operation for pure SLP statements.
27578         (vectorizable_live_operation): Handle PHIs.
27580 2018-01-16  Richard Biener  <rguenther@suse.de>
27582         PR tree-optimization/83867
27583         * tree-vect-stmts.c (vect_transform_stmt): Precompute
27584         nested_in_vect_loop_p since the scalar stmt may get invalidated.
27586 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27588         PR c/83844
27589         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
27590         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
27591         If off is not INTEGER_CST, issue a may not be aligned warning
27592         rather than isn't aligned.  Use isn%'t rather than isn't.
27593         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
27594         into MULT_EXPR.
27595         <case MULT_EXPR>: Improve the case when bottom and one of the
27596         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
27597         operand, in that case check if the other operand is multiple of
27598         bottom divided by the INTEGER_CST operand.
27600 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27602         PR target/83858
27603         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
27604         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
27605         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
27606         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
27607         * config/pa/pa.c (pa_function_arg_advance): Likewise.
27608         (pa_function_arg, pa_arg_partial_bytes): Likewise.
27609         (pa_function_arg_size): New function.
27611 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27613         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
27614         in a separate statement.
27616 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27618         PR tree-optimization/83847
27619         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
27620         group gathers and scatters.
27622 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27624         PR rtl-optimization/86620
27625         * params.def (max-sched-ready-insns): Bump minimum value to 1.
27627         PR rtl-optimization/83213
27628         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
27629         to last if both are JUMP_INSNs.
27631         PR tree-optimization/83843
27632         * gimple-ssa-store-merging.c
27633         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
27634         store_immediate_info for bswap/nop orig_stores.
27636 2018-01-15  Andrew Waterman  <andrew@sifive.com>
27638         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
27639         !TARGET_MUL.
27640         <UDIV>: Increase cost if !TARGET_DIV.
27642 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
27644         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
27645         (define_attr "cr_logical_3op"): New.
27646         (cceq_ior_compare): Adjust.
27647         (cceq_ior_compare_complement): Adjust.
27648         (*cceq_rev_compare): Adjust.
27649         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
27650         (is_cracked_insn): Adjust.
27651         (insn_must_be_first_in_group): Adjust.
27652         * config/rs6000/40x.md: Adjust.
27653         * config/rs6000/440.md: Adjust.
27654         * config/rs6000/476.md: Adjust.
27655         * config/rs6000/601.md: Adjust.
27656         * config/rs6000/603.md: Adjust.
27657         * config/rs6000/6xx.md: Adjust.
27658         * config/rs6000/7450.md: Adjust.
27659         * config/rs6000/7xx.md: Adjust.
27660         * config/rs6000/8540.md: Adjust.
27661         * config/rs6000/cell.md: Adjust.
27662         * config/rs6000/e300c2c3.md: Adjust.
27663         * config/rs6000/e500mc.md: Adjust.
27664         * config/rs6000/e500mc64.md: Adjust.
27665         * config/rs6000/e5500.md: Adjust.
27666         * config/rs6000/e6500.md: Adjust.
27667         * config/rs6000/mpc.md: Adjust.
27668         * config/rs6000/power4.md: Adjust.
27669         * config/rs6000/power5.md: Adjust.
27670         * config/rs6000/power6.md: Adjust.
27671         * config/rs6000/power7.md: Adjust.
27672         * config/rs6000/power8.md: Adjust.
27673         * config/rs6000/power9.md: Adjust.
27674         * config/rs6000/rs64.md: Adjust.
27675         * config/rs6000/titan.md: Adjust.
27677 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27679         * config/i386/predicates.md (indirect_branch_operand): Rewrite
27680         ix86_indirect_branch_register logic.
27682 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27684         * config/i386/constraints.md (Bs): Update
27685         ix86_indirect_branch_register check.  Don't check
27686         ix86_indirect_branch_register with GOT_memory_operand.
27687         (Bw): Likewise.
27688         * config/i386/predicates.md (GOT_memory_operand): Don't check
27689         ix86_indirect_branch_register here.
27690         (GOT32_symbol_operand): Likewise.
27692 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27694         * config/i386/predicates.md (constant_call_address_operand):
27695         Rewrite ix86_indirect_branch_register logic.
27696         (sibcall_insn_operand): Likewise.
27698 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27700         * config/i386/constraints.md (Bs): Replace
27701         ix86_indirect_branch_thunk_register with
27702         ix86_indirect_branch_register.
27703         (Bw): Likewise.
27704         * config/i386/i386.md (indirect_jump): Likewise.
27705         (tablejump): Likewise.
27706         (*sibcall_memory): Likewise.
27707         (*sibcall_value_memory): Likewise.
27708         Peepholes of indirect call and jump via memory: Likewise.
27709         * config/i386/i386.opt: Likewise.
27710         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27711         (GOT_memory_operand): Likewise.
27712         (call_insn_operand): Likewise.
27713         (sibcall_insn_operand): Likewise.
27714         (GOT32_symbol_operand): Likewise.
27716 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27718         PR middle-end/83837
27719         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
27720         type rather than type addr's type points to.
27721         (expand_omp_atomic_mutex): Likewise.
27722         (expand_omp_atomic): Likewise.
27724 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27726         PR target/83839
27727         * config/i386/i386.c (output_indirect_thunk_function): Use
27728         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
27729         for  __x86_return_thunk.
27731 2018-01-15  Richard Biener  <rguenther@suse.de>
27733         PR middle-end/83850
27734         * expmed.c (extract_bit_field_1): Fix typo.
27736 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27738         PR target/83687
27739         * config/arm/iterators.md (VF): New mode iterator.
27740         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
27741         Remove integer-related logic from pattern.
27742         (neon_vabd<mode>_3): Likewise.
27744 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27746         PR middle-end/82694
27747         * common.opt (fstrict-overflow): No longer an alias.
27748         (fwrapv-pointer): New option.
27749         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
27750         also for pointer types based on flag_wrapv_pointer.
27751         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
27752         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
27753         opts->x_flag_wrapv got set.
27754         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
27755         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
27756         POINTER_TYPE_OVERFLOW_UNDEFINED.
27757         * match.pd: Likewise in address comparison pattern.
27758         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
27760 2018-01-15  Richard Biener  <rguenther@suse.de>
27762         PR lto/83804
27763         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
27764         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
27765         Reset type names to their identifier if their TYPE_DECL doesn't
27766         have linkage (and thus is used for ODR and devirt).
27767         (save_debug_info_for_decl): Remove.
27768         (save_debug_info_for_type): Likewise.
27769         (add_tree_to_fld_list): Adjust.
27770         * tree-pretty-print.c (dump_generic_node): Make dumping of
27771         type names more robust.
27773 2018-01-15  Richard Biener  <rguenther@suse.de>
27775         * BASE-VER: Bump to 8.0.1.
27777 2018-01-14  Martin Sebor  <msebor@redhat.com>
27779         PR other/83508
27780         * builtins.c (check_access): Avoid warning when the no-warning bit
27781         is set.
27783 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
27785         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
27786         * ira-color (allocno_hard_regs_compare): Likewise.
27788 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
27790         PR target/83013
27791         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
27792         Use .pushsection/.popsection.
27794 2018-01-14  Martin Sebor  <msebor@redhat.com>
27796         PR c++/81327
27797         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
27799 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
27801         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
27802         entry from extra_headers.
27803         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
27804         extra_headers, make the list bitwise identical to the i?86-*-* one.
27806 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27808         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
27809         -mcmodel=large with -mindirect-branch=thunk,
27810         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
27811         -mfunction-return=thunk-extern.
27812         * doc/invoke.texi: Document -mcmodel=large is incompatible with
27813         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
27814         -mfunction-return=thunk and -mfunction-return=thunk-extern.
27816 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27818         * config/i386/i386.c (print_reg): Print the name of the full
27819         integer register without '%'.
27820         (ix86_print_operand): Handle 'V'.
27821         * doc/extend.texi: Document 'V' modifier.
27823 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27825         * config/i386/constraints.md (Bs): Disallow memory operand for
27826         -mindirect-branch-register.
27827         (Bw): Likewise.
27828         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27829         (GOT_memory_operand): Likewise.
27830         (call_insn_operand): Likewise.
27831         (sibcall_insn_operand): Likewise.
27832         (GOT32_symbol_operand): Likewise.
27833         * config/i386/i386.md (indirect_jump): Call convert_memory_address
27834         for -mindirect-branch-register.
27835         (tablejump): Likewise.
27836         (*sibcall_memory): Likewise.
27837         (*sibcall_value_memory): Likewise.
27838         Disallow peepholes of indirect call and jump via memory for
27839         -mindirect-branch-register.
27840         (*call_pop): Replace m with Bw.
27841         (*call_value_pop): Likewise.
27842         (*sibcall_pop_memory): Replace m with Bs.
27843         * config/i386/i386.opt (mindirect-branch-register): New option.
27844         * doc/invoke.texi: Document -mindirect-branch-register option.
27846 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27848         * config/i386/i386-protos.h (ix86_output_function_return): New.
27849         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
27850         set function_return_type.
27851         (indirect_thunk_name): Add ret_p to indicate thunk for function
27852         return.
27853         (output_indirect_thunk_function): Pass false to
27854         indirect_thunk_name.
27855         (ix86_output_indirect_branch_via_reg): Likewise.
27856         (ix86_output_indirect_branch_via_push): Likewise.
27857         (output_indirect_thunk_function): Create alias for function
27858         return thunk if regno < 0.
27859         (ix86_output_function_return): New function.
27860         (ix86_handle_fndecl_attribute): Handle function_return.
27861         (ix86_attribute_table): Add function_return.
27862         * config/i386/i386.h (machine_function): Add
27863         function_return_type.
27864         * config/i386/i386.md (simple_return_internal): Use
27865         ix86_output_function_return.
27866         (simple_return_internal_long): Likewise.
27867         * config/i386/i386.opt (mfunction-return=): New option.
27868         (indirect_branch): Mention -mfunction-return=.
27869         * doc/extend.texi: Document function_return function attribute.
27870         * doc/invoke.texi: Document -mfunction-return= option.
27872 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27874         * config/i386/i386-opts.h (indirect_branch): New.
27875         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
27876         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
27877         with local indirect jump when converting indirect call and jump.
27878         (ix86_set_indirect_branch_type): New.
27879         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
27880         (indirectlabelno): New.
27881         (indirect_thunk_needed): Likewise.
27882         (indirect_thunk_bnd_needed): Likewise.
27883         (indirect_thunks_used): Likewise.
27884         (indirect_thunks_bnd_used): Likewise.
27885         (INDIRECT_LABEL): Likewise.
27886         (indirect_thunk_name): Likewise.
27887         (output_indirect_thunk): Likewise.
27888         (output_indirect_thunk_function): Likewise.
27889         (ix86_output_indirect_branch_via_reg): Likewise.
27890         (ix86_output_indirect_branch_via_push): Likewise.
27891         (ix86_output_indirect_branch): Likewise.
27892         (ix86_output_indirect_jmp): Likewise.
27893         (ix86_code_end): Call output_indirect_thunk_function if needed.
27894         (ix86_output_call_insn): Call ix86_output_indirect_branch if
27895         needed.
27896         (ix86_handle_fndecl_attribute): Handle indirect_branch.
27897         (ix86_attribute_table): Add indirect_branch.
27898         * config/i386/i386.h (machine_function): Add indirect_branch_type
27899         and has_local_indirect_jump.
27900         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
27901         to true.
27902         (tablejump): Likewise.
27903         (*indirect_jump): Use ix86_output_indirect_jmp.
27904         (*tablejump_1): Likewise.
27905         (simple_return_indirect_internal): Likewise.
27906         * config/i386/i386.opt (mindirect-branch=): New option.
27907         (indirect_branch): New.
27908         (keep): Likewise.
27909         (thunk): Likewise.
27910         (thunk-inline): Likewise.
27911         (thunk-extern): Likewise.
27912         * doc/extend.texi: Document indirect_branch function attribute.
27913         * doc/invoke.texi: Document -mindirect-branch= option.
27915 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
27917         PR ipa/83051
27918         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
27920 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
27922         * ipa-inline.c (want_inline_small_function_p): Return false if
27923         inlining has already failed with CIF_FINAL_ERROR.
27924         (update_caller_keys): Call want_inline_small_function_p before
27925         can_inline_edge_p.
27926         (update_callee_keys): Likewise.
27928 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27930         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
27931         New function.
27932         (rs6000_quadword_masked_address_p): Likewise.
27933         (quad_aligned_load_p): Likewise.
27934         (quad_aligned_store_p): Likewise.
27935         (const_load_sequence_p): Add comment to describe the outer-most loop.
27936         (mimic_memory_attributes_and_flags): New function.
27937         (rs6000_gen_stvx): Likewise.
27938         (replace_swapped_aligned_store): Likewise.
27939         (rs6000_gen_lvx): Likewise.
27940         (replace_swapped_aligned_load): Likewise.
27941         (replace_swapped_load_constant): Capitalize argument name in
27942         comment describing this function.
27943         (rs6000_analyze_swaps): Add a third pass to search for vector loads
27944         and stores that access quad-word aligned addresses and replace
27945         with stvx or lvx instructions when appropriate.
27946         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
27947         New function prototype.
27948         (rs6000_quadword_masked_address_p): Likewise.
27949         (rs6000_gen_lvx): Likewise.
27950         (rs6000_gen_stvx): Likewise.
27951         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
27952         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
27953         when memory address is aligned.
27954         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
27955         this split to select lvx instruction when memory address is aligned.
27956         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
27957         instruction when memory address is aligned.
27958         (*vsx_le_perm_load_v16qi): Likewise.
27959         (four unnamed splitters): Modify to select the stvx instruction
27960         when memory is aligned.
27962 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
27964         * predict.c (determine_unlikely_bbs): Handle correctly BBs
27965         which appears in the queue multiple times.
27967 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27968             Alan Hayward  <alan.hayward@arm.com>
27969             David Sherwood  <david.sherwood@arm.com>
27971         * tree-vectorizer.h (vec_lower_bound): New structure.
27972         (_loop_vec_info): Add check_nonzero and lower_bounds.
27973         (LOOP_VINFO_CHECK_NONZERO): New macro.
27974         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
27975         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
27976         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
27977         fields.  Make seg_len the distance travelled, not including the
27978         access size.
27979         (dr_direction_indicator): Declare.
27980         (dr_zero_step_indicator): Likewise.
27981         (dr_known_forward_stride_p): Likewise.
27982         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
27983         tree-ssanames.h.
27984         (runtime_alias_check_p): Allow runtime alias checks with
27985         variable strides.
27986         (operator ==): Compare access_size and align.
27987         (prune_runtime_alias_test_list): Rework for new distinction between
27988         the access_size and seg_len.
27989         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
27990         segment lengths.
27991         (get_segment_min_max): New function.
27992         (create_intersect_range_checks): Use it.
27993         (dr_step_indicator): New function.
27994         (dr_direction_indicator): Likewise.
27995         (dr_zero_step_indicator): Likewise.
27996         (dr_known_forward_stride_p): Likewise.
27997         * tree-loop-distribution.c (data_ref_segment_size): Return
27998         DR_STEP * (niters - 1).
27999         (compute_alias_check_pairs): Update call to the dr_with_seg_len
28000         constructor.
28001         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28002         (vect_preserves_scalar_order_p): New function, split out from...
28003         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
28004         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28005         (vect_vfa_access_size): New function.
28006         (vect_vfa_align): Likewise.
28007         (vect_compile_time_alias): Take access_size_a and access_b arguments.
28008         (dump_lower_bound): New function.
28009         (vect_check_lower_bound): Likewise.
28010         (vect_small_gap_p): Likewise.
28011         (vectorizable_with_step_bound_p): Likewise.
28012         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28013         depencies if the vectorization factor is 1.  Convert the checks
28014         for nonzero steps into checks on the bounds of DR_STEP.  Try using
28015         a bunds check for variable steps if the minimum required step is
28016         relatively small. Update calls to the dr_with_seg_len
28017         constructor and to vect_compile_time_alias.
28018         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28019         function.
28020         (vect_loop_versioning): Call it.
28021         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28022         when retrying.
28023         (vect_estimate_min_profitable_iters): Account for any bounds checks.
28025 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28026             Alan Hayward  <alan.hayward@arm.com>
28027             David Sherwood  <david.sherwood@arm.com>
28029         * doc/sourcebuild.texi (vect_scatter_store): Document.
28030         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28031         optabs.
28032         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28033         Document.
28034         * genopinit.c (main): Add supports_vec_scatter_store and
28035         supports_vec_scatter_store_cached to target_optabs.
28036         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28037         IFN_MASK_SCATTER_STORE.
28038         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28039         functions.
28040         * internal-fn.h (internal_store_fn_p): Declare.
28041         (internal_fn_stored_value_index): Likewise.
28042         * internal-fn.c (scatter_store_direct): New macro.
28043         (expand_scatter_store_optab_fn): New function.
28044         (direct_scatter_store_optab_supported_p): New macro.
28045         (internal_store_fn_p): New function.
28046         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28047         IFN_MASK_SCATTER_STORE.
28048         (internal_fn_mask_index): Likewise.
28049         (internal_fn_stored_value_index): New function.
28050         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28051         for scatter stores.
28052         * optabs-query.h (supports_vec_scatter_store_p): Declare.
28053         * optabs-query.c (supports_vec_scatter_store_p): New function.
28054         * tree-vectorizer.h (vect_get_store_rhs): Declare.
28055         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28056         true for scatter stores.
28057         (vect_gather_scatter_fn_p): Handle scatter stores too.
28058         (vect_check_gather_scatter): Consider using scatter stores if
28059         supports_vec_scatter_store_p.
28060         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28061         scatter stores too.
28062         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28063         internal_fn_stored_value_index.
28064         (check_load_store_masking): Handle scatter stores too.
28065         (vect_get_store_rhs): Make public.
28066         (vectorizable_call): Use internal_store_fn_p.
28067         (vectorizable_store): Handle scatter store internal functions.
28068         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28069         when deciding whether the end of the group has been reached.
28070         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28071         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28072         (mask_scatter_store<mode>): New insns.
28074 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28075             Alan Hayward  <alan.hayward@arm.com>
28076             David Sherwood  <david.sherwood@arm.com>
28078         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28079         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28080         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28081         function.
28082         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28083         Use vect_truncate_gather_scatter_offset if we can't treat the
28084         operation as a normal gather load or scatter store.
28085         (get_group_load_store_type): Take the gather_scatter_info
28086         as argument.  Try using a gather load or scatter store for
28087         single-element groups.
28088         (get_load_store_type): Update calls to get_group_load_store_type
28089         and vect_use_strided_gather_scatters_p.
28091 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28092             Alan Hayward  <alan.hayward@arm.com>
28093             David Sherwood  <david.sherwood@arm.com>
28095         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28096         optional tree argument.
28097         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28098         null target hooks.
28099         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28100         but continue to use the current value as a fallback.
28101         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28102         to compare the updates.
28103         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28104         (get_load_store_type): Use it when handling a strided access.
28105         (vect_get_strided_load_store_ops): New function.
28106         (vect_get_data_ptr_increment): Likewise.
28107         (vectorizable_load): Handle strided gather loads.  Always pass
28108         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28110 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28111             Alan Hayward  <alan.hayward@arm.com>
28112             David Sherwood  <david.sherwood@arm.com>
28114         * doc/md.texi (gather_load@var{m}): Document.
28115         (mask_gather_load@var{m}): Likewise.
28116         * genopinit.c (main): Add supports_vec_gather_load and
28117         supports_vec_gather_load_cached to target_optabs.
28118         * optabs-tree.c (init_tree_optimization_optabs): Use
28119         ggc_cleared_alloc to allocate target_optabs.
28120         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28121         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28122         functions.
28123         * internal-fn.h (internal_load_fn_p): Declare.
28124         (internal_gather_scatter_fn_p): Likewise.
28125         (internal_fn_mask_index): Likewise.
28126         (internal_gather_scatter_fn_supported_p): Likewise.
28127         * internal-fn.c (gather_load_direct): New macro.
28128         (expand_gather_load_optab_fn): New function.
28129         (direct_gather_load_optab_supported_p): New macro.
28130         (direct_internal_fn_optab): New function.
28131         (internal_load_fn_p): Likewise.
28132         (internal_gather_scatter_fn_p): Likewise.
28133         (internal_fn_mask_index): Likewise.
28134         (internal_gather_scatter_fn_supported_p): Likewise.
28135         * optabs-query.c (supports_at_least_one_mode_p): New function.
28136         (supports_vec_gather_load_p): Likewise.
28137         * optabs-query.h (supports_vec_gather_load_p): Declare.
28138         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28139         and memory_type field.
28140         (NUM_PATTERNS): Bump to 15.
28141         * tree-vect-data-refs.c: Include internal-fn.h.
28142         (vect_gather_scatter_fn_p): New function.
28143         (vect_describe_gather_scatter_call): Likewise.
28144         (vect_check_gather_scatter): Try using internal functions for
28145         gather loads.  Recognize existing calls to a gather load function.
28146         (vect_analyze_data_refs): Consider using gather loads if
28147         supports_vec_gather_load_p.
28148         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28149         (vect_get_gather_scatter_offset_type): Likewise.
28150         (vect_convert_mask_for_vectype): Likewise.
28151         (vect_add_conversion_to_patterm): Likewise.
28152         (vect_try_gather_scatter_pattern): Likewise.
28153         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28154         (vect_vect_recog_func_ptrs): Add it.
28155         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28156         internal_fn_mask_index and internal_gather_scatter_fn_p.
28157         (check_load_store_masking): Take the gather_scatter_info as an
28158         argument and handle gather loads.
28159         (vect_get_gather_scatter_ops): New function.
28160         (vectorizable_call): Check internal_load_fn_p.
28161         (vectorizable_load): Likewise.  Handle gather load internal
28162         functions.
28163         (vectorizable_store): Update call to check_load_store_masking.
28164         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28165         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28166         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28167         (aarch64_gather_scale_operand_d): New predicates.
28168         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28169         (mask_gather_load<mode>): New insns.
28171 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28172             Alan Hayward  <alan.hayward@arm.com>
28173             David Sherwood  <david.sherwood@arm.com>
28175         * optabs.def (fold_left_plus_optab): New optab.
28176         * doc/md.texi (fold_left_plus_@var{m}): Document.
28177         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28178         * internal-fn.c (fold_left_direct): Define.
28179         (expand_fold_left_optab_fn): Likewise.
28180         (direct_fold_left_optab_supported_p): Likewise.
28181         * fold-const-call.c (fold_const_fold_left): New function.
28182         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28183         * tree-parloops.c (valid_reduction_p): New function.
28184         (gather_scalar_reductions): Use it.
28185         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28186         (vect_finish_replace_stmt): Declare.
28187         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28188         (needs_fold_left_reduction_p): New function, split out from...
28189         (vect_is_simple_reduction): ...here.  Accept reductions that
28190         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28191         (vect_force_simple_reduction): Also store the reduction type in
28192         the assignment's STMT_VINFO_REDUC_TYPE.
28193         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28194         (merge_with_identity): New function.
28195         (vect_expand_fold_left): Likewise.
28196         (vectorize_fold_left_reduction): Likewise.
28197         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28198         scalar phi in place for it.  Check for target support and reject
28199         cases that would reassociate the operation.  Defer the transform
28200         phase to vectorize_fold_left_reduction.
28201         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28202         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28203         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28205 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28207         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28208         call to ifc_temp_var.
28210 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28211             Alan Hayward  <alan.hayward@arm.com>
28212             David Sherwood  <david.sherwood@arm.com>
28214         * target.def (legitimize_address_displacement): Take the original
28215         offset as a poly_int.
28216         * targhooks.h (default_legitimize_address_displacement): Update
28217         accordingly.
28218         * targhooks.c (default_legitimize_address_displacement): Likewise.
28219         * doc/tm.texi: Regenerate.
28220         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28221         as an argument, moving assert of ad->disp == ad->disp_term to...
28222         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28223         Try calling targetm.legitimize_address_displacement before expanding
28224         the address rather than afterwards, and adjust for the new interface.
28225         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28226         Match the new hook interface.  Handle SVE addresses.
28227         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28228         new hook interface.
28230 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28232         * Makefile.in (OBJS): Add early-remat.o.
28233         * target.def (select_early_remat_modes): New hook.
28234         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28235         * doc/tm.texi: Regenerate.
28236         * targhooks.h (default_select_early_remat_modes): Declare.
28237         * targhooks.c (default_select_early_remat_modes): New function.
28238         * timevar.def (TV_EARLY_REMAT): New timevar.
28239         * passes.def (pass_early_remat): New pass.
28240         * tree-pass.h (make_pass_early_remat): Declare.
28241         * early-remat.c: New file.
28242         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28243         function.
28244         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28246 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28247             Alan Hayward  <alan.hayward@arm.com>
28248             David Sherwood  <david.sherwood@arm.com>
28250         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28251         vfm1 with a bound_epilog parameter.
28252         (vect_do_peeling): Update calls accordingly, and move the prologue
28253         call earlier in the function.  Treat the base bound_epilog as 0 for
28254         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28255         this base when peeling for gaps.
28256         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28257         with fully-masked loops.
28258         (vect_estimate_min_profitable_iters): Handle the single peeled
28259         iteration in that case.
28261 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28262             Alan Hayward  <alan.hayward@arm.com>
28263             David Sherwood  <david.sherwood@arm.com>
28265         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28266         single-element interleaving even if the size is not a power of 2.
28267         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28268         accesses for single-element interleaving if the group size is
28269         not a power of 2.
28271 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28272             Alan Hayward  <alan.hayward@arm.com>
28273             David Sherwood  <david.sherwood@arm.com>
28275         * doc/md.texi (fold_extract_last_@var{m}): Document.
28276         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28277         * optabs.def (fold_extract_last_optab): New optab.
28278         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28279         * internal-fn.c (fold_extract_direct): New macro.
28280         (expand_fold_extract_optab_fn): Likewise.
28281         (direct_fold_extract_optab_supported_p): Likewise.
28282         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28283         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28284         EXTRACT_LAST_REDUCTION.
28285         (get_initial_def_for_reduction): Do not create an initial vector
28286         for EXTRACT_LAST_REDUCTION reductions.
28287         (vectorizable_reduction): Leave the scalar phi in place for
28288         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28289         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28290         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28291         transform phase to vectorizable_condition.
28292         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28293         split out from...
28294         (vect_finish_stmt_generation): ...here.
28295         (vect_finish_replace_stmt): New function.
28296         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28297         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28298         pattern.
28299         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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         * doc/md.texi (extract_last_@var{m}): Document.
28306         * optabs.def (extract_last_optab): New optab.
28307         * internal-fn.def (EXTRACT_LAST): New internal function.
28308         * internal-fn.c (cond_unary_direct): New macro.
28309         (expand_cond_unary_optab_fn): Likewise.
28310         (direct_cond_unary_optab_supported_p): Likewise.
28311         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28312         loops using EXTRACT_LAST.
28313         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28314         (extract_last_<mode>): ...this optab.
28315         (vec_extract<mode><Vel>): Update accordingly.
28317 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28318             Alan Hayward  <alan.hayward@arm.com>
28319             David Sherwood  <david.sherwood@arm.com>
28321         * target.def (empty_mask_is_expensive): New hook.
28322         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28323         * doc/tm.texi: Regenerate.
28324         * targhooks.h (default_empty_mask_is_expensive): Declare.
28325         * targhooks.c (default_empty_mask_is_expensive): New function.
28326         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28327         if the target says that empty masks are expensive.
28328         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28329         New function.
28330         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28332 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28333             Alan Hayward  <alan.hayward@arm.com>
28334             David Sherwood  <david.sherwood@arm.com>
28336         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28337         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28338         (vect_use_loop_mask_for_alignment_p): New function.
28339         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28340         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28341         niters_skip argument.  Make sure that the first niters_skip elements
28342         of the first iteration are inactive.
28343         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28344         Update call to vect_set_loop_masks_directly.
28345         (get_misalign_in_elems): New function, split out from...
28346         (vect_gen_prolog_loop_niters): ...here.
28347         (vect_update_init_of_dr): Take a code argument that specifies whether
28348         the adjustment should be added or subtracted.
28349         (vect_update_init_of_drs): Likewise.
28350         (vect_prepare_for_masked_peels): New function.
28351         (vect_do_peeling): Skip prologue peeling if we're using a mask
28352         instead.  Update call to vect_update_inits_of_drs.
28353         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28354         mask_skip_niters.
28355         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28356         alignment.  Do not include the number of peeled iterations in
28357         the minimum threshold in that case.
28358         (vectorizable_induction): Adjust the start value down by
28359         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28360         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28361         Take the number of skipped iterations into account when calculating
28362         the loop bounds.
28363         * tree-vect-stmts.c (vect_gen_while_not): New function.
28365 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28366             Alan Hayward  <alan.hayward@arm.com>
28367             David Sherwood  <david.sherwood@arm.com>
28369         * doc/sourcebuild.texi (vect_fully_masked): Document.
28370         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28371         default value to 0.
28372         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28373         split out from...
28374         (vect_analyze_loop_2): ...here. Don't check the vectorization
28375         factor against the number of loop iterations if the loop is
28376         fully-masked.
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         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28383         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28384         (dump_groups): Update accordingly.
28385         (iv_use::mem_type): New member variable.
28386         (address_p): New function.
28387         (record_use): Add a mem_type argument and initialize the new
28388         mem_type field.
28389         (record_group_use): Add a mem_type argument.  Use address_p.
28390         Remove obsolete null checks of base_object.  Update call to record_use.
28391         (find_interesting_uses_op): Update call to record_group_use.
28392         (find_interesting_uses_cond): Likewise.
28393         (find_interesting_uses_address): Likewise.
28394         (get_mem_type_for_internal_fn): New function.
28395         (find_address_like_use): Likewise.
28396         (find_interesting_uses_stmt): Try find_address_like_use before
28397         calling find_interesting_uses_op.
28398         (addr_offset_valid_p): Use the iv mem_type field as the type
28399         of the addressed memory.
28400         (add_autoinc_candidates): Likewise.
28401         (get_address_cost): Likewise.
28402         (split_small_address_groups_p): Use address_p.
28403         (split_address_groups): Likewise.
28404         (add_iv_candidate_for_use): Likewise.
28405         (autoinc_possible_for_pair): Likewise.
28406         (rewrite_groups): Likewise.
28407         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28408         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28409         (get_alias_ptr_type_for_ptr_address): New function.
28410         (rewrite_use_address): Rewrite address uses in calls that were
28411         identified by find_address_like_use.
28413 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28414             Alan Hayward  <alan.hayward@arm.com>
28415             David Sherwood  <david.sherwood@arm.com>
28417         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28418         TARGET_MEM_REFs.
28419         * gimple-expr.h (is_gimple_addressable: Likewise.
28420         * gimple-expr.c (is_gimple_address): Likewise.
28421         * internal-fn.c (expand_call_mem_ref): New function.
28422         (expand_mask_load_optab_fn): Use it.
28423         (expand_mask_store_optab_fn): Likewise.
28425 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28426             Alan Hayward  <alan.hayward@arm.com>
28427             David Sherwood  <david.sherwood@arm.com>
28429         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28430         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28431         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28432         (cond_umax@var{mode}): Document.
28433         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28434         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28435         (cond_umin_optab, cond_umax_optab): New optabs.
28436         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28437         (COND_IOR, COND_XOR): New internal functions.
28438         * internal-fn.h (get_conditional_internal_fn): Declare.
28439         * internal-fn.c (cond_binary_direct): New macro.
28440         (expand_cond_binary_optab_fn): Likewise.
28441         (direct_cond_binary_optab_supported_p): Likewise.
28442         (get_conditional_internal_fn): New function.
28443         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28444         Cope with reduction statements that are vectorized as calls rather
28445         than assignments.
28446         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28447         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28448         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28449         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28450         (UNSPEC_COND_EOR): New unspecs.
28451         (optab): Add mappings for them.
28452         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28453         (sve_int_op, sve_fp_op): New int attributes.
28455 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28456             Alan Hayward  <alan.hayward@arm.com>
28457             David Sherwood  <david.sherwood@arm.com>
28459         * optabs.def (while_ult_optab): New optab.
28460         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28461         * internal-fn.def (WHILE_ULT): New internal function.
28462         * internal-fn.h (direct_internal_fn_supported_p): New override
28463         that takes two types as argument.
28464         * internal-fn.c (while_direct): New macro.
28465         (expand_while_optab_fn): New function.
28466         (convert_optab_supported_p): Likewise.
28467         (direct_while_optab_supported_p): New macro.
28468         * wide-int.h (wi::udiv_ceil): New function.
28469         * tree-vectorizer.h (rgroup_masks): New structure.
28470         (vec_loop_masks): New typedef.
28471         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28472         and fully_masked_p.
28473         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28474         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28475         (vect_max_vf): New function.
28476         (slpeel_make_loop_iterate_ntimes): Delete.
28477         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28478         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28479         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28480         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28481         internal-fn.h, stor-layout.h and optabs-query.h.
28482         (vect_set_loop_mask): New function.
28483         (add_preheader_seq): Likewise.
28484         (add_header_seq): Likewise.
28485         (interleave_supported_p): Likewise.
28486         (vect_maybe_permute_loop_masks): Likewise.
28487         (vect_set_loop_masks_directly): Likewise.
28488         (vect_set_loop_condition_masked): Likewise.
28489         (vect_set_loop_condition_unmasked): New function, split out from
28490         slpeel_make_loop_iterate_ntimes.
28491         (slpeel_make_loop_iterate_ntimes): Rename to..
28492         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28493         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28494         (vect_do_peeling): Update call accordingly.
28495         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28496         loops.
28497         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28498         mask_compare_type, can_fully_mask_p and fully_masked_p.
28499         (release_vec_loop_masks): New function.
28500         (_loop_vec_info): Use it to free the loop masks.
28501         (can_produce_all_loop_masks_p): New function.
28502         (vect_get_max_nscalars_per_iter): Likewise.
28503         (vect_verify_full_masking): Likewise.
28504         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28505         retries, and free the mask rgroups before retrying.  Check loop-wide
28506         reasons for disallowing fully-masked loops.  Make the final decision
28507         about whether use a fully-masked loop or not.
28508         (vect_estimate_min_profitable_iters): Do not assume that peeling
28509         for the number of iterations will be needed for fully-masked loops.
28510         (vectorizable_reduction): Disable fully-masked loops.
28511         (vectorizable_live_operation): Likewise.
28512         (vect_halve_mask_nunits): New function.
28513         (vect_double_mask_nunits): Likewise.
28514         (vect_record_loop_mask): Likewise.
28515         (vect_get_loop_mask): Likewise.
28516         (vect_transform_loop): Handle the case in which the final loop
28517         iteration might handle a partial vector.  Call vect_set_loop_condition
28518         instead of slpeel_make_loop_iterate_ntimes.
28519         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28520         (check_load_store_masking): New function.
28521         (prepare_load_store_mask): Likewise.
28522         (vectorizable_store): Handle fully-masked loops.
28523         (vectorizable_load): Likewise.
28524         (supportable_widening_operation): Use vect_halve_mask_nunits for
28525         booleans.
28526         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28527         (vect_gen_while): New function.
28528         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28529         (aarch64_uqdec<mode>): New insn.
28531 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28532             Alan Hayward  <alan.hayward@arm.com>
28533             David Sherwood  <david.sherwood@arm.com>
28535         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28536         (reduc_xor_scal_optab): New optabs.
28537         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28538         (reduc_xor_scal_@var{m}): Document.
28539         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28540         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28541         internal functions.
28542         * fold-const-call.c (fold_const_call): Handle them.
28543         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28544         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28545         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28546         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28547         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28548         (UNSPEC_XORV): New unspecs.
28549         (optab): Add entries for them.
28550         (BITWISEV): New int iterator.
28551         (bit_reduc_op): New int attributes.
28553 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28554             Alan Hayward  <alan.hayward@arm.com>
28555             David Sherwood  <david.sherwood@arm.com>
28557         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28558         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28559         * optabs.def (vec_shl_insert_optab): New optab.
28560         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28561         (duplicate_and_interleave): Likewise.
28562         * tree-vect-loop.c: Include internal-fn.h.
28563         (neutral_op_for_slp_reduction): New function, split out from
28564         get_initial_defs_for_reduction.
28565         (get_initial_def_for_reduction): Handle option 2 for variable-length
28566         vectors by loading the neutral value into a vector and then shifting
28567         the initial value into element 0.
28568         (get_initial_defs_for_reduction): Replace the code argument with
28569         the neutral value calculated by neutral_op_for_slp_reduction.
28570         Use gimple_build_vector for constant-length vectors.
28571         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28572         but the first group_size elements have a neutral value.
28573         Use duplicate_and_interleave otherwise.
28574         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28575         Update call to get_initial_defs_for_reduction.  Handle SLP
28576         reductions for variable-length vectors by creating one vector
28577         result for each scalar result, with the elements associated
28578         with other scalar results stubbed out with the neutral value.
28579         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
28580         Require IFN_VEC_SHL_INSERT for double reductions on
28581         variable-length vectors, or SLP reductions that have
28582         a neutral value.  Require can_duplicate_and_interleave_p
28583         support for variable-length unchained SLP reductions if there
28584         is no neutral value, such as for MIN/MAX reductions.  Also require
28585         the number of vector elements to be a multiple of the number of
28586         SLP statements when doing variable-length unchained SLP reductions.
28587         Update call to vect_create_epilog_for_reduction.
28588         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
28589         and remove initial values.
28590         (duplicate_and_interleave): Make public.
28591         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
28592         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
28594 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28595             Alan Hayward  <alan.hayward@arm.com>
28596             David Sherwood  <david.sherwood@arm.com>
28598         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
28599         (can_duplicate_and_interleave_p): New function.
28600         (vect_get_and_check_slp_defs): Take the vector of statements
28601         rather than just the current one.  Remove excess parentheses.
28602         Restriction rejectinon of vect_constant_def and vect_external_def
28603         for variable-length vectors to boolean types, or types for which
28604         can_duplicate_and_interleave_p is false.
28605         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
28606         (duplicate_and_interleave): New function.
28607         (vect_get_constant_vectors): Use gimple_build_vector for
28608         constant-length vectors and suitable variable-length constant
28609         vectors.  Use duplicate_and_interleave for other variable-length
28610         vectors.  Don't defer the update when inserting new statements.
28612 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28613             Alan Hayward  <alan.hayward@arm.com>
28614             David Sherwood  <david.sherwood@arm.com>
28616         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
28617         min_profitable_iters doesn't go negative.
28619 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28620             Alan Hayward  <alan.hayward@arm.com>
28621             David Sherwood  <david.sherwood@arm.com>
28623         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
28624         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
28625         * optabs.def (vec_mask_load_lanes_optab): New optab.
28626         (vec_mask_store_lanes_optab): Likewise.
28627         * internal-fn.def (MASK_LOAD_LANES): New internal function.
28628         (MASK_STORE_LANES): Likewise.
28629         * internal-fn.c (mask_load_lanes_direct): New macro.
28630         (mask_store_lanes_direct): Likewise.
28631         (expand_mask_load_optab_fn): Handle masked operations.
28632         (expand_mask_load_lanes_optab_fn): New macro.
28633         (expand_mask_store_optab_fn): Handle masked operations.
28634         (expand_mask_store_lanes_optab_fn): New macro.
28635         (direct_mask_load_lanes_optab_supported_p): Likewise.
28636         (direct_mask_store_lanes_optab_supported_p): Likewise.
28637         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
28638         parameter.
28639         (vect_load_lanes_supported): Likewise.
28640         * tree-vect-data-refs.c (strip_conversion): New function.
28641         (can_group_stmts_p): Likewise.
28642         (vect_analyze_data_ref_accesses): Use it instead of checking
28643         for a pair of assignments.
28644         (vect_store_lanes_supported): Take a masked_p parameter.
28645         (vect_load_lanes_supported): Likewise.
28646         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
28647         vect_store_lanes_supported and vect_load_lanes_supported.
28648         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
28649         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
28650         parameter.  Don't allow gaps for masked accesses.
28651         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
28652         and vect_load_lanes_supported.
28653         (get_load_store_type): Take a masked_p parameter and update
28654         call to get_group_load_store_type.
28655         (vectorizable_store): Update call to get_load_store_type.
28656         Handle IFN_MASK_STORE_LANES.
28657         (vectorizable_load): Update call to get_load_store_type.
28658         Handle IFN_MASK_LOAD_LANES.
28660 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28661             Alan Hayward  <alan.hayward@arm.com>
28662             David Sherwood  <david.sherwood@arm.com>
28664         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
28665         modes for SVE.
28666         * config/aarch64/aarch64-protos.h
28667         (aarch64_sve_struct_memory_operand_p): Declare.
28668         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
28669         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
28670         (VPRED, vpred): Handle SVE structure modes.
28671         * config/aarch64/constraints.md (Utx): New constraint.
28672         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
28673         (aarch64_sve_struct_nonimmediate_operand): New predicates.
28674         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
28675         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
28676         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
28677         structure modes.  Split into pieces after RA.
28678         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
28679         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
28680         New patterns.
28681         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
28682         SVE structure modes.
28683         (aarch64_classify_address): Likewise.
28684         (sizetochar): Move earlier in file.
28685         (aarch64_print_operand): Handle SVE register lists.
28686         (aarch64_array_mode): New function.
28687         (aarch64_sve_struct_memory_operand_p): Likewise.
28688         (TARGET_ARRAY_MODE): Redefine.
28690 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28691             Alan Hayward  <alan.hayward@arm.com>
28692             David Sherwood  <david.sherwood@arm.com>
28694         * target.def (array_mode): New target hook.
28695         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
28696         * doc/tm.texi: Regenerate.
28697         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
28698         * hooks.c (hook_optmode_mode_uhwi_none): New function.
28699         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
28700         targetm.array_mode.
28701         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
28702         type sizes.
28704 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28705             Alan Hayward  <alan.hayward@arm.com>
28706             David Sherwood  <david.sherwood@arm.com>
28708         * fold-const.c (fold_binary_loc): Check the argument types
28709         rather than the result type when testing for a vector operation.
28711 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28713         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
28714         * doc/tm.texi: Regenerate.
28716 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28717             Alan Hayward  <alan.hayward@arm.com>
28718             David Sherwood  <david.sherwood@arm.com>
28720         * doc/invoke.texi (-msve-vector-bits=): Document new option.
28721         (sve): Document new AArch64 extension.
28722         * doc/md.texi (w): Extend the description of the AArch64
28723         constraint to include SVE vectors.
28724         (Upl, Upa): Document new AArch64 predicate constraints.
28725         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
28726         enum.
28727         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
28728         (msve-vector-bits=): New option.
28729         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
28730         SVE when these are disabled.
28731         (sve): New extension.
28732         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
28733         modes.  Adjust their number of units based on aarch64_sve_vg.
28734         (MAX_BITSIZE_MODE_ANY_MODE): Define.
28735         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
28736         aarch64_addr_query_type.
28737         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
28738         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
28739         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
28740         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
28741         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
28742         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
28743         (aarch64_simd_imm_zero_p): Delete.
28744         (aarch64_check_zero_based_sve_index_immediate): Declare.
28745         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28746         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28747         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28748         (aarch64_sve_float_mul_immediate_p): Likewise.
28749         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28750         rather than an rtx.
28751         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
28752         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
28753         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
28754         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
28755         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
28756         (aarch64_regmode_natural_size): Likewise.
28757         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
28758         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
28759         left one place.
28760         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
28761         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
28762         for VG and the SVE predicate registers.
28763         (V_ALIASES): Add a "z"-prefixed alias.
28764         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
28765         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
28766         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
28767         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
28768         (REG_CLASS_NAMES): Add entries for them.
28769         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
28770         and the predicate registers.
28771         (aarch64_sve_vg): Declare.
28772         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
28773         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
28774         (REGMODE_NATURAL_SIZE): Define.
28775         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
28776         SVE macros.
28777         * config/aarch64/aarch64.c: Include cfgrtl.h.
28778         (simd_immediate_info): Add a constructor for series vectors,
28779         and an associated step field.
28780         (aarch64_sve_vg): New variable.
28781         (aarch64_dbx_register_number): Handle VG and the predicate registers.
28782         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
28783         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
28784         (VEC_ANY_DATA, VEC_STRUCT): New constants.
28785         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
28786         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
28787         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
28788         (aarch64_get_mask_mode): New functions.
28789         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
28790         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
28791         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
28792         predicate modes and predicate registers.  Explicitly restrict
28793         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
28794         to store a vector mode if it is recognized by
28795         aarch64_classify_vector_mode.
28796         (aarch64_regmode_natural_size): New function.
28797         (aarch64_hard_regno_caller_save_mode): Return the original mode
28798         for predicates.
28799         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
28800         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
28801         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
28802         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
28803         functions.
28804         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
28805         does not overlap dest if the function is frame-related.  Handle
28806         SVE constants.
28807         (aarch64_split_add_offset): New function.
28808         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
28809         them aarch64_add_offset.
28810         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
28811         and update call to aarch64_sub_sp.
28812         (aarch64_add_cfa_expression): New function.
28813         (aarch64_expand_prologue): Pass extra temporary registers to the
28814         functions above.  Handle the case in which we need to emit new
28815         DW_CFA_expressions for registers that were originally saved
28816         relative to the stack pointer, but now have to be expressed
28817         relative to the frame pointer.
28818         (aarch64_output_mi_thunk): Pass extra temporary registers to the
28819         functions above.
28820         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
28821         IP0 and IP1 values for SVE frames.
28822         (aarch64_expand_vec_series): New function.
28823         (aarch64_expand_sve_widened_duplicate): Likewise.
28824         (aarch64_expand_sve_const_vector): Likewise.
28825         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
28826         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
28827         into the register, rather than emitting a SET directly.
28828         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
28829         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
28830         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
28831         (offset_9bit_signed_scaled_p): New functions.
28832         (aarch64_replicate_bitmask_imm): New function.
28833         (aarch64_bitmask_imm): Use it.
28834         (aarch64_cannot_force_const_mem): Reject expressions involving
28835         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
28836         (aarch64_classify_index): Handle SVE indices, by requiring
28837         a plain register index with a scale that matches the element size.
28838         (aarch64_classify_address): Handle SVE addresses.  Assert that
28839         the mode of the address is VOIDmode or an integer mode.
28840         Update call to aarch64_classify_symbol.
28841         (aarch64_classify_symbolic_expression): Update call to
28842         aarch64_classify_symbol.
28843         (aarch64_const_vec_all_in_range_p): New function.
28844         (aarch64_print_vector_float_operand): Likewise.
28845         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
28846         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
28847         and the FP immediates 1.0 and 0.5.
28848         (aarch64_print_address_internal): Handle SVE addresses.
28849         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
28850         (aarch64_regno_regclass): Handle predicate registers.
28851         (aarch64_secondary_reload): Handle big-endian reloads of SVE
28852         data modes.
28853         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
28854         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
28855         (aarch64_convert_sve_vector_bits): New function.
28856         (aarch64_override_options): Use it to handle -msve-vector-bits=.
28857         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28858         rather than an rtx.
28859         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
28860         Handle SVE vector and predicate modes.  Accept VL-based constants
28861         that need only one temporary register, and VL offsets that require
28862         no temporary registers.
28863         (aarch64_conditional_register_usage): Mark the predicate registers
28864         as fixed if SVE isn't available.
28865         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
28866         Return true for SVE vector and predicate modes.
28867         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
28868         rather than an unsigned int.  Handle SVE modes.
28869         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
28870         SVE modes.
28871         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
28872         if SVE is enabled.
28873         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28874         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28875         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28876         (aarch64_sve_float_mul_immediate_p): New functions.
28877         (aarch64_sve_valid_immediate): New function.
28878         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
28879         Explicitly reject structure modes.  Check for INDEX constants.
28880         Handle PTRUE and PFALSE constants.
28881         (aarch64_check_zero_based_sve_index_immediate): New function.
28882         (aarch64_simd_imm_zero_p): Delete.
28883         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
28884         vector modes.  Accept constants in the range of CNT[BHWD].
28885         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
28886         ask for an Advanced SIMD mode.
28887         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
28888         (aarch64_simd_vector_alignment): Handle SVE predicates.
28889         (aarch64_vectorize_preferred_vector_alignment): New function.
28890         (aarch64_simd_vector_alignment_reachable): Use it instead of
28891         the vector size.
28892         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
28893         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
28894         functions.
28895         (MAX_VECT_LEN): Delete.
28896         (expand_vec_perm_d): Add a vec_flags field.
28897         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
28898         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
28899         (aarch64_evpc_ext): Don't apply a big-endian lane correction
28900         for SVE modes.
28901         (aarch64_evpc_rev): Rename to...
28902         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
28903         (aarch64_evpc_rev_global): New function.
28904         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
28905         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
28906         MAX_VECT_LEN.
28907         (aarch64_evpc_sve_tbl): New function.
28908         (aarch64_expand_vec_perm_const_1): Update after rename of
28909         aarch64_evpc_rev.  Handle SVE permutes too, trying
28910         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
28911         than aarch64_evpc_tbl.
28912         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
28913         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
28914         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
28915         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
28916         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
28917         (aarch64_expand_sve_vcond): New functions.
28918         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
28919         of aarch64_vector_mode_p.
28920         (aarch64_dwarf_poly_indeterminate_value): New function.
28921         (aarch64_compute_pressure_classes): Likewise.
28922         (aarch64_can_change_mode_class): Likewise.
28923         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
28924         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
28925         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
28926         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
28927         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
28928         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
28929         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
28930         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
28931         constraints.
28932         (Dn, Dl, Dr): Accept const as well as const_vector.
28933         (Dz): Likewise.  Compare against CONST0_RTX.
28934         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
28935         of "vector" where appropriate.
28936         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
28937         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
28938         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
28939         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
28940         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
28941         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
28942         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
28943         (v_int_equiv): Extend to SVE modes.
28944         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
28945         mode attributes.
28946         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
28947         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
28948         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
28949         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
28950         (SVE_COND_FP_CMP): New int iterators.
28951         (perm_hilo): Handle the new unpack unspecs.
28952         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
28953         attributes.
28954         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
28955         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
28956         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
28957         (aarch64_equality_operator, aarch64_constant_vector_operand)
28958         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
28959         (aarch64_sve_nonimmediate_operand): Likewise.
28960         (aarch64_sve_general_operand): Likewise.
28961         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
28962         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
28963         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
28964         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
28965         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
28966         (aarch64_sve_float_arith_immediate): Likewise.
28967         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
28968         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
28969         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
28970         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
28971         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
28972         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
28973         (aarch64_sve_float_arith_operand): Likewise.
28974         (aarch64_sve_float_arith_with_sub_operand): Likewise.
28975         (aarch64_sve_float_mul_operand): Likewise.
28976         (aarch64_sve_vec_perm_operand): Likewise.
28977         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
28978         (aarch64_mov_operand): Accept const_poly_int and const_vector.
28979         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
28980         as well as const_vector.
28981         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
28982         in file.  Use CONST0_RTX and CONSTM1_RTX.
28983         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
28984         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
28985         Use aarch64_simd_imm_zero.
28986         * config/aarch64/aarch64-sve.md: New file.
28987         * config/aarch64/aarch64.md: Include it.
28988         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
28989         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
28990         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
28991         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
28992         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
28993         (sve): New attribute.
28994         (enabled): Disable instructions with the sve attribute unless
28995         TARGET_SVE.
28996         (movqi, movhi): Pass CONST_POLY_INT operaneds through
28997         aarch64_expand_mov_immediate.
28998         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
28999         CNT[BHSD] immediates.
29000         (movti): Split CONST_POLY_INT moves into two halves.
29001         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29002         Split additions that need a temporary here if the destination
29003         is the stack pointer.
29004         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29005         (*add<mode>3_poly_1): New instruction.
29006         (set_clobber_cc): New expander.
29008 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29010         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29011         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
29012         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29013         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29014         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
29015         Change innermode from fixed_mode_size to machine_mode.
29016         (simplify_subreg): Update call accordingly.  Handle a constant-sized
29017         subreg of a variable-length CONST_VECTOR.
29019 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29020             Alan Hayward  <alan.hayward@arm.com>
29021             David Sherwood  <david.sherwood@arm.com>
29023         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29024         (add_offset_to_base): New function, split out from...
29025         (create_mem_ref): ...here.  When handling a scale other than 1,
29026         check first whether the address is valid without the offset.
29027         Add it into the base if so, leaving the index and scale as-is.
29029 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29031         PR c++/83778
29032         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29033         fold_for_warn before checking if arg2 is INTEGER_CST.
29035 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
29037         * config/rs6000/predicates.md (load_multiple_operation): Delete.
29038         (store_multiple_operation): Delete.
29039         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29040         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29041         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29042         guarded by TARGET_STRING.
29043         (rs6000_output_load_multiple): Delete.
29044         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29045         OPTION_MASK_STRING / TARGET_STRING handling.
29046         (print_operand) <'N', 'O'>: Add comment that these are unused now.
29047         (const rs6000_opt_masks) <"string">: Change mask to 0.
29048         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29049         (MASK_STRING): Delete.
29050         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29051         parts.  Simplify.
29052         (load_multiple): Delete.
29053         (*ldmsi8): Delete.
29054         (*ldmsi7): Delete.
29055         (*ldmsi6): Delete.
29056         (*ldmsi5): Delete.
29057         (*ldmsi4): Delete.
29058         (*ldmsi3): Delete.
29059         (store_multiple): Delete.
29060         (*stmsi8): Delete.
29061         (*stmsi7): Delete.
29062         (*stmsi6): Delete.
29063         (*stmsi5): Delete.
29064         (*stmsi4): Delete.
29065         (*stmsi3): Delete.
29066         (movmemsi_8reg): Delete.
29067         (corresponding unnamed define_insn): Delete.
29068         (movmemsi_6reg): Delete.
29069         (corresponding unnamed define_insn): Delete.
29070         (movmemsi_4reg): Delete.
29071         (corresponding unnamed define_insn): Delete.
29072         (movmemsi_2reg): Delete.
29073         (corresponding unnamed define_insn): Delete.
29074         (movmemsi_1reg): Delete.
29075         (corresponding unnamed define_insn): Delete.
29076         * config/rs6000/rs6000.opt (mno-string): New.
29077         (mstring): Replace by deprecation warning stub.
29078         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29080 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29082         * regrename.c (regrename_do_replace): If replacing the same
29083         reg multiple times, try to reuse last created gen_raw_REG.
29085         PR debug/81155
29086         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29087         main to workaround a bug in GDB.
29089 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29091         PR target/83737
29092         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29094 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29096         PR rtl-optimization/80481
29097         * ira-color.c (get_cap_member): New function.
29098         (allocnos_conflict_by_live_ranges_p): Use it.
29099         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29100         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29102 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29104         PR target/83628
29105         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29106         (*saddl_se_1): Ditto.
29107         (*ssubsi_1): Ditto.
29108         (*ssubl_se_1): Ditto.
29110 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29112         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29113         rather than wi::to_widest for DR_INITs.
29114         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29115         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29116         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29117         INTEGER_CSTs.
29118         (vect_analyze_group_access_1): Note that here.
29120 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29122         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29123         polynomial type sizes.
29125 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29127         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29128         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29129         (gimple_add_tmp_var): Likewise.
29131 2018-01-12  Martin Liska  <mliska@suse.cz>
29133         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29134         (gimple_alloc_sizes): Likewise.
29135         (dump_gimple_statistics): Use PRIu64 in printf format.
29136         * gimple.h: Change uint64_t to int.
29138 2018-01-12  Martin Liska  <mliska@suse.cz>
29140         * tree-core.h: Use uint64_t instead of int.
29141         * tree.c (tree_node_counts): Likewise.
29142         (tree_node_sizes): Likewise.
29143         (dump_tree_statistics): Use PRIu64 in printf format.
29145 2018-01-12  Martin Liska  <mliska@suse.cz>
29147         * Makefile.in: As qsort_chk is implemented in vec.c, add
29148         vec.o to linkage of gencfn-macros.
29149         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29150         passing the info to record_node_allocation_statistics.
29151         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29152         and pass the info.
29153         * ggc-common.c (struct ggc_usage): Add operator== and use
29154         it in operator< and compare function.
29155         * mem-stats.h (struct mem_usage): Likewise.
29156         * vec.c (struct vec_usage): Remove operator< and compare
29157         function. Can be simply inherited.
29159 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29161         PR target/81616
29162         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29163         * tree-ssa-math-opts.c: Include domwalk.h.
29164         (convert_mult_to_fma_1): New function.
29165         (fma_transformation_info): New type.
29166         (fma_deferring_state): Likewise.
29167         (cancel_fma_deferring): New function.
29168         (result_of_phi): Likewise.
29169         (last_fma_candidate_feeds_initial_phi): Likewise.
29170         (convert_mult_to_fma): Added deferring logic, split actual
29171         transformation to convert_mult_to_fma_1.
29172         (math_opts_dom_walker): New type.
29173         (math_opts_dom_walker::after_dom_children): New method, body moved
29174         here from pass_optimize_widening_mul::execute, added deferring logic
29175         bits.
29176         (pass_optimize_widening_mul::execute): Moved most of code to
29177         math_opts_dom_walker::after_dom_children.
29178         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29179         * config/i386/i386.c (ix86_option_override_internal): Added
29180         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29182 2018-01-12  Richard Biener  <rguenther@suse.de>
29184         PR debug/83157
29185         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29186         inline instance vars.
29188 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29190         PR target/81819
29191         * config/rx/rx.c (rx_is_restricted_memory_address):
29192         Handle SUBREG case.
29194 2018-01-12  Richard Biener  <rguenther@suse.de>
29196         PR tree-optimization/80846
29197         * target.def (split_reduction): New target hook.
29198         * targhooks.c (default_split_reduction): New function.
29199         * targhooks.h (default_split_reduction): Declare.
29200         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29201         target requests first reduce vectors by combining low and high
29202         parts.
29203         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29204         (get_vectype_for_scalar_type_and_size): Export.
29205         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29206         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29207         * doc/tm.texi: Regenerate.
29208         * config/i386/i386.c (ix86_split_reduction): Implement
29209         TARGET_VECTORIZE_SPLIT_REDUCTION.
29211 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29213         PR target/83368
29214         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29215         in PIC mode except for TARGET_VXWORKS_RTP.
29216         * config/sparc/sparc.c: Include cfgrtl.h.
29217         (TARGET_INIT_PIC_REG): Define.
29218         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29219         (sparc_pic_register_p): New predicate.
29220         (sparc_legitimate_address_p): Use it.
29221         (sparc_legitimize_pic_address): Likewise.
29222         (sparc_delegitimize_address): Likewise.
29223         (sparc_mode_dependent_address_p): Likewise.
29224         (gen_load_pcrel_sym): Remove 4th parameter.
29225         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29226         (sparc_expand_prologue): Do not call load_got_register here.
29227         (sparc_flat_expand_prologue): Likewise.
29228         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29229         (sparc_use_pseudo_pic_reg): New function.
29230         (sparc_init_pic_reg): Likewise.
29231         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29232         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29234 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29236         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29237         Add item for branch_cost.
29239 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29241         PR rtl-optimization/83565
29242         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29243         not extend the result to a larger mode for rotate operations.
29244         (num_sign_bit_copies1): Likewise.
29246 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29248         PR target/40411
29249         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29250         -symbolic.
29251         Use values-Xc.o for -pedantic.
29252         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29254 2018-01-12  Martin Liska  <mliska@suse.cz>
29256         PR ipa/83054
29257         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29258         New function.
29259         (possible_polymorphic_call_targets): Use it.
29260         (ipa_devirt): Likewise.
29262 2018-01-12  Martin Liska  <mliska@suse.cz>
29264         * profile-count.h (enum profile_quality): Use 0 as invalid
29265         enum value of profile_quality.
29267 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29269         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29270         -mext-string options.
29272 2018-01-12  Richard Biener  <rguenther@suse.de>
29274         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29275         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29276         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29277         Likewise.
29278         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29280 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29282         * configure.ac (--with-long-double-format): Add support for the
29283         configuration option to change the default long double format on
29284         PowerPC systems.
29285         * config.gcc (powerpc*-linux*-*): Likewise.
29286         * configure: Regenerate.
29287         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29288         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29289         used without modification.
29291 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29293         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29294         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29295         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29296         MISC_BUILTIN_SPEC_BARRIER.
29297         (rs6000_init_builtins): Likewise.
29298         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29299         enum value.
29300         (speculation_barrier): New define_insn.
29301         * doc/extend.texi: Document __builtin_speculation_barrier.
29303 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29305         PR target/83203
29306         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29307         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29308         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29309         iterators.
29310         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29311         integral modes instead of "ss" and "sd".
29312         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29313         vectors with 32-bit and 64-bit elements.
29314         (vecdupssescalarmodesuffix): New mode attribute.
29315         (vec_dup<mode>): Use it.
29317 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29319         PR target/83330
29320         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29321         frame if argument is passed on stack.
29323 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29325         PR target/82682
29326         * ree.c (combine_reaching_defs): Optimize also
29327         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29328         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29330 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29332         PR middle-end/83189
29333         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29335 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29337         PR middle-end/83718
29338         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29339         after they are computed.
29341 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29343         PR tree-optimization/83695
29344         * gimple-loop-linterchange.cc
29345         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29346         reset cached scev information after interchange.
29347         (pass_linterchange::execute): Remove call to scev_reset_htab.
29349 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29351         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29352         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29353         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29354         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29355         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29356         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29357         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29358         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29359         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29360         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29361         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29362         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29363         (V_lane_reg): Likewise.
29364         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29365         New define_expand.
29366         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29367         (vfmal_lane_low<mode>_intrinsic,
29368         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29369         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29370         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29371         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29372         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29373         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29375 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29377         * config/arm/arm-cpus.in (fp16fml): New feature.
29378         (ALL_SIMD): Add fp16fml.
29379         (armv8.2-a): Add fp16fml as an option.
29380         (armv8.3-a): Likewise.
29381         (armv8.4-a): Add fp16fml as part of fp16.
29382         * config/arm/arm.h (TARGET_FP16FML): Define.
29383         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29384         when appropriate.
29385         * config/arm/arm-modes.def (V2HF): Define.
29386         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29387         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29388         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29389         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29390         vfmsl_low, vfmsl_high): New set of builtins.
29391         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29392         (vfml_op): New code attribute.
29393         (VFMLHALVES): New int iterator.
29394         (VFML, VFMLSEL): New mode attributes.
29395         (V_reg): Define mapping for V2HF.
29396         (V_hi, V_lo): New mode attributes.
29397         (VF_constraint): Likewise.
29398         (vfml_half, vfml_half_selector): New int attributes.
29399         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29400         define_expand.
29401         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29402         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29403         New define_insn.
29404         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29405         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29406         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29407         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29408         documentation.
29409         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29410         Document new effective target and option set.
29412 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29414         * config/arm/arm-cpus.in (armv8_4): New feature.
29415         (ARMv8_4a): New fgroup.
29416         (armv8.4-a): New arch.
29417         * config/arm/arm-tables.opt: Regenerate.
29418         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29419         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29420         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29421         Add matching rules for -march=armv8.4-a and extensions.
29422         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29424 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29426         PR target/81821
29427         * config/rx/rx.md (BW): New mode attribute.
29428         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29430 2018-01-11  Richard Biener  <rguenther@suse.de>
29432         PR tree-optimization/83435
29433         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29434         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29435         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29437 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29438             Alan Hayward  <alan.hayward@arm.com>
29439             David Sherwood  <david.sherwood@arm.com>
29441         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29442         field.
29443         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29444         (aarch64_print_address_internal): Use it to check for a zero offset.
29446 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29447             Alan Hayward  <alan.hayward@arm.com>
29448             David Sherwood  <david.sherwood@arm.com>
29450         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29451         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29452         Return a poly_int64 rather than a HOST_WIDE_INT.
29453         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29454         rather than a HOST_WIDE_INT.
29455         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29456         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29457         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29458         final_offset from HOST_WIDE_INT to poly_int64.
29459         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29460         to_constant when getting the number of units in an Advanced SIMD
29461         mode.
29462         (aarch64_builtin_vectorized_function): Check for a constant number
29463         of units.
29464         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29465         GET_MODE_SIZE.
29466         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29467         attribute instead of GET_MODE_NUNITS.
29468         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29469         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29470         GET_MODE_SIZE for fixed-size registers.
29471         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29472         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29473         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29474         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29475         (aarch64_print_operand, aarch64_print_address_internal)
29476         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29477         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29478         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29479         Handle polynomial GET_MODE_SIZE.
29480         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29481         wider than SImode without modification.
29482         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29483         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29484         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29485         passing and returning SVE modes.
29486         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29487         rather than GEN_INT.
29488         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29489         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29490         (aarch64_allocate_and_probe_stack_space): Likewise.
29491         (aarch64_layout_frame): Cope with polynomial offsets.
29492         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29493         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29494         polynomial offsets.
29495         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29496         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29497         poly_int64 rather than a HOST_WIDE_INT.
29498         (aarch64_get_separate_components, aarch64_process_components)
29499         (aarch64_expand_prologue, aarch64_expand_epilogue)
29500         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29501         (aarch64_anchor_offset): New function, split out from...
29502         (aarch64_legitimize_address): ...here.
29503         (aarch64_builtin_vectorization_cost): Handle polynomial
29504         TYPE_VECTOR_SUBPARTS.
29505         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29506         GET_MODE_NUNITS.
29507         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29508         number of elements from the PARALLEL rather than the mode.
29509         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29510         rather than GET_MODE_BITSIZE.
29511         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29512         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29513         (aarch64_expand_vec_perm_const_1): Handle polynomial
29514         d->perm.length () and d->perm elements.
29515         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29516         Apply to_constant to d->perm elements.
29517         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29518         polynomial CONST_VECTOR_NUNITS.
29519         (aarch64_move_pointer): Take amount as a poly_int64 rather
29520         than an int.
29521         (aarch64_progress_pointer): Avoid temporary variable.
29522         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29523         the mode attribute instead of GET_MODE.
29525 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29526             Alan Hayward  <alan.hayward@arm.com>
29527             David Sherwood  <david.sherwood@arm.com>
29529         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29530         x exists before using it.
29531         (aarch64_add_constant_internal): Rename to...
29532         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29533         src and dest rtxes.  Handle the case in which they're different,
29534         including when the offset is zero.  Replace scratchreg with an rtx.
29535         Use 2 additions if there is no spare register into which we can
29536         move a 16-bit constant.
29537         (aarch64_add_constant): Delete.
29538         (aarch64_add_offset): Replace reg with separate src and dest
29539         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29540         Use aarch64_add_offset_1.
29541         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29542         an rtx rather than an int.  Take the delta as a poly_int64
29543         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29544         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29545         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29546         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29547         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29548         and aarch64_add_sp.
29549         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29550         aarch64_add_constant.
29552 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29554         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29555         Use scalar_float_mode.
29557 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29559         * config/aarch64/aarch64-simd.md
29560         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29561         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29562         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29563         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29564         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29565         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29566         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29567         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29568         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29569         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29571 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29573         PR target/83514
29574         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29575         targ_options->x_arm_arch_string is non NULL.
29577 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
29579         * config/aarch64/aarch64.h
29580         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
29582 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
29584         PR target/82096
29585         * expmed.c (emit_store_flag_force): Swap if const op0
29586         and change VOIDmode to mode of op0.
29588 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29590         PR rtl-optimization/83761
29591         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
29592         than bytes to mode_for_size.
29594 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29596         PR middle-end/83189
29597         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
29598         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
29599         profile.
29601 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29603         PR middle-end/83575
29604         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
29605         when in layout mode.
29606         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
29607         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
29608         partition fixup.
29610 2018-01-10  Michael Collison  <michael.collison@arm.com>
29612         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
29613         * config/aarch64/aarch64-option-extension.def: Add
29614         AARCH64_OPT_EXTENSION of 'fp16fml'.
29615         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29616         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
29617         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
29618         * config/aarch64/constraints.md (Ui7): New constraint.
29619         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
29620         (VFMLA_SEL_W): Ditto.
29621         (f16quad): Ditto.
29622         (f16mac1): Ditto.
29623         (VFMLA16_LOW): New int iterator.
29624         (VFMLA16_HIGH): Ditto.
29625         (UNSPEC_FMLAL): New unspec.
29626         (UNSPEC_FMLSL): Ditto.
29627         (UNSPEC_FMLAL2): Ditto.
29628         (UNSPEC_FMLSL2): Ditto.
29629         (f16mac): New code attribute.
29630         * config/aarch64/aarch64-simd-builtins.def
29631         (aarch64_fmlal_lowv2sf): Ditto.
29632         (aarch64_fmlsl_lowv2sf): Ditto.
29633         (aarch64_fmlalq_lowv4sf): Ditto.
29634         (aarch64_fmlslq_lowv4sf): Ditto.
29635         (aarch64_fmlal_highv2sf): Ditto.
29636         (aarch64_fmlsl_highv2sf): Ditto.
29637         (aarch64_fmlalq_highv4sf): Ditto.
29638         (aarch64_fmlslq_highv4sf): Ditto.
29639         (aarch64_fmlal_lane_lowv2sf): Ditto.
29640         (aarch64_fmlsl_lane_lowv2sf): Ditto.
29641         (aarch64_fmlal_laneq_lowv2sf): Ditto.
29642         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
29643         (aarch64_fmlalq_lane_lowv4sf): Ditto.
29644         (aarch64_fmlsl_lane_lowv4sf): Ditto.
29645         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
29646         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
29647         (aarch64_fmlal_lane_highv2sf): Ditto.
29648         (aarch64_fmlsl_lane_highv2sf): Ditto.
29649         (aarch64_fmlal_laneq_highv2sf): Ditto.
29650         (aarch64_fmlsl_laneq_highv2sf): Ditto.
29651         (aarch64_fmlalq_lane_highv4sf): Ditto.
29652         (aarch64_fmlsl_lane_highv4sf): Ditto.
29653         (aarch64_fmlalq_laneq_highv4sf): Ditto.
29654         (aarch64_fmlsl_laneq_highv4sf): Ditto.
29655         * config/aarch64/aarch64-simd.md:
29656         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
29657         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29658         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
29659         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29660         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
29661         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
29662         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
29663         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
29664         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
29665         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
29666         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
29667         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
29668         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
29669         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
29670         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
29671         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
29672         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
29673         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
29674         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
29675         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
29676         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
29677         (vfmlsl_low_u32): Ditto.
29678         (vfmlalq_low_u32): Ditto.
29679         (vfmlslq_low_u32): Ditto.
29680         (vfmlal_high_u32): Ditto.
29681         (vfmlsl_high_u32): Ditto.
29682         (vfmlalq_high_u32): Ditto.
29683         (vfmlslq_high_u32): Ditto.
29684         (vfmlal_lane_low_u32): Ditto.
29685         (vfmlsl_lane_low_u32): Ditto.
29686         (vfmlal_laneq_low_u32): Ditto.
29687         (vfmlsl_laneq_low_u32): Ditto.
29688         (vfmlalq_lane_low_u32): Ditto.
29689         (vfmlslq_lane_low_u32): Ditto.
29690         (vfmlalq_laneq_low_u32): Ditto.
29691         (vfmlslq_laneq_low_u32): Ditto.
29692         (vfmlal_lane_high_u32): Ditto.
29693         (vfmlsl_lane_high_u32): Ditto.
29694         (vfmlal_laneq_high_u32): Ditto.
29695         (vfmlsl_laneq_high_u32): Ditto.
29696         (vfmlalq_lane_high_u32): Ditto.
29697         (vfmlslq_lane_high_u32): Ditto.
29698         (vfmlalq_laneq_high_u32): Ditto.
29699         (vfmlslq_laneq_high_u32): Ditto.
29700         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
29701         (AARCH64_FL_FOR_ARCH8_4): New.
29702         (AARCH64_ISA_F16FML): New ISA flag.
29703         (TARGET_F16FML): New feature flag for fp16fml.
29704         (doc/invoke.texi): Document new fp16fml option.
29706 2018-01-10  Michael Collison  <michael.collison@arm.com>
29708         * config/aarch64/aarch64-builtins.c:
29709         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
29710         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29711         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
29712         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
29713         (AARCH64_ISA_SHA3): New ISA flag.
29714         (TARGET_SHA3): New feature flag for sha3.
29715         * config/aarch64/iterators.md (sha512_op): New int attribute.
29716         (CRYPTO_SHA512): New int iterator.
29717         (UNSPEC_SHA512H): New unspec.
29718         (UNSPEC_SHA512H2): Ditto.
29719         (UNSPEC_SHA512SU0): Ditto.
29720         (UNSPEC_SHA512SU1): Ditto.
29721         * config/aarch64/aarch64-simd-builtins.def
29722         (aarch64_crypto_sha512hqv2di): New builtin.
29723         (aarch64_crypto_sha512h2qv2di): Ditto.
29724         (aarch64_crypto_sha512su0qv2di): Ditto.
29725         (aarch64_crypto_sha512su1qv2di): Ditto.
29726         (aarch64_eor3qv8hi): Ditto.
29727         (aarch64_rax1qv2di): Ditto.
29728         (aarch64_xarqv2di): Ditto.
29729         (aarch64_bcaxqv8hi): Ditto.
29730         * config/aarch64/aarch64-simd.md:
29731         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
29732         (aarch64_crypto_sha512su0qv2di): Ditto.
29733         (aarch64_crypto_sha512su1qv2di): Ditto.
29734         (aarch64_eor3qv8hi): Ditto.
29735         (aarch64_rax1qv2di): Ditto.
29736         (aarch64_xarqv2di): Ditto.
29737         (aarch64_bcaxqv8hi): Ditto.
29738         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
29739         (vsha512h2q_u64): Ditto.
29740         (vsha512su0q_u64): Ditto.
29741         (vsha512su1q_u64): Ditto.
29742         (veor3q_u16): Ditto.
29743         (vrax1q_u64): Ditto.
29744         (vxarq_u64): Ditto.
29745         (vbcaxq_u16): Ditto.
29746         * config/arm/types.md (crypto_sha512): New type attribute.
29747         (crypto_sha3): Ditto.
29748         (doc/invoke.texi): Document new sha3 option.
29750 2018-01-10  Michael Collison  <michael.collison@arm.com>
29752         * config/aarch64/aarch64-builtins.c:
29753         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
29754         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29755         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
29756         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
29757         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
29758         (AARCH64_ISA_SM4): New ISA flag.
29759         (TARGET_SM4): New feature flag for sm4.
29760         * config/aarch64/aarch64-simd-builtins.def
29761         (aarch64_sm3ss1qv4si): Ditto.
29762         (aarch64_sm3tt1aq4si): Ditto.
29763         (aarch64_sm3tt1bq4si): Ditto.
29764         (aarch64_sm3tt2aq4si): Ditto.
29765         (aarch64_sm3tt2bq4si): Ditto.
29766         (aarch64_sm3partw1qv4si): Ditto.
29767         (aarch64_sm3partw2qv4si): Ditto.
29768         (aarch64_sm4eqv4si): Ditto.
29769         (aarch64_sm4ekeyqv4si): Ditto.
29770         * config/aarch64/aarch64-simd.md:
29771         (aarch64_sm3ss1qv4si): Ditto.
29772         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
29773         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
29774         (aarch64_sm4eqv4si): Ditto.
29775         (aarch64_sm4ekeyqv4si): Ditto.
29776         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
29777         (sm3part_op): Ditto.
29778         (CRYPTO_SM3TT): Ditto.
29779         (CRYPTO_SM3PART): Ditto.
29780         (UNSPEC_SM3SS1): New unspec.
29781         (UNSPEC_SM3TT1A): Ditto.
29782         (UNSPEC_SM3TT1B): Ditto.
29783         (UNSPEC_SM3TT2A): Ditto.
29784         (UNSPEC_SM3TT2B): Ditto.
29785         (UNSPEC_SM3PARTW1): Ditto.
29786         (UNSPEC_SM3PARTW2): Ditto.
29787         (UNSPEC_SM4E): Ditto.
29788         (UNSPEC_SM4EKEY): Ditto.
29789         * config/aarch64/constraints.md (Ui2): New constraint.
29790         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
29791         * config/arm/types.md (crypto_sm3): New type attribute.
29792         (crypto_sm4): Ditto.
29793         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
29794         (vsm3tt1aq_u32): Ditto.
29795         (vsm3tt1bq_u32): Ditto.
29796         (vsm3tt2aq_u32): Ditto.
29797         (vsm3tt2bq_u32): Ditto.
29798         (vsm3partw1q_u32): Ditto.
29799         (vsm3partw2q_u32): Ditto.
29800         (vsm4eq_u32): Ditto.
29801         (vsm4ekeyq_u32): Ditto.
29802         (doc/invoke.texi): Document new sm4 option.
29804 2018-01-10  Michael Collison  <michael.collison@arm.com>
29806         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
29807         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
29808         (AARCH64_FL_FOR_ARCH8_4): New.
29809         (AARCH64_FL_V8_4): New flag.
29810         (doc/invoke.texi): Document new armv8.4-a option.
29812 2018-01-10  Michael Collison  <michael.collison@arm.com>
29814         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29815         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
29816         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
29817         * config/aarch64/aarch64-option-extension.def: Add
29818         AARCH64_OPT_EXTENSION of 'sha2'.
29819         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
29820         (crypto): Disable sha2 and aes if crypto disabled.
29821         (crypto): Enable aes and sha2 if enabled.
29822         (simd): Disable sha2 and aes if simd disabled.
29823         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
29824         New flags.
29825         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
29826         (TARGET_SHA2): New feature flag for sha2.
29827         (TARGET_AES): New feature flag for aes.
29828         * config/aarch64/aarch64-simd.md:
29829         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
29830         conditional on TARGET_AES.
29831         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
29832         (aarch64_crypto_sha1hsi): Make pattern conditional
29833         on TARGET_SHA2.
29834         (aarch64_crypto_sha1hv4si): Ditto.
29835         (aarch64_be_crypto_sha1hv4si): Ditto.
29836         (aarch64_crypto_sha1su1v4si): Ditto.
29837         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
29838         (aarch64_crypto_sha1su0v4si): Ditto.
29839         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
29840         (aarch64_crypto_sha256su0v4si): Ditto.
29841         (aarch64_crypto_sha256su1v4si): Ditto.
29842         (doc/invoke.texi): Document new aes and sha2 options.
29844 2018-01-10  Martin Sebor  <msebor@redhat.com>
29846         PR tree-optimization/83781
29847         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
29848         as string arrays.
29850 2018-01-11  Martin Sebor  <msebor@gmail.com>
29851             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29853         PR tree-optimization/83501
29854         PR tree-optimization/81703
29856         * tree-ssa-strlen.c (get_string_cst): Rename...
29857         (get_string_len): ...to this.  Handle global constants.
29858         (handle_char_store): Adjust.
29860 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
29861             Jim Wilson  <jimw@sifive.com>
29863         * config/riscv/riscv-protos.h (riscv_output_return): New.
29864         * config/riscv/riscv.c (struct machine_function): New naked_p field.
29865         (riscv_attribute_table, riscv_output_return),
29866         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
29867         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
29868         (riscv_compute_frame_info): Only compute frame->mask if not a naked
29869         function.
29870         (riscv_expand_prologue): Add early return for naked function.
29871         (riscv_expand_epilogue): Likewise.
29872         (riscv_function_ok_for_sibcall): Return false for naked function.
29873         (riscv_set_current_function): New.
29874         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
29875         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
29876         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
29877         * doc/extend.texi (RISC-V Function Attributes): New.
29879 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
29881         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
29882         check for 128-bit long double before checking TCmode.
29883         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
29884         128-bit long doubles before checking TFmode or TCmode.
29885         (FLOAT128_IBM_P): Likewise.
29887 2018-01-10  Martin Sebor  <msebor@redhat.com>
29889         PR tree-optimization/83671
29890         * builtins.c (c_strlen): Unconditionally return zero for the empty
29891         string.
29892         Use -Warray-bounds for warnings.
29893         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
29894         for non-constant array indices with COMPONENT_REF, arrays of
29895         arrays, and pointers to arrays.
29896         (gimple_fold_builtin_strlen): Determine and set length range for
29897         non-constant character arrays.
29899 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
29901         PR middle-end/81897
29902         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
29903         empty blocks.
29905 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
29907         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
29909 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29911         PR target/83399
29912         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
29913         VECTOR_MEM_ALTIVEC_OR_VSX_P.
29914         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
29915         indexed_or_indirect_operand predicate.
29916         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
29917         (*vsx_le_perm_load_v8hi): Likewise.
29918         (*vsx_le_perm_load_v16qi): Likewise.
29919         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
29920         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
29921         (*vsx_le_perm_store_v8hi): Likewise.
29922         (*vsx_le_perm_store_v16qi): Likewise.
29923         (eight unnamed splitters): Likewise.
29925 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29927         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
29928         * config/rs6000/emmintrin.h: Likewise.
29929         * config/rs6000/mmintrin.h: Likewise.
29930         * config/rs6000/xmmintrin.h: Likewise.
29932 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
29934         PR c++/43486
29935         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
29936         "public_flag".
29937         * tree.c (tree_nop_conversion): Return true for location wrapper
29938         nodes.
29939         (maybe_wrap_with_location): New function.
29940         (selftest::check_strip_nops): New function.
29941         (selftest::test_location_wrappers): New function.
29942         (selftest::tree_c_tests): Call it.
29943         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
29944         (maybe_wrap_with_location): New decl.
29945         (EXPR_LOCATION_WRAPPER_P): New macro.
29946         (location_wrapper_p): New inline function.
29947         (tree_strip_any_location_wrapper): New inline function.
29949 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
29951         PR target/83735
29952         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
29953         stack_realign_offset for the largest alignment of stack slot
29954         actually used.
29955         (ix86_find_max_used_stack_alignment): New function.
29956         (ix86_finalize_stack_frame_flags): Use it.  Set
29957         max_used_stack_alignment if we don't realign stack.
29958         * config/i386/i386.h (machine_function): Add
29959         max_used_stack_alignment.
29961 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
29963         * config/arm/arm.opt (-mbranch-cost): New option.
29964         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
29965         account.
29967 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
29969         PR target/83629
29970         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
29971         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
29973 2018-01-10  Richard Biener  <rguenther@suse.de>
29975         PR debug/83765
29976         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
29977         early out so it also covers the case where we have a non-NULL
29978         origin.
29980 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29982         PR tree-optimization/83753
29983         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
29984         for non-strided grouped accesses if the number of elements is 1.
29986 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29988         PR target/81616
29989         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
29990         * i386.h (TARGET_USE_GATHER): Define.
29991         * x86-tune.def (X86_TUNE_USE_GATHER): New.
29993 2018-01-10  Martin Liska  <mliska@suse.cz>
29995         PR bootstrap/82831
29996         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
29997         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
29998         partitioning.
29999         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30000         CLEANUP_NO_PARTITIONING is not set.
30002 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30004         * doc/rtl.texi: Remove documentation of (const ...) wrappers
30005         for vectors, as a partial revert of r254296.
30006         * rtl.h (const_vec_p): Delete.
30007         (const_vec_duplicate_p): Don't test for vector CONSTs.
30008         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30009         * expmed.c (make_tree): Likewise.
30011         Revert:
30012         * common.md (E, F): Use CONSTANT_P instead of checking for
30013         CONST_VECTOR.
30014         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30015         checking for CONST_VECTOR.
30017 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30019         PR middle-end/83575
30020         * predict.c (force_edge_cold): Handle in more sane way edges
30021         with no prediction.
30023 2018-01-09  Carl Love  <cel@us.ibm.com>
30025         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30026         V4SI, V4SF types.
30027         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30028         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30029         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30030         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
30031         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30032         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
30033         * config/rs6000/rs6000-protos.h: Add extern defition for
30034         rs6000_generate_float2_double_code.
30035         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30036         function.
30037         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30038         (float2_v2df): Add define_expand.
30040 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
30042         PR target/83628
30043         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30044         op_mode in the force_to_mode call.
30046 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30048         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30049         instead of checking each element individually.
30050         (aarch64_evpc_uzp): Likewise.
30051         (aarch64_evpc_zip): Likewise.
30052         (aarch64_evpc_ext): Likewise.
30053         (aarch64_evpc_rev): Likewise.
30054         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30055         instead of checking each element individually.  Return true without
30056         generating rtl if
30057         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30058         whether all selected elements come from the same input, instead of
30059         checking each element individually.  Remove calls to gen_rtx_REG,
30060         start_sequence and end_sequence and instead assert that no rtl is
30061         generated.
30063 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30065         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30066         order of HIGH and CONST checks.
30068 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30070         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30071         if the destination isn't an SSA_NAME.
30073 2018-01-09  Richard Biener  <rguenther@suse.de>
30075         PR tree-optimization/83668
30076         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30077         move prologue...
30078         (canonicalize_loop_form): ... here, renamed from ...
30079         (canonicalize_loop_closed_ssa_form): ... this and amended to
30080         swap successor edges for loop exit blocks to make us use
30081         the RPO order we need for initial schedule generation.
30083 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30085         PR tree-optimization/64811
30086         * match.pd: When optimizing comparisons with Inf, avoid
30087         introducing or losing exceptions from comparisons with NaN.
30089 2018-01-09  Martin Liska  <mliska@suse.cz>
30091         PR sanitizer/82517
30092         * asan.c (shadow_mem_size): Add gcc_assert.
30094 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30096         Don't save registers in main().
30098         PR target/83738
30099         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30100         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30101         * config/avr/avr.c (avr_set_current_function): Don't error if
30102         naked, OS_task or OS_main are specified at the same time.
30103         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30104         OS_main.
30105         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30106         attribute.
30107         * common/config/avr/avr-common.c (avr_option_optimization_table):
30108         Switch on -mmain-is-OS_task for optimizing compilations.
30110 2018-01-09  Richard Biener  <rguenther@suse.de>
30112         PR tree-optimization/83572
30113         * graphite.c: Include cfganal.h.
30114         (graphite_transform_loops): Connect infinite loops to exit
30115         and remove fake edges at the end.
30117 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30119         * ipa-inline.c (edge_badness): Revert accidental checkin.
30121 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30123         PR ipa/80763
30124         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30125         symbols; not inline clones.
30127 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30129         PR target/83507
30130         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30131         hard registers.  Formatting fixes.
30133         PR preprocessor/83722
30134         * gcc.c (try_generate_repro): Pass
30135         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30136         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30137         do_report_bug.
30139 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30140             Kito Cheng  <kito.cheng@gmail.com>
30142         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30143         (riscv_leaf_function_p): Delete.
30144         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30146 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30148         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30149         function.
30150         (do_ifelse): New function.
30151         (do_isel): New function.
30152         (do_sub3): New function.
30153         (do_add3): New function.
30154         (do_load_mask_compare): New function.
30155         (do_overlap_load_compare): New function.
30156         (expand_compare_loop): New function.
30157         (expand_block_compare): Call expand_compare_loop() when appropriate.
30158         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30159         option description.
30160         (-mblock-compare-inline-loop-limit): New option.
30162 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30164         PR target/83677
30165         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30166         Reverse order of second and third operands in first alternative.
30167         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30168         of first and second elements in UNSPEC_VPERMR vector.
30169         (altivec_expand_vec_perm_le): Likewise.
30171 2018-01-08  Jeff Law  <law@redhat.com>
30173         PR rtl-optimizatin/81308
30174         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30175         (process_switch): If group_case_labels makes a change, then set
30176         cfg_altered.
30177         (pass_convert_switch::execute): If a switch is converted, then
30178         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30180         PR rtl-optimization/81308
30181         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30182         splitting insns.
30184 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30186         PR target/83663 - Revert r255946
30187         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30188         generation for cases where splatting a value is not useful.
30189         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30190         across a vec_duplicate and a paradoxical subreg forming a vector
30191         mode to a vec_concat.
30193 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30195         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30196         -march=armv8.3-a variants.
30197         * config/arm/t-multilib: Likewise.
30198         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30200 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30202         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30203         to generate rtl.
30204         (cceq_ior_compare_complement): Give it a name so I can use it, and
30205         change boolean_or_operator predicate to boolean_operator so it can
30206         be used to generate a crand.
30207         (eqne): New code iterator.
30208         (bd/bd_neg): New code_attrs.
30209         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30210         a single define_insn.
30211         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30212         decrement (bdnzt/bdnzf/bdzt/bdzf).
30213         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30214         with the new names of the branch decrement patterns, and added the
30215         names of the branch decrement conditional patterns.
30217 2018-01-08  Richard Biener  <rguenther@suse.de>
30219         PR tree-optimization/83563
30220         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30221         cache.
30223 2018-01-08  Richard Biener  <rguenther@suse.de>
30225         PR middle-end/83713
30226         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30228 2018-01-08  Richard Biener  <rguenther@suse.de>
30230         PR tree-optimization/83685
30231         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30232         references to abnormals.
30234 2018-01-08  Richard Biener  <rguenther@suse.de>
30236         PR lto/83719
30237         * dwarf2out.c (output_indirect_strings): Handle empty
30238         skeleton_debug_str_hash.
30239         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30241 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30243         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30244         (emit_store_direct): Likewise.
30245         (arc_trampoline_adjust_address): Likewise.
30246         (arc_asm_trampoline_template): New function.
30247         (arc_initialize_trampoline): Use asm_trampoline_template.
30248         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30249         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30250         * config/arc/arc.md (flush_icache): Delete pattern.
30252 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30254         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30255         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30256         munaligned-access.
30258 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30260         PR target/83681
30261         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30262         by not USED_FOR_TARGET.
30263         (make_pass_resolve_sw_modes): Likewise.
30265 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30267         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30268         USED_FOR_TARGET.
30270 2018-01-08  Richard Biener  <rguenther@suse.de>
30272         PR middle-end/83580
30273         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30275 2018-01-08  Richard Biener  <rguenther@suse.de>
30277         PR middle-end/83517
30278         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30280 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30282         PR middle-end/81897
30283         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30284         basic blocks with a small number of successors.
30285         (convert_control_dep_chain_into_preds): Improve handling of
30286         forwarder blocks.
30287         (dump_predicates): Split apart into...
30288         (dump_pred_chain): ...here...
30289         (dump_pred_info): ...and here.
30290         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30291         (can_chain_union_be_invalidated_p): Improve check for invalidation
30292         of paths.
30293         (uninit_uses_cannot_happen): Avoid unnecessary if
30294         convert_control_dep_chain_into_preds yielded nothing.
30296 2018-01-06  Martin Sebor  <msebor@redhat.com>
30298         PR tree-optimization/83640
30299         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30300         subtracting negative offset from size.
30301         (builtin_access::overlap): Adjust offset bounds of the access to fall
30302         within the size of the object if possible.
30304 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30306         PR rtl-optimization/83699
30307         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30308         extract_bit_field_as_subreg to cases in which the extracted
30309         value is also a vector.
30311         * lra-constraints.c (process_alt_operands): Test for the equivalence
30312         substitutions when detecting a possible reload cycle.
30314 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30316         PR debug/83480
30317         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30318         by default if flag_selective_schedling{,2}.  Formatting fixes.
30320         PR rtl-optimization/83682
30321         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30322         if it has non-VECTOR_MODE element mode.
30323         (vec_duplicate_p): Likewise.
30325         PR middle-end/83694
30326         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30327         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30329 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30331         PR target/83604
30332         * config/i386/i386-builtin.def
30333         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30334         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30335         Require also OPTION_MASK_ISA_AVX512F in addition to
30336         OPTION_MASK_ISA_GFNI.
30337         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30338         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30339         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30340         to OPTION_MASK_ISA_GFNI.
30341         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30342         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30343         OPTION_MASK_ISA_AVX512BW.
30344         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30345         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30346         addition to OPTION_MASK_ISA_GFNI.
30347         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30348         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30349         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30350         to OPTION_MASK_ISA_GFNI.
30351         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30352         a requirement for all ISAs rather than any of them with a few
30353         exceptions.
30354         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30355         processing.
30356         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30357         bitmasks to be enabled with 3 exceptions, instead of requiring any
30358         enabled ISA with lots of exceptions.
30359         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30360         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30361         Change avx512bw in isa attribute to avx512f.
30362         * config/i386/sgxintrin.h: Add license boilerplate.
30363         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30364         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30365         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30366         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30367         defined.
30368         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30369         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30370         temporarily sse2 rather than sse if not enabled already.
30372         PR target/83604
30373         * config/i386/sse.md (VI248_VLBW): Rename to ...
30374         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30375         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30376         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30377         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30378         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30379         mode iterator instead of VI248_VLBW.
30381 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30383         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30384         (record_modified): Skip clobbers; add debug output.
30385         (param_change_prob): Use sreal frequencies.
30387 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30389         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30390         punt for user-aligned variables.
30392 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30394         * tree-chrec.c (chrec_contains_symbols): Return true for
30395         POLY_INT_CST.
30397 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30399         PR target/82439
30400         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30401         of (x|y) == x for BICS pattern.
30403 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30405         PR tree-optimization/83605
30406         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30407         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30408         can throw.
30410 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30412         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30413         * config/epiphany/rtems.h: New file.
30415 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30416             Uros Bizjak  <ubizjak@gmail.com>
30418         PR target/83554
30419         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30420         QIreg_operand instead of register_operand predicate.
30421         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30422         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30423         comments instead of -fmitigate[-_]rop.
30425 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30427         PR bootstrap/81926
30428         * cgraphunit.c (symbol_table::compile): Switch to text_section
30429         before calling assembly_start debug hook.
30430         * run-rtl-passes.c (run_rtl_passes): Likewise.
30431         Include output.h.
30433 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30435         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30436         range_int_cst_p rather than !symbolic_range_p before calling
30437         extract_range_from_multiplicative_op_1.
30439 2018-01-04  Jeff Law  <law@redhat.com>
30441         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30442         redundant test in assertion.
30444 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30446         * doc/rtl.texi: Document machine_mode wrapper classes.
30448 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30450         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30451         using tree_to_uhwi.
30453 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30455         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30456         the VEC_PERM_EXPR fold to fail.
30458 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30460         PR debug/83585
30461         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30462         to switched_sections.
30464 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30466         PR target/83680
30467         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30468         test for d.testing.
30470 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30472         PR target/83387
30473         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30474         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30476 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30478         PR debug/83666
30479         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30480         is BLKmode and bitpos not zero or mode change is needed.
30482 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30484         PR target/83675
30485         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30486         TARGET_VIS2.
30488 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30490         PR target/83628
30491         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30492         instead of MULT rtx.  Update all corresponding splitters.
30493         (*saddl_se): Ditto.
30494         (*ssub<modesuffix>): Ditto.
30495         (*ssubl_se): Ditto.
30496         (*cmp_sadd_di): Update split patterns.
30497         (*cmp_sadd_si): Ditto.
30498         (*cmp_sadd_sidi): Ditto.
30499         (*cmp_ssub_di): Ditto.
30500         (*cmp_ssub_si): Ditto.
30501         (*cmp_ssub_sidi): Ditto.
30502         * config/alpha/predicates.md (const23_operand): New predicate.
30503         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30504         Look for ASHIFT, not MULT inner operand.
30505         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30507 2018-01-04  Martin Liska  <mliska@suse.cz>
30509         PR gcov-profile/83669
30510         * gcov.c (output_intermediate_file): Add version to intermediate
30511         gcov file.
30512         * doc/gcov.texi: Document new field 'version' in intermediate
30513         file format. Fix location of '-k' option of gcov command.
30515 2018-01-04  Martin Liska  <mliska@suse.cz>
30517         PR ipa/82352
30518         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30520 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30522         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30524 2018-01-03  Martin Sebor  <msebor@redhat.com>
30526         PR tree-optimization/83655
30527         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30528         checking calls with invalid arguments.
30530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30532         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30533         (vectorizable_mask_load_store): Delete.
30534         (vectorizable_call): Return false for masked loads and stores.
30535         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30536         instead of gimple_assign_rhs1.
30537         (vectorizable_load): Handle IFN_MASK_LOAD.
30538         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30540 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30542         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30543         split out from..,
30544         (vectorizable_mask_load_store): ...here.
30545         (vectorizable_load): ...and here.
30547 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30549         * tree-vect-stmts.c (vect_build_all_ones_mask)
30550         (vect_build_zero_merge_argument): New functions, split out from...
30551         (vectorizable_load): ...here.
30553 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30555         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30556         split out from...
30557         (vectorizable_mask_load_store): ...here.
30558         (vectorizable_store): ...and here.
30560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30562         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30563         split out from...
30564         (vectorizable_mask_load_store): ...here.
30566 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30568         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30569         (vect_model_store_cost): Take a vec_load_store_type instead of a
30570         vect_def_type.
30571         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30572         (vect_model_store_cost): Take a vec_load_store_type instead of a
30573         vect_def_type.
30574         (vectorizable_mask_load_store): Update accordingly.
30575         (vectorizable_store): Likewise.
30576         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
30578 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30580         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
30581         IFN_MASK_LOAD calls here rather than...
30582         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
30584 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30585             Alan Hayward  <alan.hayward@arm.com>
30586             David Sherwood  <david.sherwood@arm.com>
30588         * expmed.c (extract_bit_field_1): For vector extracts,
30589         fall back to extract_bit_field_as_subreg if vec_extract
30590         isn't available.
30592 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30593             Alan Hayward  <alan.hayward@arm.com>
30594             David Sherwood  <david.sherwood@arm.com>
30596         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
30597         they are variable or constant sized.
30598         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
30599         slots for constant-sized data.
30601 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30602             Alan Hayward  <alan.hayward@arm.com>
30603             David Sherwood  <david.sherwood@arm.com>
30605         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
30606         handling COND_EXPRs with boolean comparisons, try to find a better
30607         basis for the mask type than the boolean itself.
30609 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30611         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
30612         is calculated and how it can be overridden.
30613         * genmodes.c (max_bitsize_mode_any_mode): New variable.
30614         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
30615         if defined.
30616         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
30617         if nonzero.
30619 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30620             Alan Hayward  <alan.hayward@arm.com>
30621             David Sherwood  <david.sherwood@arm.com>
30623         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
30624         Remove the mode argument.
30625         (aarch64_simd_valid_immediate): Remove the mode and inverse
30626         arguments.
30627         * config/aarch64/iterators.md (bitsize): New iterator.
30628         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
30629         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
30630         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
30631         aarch64_simd_valid_immediate.
30632         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
30633         (aarch64_reg_or_bic_imm): Likewise.
30634         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
30635         with an insn_type enum and msl with a modifier_type enum.
30636         Replace element_width with a scalar_mode.  Change the shift
30637         to unsigned int.  Add constructors for scalar_float_mode and
30638         scalar_int_mode elements.
30639         (aarch64_vect_float_const_representable_p): Delete.
30640         (aarch64_can_const_movi_rtx_p)
30641         (aarch64_simd_scalar_immediate_valid_for_move)
30642         (aarch64_simd_make_constant): Update call to
30643         aarch64_simd_valid_immediate.
30644         (aarch64_advsimd_valid_immediate_hs): New function.
30645         (aarch64_advsimd_valid_immediate): Likewise.
30646         (aarch64_simd_valid_immediate): Remove mode and inverse
30647         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
30648         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
30649         and aarch64_float_const_representable_p on the result.
30650         (aarch64_output_simd_mov_immediate): Remove mode argument.
30651         Update call to aarch64_simd_valid_immediate and use of
30652         simd_immediate_info.
30653         (aarch64_output_scalar_simd_mov_immediate): Update call
30654         accordingly.
30656 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30657             Alan Hayward  <alan.hayward@arm.com>
30658             David Sherwood  <david.sherwood@arm.com>
30660         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
30661         (mode_nunits): Likewise CONST_MODE_NUNITS.
30662         * machmode.def (ADJUST_NUNITS): Document.
30663         * genmodes.c (mode_data::need_nunits_adj): New field.
30664         (blank_mode): Update accordingly.
30665         (adj_nunits): New variable.
30666         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
30667         parameter.
30668         (emit_mode_size_inline): Set need_bytesize_adj for all modes
30669         listed in adj_nunits.
30670         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
30671         listed in adj_nunits.  Don't emit case statements for such modes.
30672         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
30673         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
30674         nothing if adj_nunits is nonnull.
30675         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
30676         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
30677         (emit_mode_fbit): Update use of print_maybe_const_decl.
30678         (emit_move_size): Likewise.  Treat the array as non-const
30679         if adj_nunits.
30680         (emit_mode_adjustments): Handle adj_nunits.
30682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30684         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
30685         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
30686         (VECTOR_MODES): Use it.
30687         (make_vector_modes): Take the prefix as an argument.
30689 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30690             Alan Hayward  <alan.hayward@arm.com>
30691             David Sherwood  <david.sherwood@arm.com>
30693         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
30694         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
30695         for MODE_VECTOR_BOOL.
30696         * machmode.def (VECTOR_BOOL_MODE): Document.
30697         * genmodes.c (VECTOR_BOOL_MODE): New macro.
30698         (make_vector_bool_mode): New function.
30699         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
30700         MODE_VECTOR_BOOL.
30701         * lto-streamer-in.c (lto_input_mode_table): Likewise.
30702         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
30703         Likewise.
30704         * stor-layout.c (int_mode_for_mode): Likewise.
30705         * tree.c (build_vector_type_for_mode): Likewise.
30706         * varasm.c (output_constant_pool_2): Likewise.
30707         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
30708         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
30709         for MODE_VECTOR_BOOL.
30710         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
30711         of mode class checks.
30712         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
30713         instead of a list of mode class checks.
30714         (expand_vector_scalar_condition): Likewise.
30715         (type_for_widest_vector_mode): Handle BImode as an inner mode.
30717 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30718             Alan Hayward  <alan.hayward@arm.com>
30719             David Sherwood  <david.sherwood@arm.com>
30721         * machmode.h (mode_size): Change from unsigned short to
30722         poly_uint16_pod.
30723         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
30724         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30725         or if measurement_type is not polynomial.
30726         (fixed_size_mode::includes_p): Check for constant-sized modes.
30727         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
30728         return a poly_uint16 rather than an unsigned short.
30729         (emit_mode_size): Change the type of mode_size from unsigned short
30730         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
30731         (emit_mode_adjustments): Cope with polynomial vector sizes.
30732         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30733         for GET_MODE_SIZE.
30734         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30735         for GET_MODE_SIZE.
30736         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
30737         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
30738         * caller-save.c (setup_save_areas): Likewise.
30739         (replace_reg_with_saved_mem): Likewise.
30740         * calls.c (emit_library_call_value_1): Likewise.
30741         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
30742         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
30743         (gen_lowpart_for_combine): Likewise.
30744         * convert.c (convert_to_integer_1): Likewise.
30745         * cse.c (equiv_constant, cse_insn): Likewise.
30746         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
30747         (cselib_subst_to_values): Likewise.
30748         * dce.c (word_dce_process_block): Likewise.
30749         * df-problems.c (df_word_lr_mark_ref): Likewise.
30750         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
30751         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
30752         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
30753         (rtl_for_decl_location): Likewise.
30754         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
30755         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
30756         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
30757         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
30758         (expand_expr_real_1): Likewise.
30759         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
30760         (pad_below): Likewise.
30761         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30762         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
30763         * ira.c (get_subreg_tracking_sizes): Likewise.
30764         * ira-build.c (ira_create_allocno_objects): Likewise.
30765         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
30766         (ira_sort_regnos_for_alter_reg): Likewise.
30767         * ira-costs.c (record_operand_costs): Likewise.
30768         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
30769         (resolve_simple_move): Likewise.
30770         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
30771         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
30772         (lra_constraints): Likewise.
30773         (CONST_POOL_OK_P): Reject variable-sized modes.
30774         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
30775         (add_pseudo_to_slot, lra_spill): Likewise.
30776         * omp-low.c (omp_clause_aligned_alignment): Likewise.
30777         * optabs-query.c (get_best_extraction_insn): Likewise.
30778         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30779         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
30780         (expand_mult_highpart, valid_multiword_target_p): Likewise.
30781         * recog.c (offsettable_address_addr_space_p): Likewise.
30782         * regcprop.c (maybe_mode_change): Likewise.
30783         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
30784         * regrename.c (build_def_use): Likewise.
30785         * regstat.c (dump_reg_info): Likewise.
30786         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
30787         (find_reloads, find_reloads_subreg_address): Likewise.
30788         * reload1.c (eliminate_regs_1): Likewise.
30789         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
30790         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
30791         (simplify_binary_operation_1, simplify_subreg): Likewise.
30792         * targhooks.c (default_function_arg_padding): Likewise.
30793         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
30794         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
30795         (verify_gimple_assign_ternary): Likewise.
30796         * tree-inline.c (estimate_move_cost): Likewise.
30797         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30798         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
30799         (get_address_cost_ainc): Likewise.
30800         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
30801         (vect_supportable_dr_alignment): Likewise.
30802         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30803         (vectorizable_reduction): Likewise.
30804         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
30805         (vectorizable_operation, vectorizable_load): Likewise.
30806         * tree.c (build_same_sized_truth_vector_type): Likewise.
30807         * valtrack.c (cleanup_auto_inc_dec): Likewise.
30808         * var-tracking.c (emit_note_insn_var_location): Likewise.
30809         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
30810         (ADDR_VEC_ALIGN): Likewise.
30812 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30813             Alan Hayward  <alan.hayward@arm.com>
30814             David Sherwood  <david.sherwood@arm.com>
30816         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
30817         unsigned short.
30818         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30819         or if measurement_type is polynomial.
30820         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
30821         * combine.c (make_extraction): Likewise.
30822         * dse.c (find_shift_sequence): Likewise.
30823         * dwarf2out.c (mem_loc_descriptor): Likewise.
30824         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
30825         (extract_bit_field, extract_low_bits): Likewise.
30826         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
30827         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
30828         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
30829         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
30830         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30831         * reload.c (find_reloads): Likewise.
30832         * reload1.c (alter_reg): Likewise.
30833         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
30834         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
30835         * tree-if-conv.c (predicate_mem_writes): Likewise.
30836         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
30837         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
30838         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
30839         * valtrack.c (dead_debug_insert_temp): Likewise.
30840         * varasm.c (mergeable_constant_section): Likewise.
30841         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
30843 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30844             Alan Hayward  <alan.hayward@arm.com>
30845             David Sherwood  <david.sherwood@arm.com>
30847         * expr.c (expand_assignment): Cope with polynomial mode sizes
30848         when assigning to a CONCAT.
30850 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30851             Alan Hayward  <alan.hayward@arm.com>
30852             David Sherwood  <david.sherwood@arm.com>
30854         * machmode.h (mode_precision): Change from unsigned short to
30855         poly_uint16_pod.
30856         (mode_to_precision): Return a poly_uint16 rather than an unsigned
30857         short.
30858         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
30859         or if measurement_type is not polynomial.
30860         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
30861         in which the mode is already known to be a scalar_int_mode.
30862         * genmodes.c (emit_mode_precision): Change the type of mode_precision
30863         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
30864         initializer.
30865         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30866         for GET_MODE_PRECISION.
30867         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30868         for GET_MODE_PRECISION.
30869         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
30870         as polynomial.
30871         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
30872         (expand_field_assignment, make_extraction): Likewise.
30873         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
30874         (get_last_value): Likewise.
30875         * convert.c (convert_to_integer_1): Likewise.
30876         * cse.c (cse_insn): Likewise.
30877         * expr.c (expand_expr_real_1): Likewise.
30878         * lra-constraints.c (simplify_operand_subreg): Likewise.
30879         * optabs-query.c (can_atomic_load_p): Likewise.
30880         * optabs.c (expand_atomic_load): Likewise.
30881         (expand_atomic_store): Likewise.
30882         * ree.c (combine_reaching_defs): Likewise.
30883         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
30884         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
30885         * tree.h (type_has_mode_precision_p): Likewise.
30886         * ubsan.c (instrument_si_overflow): Likewise.
30888 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30889             Alan Hayward  <alan.hayward@arm.com>
30890             David Sherwood  <david.sherwood@arm.com>
30892         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
30893         polynomial numbers of units.
30894         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
30895         (valid_vector_subparts_p): New function.
30896         (build_vector_type): Remove temporary shim and take the number
30897         of units as a poly_uint64 rather than an int.
30898         (build_opaque_vector_type): Take the number of units as a
30899         poly_uint64 rather than an int.
30900         * tree.c (build_vector_from_ctor): Handle polynomial
30901         TYPE_VECTOR_SUBPARTS.
30902         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
30903         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
30904         (build_vector_from_val): If the number of units is variable,
30905         use build_vec_duplicate_cst for constant operands and
30906         VEC_DUPLICATE_EXPR otherwise.
30907         (make_vector_type): Remove temporary is_constant ().
30908         (build_vector_type, build_opaque_vector_type): Take the number of
30909         units as a poly_uint64 rather than an int.
30910         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
30911         VECTOR_CST_NELTS.
30912         * cfgexpand.c (expand_debug_expr): Likewise.
30913         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
30914         (store_constructor, expand_expr_real_1): Likewise.
30915         (const_scalar_mask_from_tree): Likewise.
30916         * fold-const-call.c (fold_const_reduction): Likewise.
30917         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
30918         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
30919         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
30920         (fold_relational_const): Likewise.
30921         (native_interpret_vector): Likewise.  Change the size from an
30922         int to an unsigned int.
30923         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
30924         TYPE_VECTOR_SUBPARTS.
30925         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
30926         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
30927         duplicating a non-constant operand into a variable-length vector.
30928         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
30929         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
30930         * ipa-icf.c (sem_variable::equals): Likewise.
30931         * match.pd: Likewise.
30932         * omp-simd-clone.c (simd_clone_subparts): Likewise.
30933         * print-tree.c (print_node): Likewise.
30934         * stor-layout.c (layout_type): Likewise.
30935         * targhooks.c (default_builtin_vectorization_cost): Likewise.
30936         * tree-cfg.c (verify_gimple_comparison): Likewise.
30937         (verify_gimple_assign_binary): Likewise.
30938         (verify_gimple_assign_ternary): Likewise.
30939         (verify_gimple_assign_single): Likewise.
30940         * tree-pretty-print.c (dump_generic_node): Likewise.
30941         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30942         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
30943         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
30944         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
30945         (vect_shift_permute_load_chain): Likewise.
30946         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
30947         (expand_vector_condition, optimize_vector_constructor): Likewise.
30948         (lower_vec_perm, get_compute_type): Likewise.
30949         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30950         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
30951         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
30952         (vect_recog_mask_conversion_pattern): Likewise.
30953         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
30954         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
30955         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30956         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
30957         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
30958         (vectorizable_shift, vectorizable_operation, vectorizable_store)
30959         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
30960         (supportable_widening_operation): Likewise.
30961         (supportable_narrowing_operation): Likewise.
30962         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
30963         Likewise.
30964         * varasm.c (output_constant): Likewise.
30966 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30967             Alan Hayward  <alan.hayward@arm.com>
30968             David Sherwood  <david.sherwood@arm.com>
30970         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
30971         so that both the length == 3 and length != 3 cases set up their
30972         own permute vectors.  Add comments explaining why we know the
30973         number of elements is constant.
30974         (vect_permute_load_chain): Likewise.
30976 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30977             Alan Hayward  <alan.hayward@arm.com>
30978             David Sherwood  <david.sherwood@arm.com>
30980         * machmode.h (mode_nunits): Change from unsigned char to
30981         poly_uint16_pod.
30982         (ONLY_FIXED_SIZE_MODES): New macro.
30983         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
30984         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
30985         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
30986         New typedefs.
30987         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
30988         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
30989         or if measurement_type is not polynomial.
30990         * genmodes.c (ZERO_COEFFS): New macro.
30991         (emit_mode_nunits_inline): Make mode_nunits_inline return a
30992         poly_uint16.
30993         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
30994         Use ZERO_COEFFS when emitting initializers.
30995         * data-streamer.h (bp_pack_poly_value): New function.
30996         (bp_unpack_poly_value): Likewise.
30997         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30998         for GET_MODE_NUNITS.
30999         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31000         for GET_MODE_NUNITS.
31001         * tree.c (make_vector_type): Remove temporary shim and make
31002         the real function take the number of units as a poly_uint64
31003         rather than an int.
31004         (build_vector_type_for_mode): Handle polynomial nunits.
31005         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31006         * emit-rtl.c (const_vec_series_p_1): Likewise.
31007         (gen_rtx_CONST_VECTOR): Likewise.
31008         * fold-const.c (test_vec_duplicate_folding): Likewise.
31009         * genrecog.c (validate_pattern): Likewise.
31010         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31011         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31012         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31013         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31014         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31015         * rtlanal.c (subreg_get_info): Likewise.
31016         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31017         (vect_grouped_load_supported): Likewise.
31018         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31019         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31020         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31021         (simplify_const_unary_operation, simplify_binary_operation_1)
31022         (simplify_const_binary_operation, simplify_ternary_operation)
31023         (test_vector_ops_duplicate, test_vector_ops): Likewise.
31024         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31025         instead of CONST_VECTOR_NUNITS.
31026         * varasm.c (output_constant_pool_2): Likewise.
31027         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31028         explicit-encoded elements in the XVEC for variable-length vectors.
31030 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31032         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31034 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31035             Alan Hayward  <alan.hayward@arm.com>
31036             David Sherwood  <david.sherwood@arm.com>
31038         * coretypes.h (fixed_size_mode): Declare.
31039         (fixed_size_mode_pod): New typedef.
31040         * builtins.h (target_builtins::x_apply_args_mode)
31041         (target_builtins::x_apply_result_mode): Change type to
31042         fixed_size_mode_pod.
31043         * builtins.c (apply_args_size, apply_result_size, result_vector)
31044         (expand_builtin_apply_args_1, expand_builtin_apply)
31045         (expand_builtin_return): Update accordingly.
31047 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31049         * cse.c (hash_rtx_cb): Hash only the encoded elements.
31050         * cselib.c (cselib_hash_rtx): Likewise.
31051         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31052         CONST_VECTOR encoding.
31054 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31055             Jeff Law  <law@redhat.com>
31057         PR target/83641
31058         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31059         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31060         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31061         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31063         PR target/83641
31064         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31065         explicitly probe *sp in a noreturn function if there were any callee
31066         register saves or frame pointer is needed.
31068 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31070         PR debug/83621
31071         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31072         BLKmode for ternary, binary or unary expressions.
31074         PR debug/83645
31075         * var-tracking.c (delete_vta_debug_insn): New inline function.
31076         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31077         insns from get_insns () to NULL instead of each bb separately.
31078         Use delete_vta_debug_insn.  No longer static.
31079         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31080         delete_vta_debug_insns callers.
31081         * rtl.h (delete_vta_debug_insns): Declare.
31082         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31083         instead of variable_tracking_main.
31085 2018-01-03  Martin Sebor  <msebor@redhat.com>
31087         PR tree-optimization/83603
31088         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31089         arguments past the endof the argument list in functions declared
31090         without a prototype.
31091         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31092         Avoid checking when arguments are null.
31094 2018-01-03  Martin Sebor  <msebor@redhat.com>
31096         PR c/83559
31097         * doc/extend.texi (attribute const): Fix a typo.
31098         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31099         issuing -Wsuggest-attribute for void functions.
31101 2018-01-03  Martin Sebor  <msebor@redhat.com>
31103         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31104         offset_int::from instead of wide_int::to_shwi.
31105         (maybe_diag_overlap): Remove assertion.
31106         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31107         * gimple-ssa-sprintf.c (format_directive): Same.
31108         (parse_directive): Same.
31109         (sprintf_dom_walker::compute_format_length): Same.
31110         (try_substitute_return_value): Same.
31112 2018-01-03  Jeff Law  <law@redhat.com>
31114         PR middle-end/83654
31115         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31116         non-constant residual for zero at runtime and avoid probing in
31117         that case.  Reorganize code for trailing problem to mirror handling
31118         of the residual.
31120 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31122         PR tree-optimization/83501
31123         * tree-ssa-strlen.c (get_string_cst): New.
31124         (handle_char_store): Call get_string_cst.
31126 2018-01-03  Martin Liska  <mliska@suse.cz>
31128         PR tree-optimization/83593
31129         * tree-ssa-strlen.c: Include tree-cfg.h.
31130         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31131         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31132         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31133         to false.
31134         (strlen_dom_walker::before_dom_children): Call
31135         gimple_purge_dead_eh_edges. Dump tranformation with details
31136         dump flags.
31137         (strlen_dom_walker::before_dom_children): Update call by adding
31138         new argument cleanup_eh.
31139         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31141 2018-01-03  Martin Liska  <mliska@suse.cz>
31143         PR ipa/83549
31144         * cif-code.def (VARIADIC_THUNK): New enum value.
31145         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31146         thunks.
31148 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31150         * sse.md (mov<mode>_internal): Tighten condition for when to use
31151         vmovdqu<ssescalarsize> for TI and OI modes.
31153 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31155         Update copyright years.
31157 2018-01-03  Martin Liska  <mliska@suse.cz>
31159         PR ipa/83594
31160         * ipa-visibility.c (function_and_variable_visibility): Skip
31161         functions with noipa attribure.
31163 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31165         * gcc.c (process_command): Update copyright notice dates.
31166         * gcov-dump.c (print_version): Ditto.
31167         * gcov.c (print_version): Ditto.
31168         * gcov-tool.c (print_version): Ditto.
31169         * gengtype.c (create_file): Ditto.
31170         * doc/cpp.texi: Bump @copying's copyright year.
31171         * doc/cppinternals.texi: Ditto.
31172         * doc/gcc.texi: Ditto.
31173         * doc/gccint.texi: Ditto.
31174         * doc/gcov.texi: Ditto.
31175         * doc/install.texi: Ditto.
31176         * doc/invoke.texi: Ditto.
31178 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31180         * vector-builder.h (vector_builder::m_full_nelts): Change from
31181         unsigned int to poly_uint64.
31182         (vector_builder::full_nelts): Update prototype accordingly.
31183         (vector_builder::new_vector): Likewise.
31184         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31185         (vector_builder::operator ==): Likewise.
31186         (vector_builder::finalize): Likewise.
31187         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31188         Take the number of elements as a poly_uint64 rather than an
31189         unsigned int.
31190         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31191         from unsigned int to poly_uint64.
31192         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31193         (vec_perm_indices::new_vector): Likewise.
31194         (vec_perm_indices::length): Likewise.
31195         (vec_perm_indices::nelts_per_input): Likewise.
31196         (vec_perm_indices::input_nelts): Likewise.
31197         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31198         number of elements per input as a poly_uint64 rather than an
31199         unsigned int.  Use the original encoding for variable-length
31200         vectors, rather than clamping each individual element.
31201         For the second and subsequent elements in each pattern,
31202         clamp the step and base before clamping their sum.
31203         (vec_perm_indices::series_p): Handle polynomial element counts.
31204         (vec_perm_indices::all_in_range_p): Likewise.
31205         (vec_perm_indices_to_tree): Likewise.
31206         (vec_perm_indices_to_rtx): Likewise.
31207         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31208         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31209         (tree_vector_builder::new_binary_operation): Handle polynomial
31210         element counts.  Return false if we need to know the number
31211         of elements at compile time.
31212         * fold-const.c (fold_vec_perm): Punt if the number of elements
31213         isn't known at compile time.
31215 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31217         * vec-perm-indices.h (vec_perm_builder): Change element type
31218         from HOST_WIDE_INT to poly_int64.
31219         (vec_perm_indices::element_type): Update accordingly.
31220         (vec_perm_indices::clamp): Handle polynomial element_types.
31221         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31222         (vec_perm_indices::all_in_range_p): Likewise.
31223         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31224         than shwi trees.
31225         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31226         polynomial vec_perm_indices element types.
31227         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31228         * fold-const.c (fold_vec_perm): Likewise.
31229         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31230         * tree-vect-generic.c (lower_vec_perm): Likewise.
31231         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31232         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31233         element type to HOST_WIDE_INT.
31235 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31236             Alan Hayward  <alan.hayward@arm.com>
31237             David Sherwood  <david.sherwood@arm.com>
31239         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31240         rather than an int.  Use plus_constant.
31241         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31242         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31244 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31245             Alan Hayward  <alan.hayward@arm.com>
31246             David Sherwood  <david.sherwood@arm.com>
31248         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31249         a HOST_WIDE_INT to a poly_int64.
31251 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31252             Alan Hayward  <alan.hayward@arm.com>
31253             David Sherwood  <david.sherwood@arm.com>
31255         * calls.c (load_register_parameters): Cope with polynomial
31256         mode sizes.  Require a constant size for BLKmode parameters
31257         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31258         forces a parameter to be padded at the lsb end in order to
31259         fill a complete number of words, require the parameter size
31260         to be ordered wrt UNITS_PER_WORD.
31262 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31263             Alan Hayward  <alan.hayward@arm.com>
31264             David Sherwood  <david.sherwood@arm.com>
31266         * reload1.c (spill_stack_slot_width): Change element type
31267         from unsigned int to poly_uint64_pod.
31268         (alter_reg): Treat mode sizes as polynomial.
31270 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31271             Alan Hayward  <alan.hayward@arm.com>
31272             David Sherwood  <david.sherwood@arm.com>
31274         * reload.c (complex_word_subreg_p): New function.
31275         (reload_inner_reg_of_subreg, push_reload): Use it.
31277 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31278             Alan Hayward  <alan.hayward@arm.com>
31279             David Sherwood  <david.sherwood@arm.com>
31281         * lra-constraints.c (process_alt_operands): Reject matched
31282         operands whose sizes aren't ordered.
31283         (match_reload): Refer to this check here.
31285 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31286             Alan Hayward  <alan.hayward@arm.com>
31287             David Sherwood  <david.sherwood@arm.com>
31289         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31290         that the mode size is in the set {1, 2, 4, 8, 16}.
31292 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31293             Alan Hayward  <alan.hayward@arm.com>
31294             David Sherwood  <david.sherwood@arm.com>
31296         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31297         Use plus_constant instead of gen_rtx_PLUS.
31299 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31300             Alan Hayward  <alan.hayward@arm.com>
31301             David Sherwood  <david.sherwood@arm.com>
31303         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31304         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31305         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31306         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31307         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31308         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31309         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31310         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31311         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31312         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31313         a poly_int64.
31314         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31315         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31316         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31317         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31318         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31319         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31320         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31321         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31322         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31323         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31324         function.
31325         * expr.c (emit_move_resolve_push): Treat the input and result
31326         of PUSH_ROUNDING as a poly_int64.
31327         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31328         (emit_push_insn): Likewise.
31329         * lra-eliminations.c (mark_not_eliminable): Likewise.
31330         * recog.c (push_operand): Likewise.
31331         * reload1.c (elimination_effects): Likewise.
31332         * rtlanal.c (nonzero_bits1): Likewise.
31333         * calls.c (store_one_arg): Likewise.  Require the padding to be
31334         known at compile time.
31336 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31337             Alan Hayward  <alan.hayward@arm.com>
31338             David Sherwood  <david.sherwood@arm.com>
31340         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31341         Use plus_constant instead of gen_rtx_PLUS.
31343 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31344             Alan Hayward  <alan.hayward@arm.com>
31345             David Sherwood  <david.sherwood@arm.com>
31347         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31348         rather than an int.
31350 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31351             Alan Hayward  <alan.hayward@arm.com>
31352             David Sherwood  <david.sherwood@arm.com>
31354         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31355         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31356         via stack temporaries.  Treat the mode size as polynomial too.
31358 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31359             Alan Hayward  <alan.hayward@arm.com>
31360             David Sherwood  <david.sherwood@arm.com>
31362         * expr.c (expand_expr_real_2): When handling conversions involving
31363         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31364         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31365         as a poly_uint64 too.
31367 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31368             Alan Hayward  <alan.hayward@arm.com>
31369             David Sherwood  <david.sherwood@arm.com>
31371         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31373 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31374             Alan Hayward  <alan.hayward@arm.com>
31375             David Sherwood  <david.sherwood@arm.com>
31377         * combine.c (can_change_dest_mode): Handle polynomial
31378         REGMODE_NATURAL_SIZE.
31379         * expmed.c (store_bit_field_1): Likewise.
31380         * expr.c (store_constructor): Likewise.
31381         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31382         and polynomial REGMODE_NATURAL_SIZE.
31383         (gen_lowpart_common): Likewise.
31384         * reginfo.c (record_subregs_of_mode): Likewise.
31385         * rtlanal.c (read_modify_subreg_p): Likewise.
31387 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31388             Alan Hayward  <alan.hayward@arm.com>
31389             David Sherwood  <david.sherwood@arm.com>
31391         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31392         numbers of elements.
31394 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31395             Alan Hayward  <alan.hayward@arm.com>
31396             David Sherwood  <david.sherwood@arm.com>
31398         * match.pd: Cope with polynomial numbers of vector elements.
31400 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31401             Alan Hayward  <alan.hayward@arm.com>
31402             David Sherwood  <david.sherwood@arm.com>
31404         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31405         in a POINTER_PLUS_EXPR.
31407 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31408             Alan Hayward  <alan.hayward@arm.com>
31409             David Sherwood  <david.sherwood@arm.com>
31411         * omp-simd-clone.c (simd_clone_subparts): New function.
31412         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31413         (ipa_simd_modify_function_body): Likewise.
31415 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31416             Alan Hayward  <alan.hayward@arm.com>
31417             David Sherwood  <david.sherwood@arm.com>
31419         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31420         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31421         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31422         (expand_vector_condition, vector_element): Likewise.
31423         (subparts_gt): New function.
31424         (get_compute_type): Use subparts_gt.
31425         (count_type_subparts): Delete.
31426         (expand_vector_operations_1): Use subparts_gt instead of
31427         count_type_subparts.
31429 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31430             Alan Hayward  <alan.hayward@arm.com>
31431             David Sherwood  <david.sherwood@arm.com>
31433         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31434         (vect_compile_time_alias): ...this new function.  Do the calculation
31435         on poly_ints rather than trees.
31436         (vect_prune_runtime_alias_test_list): Update call accordingly.
31438 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31439             Alan Hayward  <alan.hayward@arm.com>
31440             David Sherwood  <david.sherwood@arm.com>
31442         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31443         numbers of units.
31444         (vect_schedule_slp_instance): Likewise.
31446 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31447             Alan Hayward  <alan.hayward@arm.com>
31448             David Sherwood  <david.sherwood@arm.com>
31450         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31451         constant and extern definitions for variable-length vectors.
31452         (vect_get_constant_vectors): Note that the number of units
31453         is known to be constant.
31455 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31456             Alan Hayward  <alan.hayward@arm.com>
31457             David Sherwood  <david.sherwood@arm.com>
31459         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31460         of units as polynomial.  Choose between WIDE and NARROW based
31461         on multiple_p.
31463 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31464             Alan Hayward  <alan.hayward@arm.com>
31465             David Sherwood  <david.sherwood@arm.com>
31467         * tree-vect-stmts.c (simd_clone_subparts): New function.
31468         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31470 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31471             Alan Hayward  <alan.hayward@arm.com>
31472             David Sherwood  <david.sherwood@arm.com>
31474         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31475         vectors as polynomial.  Use build_index_vector for
31476         IFN_GOMP_SIMD_LANE.
31478 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31479             Alan Hayward  <alan.hayward@arm.com>
31480             David Sherwood  <david.sherwood@arm.com>
31482         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31483         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31484         for variable-length vectors.
31485         (vectorizable_mask_load_store): Treat the number of units as
31486         polynomial, asserting that it is constant if the condition has
31487         already been enforced.
31488         (vectorizable_store, vectorizable_load): Likewise.
31490 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31491             Alan Hayward  <alan.hayward@arm.com>
31492             David Sherwood  <david.sherwood@arm.com>
31494         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31495         of units as polynomial.  Punt if we can't tell at compile time
31496         which vector contains the final result.
31498 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31499             Alan Hayward  <alan.hayward@arm.com>
31500             David Sherwood  <david.sherwood@arm.com>
31502         * tree-vect-loop.c (vectorizable_induction): Treat the number
31503         of units as polynomial.  Punt on SLP inductions.  Use an integer
31504         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31505         cast of such a series for variable-length floating-point
31506         reductions.
31508 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31509             Alan Hayward  <alan.hayward@arm.com>
31510             David Sherwood  <david.sherwood@arm.com>
31512         * tree.h (build_index_vector): Declare.
31513         * tree.c (build_index_vector): New function.
31514         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31515         of units as polynomial, forcibly converting it to a constant if
31516         vectorizable_reduction has already enforced the condition.
31517         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31518         to create a {1,2,3,...} vector.
31519         (vectorizable_reduction): Treat the number of units as polynomial.
31520         Choose vectype_in based on the largest scalar element size rather
31521         than the smallest number of units.  Enforce the restrictions
31522         relied on above.
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-data-refs.c (vector_alignment_reachable_p): Treat the
31529         number of units as polynomial.
31531 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31532             Alan Hayward  <alan.hayward@arm.com>
31533             David Sherwood  <david.sherwood@arm.com>
31535         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31536         * target.def (autovectorize_vector_sizes): Return the vector sizes
31537         by pointer, using vector_sizes rather than a bitmask.
31538         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31539         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31540         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31541         Likewise.
31542         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31543         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31544         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31545         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31546         * omp-general.c (omp_max_vf): Likewise.
31547         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31548         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31549         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31550         * tree-vect-slp.c (vect_slp_bb): Likewise.
31551         * doc/tm.texi: Regenerate.
31552         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31553         to a poly_uint64.
31554         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31555         the vector size as a poly_uint64 rather than an unsigned int.
31556         (current_vector_size): Change from an unsigned int to a poly_uint64.
31557         (get_vectype_for_scalar_type): Update accordingly.
31558         * tree.h (build_truth_vector_type): Take the size and number of
31559         units as a poly_uint64 rather than an unsigned int.
31560         (build_vector_type): Add a temporary overload that takes
31561         the number of units as a poly_uint64 rather than an unsigned int.
31562         * tree.c (make_vector_type): Likewise.
31563         (build_truth_vector_type): Take the number of units as a poly_uint64
31564         rather than an unsigned int.
31566 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31567             Alan Hayward  <alan.hayward@arm.com>
31568             David Sherwood  <david.sherwood@arm.com>
31570         * target.def (get_mask_mode): Take the number of units and length
31571         as poly_uint64s rather than unsigned ints.
31572         * targhooks.h (default_get_mask_mode): Update accordingly.
31573         * targhooks.c (default_get_mask_mode): Likewise.
31574         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31575         * doc/tm.texi: Regenerate.
31577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31578             Alan Hayward  <alan.hayward@arm.com>
31579             David Sherwood  <david.sherwood@arm.com>
31581         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
31582         * omp-general.c (omp_max_vf): Likewise.
31583         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
31584         (expand_omp_simd): Handle polynomial safelen.
31585         * omp-low.c (omplow_simd_context): Add a default constructor.
31586         (omplow_simd_context::max_vf): Change from int to poly_uint64.
31587         (lower_rec_simd_input_clauses): Update accordingly.
31588         (lower_rec_input_clauses): Likewise.
31590 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31591             Alan Hayward  <alan.hayward@arm.com>
31592             David Sherwood  <david.sherwood@arm.com>
31594         * tree-vectorizer.h (vect_nunits_for_cost): New function.
31595         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
31596         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
31597         (vect_analyze_slp_cost): Likewise.
31598         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
31599         (vect_model_load_cost): Likewise.
31601 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31602             Alan Hayward  <alan.hayward@arm.com>
31603             David Sherwood  <david.sherwood@arm.com>
31605         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
31606         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
31607         from an unsigned int * to a poly_uint64_pod *.
31608         (calculate_unrolling_factor): New function.
31609         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
31611 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31612             Alan Hayward  <alan.hayward@arm.com>
31613             David Sherwood  <david.sherwood@arm.com>
31615         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
31616         from an unsigned int to a poly_uint64.
31617         (_loop_vec_info::slp_unrolling_factor): Likewise.
31618         (_loop_vec_info::vectorization_factor): Change from an int
31619         to a poly_uint64.
31620         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
31621         (vect_get_num_vectors): New function.
31622         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
31623         (vect_get_num_copies): Use vect_get_num_vectors.
31624         (vect_analyze_data_ref_dependences): Change max_vf from an int *
31625         to an unsigned int *.
31626         (vect_analyze_data_refs): Change min_vf from an int * to a
31627         poly_uint64 *.
31628         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31629         than an unsigned HOST_WIDE_INT.
31630         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
31631         (vect_analyze_data_ref_dependence): Change max_vf from an int *
31632         to an unsigned int *.
31633         (vect_analyze_data_ref_dependences): Likewise.
31634         (vect_compute_data_ref_alignment): Handle polynomial vf.
31635         (vect_enhance_data_refs_alignment): Likewise.
31636         (vect_prune_runtime_alias_test_list): Likewise.
31637         (vect_shift_permute_load_chain): Likewise.
31638         (vect_supportable_dr_alignment): Likewise.
31639         (dependence_distance_ge_vf): Take the vectorization factor as a
31640         poly_uint64 rather than an unsigned HOST_WIDE_INT.
31641         (vect_analyze_data_refs): Change min_vf from an int * to a
31642         poly_uint64 *.
31643         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
31644         vfm1 as a poly_uint64 rather than an int.  Make the same change
31645         for the returned bound_scalar.
31646         (vect_gen_vector_loop_niters): Handle polynomial vf.
31647         (vect_do_peeling): Likewise.  Update call to
31648         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
31649         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
31650         be constant.
31651         * tree-vect-loop.c (vect_determine_vectorization_factor)
31652         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
31653         (vect_get_known_peeling_cost): Likewise.
31654         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
31655         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
31656         (vect_transform_loop): Likewise.  Use the lowest possible VF when
31657         updating the upper bounds of the loop.
31658         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
31659         rather than an int.
31660         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
31661         polynomial unroll factors.
31662         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
31663         (vect_make_slp_decision): Likewise.
31664         (vect_supported_load_permutation_p): Likewise, and polynomial
31665         vf too.
31666         (vect_analyze_slp_cost): Handle polynomial vf.
31667         (vect_slp_analyze_node_operations): Likewise.
31668         (vect_slp_analyze_bb_1): Likewise.
31669         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31670         than an unsigned HOST_WIDE_INT.
31671         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
31672         (vectorizable_load): Handle polynomial vf.
31673         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
31674         a poly_uint64.
31675         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
31677 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31678             Alan Hayward  <alan.hayward@arm.com>
31679             David Sherwood  <david.sherwood@arm.com>
31681         * match.pd: Handle bit operations involving three constants
31682         and try to fold one pair.
31684 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31686         * tree-vect-loop-manip.c: Include gimple-fold.h.
31687         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
31688         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
31689         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
31690         Add a path that uses a step of VF instead of 1, but disable it
31691         for now.
31692         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
31693         and niters_no_overflow parameters.  Update calls to
31694         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
31695         Create a new SSA name if the latter choses to use a ste other
31696         than zero, and return it via niters_vector_mult_vf_var.
31697         * tree-vect-loop.c (vect_transform_loop): Update calls to
31698         vect_do_peeling, vect_gen_vector_loop_niters and
31699         slpeel_make_loop_iterate_ntimes.
31700         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
31701         (vect_gen_vector_loop_niters): Update declarations after above changes.
31703 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
31705         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
31706         128-bit round to integer instructions.
31707         (ceil<mode>2): Likewise.
31708         (btrunc<mode>2): Likewise.
31709         (round<mode>2): Likewise.
31711 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31713         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
31714         unaligned VSX load/store on P8/P9.
31715         (expand_block_clear): Allow the use of unaligned VSX
31716         load/store on P8/P9.
31718 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31720         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
31721         New function.
31722         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
31723         swap associated with both a load and a store.
31725 2018-01-02  Andrew Waterman  <andrew@sifive.com>
31727         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
31728         * config/riscv/riscv.md (clear_cache): Use it.
31730 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
31732         * web.c: Remove out-of-date comment.
31734 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31736         * expr.c (fixup_args_size_notes): Check that any existing
31737         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
31738         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
31739         (emit_single_push_insn): ...here.
31741 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31743         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
31744         (const_vector_encoded_nelts): New function.
31745         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
31746         (const_vector_int_elt, const_vector_elt): Declare.
31747         * emit-rtl.c (const_vector_int_elt_1): New function.
31748         (const_vector_elt): Likewise.
31749         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
31750         of CONST_VECTOR_ELT.
31752 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31754         * expr.c: Include rtx-vector-builder.h.
31755         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
31756         directly on the tree encoding.
31757         (const_vector_from_tree): Likewise.
31758         * optabs.c: Include rtx-vector-builder.h.
31759         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
31760         sequence of "u" values.
31761         * vec-perm-indices.c: Include rtx-vector-builder.h.
31762         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
31763         directly on the vec_perm_indices encoding.
31765 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31767         * doc/rtl.texi (const_vector): Describe new encoding scheme.
31768         * Makefile.in (OBJS): Add rtx-vector-builder.o.
31769         * rtx-vector-builder.h: New file.
31770         * rtx-vector-builder.c: Likewise.
31771         * rtl.h (rtx_def::u2): Add a const_vector field.
31772         (CONST_VECTOR_NPATTERNS): New macro.
31773         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
31774         (CONST_VECTOR_DUPLICATE_P): Likewise.
31775         (CONST_VECTOR_STEPPED_P): Likewise.
31776         (CONST_VECTOR_ENCODED_ELT): Likewise.
31777         (const_vec_duplicate_p): Check for a duplicated vector encoding.
31778         (unwrap_const_vec_duplicate): Likewise.
31779         (const_vec_series_p): Check for a non-duplicated vector encoding.
31780         Say that the function only returns true for integer vectors.
31781         * emit-rtl.c: Include rtx-vector-builder.h.
31782         (gen_const_vec_duplicate_1): Delete.
31783         (gen_const_vector): Call gen_const_vec_duplicate instead of
31784         gen_const_vec_duplicate_1.
31785         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
31786         (gen_const_vec_duplicate): Use rtx_vector_builder.
31787         (gen_const_vec_series): Likewise.
31788         (gen_rtx_CONST_VECTOR): Likewise.
31789         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
31790         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31791         Build a new vector rather than modifying a CONST_VECTOR in-place.
31792         (handle_special_swappables): Update call accordingly.
31793         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
31794         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31795         Build a new vector rather than modifying a CONST_VECTOR in-place.
31796         (handle_special_swappables): Update call accordingly.
31798 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31800         * simplify-rtx.c (simplify_const_binary_operation): Use
31801         CONST_VECTOR_ELT instead of XVECEXP.
31803 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31805         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
31806         the selector elements to be different from the data elements
31807         if the selector is a VECTOR_CST.
31808         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
31809         ssizetype for the selector.
31811 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31813         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
31814         before testing each element individually.
31815         * tree-vect-generic.c (lower_vec_perm): Likewise.
31817 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31819         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
31820         * selftest-run-tests.c (selftest::run_tests): Call it.
31821         * vector-builder.h (vector_builder::operator ==): New function.
31822         (vector_builder::operator !=): Likewise.
31823         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
31824         (vec_perm_indices::all_from_input_p): New function.
31825         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31826         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
31827         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
31828         instead of reading the VECTOR_CST directly.  Detect whether both
31829         vector inputs are the same before constructing the vec_perm_indices,
31830         and update the number of inputs argument accordingly.  Use the
31831         utility functions added above.  Only construct sel2 if we need to.
31833 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31835         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
31836         the broadcast of the low byte.
31837         (expand_mult_highpart): Use an explicit encoding for the permutes.
31838         * optabs-query.c (can_mult_highpart_p): Likewise.
31839         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
31840         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31841         (vectorizable_bswap): Likewise.
31842         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
31843         explicit encoding for the power-of-2 permutes.
31844         (vect_permute_store_chain): Likewise.
31845         (vect_grouped_load_supported): Likewise.
31846         (vect_permute_load_chain): Likewise.
31848 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31850         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
31851         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
31852         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
31853         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31854         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
31855         (vect_gen_perm_mask_any): Likewise.
31857 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31859         * int-vector-builder.h: New file.
31860         * vec-perm-indices.h: Include int-vector-builder.h.
31861         (vec_perm_indices): Redefine as an int_vector_builder.
31862         (auto_vec_perm_indices): Delete.
31863         (vec_perm_builder): Redefine as a stand-alone class.
31864         (vec_perm_indices::vec_perm_indices): New function.
31865         (vec_perm_indices::clamp): Likewise.
31866         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
31867         (vec_perm_indices::new_vector): New function.
31868         (vec_perm_indices::new_expanded_vector): Update for new
31869         vec_perm_indices class.
31870         (vec_perm_indices::rotate_inputs): New function.
31871         (vec_perm_indices::all_in_range_p): Operate directly on the
31872         encoded form, without computing elided elements.
31873         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
31874         encoding.  Update for new vec_perm_indices class.
31875         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
31876         the given vec_perm_builder.
31877         (expand_vec_perm_var): Update vec_perm_builder constructor.
31878         (expand_mult_highpart): Use vec_perm_builder instead of
31879         auto_vec_perm_indices.
31880         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
31881         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
31882         or double series encoding as appropriate.
31883         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
31884         vec_perm_indices instead of auto_vec_perm_indices.
31885         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31886         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31887         (vect_permute_store_chain): Likewise.
31888         (vect_grouped_load_supported): Likewise.
31889         (vect_permute_load_chain): Likewise.
31890         (vect_shift_permute_load_chain): Likewise.
31891         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31892         (vect_transform_slp_perm_load): Likewise.
31893         (vect_schedule_slp_instance): Likewise.
31894         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31895         (vectorizable_mask_load_store): Likewise.
31896         (vectorizable_bswap): Likewise.
31897         (vectorizable_store): Likewise.
31898         (vectorizable_load): Likewise.
31899         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
31900         vec_perm_indices instead of auto_vec_perm_indices.  Use
31901         tree_to_vec_perm_builder to read the vector from a tree.
31902         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
31903         vec_perm_builder instead of a vec_perm_indices.
31904         (have_whole_vector_shift): Use vec_perm_builder and
31905         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
31906         truncation to calc_vec_perm_mask_for_shift.
31907         (vect_create_epilog_for_reduction): Likewise.
31908         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
31909         from auto_vec_perm_indices to vec_perm_indices.
31910         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31911         instead of changing individual elements.
31912         (aarch64_vectorize_vec_perm_const): Use new_vector to install
31913         the vector in d.perm.
31914         * config/arm/arm.c (expand_vec_perm_d::perm): Change
31915         from auto_vec_perm_indices to vec_perm_indices.
31916         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31917         instead of changing individual elements.
31918         (arm_vectorize_vec_perm_const): Use new_vector to install
31919         the vector in d.perm.
31920         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
31921         Update vec_perm_builder constructor.
31922         (rs6000_expand_interleave): Likewise.
31923         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
31924         (rs6000_expand_interleave): Likewise.
31926 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31928         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
31929         to qimode could truncate the indices.
31930         * optabs.c (expand_vec_perm_var): Likewise.
31932 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31934         * Makefile.in (OBJS): Add vec-perm-indices.o.
31935         * vec-perm-indices.h: New file.
31936         * vec-perm-indices.c: Likewise.
31937         * target.h (vec_perm_indices): Replace with a forward class
31938         declaration.
31939         (auto_vec_perm_indices): Move to vec-perm-indices.h.
31940         * optabs.h: Include vec-perm-indices.h.
31941         (expand_vec_perm): Delete.
31942         (selector_fits_mode_p, expand_vec_perm_var): Declare.
31943         (expand_vec_perm_const): Declare.
31944         * target.def (vec_perm_const_ok): Replace with...
31945         (vec_perm_const): ...this new hook.
31946         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
31947         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
31948         * doc/tm.texi: Regenerate.
31949         * optabs.def (vec_perm_const): Delete.
31950         * doc/md.texi (vec_perm_const): Likewise.
31951         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
31952         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
31953         expand_vec_perm for constant permutation vectors.  Assert that
31954         the mode of variable permutation vectors is the integer equivalent
31955         of the mode that is being permuted.
31956         * optabs-query.h (selector_fits_mode_p): Declare.
31957         * optabs-query.c: Include vec-perm-indices.h.
31958         (selector_fits_mode_p): New function.
31959         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
31960         is defined, instead of checking whether the vec_perm_const_optab
31961         exists.  Use targetm.vectorize.vec_perm_const instead of
31962         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
31963         fit in the vector mode before using a variable permute.
31964         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
31965         vec_perm_indices instead of an rtx.
31966         (expand_vec_perm): Replace with...
31967         (expand_vec_perm_const): ...this new function.  Take the selector
31968         as a vec_perm_indices rather than an rtx.  Also take the mode of
31969         the selector.  Update call to shift_amt_for_vec_perm_mask.
31970         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
31971         Use vec_perm_indices::new_expanded_vector to expand the original
31972         selector into bytes.  Check whether the indices fit in the vector
31973         mode before using a variable permute.
31974         (expand_vec_perm_var): Make global.
31975         (expand_mult_highpart): Use expand_vec_perm_const.
31976         * fold-const.c: Includes vec-perm-indices.h.
31977         * tree-ssa-forwprop.c: Likewise.
31978         * tree-vect-data-refs.c: Likewise.
31979         * tree-vect-generic.c: Likewise.
31980         * tree-vect-loop.c: Likewise.
31981         * tree-vect-slp.c: Likewise.
31982         * tree-vect-stmts.c: Likewise.
31983         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
31984         Delete.
31985         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
31986         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
31987         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
31988         (aarch64_vectorize_vec_perm_const): ...this new function.
31989         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31990         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31991         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
31992         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
31993         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31994         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31995         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
31996         into...
31997         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
31998         check for NEON modes.
31999         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32000         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32001         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32002         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32003         into...
32004         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
32005         the old VEC_PERM_CONST conditions.
32006         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32007         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32008         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32009         (ia64_vectorize_vec_perm_const_ok): Merge into...
32010         (ia64_vectorize_vec_perm_const): ...this new function.
32011         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32012         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32013         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32014         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32015         * config/mips/mips.c (mips_expand_vec_perm_const)
32016         (mips_vectorize_vec_perm_const_ok): Merge into...
32017         (mips_vectorize_vec_perm_const): ...this new function.
32018         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32019         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32020         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32021         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32022         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32023         (rs6000_expand_vec_perm_const): Delete.
32024         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32025         Delete.
32026         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32027         (altivec_expand_vec_perm_const_le): Take each operand individually.
32028         Operate on constant selectors rather than rtxes.
32029         (altivec_expand_vec_perm_const): Likewise.  Update call to
32030         altivec_expand_vec_perm_const_le.
32031         (rs6000_expand_vec_perm_const): Delete.
32032         (rs6000_vectorize_vec_perm_const_ok): Delete.
32033         (rs6000_vectorize_vec_perm_const): New function.
32034         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32035         an element count and rtx array.
32036         (rs6000_expand_extract_even): Update call accordingly.
32037         (rs6000_expand_interleave): Likewise.
32038         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32039         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32040         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32041         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32042         (rs6000_expand_vec_perm_const): Delete.
32043         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32044         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32045         (altivec_expand_vec_perm_const_le): Take each operand individually.
32046         Operate on constant selectors rather than rtxes.
32047         (altivec_expand_vec_perm_const): Likewise.  Update call to
32048         altivec_expand_vec_perm_const_le.
32049         (rs6000_expand_vec_perm_const): Delete.
32050         (rs6000_vectorize_vec_perm_const_ok): Delete.
32051         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
32052         reference to the SPE evmerge intructions.
32053         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32054         an element count and rtx array.
32055         (rs6000_expand_extract_even): Update call accordingly.
32056         (rs6000_expand_interleave): Likewise.
32057         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32058         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32059         new function.
32060         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32062 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32064         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32065         vector mode and that that mode matches the mode of the data
32066         being permuted.
32067         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32068         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
32069         directly using expand_vec_perm_1 when forcing selectors into
32070         registers.
32071         (expand_vec_perm_var): New function, split out from expand_vec_perm.
32073 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32075         * optabs-query.h (can_vec_perm_p): Delete.
32076         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32077         * optabs-query.c (can_vec_perm_p): Split into...
32078         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32079         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32080         particular selector is valid.
32081         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32082         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32083         (vect_grouped_load_supported): Likewise.
32084         (vect_shift_permute_load_chain): Likewise.
32085         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32086         (vect_transform_slp_perm_load): Likewise.
32087         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32088         (vectorizable_bswap): Likewise.
32089         (vect_gen_perm_mask_checked): Likewise.
32090         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32091         implementations of variable permutation vectors into account
32092         when deciding which selector to use.
32093         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32094         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32095         with a false third argument.
32096         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32097         to test whether the constant selector is valid and can_vec_perm_var_p
32098         to test whether a variable selector is valid.
32100 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32102         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32103         * optabs-query.c (can_vec_perm_p): Likewise.
32104         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32105         instead of vec_perm_indices.
32106         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32107         (vect_gen_perm_mask_checked): Likewise,
32108         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32109         (vect_gen_perm_mask_checked): Likewise,
32111 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32113         * optabs-query.h (qimode_for_vec_perm): Declare.
32114         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32115         (qimode_for_vec_perm): ...this new function.
32116         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32118 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32120         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32121         does not have a conditional at the top.
32123 2018-01-02  Richard Biener  <rguenther@suse.de>
32125         * ipa-inline.c (big_speedup_p): Fix expression.
32127 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32129         PR target/81616
32130         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32131         for generic 4->6.
32133 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32135         PR target/81616
32136         Generic tuning.
32137         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32138         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32139         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32140         cond_taken_branch_cost 3->4.
32142 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32144         PR tree-optimization/83581
32145         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32146         TODO_cleanup_cfg if any changes have been made.
32148         PR middle-end/83608
32149         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32150         convert_modes if target mode has the right side, but different mode
32151         class.
32153         PR middle-end/83609
32154         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32155         last argument when extracting from CONCAT.  If either from_real or
32156         from_imag is NULL, use expansion through memory.  If result is not
32157         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32158         the parts directly to inner mode, if even that fails, use expansion
32159         through memory.
32161         PR middle-end/83623
32162         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32163         check for bswap in mode rather than HImode and use that in expand_unop
32164         too.
32166 Copyright (C) 2018 Free Software Foundation, Inc.
32168 Copying and distribution of this file, with or without modification,
32169 are permitted in any medium without royalty provided the copyright
32170 notice and this notice are preserved.