PR rtl-optimization/88018
[official-gcc.git] / gcc / ChangeLog
blob4bf430757dca783cc480176c16d876ebaff5613f
1 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
3         PR rtl-optimization/88018
4         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
5         on the presence of fallthru edge, rather than if it is a USE or not.
7 2018-11-15  Richard Biener  <rguenther@suse.de>
9         PR middle-end/87917
10         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
11         analyze_subscript_affine_affine properly.
13 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
15         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
16         (GLIBC_DYNAMIC_LINKER): Define.
17         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
19 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
21         PR tree-optimization/84648
22         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
23         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
24         by calling adjust_cond_for_loop_until_wrap.
26 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
28         PR other/56334
29         * doc/extend.texi (Common Function Attributes): Clarify linker
30         restrictions on "aligned" attribute.
31         (Common Variable Attributes): Likewise.  Mention that linker
32         restrictions don't apply to stack-allocated variables.
34 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
36         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
37         as an lvalue.
39 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
41         PR lto/55102
42         PR lto/56700
43         * doc/invoke.texi (Optimize Options): Remove bad example about
44         interaction between -flto and -O.  Replace it with a note that
45         you need to compile with -O and not just link.  Copy-edit -flto
46         discussion to reduce verbiage and improve flow.
48 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
50         PR middle-end/59658
51         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
52         also suppress many optimizations.  Alphabetize option lists for
53         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
54         correct documentation for those options to say that.
55         * opts.c (default_options_table): Sort table by level and option
56         name, to make it easier to correlate to the manual.
58 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
60         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
61         printing of the names of segment registers.
63 2018-11-14  Nathan Sidwell  <nathan@acm.org>
65         PR debug/88006
66         PR debug/87462
67         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
68         type list.
70 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
72         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
74 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
76         P1236R1 - Signed integers are two's complement
77         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
79         PR bootstrap/86739
80         * hash-map.h (hash_map::iterator::reference_pair): New class.
81         (hash_map::iterator::operator*): Return it rather than std::pair.
83 2018-11-14  Jeff Law  <law@redhat.com>
85         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
86         iff the operand is a constant.
88 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
90         * gimple-ssa-evrp-analyze.c
91         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
92         ignore_equivs_equal_p to equal_p.
93         * ipa-cp.c (meet_with_1): Use equal_p instead of
94         ignore_equivs_equal_p.
95         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
96         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
97         (value_range::operator==): Remove.
98         (value_range::operator!=): Remove.
99         (vrp_prop::visit_stmt): Use equal_p.
100         * tree-vrp.h (value_range): Remove operator==, operator!=,
101         ignore_equivs_equal_p.
102         * vr-values.c (update_value_range): Use equal_p.
104 2018-11-14  Michael Matz  <matz@suse.de>
106         PR middle-end/86575
107         * gimplify.c (collect_fallthrough_labels): Add new argument,
108         return location via that, don't modify statements.
109         (warn_implicit_fallthrough_r): Adjust call, don't use
110         statement location directly.
112 2018-11-14  Richard Biener  <rguenther@suse.de>
114         PR middle-end/87985
115         * tree-data-ref.c (split_constant_offset): Add wrapper
116         allocating a cache hash-map.
117         (split_constant_offset_1): Cache results of expanding
118         expressions from SSA def stmts.
120 2018-11-14  Richard Biener  <rguenther@suse.de>
122         PR middle-end/88021
123         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
124         to lambda_int.
125         (lambda_vector_mult_const): Likewise.
126         (lambda_matrix_right_hermite): Use lambda_int temporaries.
128 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
129             Jackson Woodruff  <jackson.woodruff@arm.com>
131         PR 71026/tree-optimization
132         * match.pd: Simplify floating point comparisons.
134 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
136         PR rtl-optimization/87817
137         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
138         last argument has low 8 bits clear, fold to 0.
140 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
142         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
143         MAX_OFILE_ALIGNMENT as needed.
145 2018-11-14  Richard Biener  <rguenther@suse.de>
147         PR tree-optimization/88019
148         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
149         COND_EXPR arguments to non-trapping overflow.
151 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
153         * config/arc/arc.h (reg_class): Reorder registers classes, remove
154         unused register classes.
155         (REG_CLASS_NAMES): Likewise.
156         (REG_CLASS_CONTENTS): Likewise.
157         (FIXED_REGISTERS): Make lp_count fixed.
158         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
159         (PROGRAM_COUNTER_REGNO): Remove.
160         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
161         register classes, use constants for register numbers, remove
162         reg_class_contents references.
163         (arc_process_double_reg_moves): Add asserts.
164         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
165         lra_in_progress predicate.
166         (arc_init_reg_tables): Remove unused register classes.
167         (arc_register_move_cost): Likewise.
168         (arc_preferred_reload_class): Likewise.
169         (hwloop_optimize): Update rtx patterns involving lp_count register.
170         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
171         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
172         CHEAP_CORE_REGS.  Former one will be used for LRA.
173         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
174         one will be used for LRA.
175         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
176         one will be used for LRA.
177         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
178         Former one will be used for LRA.
179         ("f"): Delete constraint.
180         ("k"): Likewise.
181         ("e"): Likewise.
182         ("l"): Change it from register constraint to constraint.
183         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
184         (movhi_insn): Likewise.
185         (movsi_insn): Update pattern.
186         (arc_lp): Likewise.
187         (dbnz): Likewise.
188         (stack_tie): Remove 'b' constraint letter.
189         (R4_REG): Define.
190         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
191         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
192         (R57_REG, R59_REG, PCL_REG): Likewise.
193         (ILINK1_REGNUM): Renamed to ILINK1_REG.
194         (ILINK2_REGNUM): Renamed to ILINK2_REG.
195         (Rgp): Remove.
196         (SP_REGS): Likewise.
197         (Rcw): Remove unused reg classes.
198         * config/arc/predicates.md (dest_reg_operand): Just default on
199         register_operand predicate.
200         (mpy_dest_reg_operand): Likewise.
201         (move_dest_operand): Use macros instead of constants.
202         * config/arc/arc.opt (mlra): Switch to lra as default.
204 2018-11-14  Richard Biener  <rguenther@suse.de>
206         PR tree-optimization/87974
207         * tree-vect-loop.c (vectorizable_reduction): When computing
208         the vectorized reduction PHI vector type ignore constant
209         and external defs.
211 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
213         PR tree-optimization/87977
214         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
215         stmt, build a new one and replace the old one with it.  Formatting fix.
216         Call release_ssa_name (x) if !has_other_use and !delete_div.
217         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
218         verify lhs of stmt is still def.
220 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
222         PR rtl-optimization/87507
223         * lower-subreg.c (operand_for_swap_move_operator): New function.
224         (simple_move): Strip simple operators.
225         (find_pseudo_copy): Likewise.
226         (resolve_operand_for_swap_move_operator): New function.
227         (resolve_simple_move): Strip simple operators and swap operands.
229 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
231         PR tree-optimization/87898
232         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
233         (ipa_simd_modify_function_body): Remove debug stmts where the first
234         argument was changed into a non-decl.
236 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
238         PR rtl-optimization/87899
239         * lra-lives.c (start_living): Update white space in comment.
240         (enum point_type): New.
241         (sparseset_contains_pseudos_p): New function.
242         (update_pseudo_point): Likewise.
243         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
244         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
245         handling.  Move early exit after adding conflicts.
246         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
247         if regno is already live.  Remove all handling of program points.
248         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
249         after adding conflicts.  Remove all handling of program points and
250         ignore_reg_for_conflicts.
251         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
252         and do not guard call to mark_pseudo_live.
253         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
254         and do not guard call to mark_pseudo_dead.
255         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
256         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
257         Use new function update_pseudo_point.  Handle register copies by
258         removing the source register from the live set.  Handle INOUT operands.
259         Update to the next program point using the unused_set, dead_set and
260         start_dying sets.
261         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
263 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
265         * builtins.c: Replace "source_location" with "location_t".
266         * diagnostic-show-locus.c: Likewise.
267         * diagnostic.c: Likewise.
268         * dumpfile.c: Likewise.
269         * gcc-rich-location.h: Likewise.
270         * genmatch.c: Likewise.
271         * gimple.h: Likewise.
272         * gimplify.c: Likewise.
273         * input.c: Likewise.
274         * input.h: Likewise.  Eliminate the typedef.
275         * omp-expand.c: Likewise.
276         * selftest.h: Likewise.
277         * substring-locations.h (get_source_location_for_substring):
278         Rename to...
279         (get_location_within_string): ...this.
280         * tree-cfg.c: Replace "source_location" with "location_t".
281         * tree-cfgcleanup.c: Likewise.
282         * tree-diagnostic.c: Likewise.
283         * tree-into-ssa.c: Likewise.
284         * tree-outof-ssa.c: Likewise.
285         * tree-parloops.c: Likewise.
286         * tree-phinodes.c: Likewise.
287         * tree-phinodes.h: Likewise.
288         * tree-ssa-loop-ivopts.c: Likewise.
289         * tree-ssa-loop-manip.c: Likewise.
290         * tree-ssa-phiopt.c: Likewise.
291         * tree-ssa-phiprop.c: Likewise.
292         * tree-ssa-threadupdate.c: Likewise.
293         * tree-ssa.c: Likewise.
294         * tree-ssa.h: Likewise.
295         * tree-vect-loop-manip.c: Likewise.
297 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
299         * doc/extend.texi: Fix typo in the weakref description.
301 2018-11-13  Richard Biener  <rguenther@suse.de>
303         PR tree-optimization/86991
304         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
305         group building until we have successfully detected the SLP
306         reduction.
307         (vect_is_simple_reduction): Remove fixup code here.
309 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
311         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
312         (dump_gimple_stmt): Use it.
313         (dump_gimple_stmt_loc): Likewise.
314         (dump_gimple_expr): Likewise.
315         (dump_gimple_expr_loc): Likewise.
316         (dump_generic_expr): Likewise.
317         (dump_generic_expr_loc): Likewise.
318         (dump_printf): Likewise.
319         (dump_printf_loc): Likewise.
320         (dump_dec): Likewise.
321         (dump_dec): Likewise.
322         (dump_hex): Likewise.
323         (dump_symtab_node): Likewise.
324         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
325         Guard dump call with dump_enabled_p.
326         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
327         * graphite-optimize-isl.c (optimize_isl): Likewise.
328         * graphite.c (graphite_transform_loops): Likewise.
329         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
330         * tree-parloops.c (parallelize_loops): Likewise.
331         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
332         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
333         (vect_prune_runtime_alias_test_list): Likewise.
334         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
335         (vect_estimate_min_profitable_iters): Likewise.
336         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
337         (vect_build_slp_tree_2): Likewise.
338         (vect_supported_load_permutation_p): Likewise.
339         (vect_slp_analyze_operations): Likewise.
340         (vect_slp_analyze_bb_1): Likewise.
341         (vect_slp_bb): Likewise.
342         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
343         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
344         (pass_slp_vectorize::execute): Likewise.
345         (increase_alignment): Likewise.
347 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
349         PR ipa/87955
350         * ipa-inline.c (report_inline_failed_reason): Guard calls to
351         cl_target_option_print_diff and cl_optimization_print_diff with
352         if (dump_file).
354 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
356         * doc/invoke.texi (-fsave-optimization-record): Note that the
357         output is compressed.
358         * optinfo-emit-json.cc: Include <zlib.h>.
359         (optrecord_json_writer::write): Compress the output.
361 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
363         * tree-vrp.c (value_range_base::dump): Dump type.
364         Do not use INF nomenclature for 1-bit types.
365         (dump_value_range): Group all variants to common dumping code.
366         (debug): New overloaded functions for value_ranges.
367         (value_range_base::dump): Remove no argument version.
368         (value_range::dump): Same.
370 2018-11-13  Richard Biener  <rguenther@suse.de>
372         PR tree-optimization/87931
373         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
374         nested cycles we support to latch computations vectorizable_reduction
375         handles.
377 2018-11-13  Martin Liska  <mliska@suse.cz>
379         PR tree-optimization/87885
380         * cfghooks.c (account_profile_record): Rename
381         to ...
382         (profile_record_check_consistency): ... this.
383         Calculate missing num_mismatched_freq_in.
384         (profile_record_account_profile): New function
385         that calculates time and size of a function.
386         * cfghooks.h (struct profile_record): Remove
387         all tuples.
388         (struct cfg_hooks): Remove after_pass flag.
389         (account_profile_record): Rename to ...
390         (profile_record_check_consistency): ... this.
391         (profile_record_account_profile): New.
392         * cfgrtl.c (rtl_account_profile_record): Remove
393         after_pass flag.
394         * passes.c (check_profile_consistency): Do only
395         checking.
396         (account_profile): Calculate size and time of
397         function only.
398         (pass_manager::dump_profile_report): Reformat
399         output.
400         (execute_one_ipa_transform_pass): Call
401         consistency check before clean upand call account_profile
402         after a clean up is done.
403         (execute_one_pass): Call check_profile_consistency and
404         account_profile instead of using after_pass flag..
405         * tree-cfg.c (gimple_account_profile_record): Likewise.
407 2018-11-13  Martin Liska  <mliska@suse.cz>
409         PR sanitizer/87930
410         * config/i386/i386.c (ix86_option_override_internal): Error
411         about usage -mabi=ms and -fsanitize={,kernel-}address.
413 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
415         * config/aarch64/aarch64.c
416         (aarch64_vectorize_preferred_vector_alignment): Change return type to
417         poly_uint64.
418         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
419         alignment being a poly int.
420         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
421         return type to poly_uint64.
422         * target.def (default_preferred_vector_alignment): Likewise.
423         * targhooks.c (default_preferred_vector_alignment): Likewise.
424         * targhooks.h (default_preferred_vector_alignment): Likewise.
425         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
426         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
427         poly int.
428         (vect_update_misalignment_for_peel): Likewise.
429         (vect_enhance_data_refs_alignment): Likewise.
430         (vect_find_same_alignment_drs): Likewise.
431         (vect_duplicate_ssa_name_ptr_info): Likewise.
432         (vect_setup_realignment): Likewise.
433         (vect_can_force_dr_alignment_p): Change alignment parameter type to
434         poly_uint64.
435         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
436         mask with a compile time variable vector alignment.
437         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
438         int.
439         (vect_do_peeling): Exit early if vector alignment is not constant.
440         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
441         a poly int.
442         (vectorizable_store): Likewise.
443         (vectorizable_load): Likweise.
444         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
445         poly_uint64.
446         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
447         poly int.
448         (vect_can_force_dr_alignment_p): Change alignment parameter type to
449         poly_uint64.
451 2018-11-13  Richard Biener  <rguenther@suse.de>
453         PR tree-optimization/87962
454         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
455         detect outer reduction for disqualifying in-loop uses.
457 2018-11-13  Richard Biener  <rguenther@suse.de>
459         PR tree-optimization/87967
460         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
461         for constants for the scalar loop.
463 2018-11-13  Alan Modra  <amodra@gmail.com>
465         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
466         integer constants when -mcmodel=medium.
468 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
470         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
471         * config/arc/elf.h (PROFILE_HOOK): Define.
472         * config/arc/linux.h (PROFILE_HOOK): Likewise.
474 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
476         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
477         jump table data in the text section.
479 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
481         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
482         to fit the eh_return pattern.
483         * config/arc/arc.md (eh_return): Define.
484         (VUNSPEC_ARC_EH_RETURN): Likewise.
485         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
486         new implementation.
487         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
490 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
492         * common/config/arc/arc-common.c (arc_option_optimization_table):
493         Millicode optimization is default on for size optimizations.
494         * config/arc/arc-protos.h (arc_check_multi): New function.
495         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
496         (ENTER_LEAVE_START_REG): Define.
497         (ENTER_LEAVE_END_REG): Likewise.
498         (arc_override_options): Disable millicode when long calls option
499         is on.
500         (arc_frame_info): Change it from int to bool.
501         (arc_compute_frame_size): Clean up.
502         (arc_save_restore): Remove.
503         (frame_save_reg): New function.
504         (frame_restore_reg): Likewise.
505         (arc_enter_leave_p): Likewise.
506         (arc_save_callee_saves): Likewise.
507         (arc_restore_callee_saves): Likewise.
508         (arc_save_callee_enter): Likewise.
509         (arc_restore_callee_leave): Likewise.
510         (arc_save_callee_milli): Likewise.
511         (arc_restore_callee_milli): Likewise.
512         (arc_expand_prologue): Reimplement to emit enter/leave
513         instructions.
514         (arc_expand_epilogue): Likewise.
515         (arc_check_multi): New function.
516         * config/arc/arc.md (push_multi_fp): New pattern.
517         (push_multi_fp_blink): Likewise.
518         (pop_multi_fp): Likewise.
519         (pop_multi_fp_blink): Likewise.
520         (pop_multi_fp_ret): Likewise.
521         (pop_multi_fp_blink_ret): Likewise.
522         * config/arc/arc.opt (mmillicode): Update option.
523         (mcode-density-frame): New option.
524         * config/arc/predicates.md (push_multi_operand): New predicate.
525         (pop_multi_operand): Likewise.
526         * doc/invoke.texi (ARC): Update ARC options information.
529 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
531         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
532         * config/arc/arc.c (operands_ok_ldd_std): New function.
533         (mem_ok_for_ldd_std): Likewise.
534         (gen_operands_ldd_std): Likewise.
535         * config/arc/arc.md: Add peephole2 rules for std/ldd.
537 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
539         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
540         test on stack_usage_file.
541         (lang_dependent_init): Do not open the .su file if generating LTO.
543 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
545         PR rtl-optimization/87918
546         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
547         simplify_gen_relational rather than simplify_gen_binary.
549 2018-11-13  Richard Biener  <rguenther@suse.de>
551         * tree-ssanames.h (set_range_info): Use value_range_base.
552         (get_range_info): Likewise.
553         * tree-ssanames.c (set_range_info): Likewise.
554         (get_range_info): Likewise.
555         * tree-vrp.c (value_range_base::union_helper): Split
556         out common parts of value_range[_base]::union_.
557         (value_range_base::union_): Update.
558         (value_range::union_): Likewise.
559         (determine_value_range_1): Use value_range_base.
560         (determine_value_range): Likewise.
561         * tree-vrp.h (value_range_base::union_helper): Move ...
562         (value_range::union_helper): ... from here.
564 2018-11-13  Alan Modra  <amodra@gmail.com>
566         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
567         offset for PRE_DEC.
568         (rs6000_secondary_reload_gpr): Don't call find_replacement.
570 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
572         PR middle-end/59634
573         * doc/invoke.texi (Optimize Options): Clarify that the
574         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
575         apply to data cache size.
577 2018-11-13  Alan Modra  <amodra@gmail.com>
579         * config/rs6000/predicates.md (logical_const_operand),
580         (logical_operand): Correct comment.
581         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
583 2018-11-13  Alan Modra  <amodra@gmail.com>
585         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
587 2018-11-13  Alan Modra  <amodra@gmail.com>
589         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
590         description.
591         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
592         covered by alternative.
593         (movcc_internal1): Ignore h for register preference.
594         (mov<mode>_hardfloat64): Likewise.
595         (mov<mode>_softfloat): Ignore c, l, h for register preference.
597 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
599         PR preprocessor/47823
600         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
601         documentation to...
602         (Pragmas): ...here.
603         * doc/extend.texi (Pragmas): Note additional pragmas documented
604         in the CPP manual.
606 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
608         PR middle-end/86677
609         PR middle-end/87528
610         * tree-scalar-evolution.c (expression_expensive_p): Make
611         BUILTIN POPCOUNT as expensive when backend does not define it.
613 2018-11-12  Fredrik Noring  <noring@nocrew.org>
615         * config.gcc: Update with-llsc defaults for MIPS r5900.
617 2018-11-12  Martin Liska  <mliska@suse.cz>
619         PR gcov-profile/87442
620         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
621         options.
622         * doc/invoke.texi: Document them.
623         * tree-profile.c (parse_profile_filter): New.
624         (parse_profile_file_filtering): Likewise.
625         (release_profile_file_filtering): Likewise.
626         (include_source_file_for_profile): Likewise.
627         (tree_profiling): Filter source files based on the
628         newly added options.
630 2018-11-12  Martin Liska  <mliska@suse.cz>
632         PR target/87903
633         * doc/extend.texi: Add missing values for __builtin_cpu_is and
634         __builtin_cpu_supports for x86 target.
636 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
638         PR middle-end/21110
639         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
640         have to be Pmode.
641         (Arithmetic): Likewise for "lo_sum".
643 2018-11-12  Renlin Li  <renlin.li@arm.com>
645         PR target/87815
646         * dse.c (get_stored_val): Add check for compile-time constantness
647         of gap.
649 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
651         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
652         (ARMv8_5a): New fgroup.
653         (armv8.5-a): New arch.
654         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
655         sb and predres.
656         * config/arm/arm-tables.opt: Regenerate.
657         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
658         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
659         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
660         Add matching rules for -march=armv8.5-a and extensions.
661         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
662         Add sb and predres to all armv8-a except armv8.5-a.
664 2018-11-12  Richard Biener  <rguenther@suse.de>
666         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
667         overload for single value.
668         (value_range[_base]::set_nonnull): New.
669         (value_range[_base]::set_null): Likewise.
670         (value_range): Document bitmap copying behavior, mark
671         copy constructor and assignment operator deleted.
672         (value_range::move): New.
673         (value_range::set_and_canonicalize): Default bitmap to zero.
674         (set_value_range_to_nonnull): Remove.
675         (set_value_range_to_null): Likewise.
676         (set_value_range): Likewise.
677         (set_value_range_to_value): Likewise.
678         (extract_range_from_unary_expr): Work on value_range_base.
679         (extract_range_from_binary_expr_1): Likewise.  Rename to...
680         (extract_range_from_binary_expr): ... this.
681         * tree-vrp.c (value_range::update): Clear equiv bitmap
682         if required.
683         (value_range::move): New, move equiv bitmap.
684         (value_range_base::set_undefined): Avoid assignment.
685         (value_range::set_undefined): Likewise.
686         (value_range_base::set_varying): Likewise.
687         (value_range::set_varying): Likewise.
688         (set_value_range): Remove.
689         (value_range_base::set): New overload for value.
690         (value_range::set): Likewise.
691         (set_value_range_to_nonnull): Remove.
692         (value_range_base::set_nonnull): New.
693         (value_range::set_nonnull): Likewise.
694         (set_value_range_to_null): Remove.
695         (value_range_base::set_null): New.
696         (value_range::set_null): Likewise.
697         (range_is_null): Work on value_range_base.
698         (range_is_nonnull): Likewise.
699         (ranges_from_anti_range): Likewise.
700         (extract_range_into_wide_ints): Likewise.
701         (extract_range_from_multiplicative_op): Likewise.
702         (extract_range_from_binary_expr): Likewise.  Update for API changes.
703         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
704         handling.
705         (value_range::intersect_helper): Avoid copy and assignment.
706         (value_range::union_helper): Likewise.
707         (determine_value_range_1): Adjust.
708         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
709         Avoid assignment by using move.
710         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
711         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
712         Likewise.
713         * tree-ssanames.c (get_range_info): Likewise.
714         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
715         * vr-values.c (vr_values::get_value_range): Adjust.
716         (vr_values::update_value_range): Likewise.
717         (symbolic_range_based_on_p): Work on value_range_base.
718         (vr_values::extract_range_from_binary_expr): Use value_range_base.
719         (vr_values::extract_range_from_unary_expr): Likewise.
720         (vr_values::extract_range_from_cond_expr): Avoid assignment.
721         (vr_values::extract_range_from_comparison): Adjust.
722         (vr_values::check_for_binary_op_overflow): Use value_range_base.
723         (vr_values::extract_range_basic): Adjust.
724         (vr_values::adjust_range_with_scev): Likewise.
725         (vr_values::vrp_visit_assignment_or_call): Likewise.
726         (vr_values::get_vr_for_comparison): Change API to avoid
727         assignment and copy construction.
728         (vr_values::compare_name_with_value): Adjust accordingly.
729         (vr_values::compare_names): Likewise.
730         (vr_values::extract_range_from_phi_node): Avoid assignment and
731         bogus in-place modify of equiv bitmap.
732         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
733         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
734         for extract_range_from_unary_expr API change.
735         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
737 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
739         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
741 2018-11-12  Richard Biener  <rguenther@suse.de>
743         * tree-vrp.h (value_range_base::symbolic_p,
744         value_range_base::constant_p, value_range_base::zero_p,
745         value_range_base::singleton_p): Move from value_range.
746         (value_range::dump): Add.
747         * gimple-ssa-evrp-analyze.c
748         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
749         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
750         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
751         Use set_varying.
752         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
753         (value_range::constant_p): Likewise.
754         (value_range::singleton_p): Likewise.
755         (value_range_base::dump): Add.
756         (set_value_range_to_undefined): Remove.
757         (set_value_range_to_varying): Likewise.
758         (range_int_cst_p): Take value_range_base argument.
759         (range_int_cst_singleton_p): Likewise.
760         (value_range_constant_singleton): Likewise.
761         (vrp_set_zero_nonzero_bits): Likewise.
762         (extract_range_from_multiplicative_op): Use set_varying.
763         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
764         (extract_range_from_unary_expr): Likewise.
765         (dump_value_range_base): Change to overload of dump_value_range.
766         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
767         (vrp_prop::visit_stmt): Likewise.
768         (value_range::intersect_helper): Likewise.
769         (value_range::union_helper): Likewise.
770         (determine_value_range_1): Likewise.
772 2018-11-12  Richard Biener  <rguenther@suse.de>
774         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
775         (set_value_range_to_null): Likewise.
776         * vr-values.c (vr_values::extract_range_from_comparison):
777         Clear equiv for constant singleton ranges.
779 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
781         * config/i386/sse.md: Combine VFIXUPIMM* patterns
782         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
783         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
784         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
785         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
786         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
787         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
789 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
791         PR c/69502
792         * doc/extend.texi (Common Type Attributes): For the align type
793         attribute, copy language about decreasing alignment from the
794         corresponding variable attribute.
796 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
798         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
799         -frounding-math.
801 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
803         PR c++/43105
804         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
805         -frtti and -fno-rtti code.
807 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
809         PR c/26366
810         * doc/extend.texi (Other Builtins): Document probability associated
811         with __builtin_expect.
813 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
815         PR target/87928
816         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
817         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
818         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
819         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
821 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
823         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
825 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
827         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
829 2018-11-11  Richard Biener  <rguenther@suse.de>
831         * tree-vrp.h (class value_range_base): New base class for
832         value_range containing all but the m_equiv member.
833         (dump_value_range_base): Add.
834         (range_includes_zero_p): Work on value_range_base.
835         * tree-vrp.c (value_range_base::set): Split out base handling
836         from...
837         (value_range::set): this.
838         (value_range::set_equiv): New.
839         (value_range_base::value_range_base): New constructors.
840         (value_range_base::check): Split out base handling from...
841         (value_range::check): this.
842         (value_range::equal_p): Refactor in terms of
843         ignore_equivs_equal_p which is now member of the base.
844         (value_range_base::set_undefined): New.
845         (value_range_base::set_varying): Likewise.
846         (value_range_base::dump):Split out base handling from...
847         (value_range::dump): this.
848         (value_range_base::set_and_canonicalize): Split out base handling
849         from...
850         (value_range::set_and_canonicalize): this.
851         (value_range_base::union_): New.
852         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
853         for m_vr.
854         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
855         instead of value_range everywhere.
856         (ipcp_vr_lattice::print): Use dump_value_range_base.
857         (ipcp_vr_lattice::meet_with): Adjust.
858         (ipcp_vr_lattice::meet_with_1): Likewise.
859         (ipa_vr_operation_and_type_effects): Likewise.
860         (propagate_vr_across_jump_function): Likewise.
861         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
862         (ipa_get_value_range): Likewise.
863         (ipa_set_jfunc_vr): Likewise.
864         (ipa_compute_jump_functions_for_edge): Likewise.
866 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
868         PR middle-end/65703
869         * doc/invoke.texi (Optimize Options): Add @opindex entries
870         for the positive forms of -fno-xxx and -mno-xxx options
871         that were lacking them.
873 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
875         * combine.c (make_more_copies): Only make an intermediate copy if the
876         dest of a move is a pseudo.
878 2018-11-09  Maya Rashish  <coypu@sdf.org>
880         PR target/87221
881         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
882         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
884 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
886         PR driver/41179
887         PR middle-end/65703
888         * doc/invoke.texi (Optimize Options): Clarify default behavior
889         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
891 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
893         PR tree-optimization/87940
894         * expr.c (string_constant): Don't strip NOPS in subexpressions.
895         Fold PLUS_EXPR correctly.
897 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
899         PR target/87762
900         * config/s390/s390.c (s390_safe_relative_long_p): New function.
901         (annotate_constant_pool_refs): Skip insns which support
902         relative addressing.
903         (annotate_constant_pool_refs_1): New helper function.
904         (find_constant_pool_ref): Skip insns which support relative
905         addression.
906         (find_constant_pool_ref_1): New helper function.
907         (replace_constant_pool_ref): Skip insns which support
908         relative addressing.
909         (replace_constant_pool_ref_1): New helper function.
910         (s390_mainpool_start): Adapt to the new signature.
911         (s390_mainpool_finish): Likewise.
912         (s390_chunkify_start): Likewise.
913         (s390_chunkify_finish): Likewise.
914         (pass_s390_early_mach::execute): Likewise.
915         (s390_prologue_plus_offset): Likewise.
916         (s390_emit_prologue): Likewise.
917         (s390_emit_epilogue): Likewise.
919 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
921         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
922         but unsupported lastprivate with conditional modifier.
924 2018-11-09  Jeff Law  <law@redhat.com>
926         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
927         unused argument better.  Add gcc_unreachable to silence warning.
929 2018-11-09  Martin Sebor  <msebor@redhat.com>
931         PR middle-end/81824
932         * attribs.c (has_attribute): New helper function.
933         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
934         * attribs.h (decls_mismatched_attributes): Declare.
935         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
936         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
937         * common.opt (-Wattribute-alias): Take an argument.
938         (-Wno-attribute-alias): New option.
939         * doc/extend.texi (Common Function Attributes): Document copy.
940         (Common Variable Attributes): Same.
941         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
942         (-Wattribute-alias): Document new option argument.
944 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
946         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
947         matching criteria.  Remove unused array initializer.
949 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
950             Jinsong Ji  <jji@us.ibm.com>
952         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
953         constraints by introducing a new temporary.
954         (_mm_cvtss_si64): Likewise.
956 2018-11-09  Martin Liska  <mliska@suse.cz>
958         * common.opt: Add -fipa-stack-alignment flag.
959         * doc/invoke.texi: Document it.
960         * final.c (rest_of_clean_state): Guard stack
961         shrinking with flag.
963 2018-11-09  Martin Liska  <mliska@suse.cz>
965         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
966         to ...
967         (ipa_discover_variable_flags): ... this.
968         * common.opt: Come up with new flag -fipa-reference-addressable.
969         * doc/invoke.texi: Document it.
970         * ipa-reference.c (propagate): Call the renamed fn.
971         * ipa-visibility.c (whole_program_function_and_variable_visibility):
972         Likewise.
973         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
974         ...
975         (ipa_discover_variable_flags): ... this.  Discover
976         non-addressable variables only with the newly added flag.
977         * opts.c: Enable the newly added flag with -O1 and higher
978         optimization level.
980 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
982         * json.cc (selftest::test_writing_literals): Fix comment.
984 2018-11-09  Martin Liska  <mliska@suse.cz>
986         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
987         string to a stack buffer.
988         (aarch64_parse_cpu): Likewise.
989         (aarch64_parse_tune): Likewise.
991 2018-11-09  Richard Biener  <rguenther@suse.de>
993         PR tree-optimization/87953
994         * tree-vect-loop.c (vectorizable_reduction): For analysis
995         always pass ops[0] to vectorizable_condition.
997 2018-11-09  Stafford Horne  <shorne@gmail.com>
998             Richard Henderson  <rth@twiddle.net>
999             Joel Sherrill  <joel@rtems.org>
1001         * common/config/or1k/or1k-common.c: New file.
1002         * config/or1k/*: New.
1003         * config.gcc (or1k*-*-*): New.
1004         * configure.ac (or1k*-*-*): New test for openrisc tls.
1005         * configure: Regenerated.
1006         * doc/install.texi: Document OpenRISC triplets.
1007         * doc/invoke.texi: Document OpenRISC arguments.
1008         * doc/md.texi: Document OpenRISC.
1010 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1012         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1013         (arm7tdmi-s): Delete CPU.
1014         (arm710t): Add aliases for arm720t and arm740t.
1015         (arm720t, arm740t): Delete CPUs.
1016         (arm920t): Add aliases for arm920, arm922t and arm940t.
1017         (arm920, arm922t, arm940t): Delete CPUs.
1018         (arm10tdmi): Add alias for arm1020t.
1019         (arm1020t): Delete CPU.
1020         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1021         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1022         (arm10e): Add aliases for arm1020e and arm1022e.
1023         (arm1020e, arm1022e): Delete CPU.
1024         * config/arm/arm.md (generic_sched): Remove entries that are now
1025         handled by aliases.
1026         (generic_vfp): Likewise.
1027         * config/arm/arm1020e.md: Simplify tuning selection based on alias
1028         changes.
1029         * config/arm/arm-tune.md: Regenerated.
1030         * config/arm/arm-tables.opt: Regenerated.
1032 2018-11-09  Richard Biener  <rguenther@suse.de>
1034         PR tree-optimization/87621
1035         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1036         op with only phi inputs.
1037         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1038         (ch_base::copy_headers): Run CSE on copied loop headers.
1039         (pass_ch_vect::process_loop_p): Simplify.
1041 2018-11-09  Alexandre Oliva <oliva@adacore.com>
1043         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1044         for non-w64 x86_64 biarch.
1046 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
1048         PR rtl-optimization/86438
1049         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1050         of in_b for the compare if in_b is SET_DEST.
1052         PR target/87793
1053         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1054         non-toplevel UNSPEC.
1056 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1058         * tree-vrp.c (value_range::check): Do not access internals
1059         directly.
1060         (value_range::singleton_p): Same.
1061         (value_range::type): Same.
1062         (vrp_finalize): Use value_range API.
1064 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1066         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1068 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1070         * vr-values.c (vr_values::get_value_range): Use value_range API
1071         instead of piecing together ranges.
1072         (vr_values::update_value_range): Same.
1074 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1076         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
1077         of performing ad-hoc calculations.
1078         * tree-ssanames.c (set_range_info): New overloaded function
1079         accepting value_range &.
1080         (get_range_info): Same.
1081         * tree-ssanames.h (set_range_info_raw): Remove.
1082         (set_range_info): New prototype.
1083         (get_range_info): Same.
1084         * tree-vrp.h (value_range::null_p): Rename to zero_p.
1085         * tree-vrp.c (value_range::null_p): Same.
1087 2018-11-09  Jan Hubicka  <jh@suse.cz>
1089         * tree.c (fld_type_variant_equal_p): Test user align flag.
1090         (flt_type_variant): Copy user align flag.
1091         (fld_incomplete_type_of): Clear it.
1093 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1095         * config/arm/neon.md (div<mode>3): New pattern.
1097 2018-11-08  Andi Kleen  <ak@linux.intel.com>
1099         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
1100         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
1101         (ix86_handle_option): Handle OPT_mptwrite.
1102         * config/i386/cpuid.h (bit_PTWRITE): Add.
1103         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
1104         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
1105         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
1106         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
1107         (ix86_option_override_internal): Handle PTA_PTWRITE.
1108         (ix86_valid_target_attribute_inner_p): Define ptwrite.
1109         (def_builtin2): Force UINT64 to be 64bit only.
1110         * config/i386/i386.h (TARGET_PTWRITE): Add.
1111         (TARGET_PTWRITE_P): Add.
1112         (PTA_PTWRITE): Add.
1113         * config/i386/i386.md: Define ptwrite.
1114         * config/i386/i386.opt: Add -mptwrite.
1115         * config/i386/immintrin.h (_ptwrite64): Add.
1116         (_ptwrite32): Add
1117         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
1118         * doc/invoke.texi: Document -mptwrite.
1120 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
1122         PR rtl-optimization/87600
1123         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
1124         * lra-constraints.c (process_alt_operands): Skip illegal hard
1125         register usage.  Prefer reloading non hard register operands.
1127 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
1129         PR other/36572
1130         * doc/invoke.texi (Optimize Options): Clarify default behavior
1131         for -fno-sched-interblock and -fno-sched-spec.
1133 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
1135         * collect2.c (linker_select):  Add USE_LLD_LD.
1136         (ld_suffixes): Add ld.lld.
1137         (main): Handle -fuse-ld=lld.
1138         * common.opt (-fuse-ld=lld): New option.
1139         * doc/invoke.texi (-fuse-ld=lld): Document.
1140         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1142 2018-11-08  Paul Koning  <ni1d@arrl.net>
1144         * config/pdp11/constraints.md: Add "Z" series constraints for use
1145         with pre-dec and post-inc addressing.
1146         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1147         (pdp11_expand_operands): Add int argument (word count).
1148         (pdp11_sp_frame_offset): Delete.
1149         (pdp11_cmp_length): New function.
1150         (pushpop_regeq): New function.
1151         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1152         Add hook.
1153         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1154         frame layout.
1155         (pdp11_initial_elimination_offset): Ditto.
1156         (pdp11_expand_operands): Add word count argument.  Bugfixes.
1157         (output_move_multiple): Change how pointer adjustment is done.
1158         (pdp11_gen_int_label): Correct format.
1159         (output_ascii): Ditto.
1160         (pdp11_asm_output_var): Add code for DEC assembler case.
1161         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1162         value.
1163         (legitimate_const_double_p): Ditto.
1164         (pdp11_register_move_cost): Adjust for new register classes.
1165         (pdp11_regno_reg_class): Ditto.
1166         (expand_block_move): Delete.
1167         (pushpop_regeq): New function.
1168         (pdp11_legitimate_address_p): Bugfix in check for constant
1169         offset.
1170         (pdp11_sp_frame_offset): Delete.
1171         (pdp11_reg_save_size): New helper function for new frame layout.
1172         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1173         (pdp11_expand_shift): Bugfix in check for constant shift count.
1174         (pdp11_shift_length): Ditto.
1175         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1176         (pdp11_cmp_length): New function.
1177         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1178         some compile options.
1179         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1180         (CALL_USED_REGISTERS): Ditto.
1181         (ELIMINABLE_REGS): Ditto.
1182         (REGISTER_NAMES): Ditto.
1183         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1184         constraints.
1185         (REG_CLASS_NAMES): Ditto.
1186         (REG_CLASS_CONTENTS): Ditto.  Also remove
1187         HARD_FRAME_POINTER_REGNUM.
1188         (CPU_REG_CLASS): New macro.
1189         (CLASS_MAX_NREGS): Adjust for new register classes.
1190         (FUNCTION_PROFILER): Make no-op.
1191         (may_call_alloca): Remove unused declaration.
1192         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1193         (ASM_OUTPUT_SKIP): Fix format.
1194         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1195         (HARD_FRAME_POINTER_REGNUM): Remove.
1196         (return): Delete.
1197         (*rts): Rename.  Remove epilogue related checks.
1198         (cmpsi, cmpdi): New insn.
1199         (cbranch<mode>4): Change to apply to SI and DI modes as well.
1200         (mov<mode>): Change constraints to enforce that push/pop
1201         destination cannot use the same register as source.
1202         (*mov<mode><cc_cc>): Ditto.
1203         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1204         at assembly output rather than as RTL expander.
1205         (zero_extendqihi2): Bugfix in check for same registers.
1206         (adddi3_nocc): Bugfix in check for constant operand.
1207         (addsi3_nocc): Ditto.
1208         (subdi3_nocc): Ditto.
1209         (subsi3_nocc): Ditto.
1210         (negdi2_nocc): Copy input to pdp11_expand_operands.
1211         (negsi2_nocc): Ditto.
1212         (bswap2_nocc): Ditto.
1213         * config/pdp11/pdp11.opt (mlra): Fix documentation.
1214         * config/pdp11/t-pdp11: Use -Os.
1216 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
1218         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1219         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1220         (gen_comm_data): Print out CPU alias tables.
1221         (check_cpu): Match aliases when checking the CPU name.
1222         * config/arm/arm-protos.h (cpu_alias): New structure.
1223         (cpu_option): Add entry for aliases.
1224         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1225         strongarm1100 and strongarm1110.
1226         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1227         (config/arm/arm-generic.md): Remove redundant references to
1228         strongarm110, strongarm1100 and strongarm1110.
1229         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1230         Scan aliases for additional hints.
1231         (arm_parse_cpu_option_name): Also match a cpu name against the list
1232         of aliases.
1233         * config/arm/arm-tables.opt: Regenerated.
1234         * config/arm/arm-tune.md: Regenerated.
1236 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1238         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1239         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1240         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1241         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1242         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1243         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1244         * gengtype.c (open_base_files): Add omp-general.h.
1245         * gimple.c (gimple_build_omp_critical):
1246         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1247         gimple_omp_taskgroup_set_clauses.
1248         (gimple_build_omp_atomic_load): Add mo argument, call
1249         gimple_omp_atomic_set_memory_order.
1250         (gimple_build_omp_atomic_store): Likewise.
1251         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1252         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1253         instead of GSS_OMP.
1254         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1255         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1256         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1257         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1258         different value for GF_OMP_ATOMIC_NEED_VALUE.
1259         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1260         comments.
1261         (struct gimple_statement_omp_single_layout): And remove here.
1262         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1263         than gimple_statement_omp_single_layout.
1264         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1265         GIMPLE_OMP_TEAMS.
1266         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1267         (gimple_omp_subcode): Formatting fix.
1268         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1269         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1270         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1271         gimple_omp_teams_host, gimple_omp_teams_set_host,
1272         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1273         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1274         gimple_omp_taskgroup_set_clauses): New inline functions.
1275         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1276         (gimple_build_omp_atomic_store): Likewise.
1277         (gimple_omp_atomic_seq_cst_p): Remove.
1278         (gimple_omp_atomic_memory_order): New function.
1279         (gimple_omp_atomic_set_seq_cst): Remove.
1280         (gimple_omp_atomic_set_memory_order): New function.
1281         (gimple_build_omp_taskgroup): Add clauses argument.
1282         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1283         (dump_gimple_omp_task): Print taskwait with depend clauses.
1284         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1285         dump_omp_atomic_memory_order.
1286         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1287         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1288         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1289         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1290         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1291         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1292         ORT_UNTIED_TASKLOOP enumerators.
1293         (enum gimplify_defaultmap_kind): New.
1294         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1295         target_map_pointers_as_0len_arrays members, add defaultmap.
1296         (new_omp_context): Initialize defaultmap member.
1297         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1298         (maybe_fold_stmt): Don't fold even in host teams regions.
1299         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1300         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1301         ctx->omp_firstprivatize_variable.
1302         (omp_add_variable): Don't add private/firstprivate for VLAs in
1303         ORT_TASKGROUP.
1304         (omp_default_clause): Print "taskloop" rather than "task" if
1305         ORT_*TASKLOOP.
1306         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1307         Handle new defaultmap clause kinds.
1308         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1309         iterator to be lastprivate or private.  Fix up diagnostics if linear
1310         is used on collapse>1 simd iterator.
1311         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1312         (gimplify_omp_depend): New function.
1313         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1314         combined parallel master taskloop{, simd} if taskloop has
1315         firstprivate, lastprivate or reduction clause.  Handle
1316         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1317         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1318         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1319         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1320         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1321         lastprivate.
1322         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1323         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
1324         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1325         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1326         (gimplify_omp_task): Handle taskwait with depend clauses.
1327         (gimplify_omp_for): Add shared clause on parallel for combined
1328         parallel master taskloop{, simd} if taskloop has firstprivate,
1329         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1330         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1331         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1332         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1333         __for_end and __for_range temporaries on OMP_PARALLEL for
1334         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1335         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1336         sandwiched in between two taskloops.
1337         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1338         instead of ctx->omp_firstprivatize_variable.
1339         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1340         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1341         host teams, use gimplify_and_return_first etc. for body like
1342         for target or target data constructs, and at the end call
1343         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1344         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1345         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1346         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1347         gimple_omp_atomic_set_seq_cst calls.
1348         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1349         case, handle taskgroup clauses.
1350         * lto-streamer-out.c (hash_tree): Handle
1351         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1352         * Makefile.in (GTFILES): Add omp-general.h.
1353         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1354         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1355         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1356         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1357         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1358         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1359         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1360         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1361         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1362         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1363         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1364         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1365         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1366         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1367         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1368         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1369         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1370         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1371         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1372         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1373         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1374         non-worksharing loops.
1375         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1376         (determine_parallel_type): Don't combine parallel with worksharing
1377         which has _reductemp_ clause.
1378         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1379         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1380         if there is nonmonotonic modifier or if there is no modifier and no
1381         ordered clause.  For dynamic and guided schedule without monotonic
1382         and nonmonotonic modifier, default to nonmonotonic.
1383         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1384         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1385         task reductions.
1386         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1387         there are any reduction clauses.
1388         (expand_taskwait_call): New function.
1389         (expand_teams_call): New function.
1390         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1391         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1392         depend clauses.
1393         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1394         of worksharing loops with task reductions.
1395         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1396         expansion of worksharing loops with task reductions.
1397         (expand_omp_sections): Handle expansion of sections with task
1398         reductions.
1399         (expand_omp_synch): For host teams call expand_omp_taskreg.
1400         (omp_memory_order_to_memmodel): New function.
1401         (expand_omp_atomic_load, expand_omp_atomic_store,
1402         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1403         instead of gimple_omp_atomic_seq_cst_p.
1404         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1405         depend clauses as a standalone directive.
1406         * omp-general.c (enum omp_requires): New variable.
1407         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1408         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1409         GT_EXPR loops depending on incr sign.  Formatting fixes.
1410         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1411         (enum omp_requires): New enum.
1412         (omp_requires_mask): Declare.
1413         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1414         Fix comment typos.
1415         * omp-low.c (struct omp_context): Add task_reductions and
1416         task_reduction_map fields.
1417         (is_host_teams_ctx): New function.
1418         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1419         (use_pointer_for_field): Use is_global_var instead of
1420         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1421         in outer contexts.
1422         (build_outer_var_ref): Ignore taskgroup outer contexts.
1423         (delete_omp_context): Release task_reductions and task_reduction_map.
1424         (scan_sharing_clauses): Don't add any fields for reduction clause on
1425         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1426         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1427         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1428         Handle OMP_CLAUSE_NONTEMPORAL.
1429         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1430         needed.
1431         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1432         clauses with task modifier.
1433         (scan_omp_task): Handle taskwait with depend clauses.
1434         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1435         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1436         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1437         (scan_omp_for): Fix comment formatting.
1438         (scan_omp_teams): Handle host teams constructs.
1439         (check_omp_nesting_restrictions): Allow teams with no outer
1440         OpenMP context.  Adjust diagnostics for teams strictly nested into
1441         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1442         inside of simd regions.
1443         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1444         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1445         taskreg_nesting_level while scanning host teams construct.
1446         (task_reduction_read): New function.
1447         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1448         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1449         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1450         second argument create_tmp_var if it is NULL.  Don't ignore shared
1451         clauses in is_host_teams_ctx contexts.  Handle
1452         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1453         clauses.
1454         (lower_reduction_clauses): Ignore reduction clauses with task
1455         modifier.  Remove second argument create_tmp_var if it is NULL.
1456         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1457         (lower_send_clauses): Ignore reduction clauses with task modifier.
1458         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1459         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1460         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1461         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1462         contexts.
1463         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1464         lower_omp_task_reductions): New functions.
1465         (lower_omp_sections): Handle reduction clauses with taskgroup
1466         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1467         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1468         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1469         modifiers.
1470         (lower_omp_taskgroup): Handle taskgroup reductions.
1471         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1472         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1473         (lower_depend_clauses): If there are any
1474         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1475         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1476         seen, assume lowering is done already and return early.  Set kind
1477         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1478         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1479         parallel construct.  Handle reduction clause on taskloop construct.
1480         Handle taskwait with depend clauses.
1481         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1482         for host teams constructs.
1483         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1484         nontemporal and _reductemp_ clause entries.
1485         (omp_clause_code_name): Likewise.
1486         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1487         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1488         * tree-core.h (enum omp_clause_code): Add
1489         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1490         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1491         (struct tree_base): Add omp_atomic_memory_order field into union.
1492         Remove OMP_ATOMIC_SEQ_CST comment.
1493         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1494         and OMP_CLAUSE_DEPEND_DEPOBJ.
1495         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1496         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1497         OpenMP constructs with body and clauses operands.
1498         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1499         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1500         (OMP_TASKGROUP_CLAUSES): Define.
1501         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1502         OMP_CLAUSE__LOOPTEMP_.
1503         (OMP_ATOMIC_SEQ_CST): Remove.
1504         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1505         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1506         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1507         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1508         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1509         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1510         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1511         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1512         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1513         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1514         Define.
1515         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1516         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1517         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1518         (convert_local_omp_clauses): Likewise.  Remove useless test.
1519         * tree-parloops.c (create_call_for_reduction_1): Pass
1520         OMP_MEMORY_ORDER_RELAXED as new argument to
1521         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1522         * tree-pretty-print.c (dump_omp_iterators): New function.
1523         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1524         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1525         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1526         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1527         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1528         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1529         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1530         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1531         (dump_omp_atomic_memory_order): New function.
1532         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1533         taskwait with depend clauses.
1534         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1535         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1536         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1537         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1538         write_ts_omp_clause_tree_pointers): Likewise.
1540 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1542         PR ipa/86395
1543         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1544         "-missed", "-note", and "-all" sub-options.
1545         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1546         API.
1547         (can_early_inline_edge_p): Likewise.
1548         (want_early_inline_function_p): Likewise.
1549         (want_inline_self_recursive_call_p): Likewise.
1550         (recursive_inlining): Likewise.
1551         (inline_small_functions): Likewise.
1552         (flatten_function): Likewise.
1553         (ipa_inline): Likewise.
1554         (inline_always_inline_functions): Likewise.
1555         (early_inline_small_functions): Likewise.
1556         (early_inliner): Likewise.
1557         * tree-inline.c (expand_call_inline): Likewise.
1559 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1561         * pretty-print.c (pp_format): Handle %f.
1562         (selftest::test_pp_format): Add test of %f.
1563         * pretty-print.h (pp_double): New macro.
1565 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1567         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1568         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1569         (dump_pretty_printer::decode_format): Implement "%C" for
1570         cgraph_node *.
1571         (selftest::test_capture_of_dump_calls): Rename "where" to
1572         "stmt_loc".  Convert test_decl to a function decl and set its
1573         location.  Add a symbol_table_test RAII instance and a
1574         cgraph_node, using it to test "%C" and dump_symtab_node.
1576 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1578         PR middle-end/87916
1579         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1581 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1583         * cgraph.c: Include "selftest.h".
1584         (saved_symtab): New variable.
1585         (selftest::symbol_table_test::symbol_table_test): New ctor.
1586         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1587         (selftest::test_symbol_table_test): New test.
1588         (selftest::cgraph_c_tests): New.
1589         * cgraph.h (saved_symtab): New decl.
1590         (selftest::symbol_table_test): New class.
1591         * selftest-run-tests.c (selftest::run_tests): Call
1592         selftest::cgraph_c_tests.
1593         * selftest.h (selftest::cgraph_c_tests): New decl.
1595 2018-11-08  Richard Biener  <rguenther@suse.de>
1597         * tree-data-ref.h (lambda_int): New typedef.
1598         (lambda_vector_gcd): Adjust.
1599         (lambda_vector_new): Likewise.
1600         (lambda_matrix_new): Likewise.
1601         * tree-data-ref.c  (print_lambda_vector): Adjust.
1603 2018-11-08  Richard Biener  <rguenther@suse.de>
1605         PR tree-optimization/87929
1606         * tree-complex.c (expand_complex_comparison): Clean EH.
1608 2018-11-08  Martin Liska  <mliska@suse.cz>
1610         * doc/extend.texi: Reword.
1611         * predict.c (expr_expected_value_1): Likewise.
1613 2018-11-08  Richard Biener  <rguenther@suse.de>
1615         PR tree-optimization/87913
1616         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1617         of extreme values to ordered comparisons.
1619 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1621         PR middle-end/42726
1622         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1624 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1626         * doc/invoke.texi: Remove leading dash from @opindex entries
1627         throughout the file.
1629 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1631         PR driver/80828
1632         * doc/invoke.texi (Option Summary): Add -e and --entry.
1633         (Link Options): Likewise.
1635 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1637         PR 87926
1638         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1639         --disable-checking bootstrap.
1641 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1643         * configure: Regenerated.
1645 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1647         PR c/87691
1648         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1649         to the mode of the widest field iff the widest field has mode class
1650         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1651         reference.
1653 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1655         * common.opt: Add -Wattribute-warning.
1656         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1657         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1658         call to allow user configure behavior of "warning" attribute.
1660 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1662         * target.def: Put @: after every vs., e.g., and i.e. where it is
1663         followed by whitespace.
1664         * doc/extend.texi: Ditto.
1665         * doc/fragments.texi: Ditto.
1666         * doc/gimple.texi: Ditto.
1667         * doc/implement-c.texi: Ditto.
1668         * doc/install.texi: Ditto.
1669         * doc/invoke.texi: Ditto.
1670         * doc/md.texi: Ditto.
1671         * doc/plugins.texi: Ditto.
1672         * doc/rtl.texi: Ditto.
1673         * doc/sourcebuild.texi: Ditto.
1674         * doc/tm.texi.in: Ditto.
1675         * doc/ux.texi: Ditto.
1676         * doc/tm.texi: Regenerate.
1678 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1680         * config/arm/arm-cpus.in (ares): New entry.
1681         * config/arm/arm-tables.opt: Regenerate.
1682         * config/arm/arm-tune.md: Likewise.
1683         * doc/invoke.texi (ARM Options): Document ares.
1685 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1687         * config/aarch64/aarch64-cores.def (ares): Define.
1688         * config/aarch64/aarch64-tune.md: Regenerate.
1689         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1691 2018-11-07  Jan Hubicka  <jh@suse.cz>
1693         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1694         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1695         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1696         TYPE_DECL.
1698 2018-11-07  Richard Biener  <rguenther@suse.de>
1700         PR tree-optimization/87914
1701         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1702         of nested cycles.
1703         (vectorizable_reduction): Handle shifts and rotates by dispatching
1704         to vectorizable_shift.
1705         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1706         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1707         def cases.
1708         (vectorizable_shift): Export and handle being called as
1709         vect_nested_cycle.
1710         (vect_analyze_stmt): Call vectorizable_shift after
1711         vectorizable_reduction.
1712         * tree-vectorizer.h (vectorizable_shift): Declare.
1714 2018-11-07  Jan Hubicka  <jh@suse.cz>
1716         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1717         than const decls in TREE_VALUE of enum.
1718         (dump_type_inheritance_graph): Improve duplicate dumping.
1719         (free_enum_values): New.
1720         (build_type_inheritance_graph): Use it.
1721         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
1722         which are not main variants or not ODR types.
1723         (verify_type_variant): Expect variants to have no TYPE_VALUES.
1725 2018-11-07  Richard Biener  <rguenther@suse.de>
1727         * ipa-inline.c (want_inline_small_function_p): Compute
1728         big_speedup_p lazily and last.
1730 2018-11-07  Jan Hubicka  <jh@suse.cz>
1732         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
1733         building incomplete variant of complete type.
1734         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
1735         variant of complete type.
1737 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1739         * config/mips/mips.c: Fix typo in documentation of
1740         mips_loongson_ext2_prefetch_cookie.
1741         (mips_option_override): fix brain twister logical.
1742         * config/mips/mips.h: Fix typo in documentation of
1743         ISA_HAS_CTZ_CTO and define pattern.
1744         * config/mips/mips.md (prefetch): Hoist EXT2 above
1745         the 2EF/EXT block.
1746         (prefetch_indexed): Hoist EXT2 above the EXT block.
1748 2018-11-07  Jan Hubicka  <jh@suse.cz>
1750         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
1751         return and parameter types of function and method types.
1752         (free_lang_data_in_cgraph): Update.
1754 2018-11-07  Martin Liska  <mliska@suse.cz>
1756         PR rtl-optimization/87868
1757         * postreload-gcse.c (eliminate_partially_redundant_load): Set
1758         threshold to max_count if we would overflow.
1759         * profile-count.h: Make max_count a public constant.
1761 2018-11-07  Martin Liska  <mliska@suse.cz>
1763         * mem-stats.h: Fix GNU coding style.
1765 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1767         * config/mips/gs264e.md: New.
1768         * config/mips/mips-cpus.def: Define gs264e.
1769         * config/mips/mips-tables.opt: Regenerate.
1770         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1771         gs264e.
1772         (mips_issue_rate): Add support for gs264e.
1773         (mips_multipass_dfa_lookahead): Likewise.
1774         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
1775         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
1776         (MIPS_ASE_MSA_SPEC): New.
1777         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
1778         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
1779         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
1780         * config/mips/mips.md: Include gs264e.md.
1781         (processor): Add gs264e.
1782         * config/mips/mips.opt (MSA): Use Mask instead of Var.
1783         * doc/invoke.texi: Add gs264e to supported architectures.
1785 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1787         * config/mips/gs464e.md: New.
1788         * config/mips/mips-cpus.def: Define gs464e.
1789         * config/mips/mips-tables.opt: Regenerate.
1790         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1791         gs464e.
1792         (mips_issue_rate): Add support for gs464e.
1793         (mips_multipass_dfa_lookahead): Likewise.
1794         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
1795         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
1796         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
1797         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
1798         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
1799         * config/mips/mips.md: Include gs464e.md.
1800         (processor): Add gs464e.
1801         * doc/invoke.texi: Add gs464e to supported architectures.
1803 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1805         * config/mips/loongson3a.md: Rename to ...
1806         * config/mips/gs464.md: ... here.
1807         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
1808         as an alias of gs464 processor.
1809         * config/mips/mips-tables.opt: Regenerate.
1810         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
1811         instead of PROCESSOR_LOONGSON_3A.
1812         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
1813         TUNE_LOONGSON_3A.
1814         (mips_option_override): Enable MMI and EXT for gs464.
1815         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
1816         Rename TUNE_LOONGSON_3A to TUNE_GS464.
1817         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
1818         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
1819         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
1820         TARGET_LOONGSON_3A.
1821         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
1822         (processor): Add gs464;
1823         * doc/invoke.texi: Add gs464 to supported architectures.
1825 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1827         * config/mips/mips-protos.h
1828         (mips_loongson_ext2_prefetch_cookie): New prototype.
1829         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
1830         (mips_option_override): Enable TARGET_LOONGSON_EXT when
1831         TARGET_LOONGSON_EXT2 is true.
1832         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
1833         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
1834         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
1835         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
1836         TARGET_LOONGSON_EXT2.
1837         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
1838         (define_insn "ctz<mode>2"): New insn pattern.
1839         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
1840         (define_insn "prefetch_indexed_<mode>"): Include
1841         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
1842         * config/mips/mips.opt (-mloongson-ext2): Add option.
1843         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
1845 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1847         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
1848         __mips_loongson_ext.
1849         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
1850         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
1851         -mloongson-ext.
1852         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
1853         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
1854         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
1855         instead of TARGET_LOONGSON_3A.
1856         * config/mips/mips.opt (-mloongson-ext): Add option.
1857         * gcc/doc/invoke.texi (-mloongson-ext): Document.
1859 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1861         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
1862         * config/mips/loongson.md: Move to ...
1863         * config/mips/loongson-mmi.md: here; Adjustment.
1864         * config/mips/loongson.h: Move to ...
1865         State as deprecated. Include loongson-mmiintrin.h for back
1866         compatibility and warning.
1867         * config/mips/loongson-mmiintrin.h: ... here.
1868         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
1869         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
1870         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1871         (mips_option_override): Make sure MMI use hard float;
1872         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
1873         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
1874         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
1875         TARGET_LOONGSON_VECTORS.
1876         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
1877         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
1878         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
1879         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
1880         -mloongson-mmi.
1881         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
1882         TARGET_LOONGSON_VECTORS.
1883         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
1884         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1885         (Loongson MMI patterns): Include loongson-mmi.md instead of
1886         loongson.md.
1887         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
1888         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
1890 2018-11-07  Richard Biener  <rguenther@suse.de>
1892         PR lto/87906
1893         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
1894         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
1896 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
1898         PR rtl-optimization/87874
1899         * lra.c (lra_substitute_pseudo): Do not create a subreg for
1900         const wide ints.
1902 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
1904         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
1905         if not in indexed or indirect form.
1906         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
1907         (bswap<mode>2_store): Ditto.
1909 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
1911         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
1912         the UNSPEC.
1914 2018-11-06  Richard Biener  <rguenther@suse.de>
1916         PR tree-optimization/86850
1917         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
1918         instead of src.m_vec.
1920 2018-11-06  Jan Hubicka  <jh@suse.cz>
1922         * tree.c (fld_simplified_type_name): Break out form ...
1923         (free_lang_data_in_type): ... here.
1924         (fld_type_variant_equal_p): Use it.
1926 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1928         * config/default-d.c: Include memmodel.h.
1930         * config/sol2-d.c: New file.
1931         * config/t-sol2 (sol2-d.o): New rule.
1932         * config.gcc <*-*-solaris2*>: Set d_target_objs,
1933         target_has_targetdm.
1935 2018-11-06  Jan Hubicka  <jh@suse.cz>
1937         * tree.c (fld_type_variant): Also copy alignment; be sure that
1938         new variant is equal.
1940 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1942         PR target/87762
1943         * config/s390/s390.md: Add relative_long attribute.
1945 2018-11-06  Jan Hubicka  <jh@suse.cz>
1947         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
1948         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
1949         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
1950         stream TYPE_NEEDS_CONSTRUCTING.
1951         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
1952         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
1954 2018-11-06  Richard Biener  <rguenther@suse.de>
1956         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
1957         dump-scope ...
1958         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
1960 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1962         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
1963         REG_CLASS_CONTENTS[GEN_REGS].
1964         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
1966 2018-11-06  Jan Hubicka  <jh@suse.cz>
1968          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
1970 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1972         PR sanitizer/80953
1973         * config/sol2.h (ASAN_CC1_SPEC): Define.
1974         (LD_WHOLE_ARCHIVE_OPTION): Define.
1975         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
1976         (ASAN_REJECT_SPEC): Provide default.
1977         (LIBASAN_EARLY_SPEC): Define.
1978         (LIBTSAN_EARLY_SPEC): Define.
1979         (LIBLSAN_EARLY_SPEC): Define.
1980         * config/i386/sol2.h (CC1_SPEC): Redefine.
1981         (ASAN_REJECT_SPEC): Define.
1983         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
1984         (TARGET_ASAN_SHADOW_OFFSET): Define.
1985         (sparc_asan_shadow_offset): New function.
1986         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
1987         (ASAN_REJECT_SPEC): Define.
1989 2018-11-06  Jan Hubicka  <jh@suse.cz>
1991         * tree.c (fld_type_variant): Copy canonical type.
1992         (fld_incomplete_type_of): Check that canonical types looks sane;
1993         copy canonical type.
1994         (verify_type): Accept when incomplete type has complete canonical type.
1996 2018-11-06  Jan Hubicka  <jh@suse.cz>
1998         * tree.c (free_lang_data): Reset overwite_assembler_name,
1999         print_xnode, print_decl, print_type and print_identifier of
2000         langhooks.
2002 2018-11-06  Richard Biener  <rguenther@suse.de>
2004         PR tree-optimization/87889
2005         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2006         Do nothing if old and new arg are the same
2008 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
2010         PR target/87723
2011         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2012         attributes for operands 3 and 4.
2014 2018-11-06  Richard Biener  <rguenther@suse.de>
2016         PR middle-end/18041
2017         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2018         matching bitfield insertion.
2020 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
2022         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
2023         (reg_next_debug_use): New.
2024         (attempt_change): Propagate adjusted expression into affected
2025         debug insns.
2026         (merge_in_block): Track uses in debug insns.
2027         (pass_inc_dec::execute): Allocate and release
2028         reg_next_debug_use.
2030 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2032         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2033         (_mm512_fixupimm_round_pd): Update parameters and builtin.
2034         (_mm512_maskz_fixupimm_round_pd): Ditto.
2035         (_mm512_fixupimm_round_ps): Ditto.
2036         (_mm512_maskz_fixupimm_round_ps): Ditto.
2037         (_mm_fixupimm_round_sd): Ditto.
2038         (_mm_maskz_fixupimm_round_sd): Ditto.
2039         (_mm_fixupimm_round_ss): Ditto.
2040         (_mm_maskz_fixupimm_round_ss): Ditto.
2041         (_mm512_fixupimm_pd): Ditto.
2042         (_mm512_maskz_fixupimm_pd): Ditto.
2043         (_mm512_fixupimm_ps): Ditto.
2044         (_mm512_maskz_fixupimm_ps): Ditto.
2045         (_mm_fixupimm_sd): Ditto.
2046         (_mm_maskz_fixupimm_sd): Ditto.
2047         (_mm_fixupimm_ss): Ditto.
2048         (_mm_maskz_fixupimm_ss): Ditto.
2049         (_mm512_mask_fixupimm_round_pd): Update builtin.
2050         (_mm512_mask_fixupimm_round_ps): Ditto.
2051         (_mm_mask_fixupimm_round_sd): Ditto.
2052         (_mm_mask_fixupimm_round_ss): Ditto.
2053         (_mm512_mask_fixupimm_pd): Ditto.
2054         (_mm512_mask_fixupimm_ps): Ditto.
2055         (_mm_mask_fixupimm_sd): Ditto.
2056         (_mm_mask_fixupimm_ss): Ditto.
2057         * config/i386/avx512vlintrin.h:
2058         (_mm256_fixupimm_pd): Update parameters and builtin.
2059         (_mm256_maskz_fixupimm_pd): Ditto.
2060         (_mm256_fixupimm_ps): Ditto.
2061         (_mm256_maskz_fixupimm_ps): Ditto.
2062         (_mm_fixupimm_pd): Ditto.
2063         (_mm_maskz_fixupimm_pd): Ditto.
2064         (_mm_fixupimm_ps): Ditto.
2065         (_mm_maskz_fixupimm_ps): Ditto.
2066         (_mm256_mask_fixupimm_pd): Update builtin.
2067         (_mm256_mask_fixupimm_ps): Ditto.
2068         (_mm_mask_fixupimm_pd): Ditto.
2069         (_mm_mask_fixupimm_ps): Ditto.
2070         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
2071         * config/i386/i386-builtin.def: Update builtin definitions.
2072         * config/i386/i386.c: Handle new builtin types and remove useless ones.
2073         * config/i386/sse.md: Update VFIXUPIMM* patterns.
2074         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2075         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2076         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
2077         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2078         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2079         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
2080         * config/i386/subst.md:
2081         (round_saeonly_sd_mask_operand4): Add new subst_attr.
2082         (round_saeonly_sd_mask_op4): Ditto.
2083         (round_saeonly_expand_operand5): Ditto.
2084         (round_saeonly_expand): Update.
2086 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
2088         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2090 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
2092         PR rtl-optimization/87871
2093         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
2095 2018-11-05  Paul Koning  <ni1d@arrl.net>
2097         * doc/sourcebuild.texi (target attributes): Document new "inf"
2098         effective target keyword.
2100 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2102         * config/s390/s390.c (s390_register_move_cost): Increase costs for
2103         moves involving the CC reg.
2105 2018-11-05  Richard Biener  <rguenther@suse.de>
2107         PR tree-optimization/87873
2108         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
2109         argument.
2110         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
2111         * tree-vect-loop.c (vect_transform_loop): When splitting the
2112         loop exit also create forwarder PHIs for constants.
2113         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2114         Handle constant to_arg, add extra checking we match up the correct
2115         PHIs.
2117 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2119         * config/s390/s390.md: QImode and HImode for load on condition.
2121 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2123         * config/s390/predicates.md: Fix typo.
2124         * config/s390/s390.md: Allow immediates for load on condition.
2126 2018-11-05  Martin Liska  <mliska@suse.cz>
2128         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
2129         * bitmap.h (struct bitmap_usage): Likewise.
2130         * ggc-common.c (SCALE): Remove.
2131         (LABEL): Likewise.
2132         (struct ggc_usage): Use SIZE_AMOUNT. And update
2133         compare method.
2134         * ggc-page.c (SCALE): Remove.
2135         (STAT_LABEL): Remove.
2136         (ggc_print_statistics): Use SIZE_AMOUNT.
2137         * gimple.h (SCALE): Remove.
2138         (LABEL): Likewise.
2139         * input.c (ONE_K): Remove.
2140         (ONE_M): Likewise.
2141         (SCALE): Likewise.
2142         (STAT_LABEL): Likewise.
2143         (FORMAT_AMOUNT): Likewise.
2144         (dump_line_table_statistics): Use SIZE_AMOUNT.
2145         * mem-stats.h (struct mem_usage): Likewise.
2146         * rtl.c (dump_rtx_statistics): Likewise.
2147         (rtx_alloc_counts): Change type to size_t.
2148         (rtx_alloc_sizes): Likewise.
2149         (rtx_count_cmp): New.
2150         (dump_rtx_statistics): Sort first based on counts.
2151         * tree.c (tree_nodes_cmp): New.
2152         (tree_codes_cmp): New.
2153         (dump_tree_statistics): Sort first based on counts.
2154         * system.h (ONE_K): New.
2155         (ONE_M): Likewise.
2156         (SIZE_SCALE): Likewise.
2157         (SIZE_LABEL): Likewise.
2158         (SIZE_AMOUNT): Likewise.
2159         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2160         * tree-dfa.c (dump_dfa_stats): Likewise.
2161         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2162         * tree-ssanames.c (ssanames_print_statistics): Likewise.
2163         * tree.c (dump_tree_statistics): Likewise.
2164         * vec.c (struct vec_usage): Likewise.
2165         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2166         trigger a -Werror=format-overflow with
2167         --enable-gather-detailed-stats.
2169 2018-11-05  Martin Liska  <mliska@suse.cz>
2171         * mem-stats.h (mem_alloc_description::release_instance_overhead):
2172         Return T *.
2173         * vec.c (struct vec_usage): Register m_element_size.
2174         (vec_prefix::register_overhead): New arguments: elements and
2175         element_size.
2176         (vec_prefix::release_overhead): Subtract elements.
2177         * vec.h (struct vec_prefix): Change signature.
2178         (va_heap::reserve): Pass proper arguments.
2179         (va_heap::release): Likewise.
2181 2018-11-05  Martin Liska  <mliska@suse.cz>
2183         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2184         style.
2185         * vec.c: Likewise.
2187 2018-11-05  Richard Biener  <rguenther@suse.de>
2189         * tree-scalar-evolution.h (final_value_replacement_loop): Update
2190         prototype.
2191         * tree-scalar-evolution.c (final_value_replacement_loop): Return
2192         whether anything was done.
2193         (scev_const_prop): Remove constant propagation part, fold
2194         remains into ...
2195         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2196         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2197         conditionally.
2199 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
2201         PR tree-optimization/87859
2202         * gimple-ssa-store-merging.c (struct merged_store_group): Add
2203         only_constants and first_nonmergeable_order members.
2204         (merged_store_group::merged_store_group): Initialize them.
2205         (merged_store_group::do_merge): Clear only_constants member if
2206         adding something other than INTEGER_CST store.
2207         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2208         stores with order >= first_nonmergeable_order.  Use
2209         merged_store->only_constants instead of always recomputing it.
2210         Set merged_store->first_nonmergeable_order if we've skipped any
2211         stores.  Attempt to merge overlapping INTEGER_CST stores that
2212         we would otherwise skip.
2214         PR sanitizer/87837
2215         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2217 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2219         PR target/87853
2220         * config/i386/emmintrin.h (__v16qs): New to cope with option
2221         -funsigned-char.
2222         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2223         (_mm_cmplt_epi8): Likewise.
2224         (_mm_cmpgt_epi8): Likewise.
2226 2018-11-05  Richard Biener  <rguenther@suse.de>
2228         PR rtl-optimization/87852
2229         * fwprop.c (use_killed_between): Only consider single-defs of the
2230         use whose definition statement dominates the use.
2232 2018-11-05  Martin Liska  <mliska@suse.cz>
2234         PR web/87829
2235         * doc/invoke.texi: Remove options that are
2236         not disabled with -Os.
2238 2018-11-05  Martin Liska  <mliska@suse.cz>
2240         PR c/87811
2241         * doc/extend.texi: Update constrain about the last argument
2242         of __builtin_expect_with_probability.
2244 2018-11-05  Martin Liska  <mliska@suse.cz>
2246         PR c/87811
2247         * predict.c (expr_expected_value_1): Verify
2248         that last argument is a real constants and emit
2249         error.
2251 2018-11-05  Martin Liska  <mliska@suse.cz>
2253         PR gcov-profile/77698
2254         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2255         only in LTO mode.
2257 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2259         PR tree-optimization/86572
2260         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2262 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2264         PR tree-optimization/87672
2265         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2266         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2268 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2270         PR middle-end/58372
2271         * cfgexpand.c (pass_expand::execute): Move the call to
2272         finish_eh_generation in front of the call to expand_stack_alignment.
2274 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2276         * common/config/i386/i386-common.c (processor_alias_table): Add
2277         znver2 entry.
2278         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2279         (case ${target}): Add znver2.
2280         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2281         -march=native recognize znver2 processors.
2282         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2283         * config/i386/i386.c (m_znver2): New definition.
2284         (m_ZNVER): New definition.
2285         (m_AMD_MULTIPLE): Includes m_znver2.
2286         (processor_cost_table): Add znver2 entry.
2287         (processor_target_table): Add znver2 entry.
2288         (get_builtin_code_for_version): Set priority for
2289         PROCESSOR_ZNVER2.
2290         (processor_model): Add M_AMDFAM17H_ZNVER2.
2291         (arch_names_table): Ditto.
2292         (ix86_reassociation_width): Include znver2.
2293         * config/i386/i386.h (TARGET_znver2): New definition.
2294         (struct ix86_size_cost): Add TARGET_ZNVER2.
2295         (enum processor_type): Add PROCESSOR_ZNVER2.
2296         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2297         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2298         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2299         (ix86_adjust_cost): Add znver2.
2300         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2301         * gcc/doc/extend.texi: Add details about znver2.
2302         * gcc/doc/invoke.texi: Add details about znver2.
2304 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2306         PR target/87079
2308         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2309         pattern.
2311 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2313         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2314         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2316 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2318         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2319         shrink-wrapping when -mtrack-speculation.
2321 2018-11-02  Richard Biener  <rguenther@suse.de>
2323         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2324         (pop_cost_one_pair): Do not free pair.
2325         (pop_best_coalesce): Likewise.
2326         (create_coalesce_list): Initialize obstack.
2327         (delete_coalesce_list): Free obstack.
2328         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2329         (add_cost_one_coalesce): Likewise.
2330         (struct live_track): Remove bitmap pointer indirections.
2331         (new_live_track): Adjust.
2332         (delete_live_track): Likewise.
2333         (live_track_remove_partition): Likewise.
2334         (live_track_add_partition): Likewise.
2335         (live_track_live_p): Likewise.
2336         (live_track_process_def): Likewise.
2337         (live_track_clear_base_vars): Likewise.
2339 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2341         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2342         * configure: Regenerate.
2343         * config.in: Regenerate.
2344         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2345         HAVE_AS_REGISTER_PSEUDO_OP guard.
2346         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2347         !HAVE_AS_REGISTER_PSEUDO_OP support.
2348         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2349         guard.
2351 2018-11-02  Richard Biener  <rguenther@suse.de>
2353         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2354         with is_type_die.
2356 2018-11-02  Richard Biener  <rguenther@suse.de>
2358         PR tree-optimization/87776
2359         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2360         executable when iterating but running into rpo-vn-max-loop-depth
2361         and not eliding the iteration.
2363 2018-11-30  Jan Hubicka  <jh@suse.cz>
2365         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2366         TYPE_DECL.
2368 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2370         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2371         prototype.
2372         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2373         Combine with rs6000_address_for_fpconvert.
2374         (rs6000_address_for_fpconvert): Combine with
2375         rs6000_force_indexed_or_indirect_mem.
2376         (rs6000_expand_vector_init): Change function call from
2377         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2378         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2379         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2380         (floatsi<mode>2_lfiwax_mem): Ditto.
2381         (floatunssi<mode>2_lfiwzx): Ditto.
2382         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2383         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2384         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2385         (fix_trunc<mode>si2_stfiwx): Ditto.
2386         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2387         (float_<mode>si2_hw): Ditto.
2388         (floatuns_<mode>si2_hw): Ditto.
2389         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2390         (vsx_splat_<mode>): Ditto.
2392 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2394         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2395         at end of assembler input text.
2396         * configure: Regenerate.
2398 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2400         PR tree-optimization/87826
2401         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2402         negative or larger or equal to type's precision.
2404 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2406         * opts.c (default_options_table): Do not enable
2407         OPT_fdelayed_branch at -Og.
2408         * doc/invoke.texi (-fdelayed-branch): Document it.
2410 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2412         * optabs-libfuncs.c (build_libfunc_function_visibility):
2413         New, split out from...
2414         (build_libfunc_function): ... here.
2415         (init_one_libfunc_visibility): New, split out from ...
2416         (init_one_libfunc): ... here.
2418         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2419         scratch register need not be early-clobber.  Document the reason
2420         why we cannot use ST<OP>.
2422 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2424         PR bootstrap/82856
2425         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2426         line for second argument of AC_DEFINE_UNQUOTED.
2427         * doc/install.texi (Tools/packages necessary for modifying GCC):
2428         Update to autoconf 2.69 and automake 1.15.1.
2429         * aclocal.m4, config.in, configure: Regenerate.
2431 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2433         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2434         initialization.
2436 2018-10-31  Martin Liska  <mliska@suse.cz>
2438         PR driver/83193
2439         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2440         Add new argument invalid_extension.
2441         (aarch64_get_all_extension_candidates): New function.
2442         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2443         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2444         new argument.
2445         (aarch64_get_all_extension_candidates): New function.
2446         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2447         argument invalid_extension.
2448         (aarch64_parse_cpu): Likewise.
2449         (aarch64_print_hint_for_extensions): New function.
2450         (aarch64_validate_mcpu): Provide hint about invalid extension.
2451         (aarch64_validate_march): Likewise.
2452         (aarch64_handle_attr_arch): Pass new argument.
2453         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2454         (aarch64_handle_attr_isa_flags): Likewise.
2456 2018-10-31  Richard Biener  <rguenther@suse.de>
2458         PR middle-end/70359
2459         PR middle-end/86270
2460         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2461         copy generation to useful cases.  Place the copy before
2462         the definition of the backedge value when possible.
2464 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2466         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2467         * config/arc/arc.c (arc_active_insn): New function.
2468         (check_store_cacheline_hazard): Likewise.
2469         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2470         (arc_override_options): Disable delay slot scheduler for older
2471         A7.
2472         (arc_store_addr_hazard_p): New implementation, old one renamed to
2473         ...
2474         (arc_store_addr_hazard_internal_p): Renamed.
2475         (arc_reorg): Don't combine into brcc instructions which are part
2476         of hardware hazard solution.
2477         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2478         (tune_arc700): Likewise.
2479         * config/arc/arc.opt (arc7xx): New tune value.
2480         * config/arc/arc700.md: Improve A7 scheduler.
2482 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2484         * config/arc/arc.c (arc_override_options): Remove
2485         TARGET_COMPACT_CASESI.
2486         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2487         (CASE_VECTOR_MODE): Likewise.
2488         (CASE_VECTOR_PC_RELATIVE): Likewise.
2489         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2490         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2491         (ADDR_VEC_ALIGN): Update.
2492         (ASM_OUTPUT_CASE_LABEL): Undefine.
2493         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2494         (TARGET_BI_BIH): Define.
2495         (DEFAULT_BRANCH_INDEX): Likewise.
2496         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2497         instructions, remove compact_casesi use case.
2498         (casesi_compact_jump): Remove.
2499         (casesi_dispatch): New pattern.
2500         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2501         compact_casesi option.
2502         * doc/invoke.texi: Document mbranch-index option.
2504 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2506         * config/arc/arc.c (arc_get_tp): Remove function.
2507         (arc_emit_call_tls_get_addr): Likewise.
2508         (arc_call_tls_get_addr): New function.
2509         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2510         * config/arc/arc.md (tls_load_tp_soft): Remove.
2511         (tls_gd_get_addr): Likewise.
2513 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2515         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2516         (mulsi3_600_lib): Remove pattern.
2517         (umulsi3_highpart_600_lib_le): Likewise.
2518         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2519         (umulsidi3): Remove call to umulsidi3_600_lib.
2520         (umulsidi3_600_lib): Remove pattern.
2521         (peephole2): Remove peephole using the above deprecated patterns.
2523 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2525         PR target/87374
2526         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2527         use of -mslow-flash-data and -mword-relocations.
2528         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2529         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2530         flag_pic.
2531         * doc/invoke.texi (-mword-relocations): Mention conflict with
2532         -mslow-flash-data.
2533         (-mslow-flash-data): Reciprocally.
2535 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2537         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2538         16-byte modes held in GP registers to use an even regno.
2540         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2541         (aarch64_atomic_ldop_supported_p): Remove.
2542         (aarch64_gen_atomic_ldop): Remove.
2543         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2544         Fully expand LSE operations here.
2545         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2546         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2547         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2548         and use ATOMIC_LDOP instead; use register_operand for the input;
2549         drop the split and emit insns directly.
2550         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2551         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2552         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2554         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2555         (aarch64_gen_atomic_ldop): Don't call it.
2556         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2557         Use aarch64_reg_or_zero.
2558         (aarch64_atomic_exchange<ALLI>): Likewise.
2559         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2560         operand 0; use aarch64_reg_or_zero for input; merge ...
2561         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2563         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2564         (aarch64_split_compare_and_swap): Use it.
2565         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2566         test oldval against the proper predicate.
2567         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2568         Use nonmemory_operand for expected.
2569         (cas_short_expected_pred): New.
2570         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2571         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2572         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2573         (aarch64_plushi_operand): New.
2575         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2576         Force oldval into the rval register for TARGET_LSE; emit the compare
2577         during initial expansion so that it may be deleted if unused.
2578         (aarch64_gen_atomic_cas): Remove.
2579         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2580         Change =&r to +r for operand 0; use match_dup for operand 2;
2581         remove is_weak and mod_f operands as unused.  Drop the split
2582         and merge with...
2583         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2584         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2585         (@aarch64_atomic_cas<GPI>): Similarly.
2587 2018-10-31  Richard Biener  <rguenther@suse.de>
2589         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2590         using ABSU_EXPR.
2592 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2594         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2595         * config/aarch64/aarch64.md: Include saphira.md
2596         * config/aarch64/saphira.md: New file for pipeline description.
2598 2018-10-30  Martin Sebor  <msebor@redhat.com>
2600         PR middle-end/87041
2601         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2602         inlining context.
2603         (sprintf_dom_walker::compute_format_length):
2604         Avoid setting POSUNDER4K here.
2605         (get_destination_size): Handle null argument values.
2606         (get_user_idx_format): New function.
2607         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2608         functions, including user-defined with attribute format printf.
2609         Use %G to include inlining context.
2610         Set POSUNDER4K here.
2612 2018-10-30  Jan Hubicka  <jh@suse.cz>
2614         * params.def (lto-partitions): Bump from 32 to 128.
2616 2018-10-30  Jan Hubicka  <jh@suse.cz>
2618         * tree.c
2619         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2620         head in file.
2621         (free_lang_data_in_type): Forward declare.
2622         (fld_type_variant_equal_p): New function.
2623         (fld_type_variant): New function
2624         (fld_incomplete_types): New hash.
2625         (fld_incomplete_type_of): New function
2626         (fld_simplfied-type): New function.
2627         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2628         (free_lang_data): Allocate and free fld_incomplete_type; update call
2629         of free_lang_data_in_decl.
2631 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2633         * gcov.c (output_lines): Remove duplicate line.
2635 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2637         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2638         if not in indexed or indirect form.
2639         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2640         (bswapdi2_store): Ditto.
2641         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2642         helper function.
2643         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2644         Prototype for helper function.
2646 2018-10-30  Martin Sebor  <msebor@redhat.com>
2648         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2649         (target, pragma GCC optimize, pragma GCC target): Ditto.
2651 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2653         * doc/extend.texi: Fix prototype and description of
2654         __builtin_expect_with_probability.
2656 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2658         * cgraph.h (clone_function_name_1): Replaced by new
2659           clone_function_name_numbered that takes name as string; for
2660           privatize_symbol_name_1 use only.
2661           (clone_function_name): Renamed to
2662           clone_function_name_numbered to be explicit about numbering.
2663           (clone_function_name): New two-argument function that does
2664           not number its output.
2665           (clone_function_name): New three-argument function that
2666           takes a number to append to its output.
2667         * cgraphclones.c (duplicate_thunk_for_node):
2668           (clone_function_name_1): Renamed.
2669           (clone_function_name_numbered): Two new functions.
2670           (clone_function_name): Improved documentation.
2671           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2672         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2673         * final.c (final_scan_insn_1): Use the new clone_function_name
2674           without numbering.
2675         * multiple_target.c (create_dispatcher_calls): Ditto.
2676           (create_target_clone): Ditto.
2677         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2678         * omp-low.c (create_omp_child_function_name): Ditto.
2679         * omp-simd-clone.c (simd_clone_create): Ditto.
2680         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2681           new clone_function_name without numbering.
2683 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2685         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2686         Assert that the allocation size is not zero.
2688 2018-10-30  Richard Biener  <rguenther@suse.de>
2690         PR tree-optimization/87800
2691         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2692         non-induction or reduction PHIs.
2694 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2696         * config/aarch64/falkor-tag-collision-avoidance.c
2697         (execute_tag_collision_avoidance): Call df_note_add_problem.
2699 2018-10-30  Martin Liska  <mliska@suse.cz>
2701         * doc/extend.texi: Fix typo in documentation
2702         of __builtin_expect_with_probability.
2704 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2706         PR c++/87721
2707         * input.c (get_substring_ranges_for_loc): Detect if
2708         linemap_resolve_location gives us a NULL map, and reject
2709         this case.
2711 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2713         * config.gcc (xstormy16-*-elf): Set tm_d_file.
2715 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2716             Martin Sebor  <msebor@redhat.com>
2717             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2719         * doc/ux.texi (Quoting): New subsection, adapted from material at
2720         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
2721         MartinSebor and ManuelLopezIbanez.
2722         (Fix-it hints): Note that fix-it hints shouldn't be marked for
2723         translation.
2725 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
2727         PR middle-end/87469
2728         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
2729         max value.
2731 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2733         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
2735 2018-10-29  Paul Koning  <ni1d@arrl.net>
2737         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
2739 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2741         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
2742         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
2743         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
2744         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
2745         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
2746         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
2747         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
2748         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
2749         Change 'vector' to '__vector'.
2750         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
2751         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
2752         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
2753         _mm_avg_pu16): Likewise.  And, whitespace corrections.
2755 2018-10-29  Richard Biener  <rguenther@suse.de>
2757         PR tree-optimization/87785
2758         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
2759         internal defs.
2761 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2763         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
2765 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2767         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
2768         ports configuration macro, defaults to "gnu".
2769         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
2770         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
2772 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2774         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
2776 2018-10-29  Richard Biener  <rguenther@suse.de>
2778         PR tree-optimization/87790
2779         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
2780         (vect_make_slp_decision): Adjust.
2781         (vect_slp_analyze_bb_1): Likewise.
2782         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
2783         edges.
2785 2018-10-29  Richard Biener  <rguenther@suse.de>
2787         PR tree-optimization/87785
2788         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
2789         and processing.
2790         (vect_build_slp_tree): Likewise.
2791         (vect_gather_slp_loads): New function.
2792         (vect_analyze_slp_instance): Gather loads separately from the
2793         SLP tree build.
2795 2018-10-29  Martin Liska  <mliska@suse.cz>
2797         * Makefile.in: Make dependency to json.o.
2798         * doc/gcov.texi: Document new JSON format, remove
2799         old intermediate format documentation.
2800         * gcov.c (struct function_info): Come up with m_name and
2801         m_demangled_name.
2802         (function_info::function_info): Initialize it.
2803         (function_info::~function_info): Release it.
2804         (main): Rename flag_intermediate_format to flag_json_format.
2805         (print_usage): Describe --json-format.
2806         (process_args): Set flag_json_format.
2807         (output_intermediate_line): Remove.
2808         (output_intermediate_json_line): Likewise.
2809         (get_gcov_intermediate_filename): Return new extension
2810         ".gcov.json.gz".
2811         (output_intermediate_file): Implement JSON emission.
2812         (output_json_intermediate_file): Implement JSON emission.
2813         (generate_results): Use ::get_name for function name.
2814         Handle JSON output file.
2815         (read_graph_file): Use ::get_name instead of cplus_demangle.
2816         (read_count_file): Likewise.
2817         (solve_flow_graph): Likewise.
2818         (add_line_counts): Likewise.
2819         (accumulate_line_counts): Use new flag_json_format.
2820         (output_function_details): Use ::get_name instead of cplus_demangle.
2821         (output_lines): Likewise.
2822         * json.cc (test_writing_literals): Add new tests.
2823         * json.h (class literal): Add new boolean constructor.
2825 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
2827         PR rtl-optimization/87701
2828         PR rtl-optimization/87780
2829         * combine.c (make_more_copies): Rewrite.
2831 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
2833         * doc/generic.texi (ABSU_EXPR): Document.
2834         * match.pd (absu(x)*absu(x) -> x*x): Handle.
2835         (absu(absu(X)) -> absu(X)): Likewise.
2836         (absu(-X) -> absu(X)): Likewise.
2837         (absu(X)  where X is nonnegative -> X): Likewise.
2839 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
2841         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
2842         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
2843         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
2844         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
2845         (s-tm-texi): Also check timestamp on d-target.def.
2846         (generated_files): Add TM_D_H and d-target-hooks-def.h.
2847         (build/genhooks.o): Also depend on D_TARGET_DEF.
2848         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
2849         variables.
2850         * config/aarch64/aarch64-d.c: New file.
2851         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
2852         Define.
2853         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
2854         prototype.
2855         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
2856         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
2857         * config/arm/arm-d.c: New file.
2858         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
2859         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
2860         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2861         * config/arm/t-arm (arm-d.o): New rule.
2862         * config/default-d.c: New file.
2863         * config/glibc-d.c: New file.
2864         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2865         * config/i386/i386-d.c: New file.
2866         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
2867         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
2868         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2869         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
2870         * config/i386/t-i386 (i386-d.o): New rule.
2871         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2872         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2873         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
2874         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2875         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2876         * config/mips/mips-d.c: New file.
2877         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
2878         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
2879         * config/mips/t-mips (mips-d.o): New rule.
2880         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2881         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2882         * config/powerpcspe/powerpcspe-d.c: New file.
2883         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
2884         New prototype.
2885         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2886         Support GNU D by using 0 as the language type.
2887         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
2888         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
2889         * config/riscv/riscv-d.c: New file.
2890         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
2891         prototype.
2892         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
2893         * config/riscv/t-riscv (riscv-d.o): New rule.
2894         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2895         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2896         * config/rs6000/rs6000-d.c: New file.
2897         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
2898         prototype.
2899         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2900         Support GNU D by using 0 as the language type.
2901         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
2902         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
2903         * config/s390/s390-d.c: New file.
2904         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
2905         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
2906         * config/s390/t-s390 (s390-d.o): New rule.
2907         * config/sparc/sparc-d.c: New file.
2908         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
2909         prototype.
2910         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
2911         * config/sparc/t-sparc (sparc-d.o): New rule.
2912         * config/t-glibc (glibc-d.o): New rule.
2913         * configure: Regenerated.
2914         * configure.ac (tm_d_file): New variable.
2915         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
2916         * doc/contrib.texi (Contributors): Add self for the D frontend.
2917         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
2918         * doc/install.texi (Configuration): Mention libphobos as an option for
2919         --enable-shared.  Mention d as an option for --enable-languages.
2920         (Testing): Mention check-d as a target.
2921         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
2922         name suffixes.  Mention d as a -x option.
2923         * doc/sourcebuild.texi (Top Level): Mention libphobos.
2924         * doc/standards.texi (Standards): Add section on D language.
2925         * doc/tm.texi: Regenerated.
2926         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
2927         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
2928         * dwarf2out.c (is_dlang): New function.
2929         (gen_compile_unit_die): Use DW_LANG_D for D.
2930         (declare_in_namespace): Return module die for D, instead of adding
2931         extra declarations into the namespace.
2932         (gen_namespace_die): Generate DW_TAG_module for D.
2933         (gen_decl_die): Handle CONST_DECLSs for D.
2934         (dwarf2out_decl): Likewise.
2935         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
2936         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
2937         kinds of aggregates.
2938         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
2939         * genhooks.c: Include d/d-target.def.
2941 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
2943         PR target/85669
2944         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2945         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2946         (STACK_DYNAMIC_OFFSET): Likewise.
2948 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
2950         PR target/80024
2951         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
2952         error message.
2954 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
2956         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
2957         return type and other typos.
2959 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
2961         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2962         a shorter sequence with fewer branches.
2963         (emit_final_str_compare_gpr): Ditto.
2965 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2967         * config/rs6000/tmmintrin.h: New file.
2968         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
2970 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2972         * config/rs6000/mmintrin.h: Enable 32bit compilation.
2973         * config/rs6000/xmmintrin.h: Likewise.
2975 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2977         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2979 2018-10-26  Richard Biener  <rguenther@suse.de>
2981         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
2982         and wrapper.
2983         (vect_mark_slp_stmts_relevant): Likewise.
2984         (vect_detect_hybrid_slp_stmts): Likewise.
2985         (vect_bb_slp_scalar_cost): Likewise.
2986         (vect_remove_slp_scalar_calls): Likewise.
2988 2018-10-26  Jan Hubicka  <jh@suse.cz>
2990         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
2991         (warn_types_mismatch): Fix walk of DECL_NAME.
2992         (odr_types_equivalent_p): Fix overactive assert.
2994 2018-10-26  Richard Biener  <rguenther@suse.de>
2996         PR tree-optimization/87105
2997         * tree-vectorizer.h (_slp_tree::refcnt): New member.
2998         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
2999         refcnt.
3000         (vect_create_new_slp_node): Initialize refcnt to one.
3001         (bst_traits): Move.
3002         (scalar_stmts_set_t, bst_fail): Remove.
3003         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3004         (vect_build_slp_tree): Add bst_map argument and lookup
3005         already created SLP nodes.
3006         (vect_print_slp_tree): Handle a SLP graph, print SLP node
3007         addresses.
3008         (vect_slp_rearrange_stmts): Handle a SLP graph.
3009         (vect_analyze_slp_instance): Adjust and free SLP nodes from
3010         the CSE map.  Fix indenting.
3011         (vect_schedule_slp_instance): Add short-cut.
3013 2018-10-26  Martin Liska  <mliska@suse.cz>
3015         PR testsuite/86158
3016         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3017         addr_expr and not with pointers.
3019 2018-10-26  Jan Hubicka  <jh@suse.cz>
3021         * tree.c (free_lang_data_in_type): Only check main variants.
3022         * ipa-devirt.c (warn_odr): Make static.
3023         (types_same_for_odr): Drop strict variant.
3024         (types_odr_comparable): Likewise.
3025         (odr_or_derived_type_p): Look for main variants.
3026         (odr_name_hasher::equal): Cleanup comment.
3027         (odr_subtypes_equivalent): Add warn and warned arguments; check main
3028         variants.
3029         (type_variants_equivalent_p): break out from ...
3030         (odr_types_equivalent): ... here; go for main variants where needed.
3031         (warn_odr): ... here; turn static.
3032         (warn_types_mismatch): Compare mangled names of main variants.
3033         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3034         (type_with_linkage_p): Sanity check that we look at main variant.
3035         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3036         * tree.h (types_same_for_odr): Drop strict argument.
3038 2018-10-26  Richard Biener  <rguenther@suse.de>
3040         PR tree-optimization/87746
3041         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3042         Simplify and fix WRT strided store groups with size not
3043         equal to step in element count.
3044         (vect_analyze_group_access_1): Dump the whole group.
3046 2018-10-25  Carl Love  <cel@us.ibm.com>
3048         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3049         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3050         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
3051         precicion entry for each overloaded builtin.
3052         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3053         VSCEDPUO): Rename overloaded name.
3054         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3055         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3056         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3057         define_expand for xscmpexqp instruction.
3058         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3060 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3061             Jinsong Ji <jji@us.ibm.com>
3063         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3064         function with vec_sl.
3065         (_mm_slli_epi32): Likewise.
3066         (_mm_slli_epi64): Likewise.
3067         (_mm_srai_epi16): Replace deprecated function with vec_sra.
3068         (_mm_srai_epi32): Likewise.
3069         (_mm_srli_epi16): Replace deprecated function with vec_sr.
3070         (_mm_srli_epi32): Likewise.
3071         (_mm_srli_epi64): Likewise.
3072         (_mm_sll_epi16): Replace deprecated function with vec_sl.
3073         (_mm_sll_epi32): Likewise.
3074         (_mm_sll_epi64): Likewise.
3075         (_mm_sra_epi16): Replace deprecated function with vec_sra.
3076         (_mm_sra_epi32): Likewise.
3077         (_mm_srl_epi16): Replace deprecated function with vec_sr.
3078         (_mm_srl_epi32): Likewise.
3079         (_mm_srl_epi64): Likewise.
3081 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3082             Jinsong Ji <jji@us.ibm.com>
3084         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
3085         comparison operators with vec_cmp* for compatibility due to
3086         unfortunate history; clean up formatting and use types more
3087         appropriately.
3088         (_mm_sll_epi32): Likewise.
3089         (_mm_sll_epi64): Likewise.
3090         (_mm_srl_epi16): Likewise.
3091         (_mm_srl_epi32): Likewise.
3092         (_mm_srl_epi64): Likewise.
3094 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3095             Jinsong Ji <jji@us.ibm.com>
3097         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
3098         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
3099         __vector __bool int.  Use vec_cmpgt in preference to deprecated
3100         function vec_vcmpgtfp.
3101         (_mm_max_ps): Likewise.
3103 2018-10-25  Jeff Law  <law@redhat.com>
3105         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
3106         if returning false.
3108 2018-10-25  Martin Sebor  <msebor@redhat.com>
3110         * doc/extend.texi (aligned): Expand attribute description.
3111         (Alignment): Rename section.  Discuss function arguments.
3113 2018-10-25  Jan Hubicka  <jh@suse.cz>
3115         * ipa-devirt.c (main_odr_variant): Remove.
3116         (hash_odr_name, types_same_for_odr, types_odr_comparable,
3117         odr_name_hasher::equal, odr_subtypes_equivalent_p):
3118         Drop use of main_odr_variant.
3119         (add_type_duplicate): Silence confused warnings on integer types.
3120         (get_odr_type): Always look for main variant.
3121         (register_odr_type): Simplify.
3123 2018-10-25  Richard Biener  <rguenther@suse.de>
3125         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
3126         Initialize ng to silence error with release checking bootstrap.
3128 2018-10-25  Richard Biener  <rguenther@suse.de>
3130         * tree-if-conv.c: Include tree-ssa-sccvn.h.
3131         (tree_if_conversion): Run CSE on the if-converted loop body.
3133 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3135         * config/s390/constraints.md (ZL): New constraint.
3136         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3137         operands.
3138         * config/s390/s390.md (movdi_larl): Remove.
3139         (movdi_64): Add the LARL alternative.
3141 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3143         PR bootstrap/87747
3144         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3145         (RTX_CODE_HWINT_P): New macro.
3146         (rtx_code_size): Use RTX_CODE_HWINT_P ().
3148 2018-10-25  Jan Hubicka  <jh@suse.cz>
3150         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3151         is anonymous.
3153 2018-10-25  Richard Biener  <rguenther@suse.de>
3155         PR tree-optimization/87665
3156         PR tree-optimization/87745
3157         * tree-vectorizer.h (get_earlier_stmt): Remove.
3158         (get_later_stmt): Pick up UID from the original non-pattern stmt.
3160 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
3162         * options.texi (Deprecated): Move list to Var section.
3164 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
3165             Jinsong Ji <jji@us.ibm.com>
3167         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3168         __vector long to __vector long long.
3169         (_mm_cvtpd_ps): Likewise.
3170         (_mm_cvttpd_epi32): Likewise.
3171         (_mm_cvtpi32_pd): Likewise.
3172         (_mm_unpackhi_epi64): Likewise.
3173         (_mm_unpacklo_epi64): Likewise.
3175 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
3177         PR rtl-optimization/87720
3178         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3180 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
3182         * gimple-ssa-isolate-paths.c
3183         (find_implicit_erroneous_behavior): Do not change code if the
3184         pass is running for warnings only.
3185         (find_explicit_erroneous_behavior): Likewise.
3187 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
3189         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3190         Define as rs6000_mangle_decl_assembler_name.
3191         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3192         long double to IEEE long double, switch the names of the long
3193         double built-in functions to be <func>f128 instead of <func>l.
3195 2018-10-24  Martin Sebor  <msebor@redhat.com>
3197         * doc/extend.texi (nonnull): List no-argument form.  Reference
3198         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3200 2018-10-24  Richard Biener  <rguenther@suse.de>
3202         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3204 2018-10-24  Martin Liska  <mliska@suse.cz>
3206         PR tree-optimization/84436
3207         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3208         Remove.
3209         (switch_conversion::contains_linear_function_p): New.
3210         (switch_conversion::build_one_array): Support linear
3211         transformation on input.
3212         * tree-switch-conversion.h (struct switch_conversion): Add
3213         contains_linear_function_p declaration.
3215 2018-10-24  Richard Biener  <rguenther@suse.de>
3217         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3218         if its argument is CONSTANT_CLASS_P.
3220 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3222         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3223         it is wrong for forward declarations.
3225 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3227         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3228         SYMBOL_REF_P () check.
3230 2018-10-24  Richard Biener  <rguenther@suse.de>
3232         PR tree-optimization/87105
3233         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3234         dump classification.
3235         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3236         stores by splitting the affected group after the fact.
3237         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3238         fail the SLP build because of size constraints.
3240 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3242         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3243         * configure: Regenerate.
3244         * config.in: Regenerate.
3245         * varasm.c (mergeable_string_section): Use readonly_data_section
3246         if linker doesn't support SHF_MERGE with alignment > 8.
3247         (mergeable_constant_section): Likewise.
3249 2018-10-24  Richard Biener  <rguenther@suse.de>
3251         PR tree-optimization/84013
3252         * tree-ssa-structalias.c (struct msdi_data): New struct for
3253         marshalling data to walk_stmt_load_store_ops.
3254         (maybe_set_dependence_info): Refactor as callback for
3255         walk_stmt_load_store_ops.
3256         (compute_dependence_clique): Set restrict info on all stmt kinds.
3258 2018-10-24  Martin Liska  <mliska@suse.cz>
3260         * cgraph.c (cgraph_node::dump):
3261         Remove reduntant dumps and make tp_first_run dump more compact.
3263 2018-10-24  Richard Biener  <rguenther@suse.de>
3265         PR tree-optimization/87665
3266         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3267         to reflect reality.
3269 2018-10-12  Jeff Law  <law@redhat.com>
3271         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3272         for H8/S.
3274 2018-10-23  Richard Biener  <rguenther@suse.de>
3276         * tree-vrp.c (add_assert_info): Guard dump_printf with
3277         dump_enabled_p.
3278         * gimple-ssa-evrp-analyze.c
3279         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3280         Use value_range::ignore_equivs_equal_p.
3282 2018-10-23  Richard Biener  <rguenther@suse.de>
3284         PR tree-optimization/87105
3285         PR tree-optimization/87608
3286         * passes.def (pass_all_early_optimizations): Add early phi-opt
3287         after dce.
3288         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3289         addition to debug stmts.
3290         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3291         and abs replacement early.
3292         * tree-cfg.c (gimple_empty_block_p): Likewise.
3294 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3296         PR target/86383
3297         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3298         specified to configure.
3299         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3301 2018-10-23  Richard Biener  <rguenther@suse.de>
3303         PR tree-optimization/87700
3304         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3306 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3308         PR target/87674
3309         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3310         second argument from __mmask16 to __mmask8.
3311         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3312         _mm_mask_packs_epi32): Likewise.
3313         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3314         Likewise.
3315         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3317 2018-10-23  Richard Biener  <rguenther@suse.de>
3319         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3321 2018-10-23  Richard Biener  <rguenther@suse.de>
3323         PR tree-optimization/86144
3324         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3325         over simd attribute.
3327 2018-10-23  Richard Biener  <rguenther@suse.de>
3329         PR tree-optimization/87693
3330         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3331         the case we do not find the taken edge.
3333 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3334             Jinsong Ji  <jji@us.ibm.com>
3336         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3337         (_mm_store_pd): Use unaligned vector type for pointer cast.
3338         (_mm_maskmoveu_si128): Likewise.
3339         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3340         (_mm_store_ps): Use unaligned vector type for pointer cast.
3342 2018-10-22  Paul Koning  <ni1d@arrl.net>
3344         * symtab.c (symtab_node::increase_alignment): Correct max
3345         alignment check.
3347 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3349         PR tree-optimization/87633
3350         * match.pd: Do not generate unordered integer comparisons.
3352 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3354         PR rtl-optimization/87600
3355         * combine.c: Add include of expr.h.
3356         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3357         register to a pseudo.
3358         (make_more_copies): New function, add a copy to a new pseudo after
3359         the moves from hard registers into pseudos.
3360         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3361         later.  Call make_more_copies.
3363 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3365         * lra-constraints.c (process_alt_operands): New local array,
3366         matching_early_clobber.  Check matching_early_clobber before
3367         decrementing reject, and set matching_early_clobber after.
3369 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3371         PR target/87598
3372         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3373         call output_addr_const and hope for the best.
3375 2018-10-22  Richard Biener  <rguenther@suse.de>
3377         * gimple-ssa-evrp-analyze.c
3378         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3379         smarter about what ranges to use.
3380         * tree-vrp.c (add_assert_info): Dump here.
3381         (register_edge_assert_for_2): Instead of here at multiple but
3382         not all places.
3384         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3385         * gcc.dg/predict-6.c: Adjust.
3386         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3387         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3388         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3390 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3391         Richard Biener  <rguenther@suse.de>
3393         * bitmap.h: Update data structure documentation, including a
3394         description of bitmap views as either linked-lists or splay trees.
3395         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3396         (struct bitmap_head_def): Likewise.
3397         (bitmap_list_view, bitmap_tree_view): New prototypes.
3398         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3399         tree_form fields.
3400         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3401         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3402         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3403         released bitmap element here.
3404         (bitmap_element_free): Remove.
3405         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3406         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3407         this function similar ones such that linked-list bitmap implementation
3408         functions are grouped.
3409         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3410         and moved for grouping.
3411         (bitmap_list_insert_element_after): Renamed from
3412         bitmap_elt_insert_after, and moved for grouping.
3413         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3414         (bitmap_tree_link_left, bitmap_tree_link_right,
3415         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3416         bitmap_tree_link_element, bitmap_tree_unlink_element,
3417         bitmap_tree_find_element): New functions for splay-tree bitmap
3418         implementation.
3419         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3420         Renamed and moved, see above entries.
3421         (bitmap_tree_listify_from): New function to convert part of a splay
3422         tree bitmap to a linked-list bitmap.
3423         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3424         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3425         (bitmap_find_bit): Remove.
3426         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3427         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3428         Handle splay tree bitmaps.
3429         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3430         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3431         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3432         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3433         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3434         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3435         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3436         corresponding changes to use linked-list specific bitmap_element
3437         manipulation functions as applicable for efficiency.
3438         (bitmap_tree_to_vec): New function.
3439         (debug_bitmap_elt_file): New function split out from ...
3440         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3441         (bitmap_print): Likewise.
3443         PR tree-optimization/63155
3444         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3445         SSA edge worklists.
3446         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3447         in tree-view.
3449 2018-10-22  Martin Liska  <mliska@suse.cz>
3451         PR tree-optimization/87686
3452         Revert
3453         2018-08-29  Martin Liska  <mliska@suse.cz>
3455         * tree-switch-conversion.c (switch_conversion::expand):
3456         Strenghten assumption about gswitch statements.
3458 2018-10-22  Martin Liska  <mliska@suse.cz>
3460         * ipa-icf.c (sem_item::compare_attributes): Remove.
3461         (sem_item::compare_referenced_symbol_properties): Use
3462         attribute_list_equal instead.
3463         (sem_function::equals_wpa): Likewise.
3464         * ipa-icf.h: Remove compare_attributes.
3466 2018-10-22  Richard Biener  <rguenther@suse.de>
3468         PR middle-end/87682
3469         * mem-stats.h (mem_usage::operator==): Fix pasto.
3471 2018-10-22  Richard Biener  <rguenther@suse.de>
3473         PR tree-optimization/87640
3474         * tree-vrp.c (set_value_range_with_overflow): Decompose
3475         incomplete result.
3476         (extract_range_from_binary_expr_1): Adjust.
3478 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3480         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3481         (stmt_can_throw_external): Likewise.
3482         (stmt_can_throw_internal): Likewise.
3483         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3484         (lower_eh_constructs_2): Likewise.
3485         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3486         (stmt_can_throw_external): Likewise.
3487         (stmt_can_throw_internal): Likewise.
3488         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3489         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3490         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3491         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3492         (pass_lower_eh_dispatch::execute): Pass cfun to
3493         stmt_can_throw_external.
3494         (cleanup_empty_eh): Likewise.
3495         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3496         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3497         stmt_can_throw_external instead of pushing it to cfun.
3498         (symbol_table::create_edge): Likewise.
3499         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3500         stmt_can_throw_internal.
3501         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3502         to stmt_could_throw_p.
3503         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3504         stmt_can_throw_internal.
3505         (pass_store_merging::execute): Likewise.
3506         * gimple-ssa-strength-reduction.c
3507         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3508         stmt_could_throw_p.
3509         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3510         stmt_can_throw_internal.
3511         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3512         to stmt_can_throw_external.
3513         (check_stmt): Pass cfun to stmt_could_throw_p.
3514         (check_stmt): Pass cfun to stmt_can_throw_external.
3515         (pass_nothrow::execute): Likewise.
3516         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3517         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3518         stmt_can_throw_internal.
3519         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3520         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3521         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3522         * tree-complex.c (expand_complex_libcall): Pass cfun to
3523         stmt_could_throw_p and to stmt_can_throw_internal.
3524         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3525         * tree-inline.c (copy_edges_for_bb): Likewise.
3526         (maybe_move_debug_stmts_to_successors): Likewise.
3527         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3528         stmt_could_throw_p.
3529         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3530         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3531         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3532         stmt_can_throw_internal.
3533         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3534         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3535         stmt_could_throw_p.
3536         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3537         stmt_can_throw_internal.
3538         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3539         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3540         stmt_could_throw_p.
3541         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3542         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3543         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3544         (convert_mult_to_fma_1): Likewise.
3545         (convert_to_divmod): Likewise.
3546         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3547         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3548         * tree-ssa-propagate.c
3549         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3550         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3551         (maybe_optimize_range_tests): Likewise.
3552         (linearize_expr_tree): Likewise.
3553         (reassociate_bb): Likewise.
3554         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3555         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3556         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3557         (handle_char_store): Likewise.
3558         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3559         stmt_can_throw_internal.
3560         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3561         stmt_could_throw_p.
3562         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3563         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3564         (vectorizable_simd_clone_call): Likewise.
3565         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3566         (gimple_stringop_fixed_value): Likewise.
3568 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3570         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3571         literal pool references.
3572         (s390_check_qrst_address): Adapt to the new behavior of
3573         s390_loadrelative_operand_p ().
3575 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3577         PR target/72782
3578         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3580 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3582         PR target/72782
3583         * config/i386/sse.md (*<code><mode>3_bcst): New.
3585 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3587         PR target/72782
3588         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3589         V4DI, V16SI and V8DI.
3590         (*sub<mode>3<mask_name>_bcst): New.
3591         (*add<mode>3<mask_name>_bcst): Likewise.
3593 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3594             Jinsong Ji  <jji@us.ibm.com>
3596         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3597         __m64 with __vector unsigned long long for compatibility.
3598         (_mm_movemask_epi8): Likewise.
3599         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3600         (_mm_cvttps_pi32): Likewise.
3601         (_mm_cvtpi32_ps): Likewise.
3602         (_mm_cvtps_pi16): Likewise.
3603         (_mm_loadh_pi): Likewise.
3604         (_mm_storeh_pi): Likewise.
3605         (_mm_movehl_ps): Likewise.
3606         (_mm_movelh_ps): Likewise.
3607         (_mm_loadl_pi): Likewise.
3608         (_mm_storel_pi): Likewise.
3609         (_mm_movemask_ps): Likewise.
3610         (_mm_shuffle_pi16): Likewise.
3612 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3614         PR target/72782
3615         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3616         __builtin_ia32_vfnmsubpd512_mask.
3617         (_mm512_mask_fnmsub_round_pd): Likewise.
3618         (_mm512_fnmsub_pd): Likewise.
3619         (_mm512_mask_fnmsub_pd): Likewise.
3620         (_mm512_maskz_fnmsub_round_pd): Use
3621         __builtin_ia32_vfnmsubpd512_maskz.
3622         (_mm512_maskz_fnmsub_pd): Likewise.
3623         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3624         (_mm512_mask_fnmsub_round_ps): Likewise.
3625         (_mm512_fnmsub_ps): Likewise.
3626         (_mm512_mask_fnmsub_ps): Likewise.
3627         (_mm512_maskz_fnmsub_round_ps): Use
3628         __builtin_ia32_vfnmsubps512_maskz.
3629         (_mm512_maskz_fnmsub_ps): Likewise.
3630         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3631         __builtin_ia32_vfnmsubpd256_mask.
3632         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3633         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3634         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3635         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3636         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3637         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3638         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3639         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3640         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3641         __builtin_ia32_vfnmsubpd.
3642         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3643         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3644         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3645         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3646         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3647         * config/i386/i386-builtin.def: Add
3648         __builtin_ia32_vfnmsubpd256_mask,
3649         __builtin_ia32_vfnmsubpd256_maskz,
3650         __builtin_ia32_vfnmsubpd128_mask,
3651         __builtin_ia32_vfnmsubpd128_maskz,
3652         __builtin_ia32_vfnmsubps256_mask,
3653         __builtin_ia32_vfnmsubps256_maskz,
3654         __builtin_ia32_vfnmsubps128_mask,
3655         __builtin_ia32_vfnmsubps128_maskz,
3656         __builtin_ia32_vfnmsubpd512_mask,
3657         __builtin_ia32_vfnmsubpd512_maskz,
3658         __builtin_ia32_vfnmsubps512_mask,
3659         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3660         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3661         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3662         __builtin_ia32_vfnmsubpd256.
3663         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3664         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3665         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3666         Likewise.
3667         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3668         Likewise.
3669         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3670         Likewise.
3671         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3673 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3675         PR target/72782
3676         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3677         __builtin_ia32_vfnmaddpd512_mask.
3678         (_mm512_mask_fnmadd_round_pd): Likewise.
3679         (_mm512_fnmadd_pd): Likewise.
3680         (_mm512_mask_fnmadd_pd): Likewise.
3681         (_mm512_maskz_fnmadd_round_pd): Use
3682         __builtin_ia32_vfnmaddpd512_maskz.
3683         (_mm512_maskz_fnmadd_pd): Likewise.
3684         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3685         (_mm512_mask_fnmadd_round_ps): Likewise.
3686         (_mm512_fnmadd_ps): Likewise.
3687         (_mm512_mask_fnmadd_ps): Likewise.
3688         (_mm512_maskz_fnmadd_round_ps): Use
3689         __builtin_ia32_vfnmaddps512_maskz.
3690         (_mm512_maskz_fnmadd_ps): Likewise.
3691         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3692         __builtin_ia32_vfnmaddpd256_mask.
3693         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3694         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3695         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3696         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3697         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3698         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3699         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3700         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3701         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3702         __builtin_ia32_vfnmaddpd.
3703         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3704         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3705         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3706         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3707         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3708         * config/i386/i386-builtin.def: Add
3709         __builtin_ia32_vfnmaddpd256_mask,
3710         __builtin_ia32_vfnmaddpd256_maskz,
3711         __builtin_ia32_vfnmaddpd128_mask,
3712         __builtin_ia32_vfnmaddpd128_maskz,
3713         __builtin_ia32_vfnmaddps256_mask,
3714         __builtin_ia32_vfnmaddps256_maskz,
3715         __builtin_ia32_vfnmaddps128_mask,
3716         __builtin_ia32_vfnmaddps128_maskz,
3717         __builtin_ia32_vfnmaddpd512_mask,
3718         __builtin_ia32_vfnmaddpd512_maskz,
3719         __builtin_ia32_vfnmaddps512_mask,
3720         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
3721         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
3722         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
3723         __builtin_ia32_vfnmaddpd256.
3724         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
3725         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
3726         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
3727         Likewise.
3728         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
3729         Likewise.
3730         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
3731         Likewise.
3732         (fmai_vmfnmadd_<mode><round_name>): Likewise.
3734 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3736         PR target/72782
3737         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
3738         __builtin_ia32_vfmsubpd512_mask.
3739         (_mm512_mask_fmsub_round_pd): Likewise.
3740         (_mm512_fmsub_pd): Likewise.
3741         (_mm512_mask_fmsub_pd): Likewise.
3742         (_mm512_maskz_fmsub_round_pd): Use
3743         __builtin_ia32_vfmsubpd512_maskz.
3744         (_mm512_maskz_fmsub_pd): Likewise.
3745         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
3746         (_mm512_mask_fmsub_round_ps): Likewise.
3747         (_mm512_fmsub_ps): Likewise.
3748         (_mm512_mask_fmsub_ps): Likewise.
3749         (_mm512_maskz_fmsub_round_ps): Use
3750         __builtin_ia32_vfmsubps512_maskz.
3751         (_mm512_maskz_fmsub_ps): Likewise.
3752         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
3753         __builtin_ia32_vfmsubpd256_mask.
3754         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
3755         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3756         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
3757         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3758         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3759         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
3760         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
3761         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
3762         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
3763         __builtin_ia32_vfmsubpd.
3764         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
3765         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
3766         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
3767         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
3768         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
3769         * config/i386/i386-builtin.def: Add
3770         __builtin_ia32_vfmsubpd256_mask,
3771         __builtin_ia32_vfmsubpd256_maskz,
3772         __builtin_ia32_vfmsubpd128_mask,
3773         __builtin_ia32_vfmsubpd128_maskz,
3774         __builtin_ia32_vfmsubps256_mask,
3775         __builtin_ia32_vfmsubps256_maskz,
3776         __builtin_ia32_vfmsubps128_mask,
3777         __builtin_ia32_vfmsubps128_maskz,
3778         __builtin_ia32_vfmsubpd512_mask,
3779         __builtin_ia32_vfmsubpd512_maskz,
3780         __builtin_ia32_vfmsubps512_mask,
3781         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
3782         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
3783         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
3784         __builtin_ia32_vfmsubpd256.
3785         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
3786         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
3787         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
3788         Likewise.
3789         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
3790         Likewise.
3791         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
3792         Likewise.
3793         (fmai_vmfmsub_<mode><round_name>): Likewise.
3795 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3797         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
3798         Remove plus.  Renamed to ...
3799         (*sub<mode>3<mask_name>_bcst): This.
3800         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
3801         (*add<mode>3<mask_name>_bcst): This.
3803 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3805         PR target/72782
3806         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
3808 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3810         PR target/87662
3811         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
3812         (_mm_or_epi32): Likewise.
3813         (_mm256_xor_epi32): Likewise.
3814         (_mm_xor_epi32): Likewise.
3815         (_mm256_or_epi64): Likewise.
3816         (_mm_or_epi64): Likewise.
3817         (_mm256_xor_epi64): Likewise.
3818         (_mm_xor_epi64): Likewise.
3820 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
3822         PR target/72782
3823         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
3825 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
3827         PR middle-end/87647
3828         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
3830 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
3832         * doc/ux.texi: Move @section directly after @node.
3834 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
3836         PR middle-end/85488
3837         PR middle-end/87649
3838         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
3839         depend closely nested inside of loop with ordered clause with
3840         a parameter.
3842 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
3844         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
3845         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
3846         * doc/ux.texi: New file.
3848 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
3850         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
3851         be the first CR field allocated.
3853 2018-10-19  Richard Biener  <rguenther@suse.de>
3855         PR target/87657
3856         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
3857         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
3859 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3861         PR target/72782
3862         * config/i386/sse.md
3863         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
3864         (*add<mode>3<mask_name>_bcst_2): Likewise.
3866 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3868         * config/i386/sse.md
3869         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3870         Replace nonimmediate_operand with register_operand.
3871         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3872         Likewise.
3873         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3874         Likewise.
3876 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3878         PR rtl-optimization/87596
3879         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
3880         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
3881         for instructions in FROM..TO range.
3883 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
3885         * cfgexpand.c (expand_one_var): Use specific wording in error message
3886         for non-local frame variables.
3887         * stor-layout.c (layout_decl): Do not issue a warning for them.
3889 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
3891         * haifa-sched.c (priority): Add force_recompute parameter.
3892         (apply_replacement): Call priority () with force_recompute = true.
3893         (restore_pattern): Likewise.
3895 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3897         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
3898         HOST_BITS_PER_WIDE_INT.
3899         (test_vector_ops_duplicate): Likewise.
3901 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3903         PR target/72782
3904         * config/i386/sse.md (VF_AVX512): New.
3905         (avx512bcst): Likewise.
3906         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3907         Likewise.
3908         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3909         Likewise.
3910         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3911         Likewise.
3913 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
3915         * doc/invoke.texi (-dumpversion): Improve grammar.
3916         (-dumpfullversion): Make more consistent with -dumpversion.
3918 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
3920         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
3921         Set JUMP_LABEL to the jump insn.
3922         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
3923         Predict emitted jump and add label to jump insn.
3925 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
3927         PR tree-optimization/87562
3928         * input.c (get_substring_ranges_for_loc): Use
3929         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
3930         getting the linemap for the endpoint.  Verify that it's either
3931         in the same linemap as the start point's spelling location, or
3932         at least in the same file.
3934 2018-10-18  Richard Biener  <rguenther@suse.de>
3936         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
3937         feed width-specific load/store costs through ix86_vec_cost.
3938         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
3939         (k8_cost): Likewise.
3940         (bdver_cost): Likewise.
3941         (znver1_cost): Likewise.
3942         (btver1_cost): Likewise.
3943         (btver2_cost): Likewise.
3945 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3947         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
3948         to simplify subreg of vec_merge.
3950 2018-10-18  Richard Biener  <rguenther@suse.de>
3952         * config/i386/i386.c: Fix costing of vector FMA.
3954 2018-10-18  Richard Biener  <rguenther@suse.de>
3956         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
3957         and argument.
3958         (ix86_builtin_vectorization_cost): For vec_construct properly
3959         cost insertion into SSE regs.
3960         (...): Adjust calls to ix86_vec_cost.
3962 2018-10-18  Richard Biener  <rguenther@suse.de>
3964         PR middle-end/87087
3965         Revert
3966         2018-02-07  Richard Biener  <rguenther@suse.de>
3968         PR tree-optimization/84204
3969         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3970         this place.
3972 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3974         PR target/87537
3975         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
3976         of vec_duplicate.
3977         (test_vector_ops_duplicate): Add test for a scalar subreg of a
3978         VEC_MERGE of a VEC_DUPLICATE.
3980 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3982         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
3983         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
3984         * doc/standards.texi (C Language): Document C2X.
3985         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
3986         (rl78_option_override): Handle "GNU C2X" language name.
3988 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3990         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
3991         Document C17 as published in 2018.
3993 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
3995         PR middle-end/87623
3996         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
3997         bail out if both sides do not have the same storage order.
3999 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4001         * bitmap.c (bitmap_head::dump): New.
4002         * bitmap.h (bitmap_head): Add dump().
4003         * gimple-ssa-evrp-analyze.c
4004         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4005         (evrp_range_analyzer::set_ssa_range_info): Same.
4006         (evrp_range_analyzer::record_ranges_from_phis): Same.
4007         (evrp_range_analyzer::record_ranges_from_stmt): Same.
4008         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4009         * gimple-ssa-sprintf.c (get_int_range): Same.
4010         (format_integer): Same.
4011         (sprintf_dom_walker::handle_gimple_call): Same.
4012         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4013         (ipcp_vr_lattice::top_p): Same.
4014         (ipcp_vr_lattice::bottom_p): Same.
4015         (ipcp_vr_lattice::set_to_bottom): Same.
4016         (ipa_vr_operation_and_type_effects): Same.
4017         (propagate_vr_across_jump_function): Same.
4018         (ipcp_store_vr_results): Same.
4019         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4020         (ipa_print_node_jump_functions_for_edge): Same.
4021         (ipa_get_value_range): Same.
4022         (ipa_compute_jump_functions_for_edge): Same.
4023         (ipa_write_jump_function): Same.
4024         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4025         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4026         Same.
4027         * vr-values.c (set_value_range_to_nonnegative): Same.
4028         (set_value_range_to_truthvalue): Same.
4029         (vr_values::get_value_range): Same.
4030         (vr_values::set_defs_to_varying): Same.
4031         (vr_values::update_value_range): Same.
4032         (symbolic_range_based_on_p): Same.
4033         (vr_values::op_with_boolean_value_range_p): Same.
4034         (vr_values::extract_range_for_var_from_comparison_expr): Same.
4035         (vr_values::extract_range_from_ssa_name): Same.
4036         (vr_values::extract_range_from_binary_expr): Same.
4037         (vr_values::extract_range_from_unary_expr): Same.
4038         (vr_values::extract_range_from_cond_expr): Same.
4039         (vr_values::extract_range_from_comparison): Same.
4040         (vr_values::check_for_binary_op_overflow): Same.
4041         (vr_values::extract_range_basic): Same.
4042         (vr_values::extract_range_from_assignment): Same.
4043         (compare_ranges): Same.
4044         (compare_range_with_value): Same.
4045         (vr_values::adjust_range_with_scev): Same.
4046         (vrp_valueize): Same.
4047         (vrp_valueize_1): Same.
4048         (vr_values::get_vr_for_comparison): Same.
4049         (vr_values::compare_name_with_value): Same.
4050         (vr_values::compare_names): Same.
4051         (vr_values::vrp_evaluate_conditional): Same.
4052         (find_case_label_ranges): Same.
4053         (vr_values::vrp_visit_switch_stmt): Same.
4054         (vr_values::extract_range_from_phi_node): Same.
4055         (vr_values::simplify_div_or_mod_using_ranges): Same.
4056         (vr_values::simplify_bit_ops_using_ranges): Same.
4057         (test_for_singularity): Same.
4058         (range_fits_type_p): Same.
4059         (vr_values::simplify_cond_using_ranges_1): Same.
4060         (vr_values::simplify_switch_using_ranges): Same.
4061         (vr_values::simplify_float_conversion_using_ranges): Same.
4062         (vr_values::two_valued_val_range_p): Same.
4063         (vr_values::add_equivalence): Move to value_range::equiv_add.
4064         * vr-values.h (vr_values::add_equivalence): Remove.
4065         (VR_INITIALIZER): Remove.
4066         * tree-vrp.c (value_range::set): New.
4067         (value_range::equiv_add): New.
4068         (value_range::value_range): New.
4069         (value_range::deep_copy): New.
4070         (value_range::check): New.
4071         (value_range::equal_p): New.
4072         (value_range::ignore_equivs_equal_p): New.
4073         (value_range::operator==): New.
4074         (value_range::operator!=): New.
4075         (value_range::symbolic_p): New.
4076         (value_range::numeric_p): New.
4077         (value_range::set_undefined): New.
4078         (value_range::set_varying): New.
4079         (value_range::may_contain_p): New.
4080         (value_range::equiv_clear): New.
4081         (value_range::singleton_p): New.
4082         (value_range::intersect): New.
4083         (value_range::dump): New.
4084         (value_range::set_and_canonicalize): New.
4085         (set_value_range): Adjust for value_range API.
4086         (set_value_range_to_undefined): Same.
4087         (set_value_range_to_varying): Same.
4088         (set_and_canonicalize_value_range): Same.
4089         (set_value_range_to_nonnull): Same.
4090         (set_value_range_to_null): Same.
4091         (range_is_null): Same.
4092         (range_is_nonnull): Same.
4093         (range_int_cst_p): Same.
4094         (range_int_cst_singleton_p): Same.
4095         (symbolic_range_p): Same.
4096         (range_includes_zero_p): Same.
4097         (value_range_constant_singleton): Same.
4098         (vrp_set_zero_nonzero_bits): Same.
4099         (ranges_from_anti_range): Same.
4100         (extract_range_into_wide_ints): Same.
4101         (extract_range_from_multiplicative_op): Same.
4102         (set_value_range_with_overflow): Same.
4103         (extract_range_from_binary_expr_1): Same.
4104         (extract_range_from_unary_expr): Same.
4105         (dump_value_range): Same.
4106         (debug_value_range): Same.
4107         (vrp_prop::check_array_ref): Same.
4108         (vrp_prop::check_mem_ref): Same.
4109         (vrp_prop::vrp_initialize): Same.
4110         (vrp_prop::visit_stmt): Same.
4111         (intersect_ranges): Same.
4112         (vrp_prop::visit_phi): Same.
4113         (vrp_prop::vrp_finalize): Same.
4114         (determine_value_range_1): Same.
4115         (determine_value_range): Same.
4116         (vrp_intersect_ranges_1): Rename to...
4117         (vrp_intersect_1): this.
4118         (vrp_intersect_ranges): Rename to...
4119         (value_range::intersect_helper): ...this.
4120         (vrp_meet_1): Rename to...
4121         (value_range::union_helper): ...this.
4122         (vrp_meet): Rename to...
4123         (value_range::union_): ...this.
4124         (copy_value_range): Remove.
4125         * tree-vrp.h (struct value_range): Rewrite into a proper class.
4126         (value_range::vrtype): New.
4127         (value_range::type): New.
4128         (value_range::equiv): New.
4129         (value_range::min): New.
4130         (value_range::max): New.
4131         (value_range::varying_p): New.
4132         (value_range::undefined_p): New.
4133         (value_range::null_p): New.
4134         (value_range::equiv_add): New.
4135         (copy_value_range): Remove.
4137 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
4139         * Makefile.in (SELFTEST_TARGETS): New.
4140         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4141         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4142         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4143         c/Make-lang.in.
4144         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4145         (selftest-c++-gdb, selftest-c++-valgrind): Move to
4146         cp/Make-lang.in.
4147         * configure: Regenerate.
4148         * configure.ac (selftest_languages): New.
4150 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4152         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4153         overflow wraps argument.
4154         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4155         wide_int_range_multiplicative_op.
4156         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4157         overflow wraps argument.
4158         (wide_int_range_multiplicative_op): Same.
4159         (wide_int_range_lshift): Same.
4160         (wide_int_range_div): Same.
4161         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4162         (wide_int_range_lshift): Same.
4163         (wide_int_range_div): Same.
4165 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4167         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4168         use sign as argument.
4169         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4170         wide_int_range_shift_undefined_p.
4172 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4174         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4175         Rename to...
4176         (@despeculate_copy<ALLI_TI:mode>): ... This.
4177         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4178         switch statement.
4180 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4182         * config.gcc: Obsolete *-*-solaris2.10*.
4183         * doc/install.texi (Specific, *-*-solaris2*): Document it.
4185 2018-10-12  Jeff Law  <law@redhat.com>
4187         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4188         reg + sym +- const_int addressing modes.
4190 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4192         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4193         * diagnostic-show-locus.c (layout::layout): Apply the minimum
4194         margin width.
4195         (layout::start_annotation_line): Only print up to 3 of the
4196         margin character, to avoid touching the left-hand side.
4197         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4198         minimum margin width, as set by test_diagnostic_context's ctor.
4199         (selftest::test_fixit_insert_containing_newline): Likewise.
4200         (selftest::test_fixit_insert_containing_newline_2): Likewise.
4201         (selftest::test_line_numbers_multiline_range): Clear
4202         dc.min_margin_width.
4203         * diagnostic.c (diagnostic_initialize): Initialize
4204         min_margin_width.
4205         * diagnostic.h (struct diagnostic_context): Add field
4206         "min_margin_width".
4207         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4208         * opts.c (common_handle_option): Handle
4209         OPT_fdiagnostics_minimum_margin_width_.
4210         * selftest-diagnostic.c
4211         (selftest::test_diagnostic_context::test_diagnostic_context):
4212         Initialize min_margin_width to 6.
4213         * toplev.c (general_init): Initialize global_dc->min_margin_width.
4215 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4217         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4218         Fix usage of "error_at_rich_loc" in the comment.
4220 2018-10-15  Renlin Li  <renlin.li@arm.com>
4222         PR target/87563
4223         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4224         if-conversioned loop when it contains ifn with types not
4225         supported by backend.
4226         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4227         (direct_internal_fn_supported_p): New helper function.
4228         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4230 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4232         PR target/87572
4233         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4234         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4235         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4237 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4239         PR tree-optimization/87022
4240         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4241         bits in dist vector rather than the first one.
4243 2018-10-15  Richard Biener  <rguenther@suse.de>
4245         PR middle-end/87610
4246         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4247         (visit_loadstore): When a used restrict tag escaped verify that
4248         the points-to solution of "other" pointers do not include
4249         escaped.
4250         (compute_dependence_clique): If a used restrict tag escaped
4251         communicated that down to visit_loadstore.
4253 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4255         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4256         into reg if it isn't a general operand.
4258 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4260         PR target/87599
4261         * config/i386/sse.md (*vec_dupv2di): Add register source to
4262         movddup.
4264 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4266         PR target/87572
4267         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4268         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4269         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4270         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4272 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4274         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4275         (notice_args_size): Set it in the current trace if no insn that can
4276         throw internally has been seen yet.
4277         (connect_traces): When connecting args_size between traces, allow the
4278         incoming values not to match if there is an insn setting it before the
4279         first insn that can throw internally; in that case, force the creation
4280         of a CFI note on this latter insn.
4282 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4284         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4285         the base class, because using the injected-class-name was not clearly
4286         specified until DR 176.
4288 2018-10-12  Paul Koning  <ni1d@arrl.net>
4290         * config/pdp11/pdp11.md (doloop_end): New expander.
4291         (doloop_end_insn): renamed from "doloop_end".
4292         (addqi3): New pattern.
4293         (subqi3): New pattern.
4294         * config/pdp11/predicates.md (incdec_operand): New predicate.
4296 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4298         PR middle-end/81376
4299         * real.c (format_helper::can_represent_integral_type_p): New function
4300         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4301         * match.pd: New pattern.
4303 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4305         * configure.ac: Introduce --enable-large-address-aware
4306         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4307         * doc/install.texi: Document it.
4308         * configure, config.in: Rebuilt.
4309         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4310         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4311         (LINK_SPEC): Insert it.
4312         * config/i386/mingw-w64.h: Likewise.
4314         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4316 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4318         PR rtl-optimization/87600
4319         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4321 2018-10-12  Paul Koning  <ni1d@arrl.net>
4323         * doc/md.texi (doloop_end): Document that the pattern code may
4324         need to check operand mode.
4326 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4328         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4329         to zero-extend between int and floating-point registers.
4330         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4331         ldp into floating-point registers.  Add type and arch attributes.
4332         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4333         Use f_loads for type attribute.
4335 2018-10-11  Martin Sebor  <msebor@redhat.com>
4337         * doc/extend.texi (attribute packed): Correct typos.
4339 2018-10-11  Martin Sebor  <msebor@redhat.com>
4341         * doc/extend.texi (attribute flatten): Mention interaction with
4342         noinline.
4344 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4346         PR target/87156
4347         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4348         Set new_decl virtual flag to zero.
4350 2018-10-11  Martin Sebor  <msebor@redhat.com>
4352         PR middle-end/87593
4353         * doc/extend.texi (attribute format_arg): Discuss using multiple
4354         attributes on a single function.
4356 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4358         PR tree-optimization/86829
4359         * match.pd (sin (atan (x))): New simplification rules.
4360         (cos (atan (x))): Likewise.
4361         * real.c (build_sinatan_real): New function.
4362         * real.h (build_sinatan_real): Prototype.
4364 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4366         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4367         function.
4368         (fold_mergeeo_helper): New helper function.
4369         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4370         intrinsics.  Correct some whitespace indentation issues.
4372 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4374         PR target/87511
4375         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4376         Use HOST_WIDE_INT_1U for shift.
4378 2018-10-11  Doug Rupp  <rupp@adacore.com>
4379             Olivier Hainque  <hainque@adacore.com>
4381         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4382         Pass --relax to the linker for RTPs.
4383         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4385 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4386             Jan Hubicka  <jh@suse.cz>
4387             Martin Jambor  <mjambor@suse.cz>
4389         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4390         the same elements are repeated rather than printing all of them.
4391         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4392         "repeated" elements.
4393         * read-rtl-function.c (test_loading_repeat): New function.
4394         (read_rtl_function_c_tests): Call test_loading_repeat.
4395         * rtl-tests.c (test_dumping_repeat): New function.
4396         (rtl_tests_c_tests): Call test_dumping_repeat.
4398 2018-10-11  Richard Biener  <rguenther@suse.de>
4400         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4401         bdver?_cost): Unify to ...
4402         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4403         * config/i386/i386.c (processor_cost_table): Adjust.
4405 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4407         PR middle-end/87574
4408         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4409         the thunk when expanding to GIMPLE.
4411 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4413         * varasm.c (mergeable_string_section): Don't try to move zero-length
4414         strings to the merge section.
4416 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4418         PR target/87573
4419         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4421 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4423         PR target/87550
4424         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4425         to special_args set.
4427 2018-10-10  Richard Biener  <rguenther@suse.de>
4429         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4430         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4431         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4432         and recursing and pattern terminating the recursion on SSE
4433         vector width using ix86_expand_reduc.
4434         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4435         width and recursing and SSE2 vector variant doing the final
4436         reduction with ix86_expand_reduc.
4437         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4438         with terminating the recursion at AVX level, splitting that
4439         to SSE there.
4441 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4443         * genmatch.c (error_cb): Rename to...
4444         (diagnostic_cb): ...this, converting int params to enums.
4445         (fatal_at): Update for renaming.
4446         (warning_at): Likewise.
4447         (main): Likewise.
4448         * input.c (selftest::ebcdic_execution_charset::apply):
4449         Update for renaming of...
4450         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4451         to...
4452         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4453         converting level and reason to enums.
4454         (class selftest::lexer_error_sink): Rename to...
4455         (class selftest::lexer_test_options): ...this, renaming field
4456         "m_errors" to "m_diagnostics".
4457         (selftest::lexer_test_options::apply): Update for renaming of...
4458         (selftest::lexer_test_options::on_error): ...this, renaming to...
4459         (selftest::lexer_test_options::on_diagnostic): ...this
4460         converting level and reason to enums.
4461         (selftest::test_lexer_string_locations_raw_string_unterminated):
4462         Update for renamings.
4463         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4464         "reason".
4466 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4468         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4469         * config/rs6000/pmmintrin.h: New file.
4471 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4473         PR tree-optimization/86659
4474         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4476 2018-10-09  Richard Biener  <rguenther@suse.de>
4478         PR tree-optimization/63155
4479         * tree-ssa-structalias.c: Include tree-ssa.h.
4480         (get_constraint_for_ssa_var): For undefs return nothing_id.
4481         (find_func_aliases): Cleanup PHI handling.
4483 2018-10-09  Richard Biener  <rguenther@suse.de>
4485         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4486         replacements.
4488 2018-10-09  Martin Liska  <mliska@suse.cz>
4490         * asan.c (asan_emit_stack_protection): If a stack variable
4491         is located in a same file as current function, then emit
4492         line info into variable definition string.
4494 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4496         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4497         information.
4499 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4501         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4502         on the thunk.
4504 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4506         PR tree-optimization/86659
4507         * gimple-match.h (struct gimple_match_op): Add reverse field.
4508         (gimple_match_op::set_op): New overloaded method.
4509         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4510         the REF_REVERSE_STORAGE_ORDER flag on the value.
4511         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4512         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4513         
4514 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4516         PR middle-end/63155
4517         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4518         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4520 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4522         PR target/87517
4523         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4524         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4526 2018-10-08  Richard Biener  <rguenther@suse.de>
4528         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4529         cost the same as AVX128 ones.
4531 2018-10-08  Paul Koning  <ni1d@arrl.net>
4533         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4534         (expand_block_move): New function.
4535         * config/pdp11/pdp11.c (output_block_move): Remove.
4536         (expand_block_move): New function.
4537         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4538         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4539         (*movmemhi1): Remove.
4541 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4543         * config/s390/2827.md: Increase latencies for some FP instructions.
4545 2018-10-08  Richard Biener  <rguenther@suse.de>
4547         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4548         Open a dump scope.
4549         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4550         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4551         (add_stmt_cost): Dump return value of the hook.
4553 2018-10-08  Richard Biener  <rguenther@suse.de>
4555         PR tree-optimization/63155
4556         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4557         (ssa_propagation_engine::ssa_propagate): Remove redundant
4558         bitmap bit clearing.
4560 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4562         PR rtl-optimization/86939
4563         PR rtl-optimization/87479
4564         * ira.h (non_conflicting_reg_copy_p): New prototype.
4565         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4566         (make_hard_regno_dead): Don't add conflicts for register
4567         ignore_reg_for_conflicts.
4568         (make_object_dead): Likewise.
4569         (non_conflicting_reg_copy_p): New function.
4570         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4571         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4572         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4573         (make_hard_regno_dead): Don't add conflicts for register
4574         ignore_reg_for_conflicts.  Remove special conflict handling of
4575         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4576         check_pic_pseudo_p and update callers.
4577         (mark_pseudo_dead): Don't add conflicts for register
4578         ignore_reg_for_conflicts.
4579         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4581 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4582             Jim Wilson  <jimw@sifive.com>
4584         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4585         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4586         new pattern using HONOR_SNANS that emits one extra instruction.
4588 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4590         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4591         patterns): Merge SI and DI patterns to a GPR pattern.
4592         (unnamed define_insn and define_split for record form of that): Merge
4593         to a single define_insn_and_split pattern.
4595 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4597         PR c++/56856
4598         * input.c (expand_location_to_spelling_point): Add param "aspect"
4599         and use rather than hardcoding LOCATION_ASPECT_CARET.
4600         (get_substring_ranges_for_loc): Handle the case of a single token
4601         within a macro expansion.
4602         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4603         defaulting to LOCATION_ASPECT_CARET.
4605 2018-10-05  Paul Koning  <ni1d@arrl.net>
4607         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4608         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4609         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4610         (pdp11_guard_type): New function.
4612 2018-10-05  Paul Koning  <ni1d@arrl.net>
4614         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4615         * config/pdp11/pdp11.opt (mfloat32): Remove.
4616         (mfloat64): Remove.
4617         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4618         (pdp11 -mfloat64): Remove.
4620 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4622         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4623         (*cmp<mode>_cc_i387): Ditto.
4624         (*cmpu<mode>_cc_i387): Ditto.
4625         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4626         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4627         "scratch" argument.
4628         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4629         Emit x86_sahf_1 pattern.
4630         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4631         (ix86_expand_carry_flag_compare): Ditto.
4633 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4635         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4636         to reg_or_0_operand.  Add "C" constraint.
4637         (*cmpxf_cc_i387): Ditto.
4638         (*cmp<mode>_i387): Change operand 2 predicate
4639         to nonimm_or_0_operand.  Add "C" constraint.
4640         (*cmp<mode>_cc_i387): Ditto.
4641         (*cmp<mode>_0_i387): Remove insn pattern.
4642         (*cmp<mode>_0_cc_i387): Ditto.
4644 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4646         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4647         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4648         from vector_move_operand.  Update all uses.
4650 2018-10-05  Martin Sebor  <msebor@redhat.com>
4652         PR tree-optimization/87490
4653         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4654         consistently.
4656 2018-10-05  Richard Biener  <rguenther@suse.de>
4658         PR tree-optimization/63155
4659         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4660         vertical space in dumpfiles.
4661         * tree-ssa-propagate.h
4662         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4663         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4664         (ssa_edge_worklist_back): Likewise.
4665         (curr_order): Likewise.
4666         (cfg_blocks_get): Remove abstraction.
4667         (cfg_blocks_add): Likewise.
4668         (cfg_blocks_empty_p): Likewise.
4669         (add_ssa_edge): Add to current or next worklist based on
4670         RPO index.
4671         (add_control_edge): Likewise.
4672         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4673         into ...
4674         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4675         iteration from CFG and SSA edge worklist so we process
4676         everything in RPO order, prioritizing forward progress
4677         over iteration.
4678         (ssa_prop_init): Allocate new worklists, do not dump
4679         immediate uses.
4680         (ssa_prop_fini): Free new worklists.
4682 2018-10-05  Richard Biener  <rguenther@suse.de>
4684         * tree-core.h (tree_block::abstract_flag): Remove.
4685         (tree_block::block_num): Make full 32bits.
4686         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4687         * tree.h (BLOCK_ABSTRACT): Remove.
4688         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4689         resulting from BLOCK_ABSTRACT being always false.
4690         (gen_inlined_subroutine_die): Likewise.
4691         (gen_block_die): Likewise.
4692         * tree.c (block_ultimate_origin): Likewise.
4693         * tree-pretty-print.c (dump_block_node): Remove code dealing
4694         with BLOCK_ABSTRACT.
4695         * tree-ssa-live.c (dump_scope_block): Likewise.
4696         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4697         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4699 2018-10-05   Richard Biener  <rguenther@suse.de>
4701         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4702         is asked for initialize mode to the component mode of the
4703         vector type.
4705 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4707         PR target/87522
4708         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4709         assembler for -mavx.
4710         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4712 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4714         PR target/87509
4715         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4716         RS6000_BTM_DFP.
4717         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4718         to be DImode.  When using mffscrn, force the operand to a register.
4720 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
4722         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
4723         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
4724         X87MODEF mode iterator.
4725         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
4726         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
4727         X87MODEF mode iterator.
4729 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
4731         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
4732         -Wno-prio-ctor-dtor.
4734 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4736         * Makefile.in (OBJS): Add opt-problem.o.
4737         * dump-context.h: Include "selftest.h.
4738         (selftest::temp_dump_context): New forward decl.
4739         (class dump_context): Make friend of class
4740         selftest::temp_dump_context.
4741         (dump_context::dump_loc_immediate): New decl.
4742         (class dump_pretty_printer): Move here from dumpfile.c.
4743         (class temp_dump_context): Move to namespace selftest.
4744         (temp_dump_context::temp_dump_context): Add param
4745         "forcibly_enable_dumping".
4746         (selftest::verify_dumped_text):
4747         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
4748         (selftest::verify_item):
4749         (ASSERT_IS_TEXT): Move here from dumpfile.c.
4750         (ASSERT_IS_TREE): Likewise.
4751         (ASSERT_IS_GIMPLE): Likewise.
4752         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
4753         to...
4754         (dump_context::dump_loc_immediate): ...this new function.
4755         (class dump_pretty_printer): Move to dump-context.h.
4756         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
4757         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
4758         (temp_dump_context::temp_dump_context): Move to "selftest"
4759         namespace.  Add param "forcibly_enable_dumping", and use it to
4760         conditionalize the use of m_pp;
4761         (selftest::verify_dumped_text): Make non-static.
4762         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
4763         (selftest::verify_item): Make non-static.
4764         (ASSERT_IS_TEXT): Move to dump-context.h.
4765         (ASSERT_IS_TREE): Likewise.
4766         (ASSERT_IS_GIMPLE): Likewise.
4767         (selftest::test_capture_of_dump_calls): Pass "true" for new
4768         param of temp_dump_context.
4769         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
4770         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
4771         TDF_COMPARE_DEBUG.
4772         * opt-problem.cc: New file.
4773         * opt-problem.h: New file.
4774         * optinfo-emit-json.cc
4775         (selftest::test_building_json_from_dump_calls): Pass "true" for
4776         new param of temp_dump_context.
4777         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
4778         (optinfo::emit_for_opt_problem): New function.
4779         (optinfo::emit): Clarity which emit_item is used.
4780         * optinfo.h (optinfo::get_dump_location): New accessor.
4781         (optinfo::emit_for_opt_problem): New decl.
4782         (optinfo::emit): Make const.
4783         * selftest-run-tests.c (selftest::run_tests): Call
4784         selftest::opt_problem_cc_tests.
4785         * selftest.h (selftest::opt_problem_cc_tests): New decl.
4786         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
4787         bool to opt_result, converting fprintf messages to
4788         opt_result::failure_at calls.  Add "stmt" param for use by the
4789         failure_at calls.
4790         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
4791         (runtime_alias_check_p): Convert return type from bool to
4792         opt_result, converting dump_printf calls to
4793         opt_result::failure_at, using the statement DDR_A for their
4794         location.
4795         (find_data_references_in_stmt): Convert return type from bool to
4796         opt_result, converting "return false" to opt_result::failure_at
4797         with a new message.
4798         * tree-data-ref.h: Include "opt-problem.h".
4799         (dr_analyze_innermost): Convert return type from bool to opt_result,
4800         and add a const gimple * param.
4801         (find_data_references_in_stmt): Convert return type from bool to
4802         opt_result.
4803         (runtime_alias_check_p): Likewise.
4804         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
4805         dr_analyze_innermost.
4806         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
4807         Convert return type from bool to opt_result, adding a message for
4808         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
4809         (vect_analyze_data_ref_dependence): Convert return type from bool
4810         to opt_result.  Change sense of return type from "false"
4811         effectively meaning "no problems" to "false" meaning a problem,
4812         so that "return false" becomes "return opt_result::success".
4813         Convert "return true" calls to opt_result::failure_at, using
4814         the location of statement A rather than vect_location.
4815         (vect_analyze_data_ref_dependences): Convert return type from bool
4816         to opt_result.
4817         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
4818         calls to opt_result::failure_at, using the stmt location rather
4819         than vect_location.
4820         (vect_verify_datarefs_alignment): Convert return type from bool
4821         to opt_result.
4822         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
4823         into multiple more-tightly-scoped copies.
4824         (vect_analyze_data_refs_alignment): Convert return type from bool
4825         to opt_result.
4826         (vect_analyze_data_ref_accesses): Likewise, converting a
4827         "return false" to a "return opt_result::failure_at", adding a
4828         new message.
4829         (vect_prune_runtime_alias_test_list): Convert return type from
4830         bool to opt_result, converting dump_printf_loc to
4831         opt_result::failure_at.  Add a %G to show the pertinent statement,
4832         and use the stmt's location rather than vect_location.
4833         (vect_find_stmt_data_reference): Convert return type from
4834         bool to opt_result, converting dump_printf_loc to
4835         opt_result::failure_at, using stmt's location.
4836         (vect_analyze_data_refs):  Convert return type from bool to
4837         opt_result.  Convert "return false" to "return
4838         opt_result::failure_at", adding messages as needed.
4839         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
4840         type from bool to opt_result.
4841         (vect_determine_vf_for_stmt): Likewise.
4842         (vect_determine_vectorization_factor): Likewise, converting
4843         dump_printf_loc to opt_result::failure_at, using location of phi
4844         rather than vect_location.
4845         (vect_analyze_loop_form_1): Convert return type from bool to
4846         opt_result, converting dump_printf_loc calls, retaining the use of
4847         vect_location.
4848         (vect_analyze_loop_form): Convert return type from loop_vec_info
4849         to opt_loop_vec_info.
4850         (vect_analyze_loop_operations): Convert return type from bool to
4851         opt_result, converting dump_printf_loc calls, using the location
4852         of phi/stmt rather than vect_location where available.  Convert
4853         various "return false" to "return opt_result::failure_at" with
4854         "unsupported phi" messages.
4855         (vect_get_datarefs_in_loop): Convert return type from bool to
4856         opt_result.  Add a message for the
4857         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
4858         (vect_analyze_loop_2): Convert return type from bool to
4859         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
4860         each "goto again;", adding new messages where needed.
4861         Add "unsupported grouped {store|load}" messages.
4862         (vect_analyze_loop): Convert return type from loop_vec_info to
4863         opt_loop_vec_info.
4864         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
4865         bool to opt_result.
4866         * tree-vect-stmts.c (process_use): Likewise, converting
4867         dump_printf_loc call and using stmt location, rather than
4868         vect_location.
4869         (vect_mark_stmts_to_be_vectorized): Likeise.
4870         (vect_analyze_stmt): Likewise, adding a %G.
4871         (vect_get_vector_types_for_stmt): Convert return type from bool to
4872         opt_result, converting dump_printf_loc calls and using stmt
4873         location, rather than vect_location.
4874         (vect_get_mask_type_for_stmt): Convert return type from tree to
4875         opt_tree, converting dump_printf_loc calls and using stmt location.
4876         * tree-vectorizer.c: Include "opt-problem.h.
4877         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
4878         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
4879         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
4880         enabled, use it to report at the top level "couldn't vectorize
4881         loop" followed by the problem.
4882         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
4883         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
4884         to opt_result.
4885         (vect_analyze_stmt): Likewise.
4886         (vect_get_vector_types_for_stmt): Likewise.
4887         (tree vect_get_mask_type_for_stmt): Likewise.
4888         (vect_analyze_data_ref_dependences): Likewise.
4889         (vect_enhance_data_refs_alignment): Likewise.
4890         (vect_analyze_data_refs_alignment): Likewise.
4891         (vect_verify_datarefs_alignment): Likewise.
4892         (vect_analyze_data_ref_accesses): Likewise.
4893         (vect_prune_runtime_alias_test_list): Likewise.
4894         (vect_find_stmt_data_reference): Likewise.
4895         (vect_analyze_data_refs): Likewise.
4896         (vect_analyze_loop): Convert return type from loop_vec_info to
4897         opt_loop_vec_info.
4898         (vect_analyze_loop_form): Likewise.
4899         (vect_analyze_slp): Convert return type from bool to opt_result.
4901 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4903         * doc/invoke.texi (-fopt-info): Document new "internals"
4904         sub-option.
4905         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
4906         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
4907         MSG_ALL_KINDS.
4908         (optinfo_verbosity_options): Add "internals".
4909         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
4910         (dump_context::apply_dump_filter_p): New member function.
4911         (dump_context::dump_loc): Use apply_dump_filter_p rather than
4912         explicitly masking the dump_kind.
4913         (dump_context::begin_scope): Increment the scope depth first.  Use
4914         apply_dump_filter_p rather than explicitly masking the dump_kind.
4915         (dump_context::emit_item): Use apply_dump_filter_p rather than
4916         explicitly masking the dump_kind.
4917         (dump_dec): Likewise.
4918         (dump_hex): Likewise.
4919         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
4920         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
4921         (opt_info_switch_p): Update handling of default
4922         MSG_OPTIMIZED_LOCATIONS to cope with default of
4923         MSG_PRIORITY_USER_FACING.
4924         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
4925         masking the dump_kind.
4926         (selftest::test_capture_of_dump_calls): Update test_dump_context
4927         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
4928         than MSG_ALL.  Generalize scope test to be run at all four
4929         combinations of with/without MSG_PRIORITY_USER_FACING and
4930         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
4931         for each of the two values.
4932         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
4933         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
4934         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
4935         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
4936         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
4937         with MSG_PRIORITY_*.
4938         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
4939         dump messages as MSG_PRIORITY_USER_FACING.
4940         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
4941         about the interaction with MSG_PRIORITY_*.
4943 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4945         * varasm.c (output_constant): Add new parameter merge_strings.
4946         Make strings properly zero terminated in merge string sections.
4947         (mergeable_string_section): Don't fail if the last char is non-zero.
4948         (assemble_variable_contents): Handle merge string sections.
4949         (assemble_variable): Likewise.
4950         (assemble_constant_contents): Likewise.
4951         (output_constant_def_contents): Likewise.
4952         (output_constructor_array_range,
4953         output_constructor_regular_field): Adjust call to output_constant.
4954         (output_object_block): Adjust call to assemble_constant_contents
4955         and assemble_variable_contents.
4957 2018-10-04  Martin Liska  <mliska@suse.cz>
4959         PR c/87483
4960         * cgraphunit.c (process_function_and_variable_attributes):
4961         Warn about a function with alias attribute and a body.
4963 2018-10-04  Martin Liska  <mliska@suse.cz>
4965         PR ipa/82625
4966         * multiple_target.c (redirect_to_specific_clone): New function.
4967         (ipa_target_clone): Use it.
4968         * tree-inline.c: Fix comment.
4970 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4972         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
4973         fields.
4974         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
4975         (gcc::dump_manager::register_pass): New member function, adapted
4976         from loop body in gcc::pass_manager::register_pass, adding a
4977         call to update_dfi_for_opt_info.
4978         (gcc::dump_manager::opt_info_enable_passes): Store the
4979         -fopt-info options into the new fields.  Move the loop
4980         bodies into...
4981         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
4982         function.
4983         * dumpfile.h (struct opt_pass): New forward decl.
4984         (gcc::dump_manager::register_pass): New decl.
4985         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
4986         (class gcc::dump_manager): Add fields "m_optgroup_flags",
4987         "m_optinfo_flags", and "m_optinfo_filename".
4988         * passes.c (gcc::pass_manager::register_pass): Move all of the
4989         dump-handling code to gcc::dump_manager::register_pass.
4991 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
4993         PR rtl-optimization/87466
4994         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
4995         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
4996         * doc/tm.texi: Regenerate.
4997         * ira-lives.c (process_bb_node_lives): Use the new target hook.
4998         * lra-lives.c (process_bb_lives): Likewise.
4999         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5000         Define.
5002 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
5004         * params.c (add_params): Fix initialization.
5006 2018-10-04  Martin Liska  <mliska@suse.cz>
5008         PR gcov-profile/84107
5009         * tree-profile.c (init_ic_make_global_vars):
5010         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5011         Come up with new ic_tuple* variables.  Emit
5012         __gcov_indirect_call{,_topn} variables.
5013         (gimple_gen_ic_profiler): Access the variable
5014         and emit gimple.
5015         (gimple_gen_ic_func_profiler): Access
5016         __gcov_indirect_call.callee field.
5017         (gimple_init_gcov_profiler): Use ptr_type_node.
5018         * value-prof.c (gimple_ic): Use ptr_type_node.
5020 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5022         PR tree-optimization/85787
5023         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5024         malloc_candidate_p into this function and add support for
5025         detecting multiple phis.
5026         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5028 2018-10-04  Martin Liska  <mliska@suse.cz>
5030         PR ipa/87491
5031         * ipa-inline.c (inline_to_all_callers_1):
5032         Call ultimate_alias_target for node being inlined.
5034 2018-10-03  Jeff Law  <law@redhat.com>
5036         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5037         target's wchar_t.
5038         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5039         * tree.h (get_typenode_from_name): Prototype.
5041 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5043         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5044         Change operand 2 predicate to nonimmediate_operand.
5045         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5047 2018-10-03  Martin Sebor  <msebor@redhat.com>
5048             Jeff Law  <law@redhat.com>
5050         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5051         initialize it.
5052         (get_string_length): Detect unterminated arrays.
5053         (format_string): Same.
5054         (format_directive): Warn about unterminated arrays.
5055         (handle_gimple_call): Mark statements with no_warning as needed.
5057 2018-10-03  Jim Wilson  <jimw@sifive.com>
5059         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5060         also define __riscv_abi_rve.  Delete trailing white space.
5062 2018-10-03  Paul Koning  <ni1d@arrl.net>
5064         Enable LRA register allocator for PDP11.
5065         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5066         (R): Likewise.
5067         (D): Likewise.
5068         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5069         * config/pdp11/pdp11.opt (-mlra): New option.
5070         * doc/invoke.texi (PDP-11 Options): Document -mlra.
5072 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5074         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
5075         (*<absneg:code>extend<mode>xf2): Ditto.
5077 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5079         PR tree-optimization/87415
5080         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
5081         precision fields.
5083 2018-10-02  Jeff Law  <law@redhat.com>
5085         * gimple-fold.c (get_range_strlen): Only set *nonstr when
5086         an unterminated string is discovered.  Bubble up range
5087         even for unterminated strings.
5088         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
5089         indicates the string was not terminated via NONSTR.
5091 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5093         * tree-vrp.c (extract_range_from_unary_expr): Special case all
5094         pointer conversions.
5095         Do not do anything special for anti-ranges.
5097 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
5099         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
5100         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
5102 2018-10-03  Martin Liska  <mliska@suse.cz>
5104         PR gcov-profile/86109
5105         * coverage.c (coverage_begin_function): Do not
5106         mark lambdas as artificial.
5107         * tree-core.h (struct GTY): Remove tm_clone_flag
5108         and introduce new lambda_function.
5109         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
5111 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5113         PR target/87474
5114         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
5115         P8_VECTOR and VSX are enabled.
5117 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5119         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
5120         0x3907 as CPU model number.
5122 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5124         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
5125         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
5126         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
5127         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
5128         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
5129         * config/s390/s390.md: Likewise. Rename also the cpu attribute
5130         value from arch12 to z14.
5132 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5134         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5135         (isinfxf2): Ditto.
5136         (isinf<mode>2): Ditto.
5138 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5140         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5141         before emitting fxam.  Perform calculations in XFmode.
5143 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5145         * match.pd (((X /[ex] A) +- B) * A): New transformation.
5147 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5149         PR middle-end/87319
5150         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5151         * tree.c (signed_or_unsigned_type_for): Handle complex.
5153 2018-10-02  Jeff Law  <law@redhat.com>
5155         * gimple-fold.c (get_range_strlen): Remove dead code.
5157 2018-10-02  Martin Sebor  <msebor@redhat.com>
5158             Jeff Law  <law@redhat.com>
5160         * builtins.c (unterminated_array): Add new arguments.
5161         If argument is not terminated, bubble up size and exact
5162         state to callers.
5163         (expand_builtin_strnlen): Detect, avoid expanding
5164         and diagnose unterminated arrays.
5165         (c_strlen): Fill in offset of start of unterminated strings.
5166         * builtins.h (unterminated_array): Update prototype.
5168 2018-10-02  Richard Biener  <rguenther@suse.de>
5170         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5171         of haddv4df, first reduce to SSE width and exploit the fact
5172         that we only need element zero with the reduction result.
5173         (reduc_plus_scal_v2df): Likewise.
5175 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5177         * dojump.h (do_jump): Delete.
5178         (do_jump_1): Likewise.
5179         (split_comparison): Move around.
5180         * dojump.c (do_jump): Make static.
5181         (do_jump_1): Likewise.
5182         (jumpifnot): Move around.
5183         (jumpifnot_1): Likewise.
5184         (jumpif): Likewise.
5185         (jumpif_1): Likewise.
5186         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5188 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5190         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5191         insns in the delay slot and add_insn_after for the jump insn.
5193 2018-10-02  Richard Biener  <rguenther@suse.de>
5195         * tree-inline.c (expand_call_inline): Use the location of
5196         the callee declaration for the inline-entry marker.
5197         * final.c (notice_source_line): Remove special-casing of
5198         NOTE_INSN_INLINE_ENTRY.
5200 2018-10-01  Carl Love  <cel@us.ibm.com>
5202         PR 69431
5203         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5204         (__builtin_mtfsb0): New.
5205         (__builtin_mtfsb1): New.
5206         ( __builtin_set_fpscr_rn): New.
5207         (__builtin_set_fpscr_drn): New.
5208         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5209         (rs6000_expand_set_fpscr_rn_builtin): Add.
5210         (rs6000_expand_set_fpscr_drn_builtin): Add.
5211         (rs6000_expand_builtin): Add case statement entries for
5212         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5213         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5214         RS6000_BUILTIN_MFFSL.
5215         (rs6000_init_builtins): Add ftype initialization and def_builtin
5216         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5217         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5218         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5219         rs6000_mffscdrn): Add define_insn.
5220         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5221         * doc/extend.texi: Add documentation for the builtins.
5223 2018-10-01  Richard Biener  <rguenther@suse.de>
5225         PR tree-optimization/87465
5226         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5227         causing branch miscounts.
5229 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5231         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5232         aarch64_option_default_param):  New.
5233         (params.h): Include.
5234         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5235         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5236         stack-clash protection validation code.
5238 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5240         * params.c (validate_param): New.
5241         (add_params): Use it.
5242         (set_param_value): Refactor param validation into validate_param.
5243         (diagnostic.h): Include.
5244         * diagnostic.h (diagnostic_ready_p): New.
5246 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5248         * params.c (set_param_value):
5249         Add index of parameter being validated.
5250         * common/common-target.def (option_validate_param): New.
5251         * common/common-targhooks.h (default_option_validate_param): New.
5252         * common/common-targhooks.c (default_option_validate_param): New.
5253         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5254         * doc/tm.texi: Regenerate.
5256 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5258         PR target/86486
5259         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5260         Add validation for stack-clash parameters and set defaults.
5262 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5264         PR target/86486
5265         * configure.ac: Add stack-clash-protection-guard-size.
5266         * doc/install.texi: Document it.
5267         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5268         * params.def: Update comment for guard-size.
5269         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5270         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5271         * configure: Regenerate.
5273 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5275         PR target/86486
5276         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5277         STACK_DYNAMIC_OFFSET): New.
5278         * config/aarch64/aarch64.c (aarch64_layout_frame):
5279         Update outgoing args size.
5280         (aarch64_stack_clash_protection_alloca_probe_range,
5281         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5283 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5285         PR target/86486
5286         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5287         probe ranges.
5288         * target.def (stack_clash_protection_alloca_probe_range): New.
5289         (stack_clash_protection_final_dynamic_probe): Remove.
5290         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5291         (default_stack_clash_protection_final_dynamic_probe): Remove.
5292         * targhooks.c: Likewise.
5293         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5294         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5295         * doc/tm.texi: Regenerate.
5297 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5299         PR target/86486
5300         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5301         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5302         aarch64_clamp_to_uimm12_shift): New.
5303         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5304         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5306 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5308         PR target/86486
5309         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5311 2018-10-01  Jeff Law  <law@redhat.com>
5312             Richard Sandiford <richard.sandiford@linaro.org>
5313             Tamar Christina  <tamar.christina@arm.com>
5315         PR target/86486
5316         * config/aarch64/aarch64.md
5317         (probe_stack_range): Add k (SP) constraint.
5318         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5319         STACK_CLASH_MAX_UNROLL_PAGES): New.
5320         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5321         stack probes for stack clash.
5322         (aarch64_allocate_and_probe_stack_space): New.
5323         (aarch64_expand_prologue): Use it.
5324         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5325         (aarch64_sub_sp): Add emit_move_imm optional param.
5327 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5329         PR tree-optimization/87261
5330         * match.pd: Remove trailing whitespace.
5331         Add (x & y) | ~(x | y) -> ~(x ^ y),
5332         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5334 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5336         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5337         constraints.
5338         (ashlsi3_insn): Update instruction constraints.
5339         (ashrsi3_insn): Likewise.
5340         (rotrsi3): Likewise.
5341         (add_shift): Likewise.
5342         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5343         avoids placing in the limm field small constants which, otherwise,
5344         could end into a small instruction.
5346 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5348         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5349         destination register is not odd-even.
5350         (umaddsidi4_split): Likewise.
5352 2018-10-01  Richard Biener  <rguenther@suse.de>
5354         * tree-inline.c (expand_call_inline): Store origin of fn
5355         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5356         * tree.c (block_ultimate_origin): Simplify and do some
5357         checking.
5359 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5361         * config/i386/mmx.md (EMMS): New int iterator.
5362         (emms): New int attribute.
5363         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5364         EMMS int iterator.  Explicitly declare clobbers.
5365         (mmx_emms): Remove expander.
5366         (mmx_femms): Ditto.
5367         * config/i386/predicates.md (emms_operation): Remove predicate.
5368         (vzeroall_pattern): New predicate.
5369         (vzeroupper_pattern): Rename from vzeroupper_operation.
5370         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5371         vzeroupper_pattern and vzeroall_pattern predicates.
5373 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5375         PR rtl-optimization/86939
5376         * ira-lives.c (make_hard_regno_born): Rename from this...
5377         (make_hard_regno_live): ... to this.  Remove update to conflict
5378         information.  Update function comment.
5379         (make_hard_regno_dead): Add conflict information update.  Update
5380         function comment.
5381         (make_object_born): Rename from this...
5382         (make_object_live): ... to this.  Remove update to conflict information.
5383         Update function comment.
5384         (make_object_dead):  Add conflict information update.  Update function
5385         comment.
5386         (mark_pseudo_regno_live): Call make_object_live.
5387         (mark_pseudo_regno_subword_live): Likewise.
5388         (mark_hard_reg_dead): Update function comment.
5389         (mark_hard_reg_live): Call make_hard_regno_live.
5390         (process_bb_node_lives): Likewise.
5391         * lra-lives.c (make_hard_regno_born): Rename from this...
5392         (make_hard_regno_live): ... to this.  Remove update to conflict
5393         information.  Remove now uneeded check_pic_pseudo_p argument.
5394         Update function comment.
5395         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5396         to conflict information.  Update function comment.
5397         (mark_pseudo_live): Remove update to conflict information.  Update
5398         function comment.
5399         (mark_pseudo_dead): Add conflict information update.
5400         (mark_regno_live): Call make_hard_regno_live.
5401         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5402         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5404 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5406         PR target/87370
5407         * config/i386/i386.c (construct_container): Use TImode for
5408         BLKmode values in 2 integer registers.
5410 2018-09-29  Jeff Law  <law@redhat.com>
5412         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5413         c_strlen rather than just a tree *.
5414         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5415         Update recursive calls appropriately.  If caller did not provide a
5416         suitable data pointer, create a local one.  When a non-terminated
5417         string is discovered, bubble up information about the string via the
5418         c_strlen_data object.
5419         * builtins.h (c_strlen): Update prototype.
5420         (c_strlen_data): New structure.
5421         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5422         For a type 2 call, if c_strlen indicates a non-terminated string
5423         use the length of the non-terminated string.
5424         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5426 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5428         PR target/87467
5429         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5430         __m512d type for __A argument rather than __m512.
5432 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5434         * match.pd (simple_comparison): Don't optimize if either operand is
5435         a function pointer when target needs function pointer canonicalization.
5437 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5439         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5440         power5 .. power9 to remove indirection.
5441         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5442         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5443         ASM_CPU_476_SPEC): Delete.
5444         (ASM_CPU_SPEC): Adjust.
5445         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5446         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5448 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5450         * config.in: Delete HAVE_AS_DCI.
5451         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5452         * config/rs6000/rs6000.h: Ditto.
5453         * configure.ac: Delete HAVE_AS_DCI.
5454         * configure: Regenerate.
5456 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5458         * config.in (HAVE_AS_LWSYNC): Delete.
5459         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5460         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5461         do it as a .long .
5462         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5463         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5464         as a .long .
5465         * configure.ac: Delete HAVE_AS_LWSYNC.
5466         * configure: Regenerate.
5468 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5469             Pierre-Marie de Rodat  <derodat@adacore.com>
5471         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5472         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5473         (cgraph_node::create_thunk): Add indirect_offset parameter.
5474         (thunk_adjust): Likewise.
5475         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5476         and initialize the corresponding field with it.
5477         (cgraph_node::dump): Dump indirect_offset field.
5478         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5479         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5480         (thunk_adjust): Add indirect_offset parameter and deal with it.
5481         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5482         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5483         or if the thunk is external or local.  Fix formatting.  Do not chain
5484         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5485         if any, in the GIMPLE representation.
5486         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5487         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5488         (input_node): Read indirect_offset field.
5489         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5490         call to thunk_adjust.
5491         * tree-nested.c (struct nesting_info): Add thunk_p field.
5492         (create_nesting_tree): Set it.
5493         (convert_all_function_calls): Copy static chain from targets to thunks.
5494         (finalize_nesting_tree_1): Return early for thunks.
5495         (unnest_nesting_tree_1): Do not finalize thunks.
5496         (gimplify_all_functions): Do not gimplify thunks.
5498 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5500         * opt-suggestions.c (option_proposer::build_option_suggestions):
5501         Release "option_values".
5503 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5505         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5506         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5507         * dumpfile.c (kind_as_string): New function.
5508         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5509         kind_as_string to vary the prefix based on dump_kind.
5510         (selftest::test_capture_of_dump_calls): Update for above.
5512 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5514         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5515         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5517 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5519         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5520         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5521         INVALID_REGNUM instead of FPSR_REG.
5522         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5523         * config/i386/i386.md: Update comment of FP compares.
5524         (fldenv): Do not clobber FPSR_REG.
5526 2018-09-28  Richard Biener  <rguenther@suse.de>
5528         * tree.h (BLOCK_ORIGIN): New.
5529         * omp-expand.c (grid_expand_target_grid_body): Assign
5530         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5531         * tree-inline.c (remap_block): Likewise.
5532         * auto-profile.c (get_function_decl_from_block): Simplify
5533         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5534         * langhooks.c (lhd_print_error_function): Likewise.
5535         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5536         Likewise.
5537         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5538         * tree.c (block_nonartificial_location): Likewise.
5539         (block_ultimate_origin): Likewise.
5540         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5541         no longer needed LTO case.
5543 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5544             Jan Hubicka  <jh@suse.cz>
5545             Martin Jambor  <mjambor@suse.cz>
5547         * simplify-rtx.c (simplify_merge_mask): New function.
5548         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5549         same masks are used in op1 or op2.
5550         (test_vec_merge): New function.
5551         (test_vector_ops): Call test_vec_merge.
5553 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5555         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5556         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5557         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5558         (cypress_costs): Set it.
5559         (supersparc_costs): Likewise.
5560         (hypersparc_costs): Likewise.
5561         (leon_cost): Likewise.
5562         (leon3_costs): Likewise.
5563         (sparclet_costs): Likewise.
5564         (ultrasparc_costs): Likewise.
5565         (ultrasparc_costs): Likewise.
5566         (niagara_costs): Likewise.
5567         (niagara2_costs): Likewise.
5568         (niagara3_costs): Likewise.
5569         (niagara4_costs): Likewise.
5570         (niagara7_costs): Likewise.
5571         (m8_costs): Likewise.
5572         (TARGET_CAN_FOLLOW_JUMP): Define.
5573         (pass_work_around_errata::gate): Minor tweak.
5574         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5575         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5576         Automaitcally clear MASK_FSMULD mask for V7 processors.
5577         (sparc_can_follow_jump): New static function.
5578         (output_ubranch): Deal with CROSSING_JUMP_P.
5579         (sparc_use_sched_lookahead): Rewrite using switch statement.
5580         (sparc_issue_rate): Reorder.
5581         (sparc_branch_cost): New function.
5583 2018-09-27  Martin Sebor  <msebor@redhat.com>
5585         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5586         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5587         (int_fits_type_p): Same.
5589 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5591         * config/i386/i386.md (FPCR_REG): Remove.
5592         (UNSPEC_FLDCW): Remove.
5593         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5594         (x86_fldcw_1): Remove insn pattern.
5595         (fnstenv): Do not clobber FPCR_REG.
5596         (fldenv): Ditto.
5597         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5598         (CALL_USED_REGISTERS): Ditto.
5599         (REG_ALLOC_ORDER): Ditto.
5600         (REG_CLASS_CONTENTS): Ditto.
5601         (HI_REGISTER_NAMES): Ditto.
5602         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5603         of numerical constants.
5604         * config/i386/i386.c (regclass_map): Remove fpsr register.
5605         (dbx_register_map): Ditto.
5606         (dbx64_register_map): Ditto.
5607         (svr4_dbx_register_map): Ditto.
5608         (print_reg): Do not handle FPCR_REG.
5610 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5612         PR target/87149
5613         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5614         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5615         Delete, always treat as true.
5616         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5617         Ditto.  Simplify remaining code.
5618         * config/powerpcspe/powerpcspe.h: Ditto.
5619         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5620         Simplify remaining code.
5621         (rs6000_expand_builtin): Ditto.
5622         * config/rs6000/rs6000.h: Ditto.
5623         * configure.ac: Ditto.
5624         * configure: Regenerate.
5626 2018-09-27  Martin Liska  <mliska@suse.cz>
5628         * coverage.c (get_coverage_counts): Revert the formatting
5629         of missing profile opt info.
5631 2018-09-27  Richard Biener  <rguenther@suse.de>
5633         PR debug/37801
5634         PR debug/87440
5635         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5636         block as we do not output that.
5637         (gen_inlined_subroutine_die): Elide the originally outermost
5638         block, matching what we do for concrete instances.
5639         (decls_for_scope): Add parameter specifying whether to recurse
5640         to subblocks.
5642 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5643             Tom de Vries  <tom@codesourcery.com>
5645         PR 82089
5647         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5648         STORE_FLAG_VALUE == 1.
5650 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5652         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5653         constant definitions.
5654         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5655         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5656         ("speculation_barrier"): New expander definition.
5658 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5660         PR gcov-profile/86957
5661         * common.opt: New warning option -Wmissing-profile.
5662         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5663         * doc/invoke.texi: Document -Wmissing-profile.
5665 2018-09-26  Jim Wilson  <jimw@sifive.com>
5667         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5668         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5670 2018-09-26  Martin Sebor  <msebor@redhat.com>
5672         * tree.c (zerop): Change return type to bool.
5673         (integer_zerop, integer_onep, integer_each_onep): Same.
5674         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5675         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5676         (real_onep, real_minus_onep, chain_index): Same.
5677         (print_type_hash_statistics, type_list_equal): Same.
5678         * tree.h (zerop): Same.
5679         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5680         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5681         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5682         (real_onep, real_minus_onep, chain_index): Same.
5683         (print_type_hash_statistics, type_list_equal): Same.
5685 2018-09-26  Jim Wilson  <jimw@sifive.com>
5687         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5689 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5691         PR target/87414
5692         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5693         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5694         call.
5696 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5698         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5700 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5702         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5703         and FP_SECOND_SSE_REGS.
5704         (REG_CLASS_NAMES): Ditto.
5705         (REG_CLASS_CONTENTS): Ditto.
5706         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5707         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5708         (ix86_preferred_output_reload_class): Ditto.
5709         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5710         clobber constraint to "=&f".
5711         (fix_truncdi_i387): Ditto.
5712         (lrintxfdi2): Ditto.
5713         (fistdi2_<rounding>): Ditto.
5714         (fpremxf4_i387): Change "=u" constraint to "=f".
5715         (fprem1xf4_i387): Ditto.
5716         (sincosxf3): Ditto.
5717         (fptanxf4_i387): Ditto.
5718         (fxtractxf3_i387): Ditto.
5719         (fscalexf4_i387): Ditto.
5720         (atan2xf3): Change "u" constraint to "f".
5721         (fyl2xxf3_i387): Ditto.
5722         (fyl2xp1xf3_i387): Ditto.
5724 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5726         PR target/87439
5727         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5728         for removed I387_MASK_PM entity.
5731 2018-09-26  Jeff Law  <law@redhat.com>
5732         Revert
5733         2018-09-26  Alexey Neyman  <stilor@att.net>
5735         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5736         headers are no longer pulled in by <isl/val.h>.
5738 2018-09-26  Richard Biener  <rguenther@suse.de>
5740         PR debug/87443
5741         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
5742         or concrete instance DIE to the tree.  Create abstract origin
5743         attributes also for concrete instances.
5745 2018-09-26  Alexey Neyman  <stilor@att.net>
5747         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5748         headers are no longer pulled in by <isl/val.h>.
5750 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
5752         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
5753         Use new helper functions.
5754         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
5755         Use new helper functions.
5756         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
5757         aarch_mm_needs_release): New declarations.
5758         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
5759         aarch_mm_needs_release): New.
5761 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
5763         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
5764         (arm32_output_mi_thunk): Deal with long calls.
5766 2018-09-26  Richard Biener  <rguenther@suse.de>
5768         PR debug/87428
5769         PR debug/87362
5770         * tree-inline.c (expand_call_inline): When the location
5771         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
5772         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
5773         the inserted BLOCK to make inlined_function_outer_scope_p
5774         recognize it.
5775         * dwarf2out.c (add_call_src_coords_attributes): Do not add
5776         coords for reserved locations.
5778 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5780         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
5781         (*call_indirect_nonlocal_sysv<mode>): Ditto.
5782         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
5783         (*sibcall_nonlocal_sysv<mode>): Ditto.
5784         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
5785         (<bd>_<mode>): Ditto.
5786         (<bd>tf_<mode>): Ditto.
5788 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5790         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
5791         control string as a list of templates instead of as C code.
5792         (*altivec_movti): Ditto.
5793         * config/rs6000/darwin.md (movdf_low_di): Ditto.
5795 2018-09-25  Jim Wilson  <jimw@sifive.com>
5797         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
5798         when target symbol is weak.
5800 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5802         PR c/87387
5803         * builtins.c (unterminated_array): Simplify.
5804         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
5805         where pointer arithmetic is safe.
5807 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5809         PR target/86987
5810         * config/rs6000/altivec.md (altivec_vspltb): Use
5811         const_0_to_15_operand instead of u5bit_cint_operand.
5812         (*altivec_vspltb_internal): Ditto.
5813         (altivec_vspltb_direct): Ditto.
5814         (altivec_vsplth): Use const_0_to_7_operand instead of
5815         u5bit_cint_operand.
5816         (*altivec_vsplth_internal): Ditto.
5817         (altivec_vsplth_direct): Ditto.
5818         (altivec_vspltw): Use const_0_to_3_operand instead of
5819         u5bit_cint_operand.
5820         (*altivec_vspltw_internal): Ditto.
5821         (altivec_vspltw_direct): Ditto.
5822         (altivec_vspltsf): Ditto.
5823         (*altivec_vspltsf_internal): Ditto.
5824         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
5825         various splats with the proper size immediate.  Reorder the various
5826         cases by ascending size of immediate, and put all such together.
5828 2018-09-25  Richard Biener  <rguenther@suse.de>
5830         PR debug/83941
5831         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
5832         GC-ification.
5833         (maybe_create_die_with_external_ref): Do not create
5834         DW_TAG_imported_unit here.
5835         (add_abstract_origin_attribute): Handle external BLOCK refs.
5836         (dwarf2out_abstract_function): Simplify LTO case.
5837         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
5838         rather than using maybe_create_die_with_external_ref.
5840 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
5842         PR target/71278
5843         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
5844         (frndintxf2_mask_pm_i387): Ditto.
5845         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
5846         Enable for !flag_trapping_math.
5847         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
5848         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
5849         Change operand 1 predicate to nonimmediate_operand.
5850         (attr "i387_cw"): Remove mask_pm.
5851         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
5852         (enum ix86_entity): Remove I387_MASK_PM.
5853         * config/i386/i386.c (ix86_i387_mode_needed): Do not
5854         handle I387_MASK_PM.
5855         (ix86_mode_needed): Ditto.
5856         (ix86_mode_after): Ditto.
5857         (ix86_mode_entry): Ditto.
5858         (ix86_mode_exit): Ditto.
5859         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
5861 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
5863         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
5864         to_update_switch_stmts to vNULL instead of calling create on them
5865         immediately.
5867 2018-09-25  Richard Biener  <rguenther@suse.de>
5869         PR tree-optimization/87402
5870         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
5871         (visit_phi): Re-instantiate handling of supposed to be VARYING
5872         but non-VARYING backedge value.
5874 2018-09-25  Richard Biener  <rguenther@suse.de>
5876         PR debug/83941
5877         * dwarf2out.c (struct sym_off_pair): New.
5878         (external_die_map): New global.
5879         (lookup_decl_die): When in LTO create DIEs lazily from the
5880         external_die_map.
5881         (lookup_block_die): New function, create DIEs lazily in LTO.
5882         (equate_block_to_die): New function.
5883         (dwarf2out_die_ref_for_decl): During WPA get the association
5884         from the external DIE map.
5885         (dwarf2out_register_external_die): Record mapping into the
5886         external DIE map.
5887         (maybe_create_die_with_external_ref): New function split out from
5888         DIE generation part of old dwarf2out_register_external_die.
5889         (add_abstract_origin_attribute): Do not return the DIE.  When
5890         in LTO reference externals directly.
5891         (dwarf2out_abstract_function): When in LTO ignore calls for
5892         decls with external DIEs (already present abstract instances).
5893         (gen_call_site_die): Adjust.
5894         (add_high_low_attributes): Likewise.
5895         (gen_lexical_block_die): Likewise.
5896         (gen_inlined_subroutine_die): Likewie.
5897         (gen_block_die): Likewise.
5898         (dwarf2out_inline_entry): Likewise.
5899         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
5900         DIEs.
5902 2018-09-25  Martin Liska  <mliska@suse.cz>
5904         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
5905         integers and not by a float value.
5907 2018-09-25  Martin Liska  <mliska@suse.cz>
5909         PR fortran/87394
5910         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
5911         instead of NULL.
5912         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
5913         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
5914         * godump.c (go_format_type): Remove extra parenthesis.
5916 2018-09-25  Martin Liska  <mliska@suse.cz>
5918         * alias.c (set_dest_equal_p): Remove unused function.
5919         * config/i386/i386.c (def_builtin_pure2): Likewise.
5920         * diagnostic-show-locus.c (class layout): Remove
5921         unused field.
5922         (layout::layout): Likewise here.
5923         * dump-context.h (class temp_dump_context): Likewise.
5924         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
5925         (add_AT_loclistsptr): Likewise.
5926         (add_AT_offset): Likewise.
5927         (get_AT_hi_pc): Likewise.
5928         (is_comdat_die): Likewise.
5929         (type_is_enum): Likewise.
5930         (ceiling): Likewise.
5931         (add_AT_vms_delta): Likewise.
5932         (is_class_die): Likewise.
5933         * edit-context.c (class line_event): Remove unused field.
5934         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
5935         unused function.
5936         * ipa-cp.c (ipa_get_vr_lat): Likewise.
5937         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
5938         (ok_for_base_p_nonstrict): Likewise.
5939         * tree-chrec.c (is_not_constant_evolution): Likewise.
5940         (chrec_fold_poly_cst): Likewise.
5941         * tree-if-conv.c (has_pred_critical_p): Likewise.
5942         * tree-ssa-coalesce.c (print_exprs): Likewise.
5943         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
5944         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
5945         * tree-vrp.c (value_ranges_intersect_p): Likewise.
5946         (value_range_nonnegative_p): Likewise.
5948 2018-09-25  Martin Liska  <mliska@suse.cz>
5950         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5951         Do not handle "GNU Pascal".
5952         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5953         Likewise.
5954         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
5955         from documentation. Likewise.
5956         * dbxout.c (dbxout_range_type): Likewise.
5957         * doc/cpp.texi: Likewise.
5958         * doc/extend.texi: Likewise.
5959         * doc/frontends.texi: Likewise.
5960         * doc/invoke.texi: Remove Pascal entry.
5961         * tree.def (CLEANUP_POINT_EXPR): Likewise.
5962         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
5964 2018-09-25  Martin Liska  <mliska@suse.cz>
5966         PR middle-end/86078
5967         * doc/invoke.texi: Document all parameters and remove default
5968         of the parameters.
5970 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5972         PR bootstrap/87417
5973         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
5974         contains HOST_WIDE_INTs when computing its size.
5976 2018-09-24  Jim Wilson  <jimw@sifive.com>
5978         PR target/87391
5979         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
5980         not TARGET_RVE.
5981         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
5983 2018-09-24  Andrew Pinski  <apinski@marvell.com>
5985         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
5986         access prev before checking it for NULLness in the
5987         AARCH64_FUSE_CMP_BRANCH case.
5989 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
5991         PR target/82699
5992         * config/i386/i386.c (rest_of_insert_endbranch): Set
5993         endbr_queued_at_entrance to true and don't insert ENDBR if
5994         x86_function_profiler will be called.
5995         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
5996         is true.
5997         * config/i386/i386.h (machine_function): Add
5998         endbr_queued_at_entrance.
6000 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6002         * genattrtab.c (mk_attr_alt): Use alternative_mask.
6003         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6004         types.
6005         (check_attr_test): Use alternative_mask.
6006         (get_attr_value): Likewise.
6007         (compute_alternative_mask): Use alternative_mask and XWINT.
6008         (make_alternative_compare): Use alternative_mask.
6009         (attr_alt_subset_p): Use XWINT.
6010         (attr_alt_subset_of_compl_p): Likewise.
6011         (attr_alt_intersection): Use alternative_mask and XWINT.
6012         (attr_alt_union): Likewise.
6013         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6014         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6015         (simplify_test_exp): Use alternative_mask and XWINT.
6016         (write_test_expr): Use alternative_mask and XWINT, adjust bit
6017         number calculation to support 64 bits.  Generate code that
6018         checks 64-bit masks.
6019         (main): Use alternative_mask.
6020         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6022 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6024         PR target/80080
6025         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6026         RETURN+USE when returning via %r14.
6028 2018-09-24  Martin Liska  <mliska@suse.cz>
6030         * gcov.c (output_lines): Print colorization legend
6031         for both flag_use_colors and flag_use_hotness_colors.
6032         Reword the help.
6034 2018-09-24  Martin Liska  <mliska@suse.cz>
6036         * coverage.c (get_coverage_counts): Use warning_at
6037         with current_function_decl location. Use %qD in warning
6038         message.
6040 2018-09-24  Martin Liska  <mliska@suse.cz>
6042         * memory-block.h (memory_block_pool::release): Annotate with
6043         valgrind that the memory is not accessible.
6045 2018-09-24  Martin Liska  <mliska@suse.cz>
6047         PR sanitizer/85774
6048         * asan.c: Make asan_handled_variables extern.
6049         * asan.h: Likewise.
6050         * cfgexpand.c (expand_stack_vars): Make sure
6051         a representative is unpoison if another
6052         variable in the partition is handled by
6053         use-after-scope sanitization.
6055 2018-09-24  Richard Biener  <rguenther@suse.de>
6057         PR tree-optimization/63155
6058         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6059         the worklist when the edge of the respective argument isn't
6060         executable.
6062 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6064         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6065         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6066         (MASK_CLASS_P): Update for rename.
6067         (MAYBE_MASK_CLASS_P): Ditto.
6068         (REG_CLASS_NAMES): Update.
6069         (REG_CLASS_CONTENT): Update.
6070         * config/i386/i386.c (regclass_map): Update for MASK_REG
6071         and ALL_MASK_REGS rename.
6072         * config/i386/constraints.md (Yk): Update for rename.
6073         (k): Ditto.
6075 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6077         * config/i386/i386.h (enum reg_class): Remove
6078         EVEX_SSE_REGS and MOD4_SSE_REGS.
6079         (REG_CLASS_NAMES): Update.
6080         (REG_CLASS_CONTENT): Update.
6081         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
6082         registers as ALL_SSE_REGS.
6083         (ix86_additional_allocno_class_p): Remove.
6084         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
6085         (ix86_register_priority): Lower priority of EVEX SSE registers.
6086         Use IN_RANGE macro where appropriate.
6087         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
6088         AVX-5124VNNIW checks.
6089         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
6090         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
6091         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
6092         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
6093         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
6094         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
6095         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
6096         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
6097         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
6098         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
6099         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
6100         * config/i386/constraints.md (Yh): Remove.
6102 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6104         * config/i386/i386.c (regclass_map): Declare integer REX registers
6105         as GENERAL_REGS.
6107 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
6109         * doc/service.texi (Service): Switch the fsf.org link to https.
6111 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
6113         PR target/86798
6114         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6115         Define to speculation_safe_value_not_needed.
6117 2018-09-21  Florian Weimer  <fweimer@redhat.com>
6119         PR middle-end/81035
6120         * doc/extend.texi (Common Function Attributes): Mention that
6121         noreturn suppresses tail call optimization.
6123 2018-09-21  Jeff Law  <law@redhat.com>
6125         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
6126         vr_values::cleanup_edges_and_switches.
6127         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
6128         vr_values class.
6129         (identify_jump_threads): Remove EDGE_IGNORE handling.
6130         (execute_vrp): Move handling of to_remove_edges and
6131         to_update_switch_stmts into vr_values class member functions.
6132         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6133         (to_update_switch_stmts): Likewise.
6134         * vr-values.c: Include cfghooks.h.
6135         (vr_values::vr_values): Initialize to_remove_edges and
6136         to_update_switch_stmts.
6137         (vr_values::~vr_values): Verify to_remove_edges and
6138         to_update_switch_stmts are empty.
6139         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6140         (vr_values::cleanup_edges_and_switches): New member function.
6141         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6142         function.  Add new data members.
6144 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
6146         PR tree-optimization/87309
6147         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6148         calls with pflags and alt_flags.
6149         (selftest::test_capture_of_dump_calls): Add test of interaction of
6150         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6152 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6154         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6156 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6158         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6160 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6162         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6163         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6164         (VXWORKS_LIBS_RTP): Minor reordering.
6166 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6168         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6169         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6171 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6173         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6174         (PTRDIFF_TYPE): Likewise.
6176 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6178         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6179         triplet, similar to support for VxWorks7.
6180         * config/vxworks-dummy.h: Provide a default definition
6181         of TARGET_VXWORKS64 to 0.
6183 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6185         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6186         * config/vxworks-dummy.h: here.
6188 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6190         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6192 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
6193             Bo Zhou  <zbo.zhou@hisilicon.com>
6195         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6196         * config/aarch64/aarch64-tune.md: Regenerated.
6197         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6198         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6199         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6201 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
6202             Julian Brown  <julian@codesourcery.com>
6204         * builtins.c (get_builtin_sync_mem): Handle address spaces.
6206 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6208         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6209         if the call takes a static chain.
6211 2018-09-21  Martin Liska  <mliska@suse.cz>
6213         * auto-profile.c (autofdo_source_profile::read): Do not
6214         set sum_all.
6215         (read_profile): Do not add working sets.
6216         (read_autofdo_file): Remove sum_all.
6217         (afdo_callsite_hot_enough_for_early_inline): Remove const
6218         qualifier.
6219         * coverage.c (struct counts_entry): Remove gcov_summary.
6220         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6221         do not support GCOV_TAG_PROGRAM_SUMMARY.
6222         (get_coverage_counts): Remove summary and expected
6223         arguments.
6224         * coverage.h (get_coverage_counts): Likewise.
6225         * doc/gcov-dump.texi: Remove -w option.
6226         * gcov-dump.c (dump_working_sets): Remove.
6227         (main): Do not support '-w' option.
6228         (print_usage): Likewise.
6229         (tag_summary): Likewise.
6230         * gcov-io.c (gcov_write_summary): Do not dump
6231         histogram.
6232         (gcov_read_summary): Likewise.
6233         (gcov_histo_index): Remove.
6234         (gcov_histogram_merge): Likewise.
6235         (compute_working_sets): Likewise.
6236         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6237         it not obsolete.
6238         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6239         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6240         (GCOV_HISTOGRAM_SIZE): Remove.
6241         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6242         (struct gcov_summary): Simplify rapidly just
6243         to runs and sum_max fields.
6244         (gcov_histo_index): Remove.
6245         (NUM_GCOV_WORKING_SETS): Likewise.
6246         (compute_working_sets): Likewise.
6247         * gcov-tool.c (print_overlap_usage_message): Remove
6248         trailing empty line.
6249         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6250         (output_lines): Remove program related line.
6251         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6252         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6253         histogram.
6254         (input_profile_summary): Do not read it.
6255         (merge_profile_summaries): And do not merge it.
6256         (input_symtab): Do not call removed function.
6257         * modulo-sched.c (sms_schedule): Do not print sum_max.
6258         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6259         removed when histogram method was invented.
6260         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6261         mode.
6262         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6263         GCOV coding style.
6264         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6265         and dump selected value.
6266         * profile.c (add_working_set): Remove.
6267         (get_working_sets): Likewise.
6268         (find_working_set): Likewise.
6269         (get_exec_counts): Do not work with working sets.
6270         (read_profile_edge_counts): Do not inform as sum_max is removed.
6271         (compute_branch_probabilities): Likewise.
6272         (compute_value_histograms): Remove argument for call of
6273         get_coverage_counts.
6274         * profile.h: Do not make gcov_summary const.
6276 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6278         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6280 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6282         PR tree-optimization/86990
6283         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6284         Check that the entire merged store group is made of constants only for
6285         overlapping stores.
6287 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6289         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6290         (VTABLE_VERIFICATION_SPEC): Likewise.
6291         (SANITIZER_EARLY_SPEC): Likewise.
6292         (SANITIZER_SPEC): Likewise.
6293         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6294         * doc/invoke.texi (Link Options): Document -r.
6296 2018-09-20  Richard Biener <rguenther@suse.de>
6298         PR middle-end/87054
6299         * gimplify.c (gimplify_expr): Retain alignment of
6300         addressable lvalue in dereference.
6302 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6304         PR bootstrap/87013
6305         * configure.ac: Check for .loc is_stmt support.
6306         * configure, config.in: Rebuilt.
6307         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6308         if not supported.
6310 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6312         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6313         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6314         -misel=no.
6316 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6318         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6319         VECTOR_OTHER.
6320         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6321         case VECTOR_OTHER.
6323 2018-09-20  Marek Polacek  <polacek@redhat.com>
6325         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6327 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6329         PR tree-optimization/87288
6330         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6331         into account when determining PEELING_FOR_NITERS.
6333 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6335         PR tree-optimization/86877
6336         * tree-vect-loop.c (vect_analyze_loop_2): Call
6337         vect_verify_datarefs_alignment.
6339 2018-09-19  Marek Polacek  <polacek@redhat.com>
6341         * doc/invoke.texi: Document -Wclass-conversion.
6343 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6345         * config/pa/pa.c (pa_adjust_priority): Delete.
6346         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6348         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6349         (atomic_storehi): Likewise.
6350         (atomic_storesi): Likewise.
6351         (atomic_loaddi): Restore compare and swap exchange loop code.
6353 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6355         PR rtl-optimization/86902
6356         * combine.c (try_combine): When changing the CC mode used, don't change
6357         an unrelated mode in other_insn to that new CC mode.
6359 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6361         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6362         with %T in place of calls to dump_generic_expr.
6363         (prune_runtime_alias_test_list): Likewise.
6364         (create_runtime_alias_checks): Likewise.
6365         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6366         (vect_analyze_data_ref_dependence): Likewise.
6367         (vect_slp_analyze_data_ref_dependence): Likewise.
6368         (vect_record_base_alignment): Likewise.  Use %G in place of call
6369         to dump_gimple_stmt.
6370         (vect_compute_data_ref_alignment): Likewise.
6371         (verify_data_ref_alignment): Likewise.
6372         (vect_find_same_alignment_drs): Likewise.
6373         (vect_analyze_group_access_1): Likewise.
6374         (vect_analyze_data_ref_accesses): Likewise.
6375         (dependence_distance_ge_vf): Likewise.
6376         (dump_lower_bound): Likewise.
6377         (vect_prune_runtime_alias_test_list): Likewise.
6378         (vect_find_stmt_data_reference): Likewise.
6379         (vect_analyze_data_refs): Likewise.
6380         (vect_create_addr_base_for_vector_ref): Likewise.
6381         (vect_create_data_ref_ptr): Likewise.
6382         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6383         (vect_can_advance_ivs_p): Likewise.
6384         (vect_update_ivs_after_vectorizer): Likewise.
6385         (vect_gen_prolog_loop_niters): Likewise.
6386         (vect_prepare_for_masked_peels): Likewise.
6387         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6388         (vect_determine_vectorization_factor): Likewise.
6389         (vect_is_simple_iv_evolution): Likewise.
6390         (vect_analyze_scalar_cycles_1): Likewise.
6391         (vect_analyze_loop_operations): Likewise.
6392         (report_vect_op): Likewise.
6393         (vect_is_slp_reduction): Likewise.
6394         (check_reduction_path): Likewise.
6395         (vect_is_simple_reduction): Likewise.
6396         (vect_create_epilog_for_reduction): Likewise.
6397         (vect_finalize_reduction:): Likewise.
6398         (vectorizable_induction): Likewise.
6399         (vect_transform_loop_stmt): Likewise.
6400         (vect_transform_loop): Likewise.
6401         (optimize_mask_stores): Likewise.
6402         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6403         (vect_split_statement): Likewise.
6404         (vect_recog_over_widening_pattern): Likewise.
6405         (vect_recog_average_pattern): Likewise.
6406         (vect_determine_min_output_precision_1): Likewise.
6407         (vect_determine_precisions_from_range): Likewise.
6408         (vect_determine_precisions_from_users): Likewise.
6409         (vect_mark_pattern_stmts): Likewise.
6410         (vect_pattern_recog_1): Likewise.
6411         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6412         (vect_record_max_nunits): Likewise.
6413         (vect_build_slp_tree_1): Likewise.
6414         (vect_build_slp_tree_2): Likewise.
6415         (vect_print_slp_tree): Likewise.
6416         (vect_analyze_slp_instance): Likewise.
6417         (vect_detect_hybrid_slp_stmts): Likewise.
6418         (vect_detect_hybrid_slp_1): Likewise.
6419         (vect_slp_analyze_operations): Likewise.
6420         (vect_slp_analyze_bb_1): Likewise.
6421         (vect_transform_slp_perm_load): Likewise.
6422         (vect_schedule_slp_instance): Likewise.
6423         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6424         (vect_mark_stmts_to_be_vectorized): Likewise.
6425         (vect_init_vector_1): Likewise.
6426         (vect_get_vec_def_for_operand): Likewise.
6427         (vect_finish_stmt_generation_1): Likewise.
6428         (vect_check_load_store_mask): Likewise.
6429         (vectorizable_call): Likewise.
6430         (vectorizable_conversion): Likewise.
6431         (vectorizable_operation): Likewise.
6432         (vectorizable_load): Likewise.
6433         (vect_analyze_stmt): Likewise.
6434         (vect_is_simple_use): Likewise.
6435         (vect_get_vector_types_for_stmt): Likewise.
6436         (vect_get_mask_type_for_stmt): Likewise.
6437         * tree-vectorizer.c (increase_alignment): Likewise.
6439 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6441         * doc/rtl.texi: Adjust vec_select description.
6442         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6443         non-constant selectors.
6445 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6447         * config/aarch64/aarch64-protos.h
6448         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6449         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6450         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6451         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6452         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6453         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6454         (AARCH64_ISA_RCPC8_4): New flag.
6455         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6456         to aarch64_offset_9bit_signed_unscaled_p.
6457         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6458         and use stlur.
6459         * config/aarch64/constraints.md (Ust): New constraint.
6460         * config/aarch64/predicates.md.
6461         (aarch64_9bit_offset_memory_operand): New predicate.
6462         (aarch64_rcpc_memory_operand): New predicate.
6464 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6466         PR rtl-optimization/87361
6467         * rtlanal.c (nonzero_bits1): Revert accidental change.
6469 2018-09-19  Richard Biener  <rguenther@suse.de>
6471         PR tree-optimization/87349
6472         PR tree-optimization/87342
6473         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6475 2018-09-18  Marek Polacek  <polacek@redhat.com>
6477         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6478         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6480 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6482         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6484 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6486         PR rtl-optimization/86882
6487         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6489 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6491         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6492         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6494 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6496         PR other/87353
6497         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6499 2018-09-18  Richard Biener  <rguenther@suse.de>
6501         PR middle-end/63155
6502         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6503         (compute_samebase_partition_bases): Likewise.
6504         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6505         (gimple_can_coalesce_p): Simplify.
6507 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6509         Handle a library implementation of ffs calling __builtin_ffs.
6510         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6511         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6512         to __ffsdi2.
6514 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6516         * diagnostic-show-locus.c (class layout_range): Add field
6517         "m_original_idx".
6518         (layout_range::layout_range): Add "original_idx" param and use it
6519         to initialize new field.
6520         (make_range): Use 0 for original_idx.
6521         (layout::layout): Pass in index to calls to
6522         maybe_add_location_range.
6523         (layout::maybe_add_location_range): Add param "original_idx" and
6524         pass it on to layout_range.
6525         (layout::print_any_labels): Pass on range->m_original_idx to
6526         get_text call.
6527         (gcc_rich_location::add_location_if_nearby): Use 0 for
6528         original_idx.
6529         * gcc-rich-location.h (text_range_label::get_text): Update for new
6530         param.
6531         (range_label_for_type_mismatch::get_text): Likewise.
6533 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6535         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6537 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6539         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6540         format_string_diagnostic_t.
6541         (fmtwarn_n): Likewise.
6542         * substring-locations.c
6543         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6544         (format_warning_n_va): Convert to...
6545         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6546         (format_warning_va): Convert to...
6547         (format_string_diagnostic_t::emit_warning_va): ...this.
6548         (format_warning_at_substring): Convert to...
6549         (format_string_diagnostic_t::emit_warning): ...this.
6550         (format_warning_at_substring_n): Convert to...
6551         (format_string_diagnostic_t::emit_warning_n): ...this.
6552         * substring-locations.h (class format_string_diagnostic_t): New
6553         class.
6554         (format_warning_va): Convert to
6555         format_string_diagnostic_t::emit_warning_va.
6556         (format_warning_n_va): Convert to
6557         format_string_diagnostic_t::emit_warning_n_va.
6558         (format_warning_at_substring): Convert to
6559         format_string_diagnostic_t::emit_warning.
6560         (format_warning_at_substring_n): Convert to
6561         format_string_diagnostic_t::emit_warning_n.
6563 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6564             Bernd Schmidt <bernds_cb1@t-online.de>
6566         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6567         SImode args.
6569 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6571         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6572         operand 0 predicate to nonimmediate operand.
6573         (rint<mode>2_frndint): Remove insn pattern.
6574         (rint<mode>2): Change operand 1 predicate to general_operand.
6575         Extend operand 1 to XFmode and generate rintxf2 insn.
6576         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6577         Do not use X87MODEF mode macro.
6578         (frndintxf2_<rounding>_i387): Rename from
6579         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6580         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6581         to XFmode and generate significandxf3 insn.
6583 2018-09-17  Richard Biener  <rguenther@suse.de>
6585         PR tree-optimization/87328
6586         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6587         visiting unexecutable backedges when not iterating.
6588         (do_rpo_vn): Mark all edges not executable even when not
6589         iterating.
6591 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6593         PR c/63886
6594         * doc/invoke.texi (Warning Options): Likewise.
6596 2018-09-17  Richard Biener  <rguenther@suse.de>
6598         PR tree-optimization/87301
6599         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6600         clean EH info from leftover copy assignments.
6602 2018-09-17  Martin Liska  <mliska@suse.cz>
6604         PR gcov-profile/85871
6605         * gcov.c (output_intermediate_file): Fix out of bounds
6606         access.
6608 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6610         * config/arc/arc.c: Object attributes for core4 not reflected
6611         correctly.
6612         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6613         core3).
6615 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6617         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6619 2018-09-17  Martin Liska  <mliska@suse.cz>
6621         * doc/gcov.texi: Document new option --use-hotness-colors.
6622         * gcov.c (struct source_info): Declare new field.
6623         (source_info::source_info): Set default for maximum_count.
6624         (print_usage): Add new -q option.
6625         (process_args): Process it.
6626         (accumulate_line_info): Save src->maximum_count.
6627         (output_line_beginning): Make color line number if
6628         flag_use_hotness_colors is set.
6629         (output_line_details): Pass default argument value.
6630         (output_lines): Pass src->maximum_count.
6632 2018-09-17  Martin Liska  <mliska@suse.cz>
6634         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6635         Use processor_names table.
6636         * config/i386/i386.c (ix86_default_align): Use
6637         processor_cost_table for alignment values.
6638         (ix86_option_override_internal): Use processor_names.
6639         (ix86_function_specific_print): Likewise.
6640         * config/i386/i386.h (struct processor_costs):
6641         Add alignment values.
6642         (struct ptt): Remove and replace with const char *.
6643         * config/i386/x86-tune-costs.h (struct processor_costs):
6644         Declare default alignments for all costs.
6646 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6648         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6649         symbolics or VR_VARYING ranges for ABS_EXPR.
6650         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6651         when range will wrap.
6653 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6655         PR middle-end/86864
6656         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6657         before and after a JUMP_TABLE_DATA.
6659 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6661         PR middle-end/87188
6662         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6663         when one operand is a function pointer.  Use POINTER_TYPE_P and
6664         FUNC_OR_METHOD_TYPE_P.
6665         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6666         * fold-const.c (build_range_check): Likewise.
6667         * match.pd (simple_comparison): Likewise.
6669 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6671         PR c/82967
6672         * spellcheck.c (get_edit_distance_cutoff): New function.
6673         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6674         (selftest::test_get_edit_distance_one_way): ...this.
6675         (selftest::test_get_edit_distance_unit): Rename to...
6676         (selftest::test_get_edit_distance_both_ways): ...this.
6677         (selftest::test_edit_distances): Move tests to this new function,
6678         and test some more pairs of strings.  Update for above renaming.
6679         (selftest::get_old_cutoff): New function.
6680         (selftest::test_get_edit_distance_cutoff): New function.
6681         (selftest::assert_suggested_for): New function.
6682         (ASSERT_SUGGESTED_FOR): New macro.
6683         (selftest::assert_not_suggested_for): New function.
6684         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6685         (selftest::test_suggestions): New function.
6686         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6687         tests to selftest::test_edit_distances and call it.  Add calls to
6688         selftest::test_get_edit_distance_cutoff and
6689         selftest::test_suggestions.
6690         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6691         (best_match::consider): Replace hard-coded cutoff calculation with
6692         a call to...
6693         (best_match::get_cutoff): New declaration.
6694         (best_match::get_best_meaningful_candidate): Likewise.
6696 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6698         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6700 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6702         revert:
6703         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6705         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6706         terminated string literal.
6708 2018-09-14  Martin Sebor  <msebor@redhat.com>
6710         * builtins.c (unterminated_array): Handle ARRAY_REF.
6711         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6712         * builtins.h (unterminated_array): Declare extern.
6713         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6714         arrays.
6715         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6716         calls.
6718 2018-09-14  Martin Sebor  <msebor@redhat.com>
6719             Jeff Law  <law@redhat.com>
6721         * builtins.c (unterminated_array): New.
6722         (expand_builtin_strcpy): Adjust.
6723         (expand_builtin_strcpy_args): Detect unterminated arrays.
6724         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
6725         unterminated arrays.
6726         * gimple-fold.h (get_maxval_strlen): Add argument.
6727         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
6729         * gimple-fold.c (get_range_strlen): Add argument.
6730         (get_maxval_strlen): Adjust.
6731         * gimple-fold.h (get_range_strlen): Add argument.
6733 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
6735         * config/i386/movdirintrin.h: Fix copyright year.
6737 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
6739         * reg-stack.c: Include regs.h.
6740         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
6741         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6742         FIRST_STACK_REG, not DFmode.
6743         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
6744         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
6745         (change stack): Default register mode to the reg_raw_mode of
6746         FIRST_STACK_REG, not DFmode.
6747         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
6748         (*swapxf): Rename from swapxf.
6750 2018-09-14  Carl Love  <cel@us.ibm.com>
6752         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
6753         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
6755 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
6757         PR target/87224
6758         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
6759         alternatives.
6761 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
6763         PR target/85628
6764         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
6766 2018-09-14  Jason Merrill  <jason@redhat.com>
6768         Fix --enable-gather-detailed-mem-stats.
6769         * hash-table.c (hash_table_usage): Change from variable to function.
6770         * hash-table.h: Adjust.
6771         * Makefile.in: Add missing dependencies on hash-table.h.
6773 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6775         PR tree-optimization/87259
6776         PR lto/87283
6777         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
6778         execute_cse_reciprocals_1 has tried transforming.
6780 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
6782         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
6783         VR_VARYING for PLUS/MINUS_EXPR.
6785 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6787         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
6788         formatting.
6790 2018-09-14  Richard Biener  <rguenther@suse.de>
6792         PR middle-end/63155
6793         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
6794         bits for the merged partition.
6796 2018-09-13  Martin Sebor  <msebor@redhat.com>
6797             Bernd Edlinger  <bernd.edlinger@hotmail.de>
6799         * builtins.h (c_srlen): Add argument.
6800         * builtins.c (warn_string_no_nul): New function.
6801         (c_strlen): Add argument and use it.  Update recursive calls.
6802         Pass DECL argument to string_constant to get info on non
6803         terminated strings.  Update *NONSTR as needed.
6804         (fold_builtin_strlen): Add argument to calls to c_strlen.
6805         Warn for unterminated arrays.
6806         (warn_string_no_null): Add prototype.
6807         * expr.c (string_constant): Update arguments.  Update recursive
6808         calls appropriately.  Detect missing NUL terminator and outermost
6809         declaration its missing in.
6810         Improve checks for arrays with nonzero lower bound or elements
6811         that are not a single byte.  Simplify offset computation.
6812         Simplify checks for non-NUL terminated strings.
6813         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
6814         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
6816 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6818         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
6819         correctly.
6820         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
6821         argument.  Fix range checks.
6822         * fold-const.h (c_getstr): Adjust protoype.
6823         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
6824         string is constant but contains no NUL byte.
6826         * expr.c (string_constant): Adjust function comment.
6827         Remove bogus check for zero termination.
6829         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
6831         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
6832         (get_constant_size): Don't make STRING_CSTs larger than they are.
6833         (check_string_literal): New check function for STRING_CSTs.
6834         (output_constant): Use it.
6836 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
6838         PR target/86812
6839         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
6841 2018-09-13  Richard Biener  <rguenther@suse.de>
6843         PR tree-optimization/87263
6844         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
6845         (struct unwind_state): Add max_rpo field.
6846         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
6847         Compute max_rpo, the max RPO number a block can be backwards reached
6848         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
6849         separating it from the iterating mode.
6851 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6853         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
6854         (rfs_decision): New scheduling decision.
6856 2018-09-13  Richard Biener  <rguenther@suse.de>
6858         PR bootstrap/87134
6859         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
6860         (vn_nary_op_insert_pieces_predicated): Do not write useless
6861         valid_dominated_by_p entry outside of the allocated storage.
6863 2018-09-13  Omar Sandoval  <osandov@osandov.com>
6864             Tom de Vries  <tdevries@suse.de>
6866         PR debug/86985
6867         * dwarf2out.c (is_c): New function.
6868         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
6870 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
6872         PR target/85628
6873         * config/aarch64/aarch64.md (*aarch64_bfxil):
6874         Define.
6875         * config/aarch64/constraints.md (Ulc): Define.
6876         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
6877         Define.
6878         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
6879         New function.
6881 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6883         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
6884         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
6885         aarch64_layout_frame call.
6886         (aarch64_expand_epilogue): Likewise.
6887         (aarch64_initial_elimination_offset): Likewise.
6888         (aarch64_get_separate_components): Likewise.
6889         (aarch64_use_return_insn_p): Likewise.
6890         (aarch64_layout_frame): Remove unneeded check.
6892 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6894         * configure.ac: Only append
6895         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
6896         gcc_config_arguments if it was never reconfigured or last reconfigure
6897         was with different arguments.
6898         * configure: Regenerated.
6900 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6901             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6903         PR middle-end/87290
6904         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
6905         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
6907 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6909         PR tree-optimization/87287
6910         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
6911         X % C == 0 to X % (unsigned) C == 0 optimization to ...
6912         * match.pd (X % C == 0): ... here.  New optimization.
6914 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6916         PR middle-end/82853
6917         * expr.h (maybe_optimize_mod_cmp): Declare.
6918         * expr.c (mod_inv): New function.
6919         (maybe_optimize_mod_cmp): New function.
6920         (do_store_flag): Use it.
6921         * cfgexpand.c (expand_gimple_cond): Likewise.
6923 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
6924             Julian Brown  <julian@codesourcery.com>
6926         PR middle-end/86336
6927         * gimplify.c (gimplify_scan_omp_clauses): Set
6928         target_firstprivatize_array_bases in OpenACC parallel and kernels
6929         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
6930         OpenACC data regions.
6932 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
6934         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
6935         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
6937 2018-09-12  Richard Biener  <rguenther@suse.de>
6939         PR tree-optimization/87280
6940         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
6941         edge but unreachable target.
6942         (do_rpo_vn): For conservatively handling a PHI only mark
6943         the backedge executable but not the block reachable.
6945 2018-09-12  Richard Biener  <rguenther@suse.de>
6947         PR tree-optimization/87266
6948         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
6949         visited blocks.
6951 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
6953         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
6954         constants.
6955         ("trunc<BFP:mode><DFP_ALL:mode>2")
6956         ("trunc<DFP_ALL:mode><BFP:mode>2")
6957         ("extend<BFP:mode><DFP_ALL:mode>2")
6958         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
6959         according to the target operand type.
6961 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6962             Andreas Krebbel  <krebbel@linux.ibm.com>
6964         PR tree-optimization/86844
6965         * gimple-ssa-store-merging.c
6966         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
6967         there are any overlapping stores in between them, make sure they are
6968         also coalesced or we give up completely.
6970 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6972         PR middle-end/87248
6973         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
6974         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
6976 2018-09-12  Tom de Vries  <tdevries@suse.de>
6978         * common.opt (gdescribe-dies): Add option.
6979         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
6980         attribute for artifical and nameless decls.
6981         (dwarf2out_register_external_die): Add description attribute to
6982         external reference die.
6983         (add_desc_attribute): New functions.
6984         (gen_subprogram_die): Add description attribute to
6985         DW_TAG_call_site_parameter.
6986         * tree-pretty-print.c (print_generic_expr_to_str): New function.
6987         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
6988         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
6989         -gno-describe-dies.
6990         (@item -gdescribe-dies): Add.
6992 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6994         * tree-vrp.c (vrp_shift_undefined_p): Remove.
6995         (extract_range_from_binary_expr_1: Call
6996         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
6997         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
6998         depend on sign.
7000 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7002         * gimple-ssa-warn-alloca.c
7003         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7004         field for ALLOCA_BOUND_*_LARGE.
7006 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7008         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7010 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7012         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7013         for clobbers.  Remove obsolete comment.
7015 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7017         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7018         mpxchk, mpxld and mpxst types.
7019         (define_attr length_immediate): Remove all processing of mpx types.
7020         (define_attr prefix_0f): Ditto.
7021         (define_attr memory): Ditto.
7023 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7025         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7026         (log<mode>2): Change operand 1 predicate to general_operand.
7027         Extend operand 1 to XFmode and generate logxf3 insn.
7028         (log10<mode>2): Change operand 1 predicate to general_operand.
7029         Extend operand 1 to XFmode and generate log10xf3 insn.
7030         (log2<mode>2): Change operand 1 predicate to general_operand.
7031         Extend operand 1 to XFmode and generate log2xf3 insn.
7032         (fyl2xp1_extend<mode>xf3_i387): Remove.
7033         (log1p<mode>2): Change operand 1 predicate to general_operand.
7034         Extend operand 1 to XFmode and generate log1pxf3 insn.
7035         (fxtract_extend<mode>xf3_i387): Remove.
7036         (logb<mode>2): Change operand 1 predicate to general_operand.
7037         Extend operand 1 to XFmode and generate logbxf3 insn.
7038         (ilogb<mode>2): Change operand 1 predicate to general_operand.
7039         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7040         (significand<mode>2): Change operand 1 predicate to general_operand.
7041         Extend operand 1 to XFmode and generate significandxf3 insn.
7043 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7045         * gcc.c (perror_with_name, pfatal_with_name): Delete.
7046         (load_specs): Use fatal_error.
7047         (DELETE_IF_ORDINARY, process_command): Use error.
7048         (execute, run_attempt): Use fatal_error.
7050 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
7052         * diagnostic-core.h (sorry_at): New prototype.
7053         * diagnostic.c (sorry_at): New function.
7055 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
7057         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7058         by zero as VR_UNDEFINED.
7060 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7062         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7063         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7064         (<sincos>mode2): New expander.
7065         (sincos_extend<mode>xf3_i387): Remove insn pattern.
7066         (sincos -> sin, cos splitters): Remove splitter patterns.
7067         (sincos<mode>3): Change operand 2 predicate to general_operand.
7068         Extend operand 2 to XFmode and generate sincosxf3 insn.
7069         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
7070         Change operand 3 predicate to const1_operand.
7071         (fptan_extend<mode>xf4_i387): Remove insn pattern.
7072         (tanxf2): Update operands in the call to fptanxf4_i387.
7073         (tan<mode>2): Change operand 1 predicate to general_operand.
7074         Extend operand 1 to XFmode and generate tanxf3 insn.
7075         (atan2xf3): Rename from *fpatanxf3_i387.
7076         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
7077         (atan2xf3): Remove expander.
7078         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
7079         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
7080         (atan<mode>2): Change operand 1 predicate to general_operand.
7081         Extend operand 1 to XFmode and generate atanxf3 insn.
7083 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7085         * config/i386/i386.md (x87/SSE constant load splitter): Use
7086         memory_operand instead of nonimmediate_operand for input operand
7087         predicate.
7089 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
7091         * config/i386/i386.md (float partial SSE register stall splitter): Move
7092         splitter near its instruction pattern.
7093         (float_extend partial SSE register stall splitter): Ditto.
7094         (float_truncate partial SSE register stall splitter): Ditto.
7096 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
7098         PR target/86794
7099         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7100         to speculation_safe_value_not_needed.
7102         PR target/85666
7103         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
7104         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
7105         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
7106         leaf_function_p, instead use has_hard_reg_initial_val.
7108 2018-09-09  Nathan Sidwell  <nathan@acm.org>
7110         * gcc.h (pfatal_with_name): Don't declare here.
7111         * gcc.c (pfatal_with_name): Make static.
7113 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
7115         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
7116         earlyclobber.
7118 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
7120         PR rtl-optimization/85458
7121         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
7122         priority hook to reduce the priority of EXPR.
7124 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
7126         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
7127         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
7128         with X87_ENABLE_FLOAT.  Remove preparation code.
7129         (*float<SWI48:mode><MODEF:mode>2): Rename from
7130         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
7131         math using "enabled" attribute.
7132         (*floatdi<MODEF:mode>2_i387): Rename from
7133         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
7134         enable for 32bit targets only.
7135         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7136         splitter.
7137         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7138         as operand 1 predicate.  Rewrite as define_insn_and_split.
7139         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7141 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7143         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7144         to fallthru to FLOAT case.
7146 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7148         PR target/86731
7149         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7150         around folding of vec_sl to handle out of range shift values.
7152 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7154         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7155         Update callers to gen_fix_trunc<mode>_i387_fisttp
7156         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7157         nonimmediate_operand.
7158         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7159         and corresponding splitters.
7160         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7161         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7162         (fix_truncdi_i387_with_temp): Remove insn pattern
7163         and corresponding splitters.
7164         (fix_trunc<mode>_i387): Change operand 0 predicate to
7165         nonimmediate_operand.
7166         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7167         and corresponding splitters.
7168         (*fistdi2_1): Remove.
7169         (fistdi2): Ditto.
7170         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7171         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
7172         (*fist<mode>2_1): Remove.
7173         (fist<mode>2): Ditto.
7174         (fist<mode>2_with_temp): Remove insn pattern and corresponding
7175         splitters.
7176         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
7177         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7178         (fistdi2_<rounding>): Change operand 0 predicate to
7179         nonimmediate_operand.
7180         (fistdi2_<rounding>_with_temp): Remove insn pattern
7181         and corresponding splitters.
7182         (fist<mode>2_<rounding>): Change operand 0 predicate to
7183         nonimmediate_operand.
7184         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7185         and corresponding splitters.
7187         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7189 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7191         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7192         the init value.
7194 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7196         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7197         early gimple folding of vec_splat().
7198         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7199         * gimple-fold.h: Add an extern define for tree_vec_extract().
7201 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7203         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7204         wrappers around TREE_TYPE comparisons.
7206 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7208         PR target/80080
7209         * config/s390/predicates.md: Add nonsym_memory_operand.
7210         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7211         contains a SYMBOL_REF, load it into an intermediate pseudo.
7212         (s390_emit_compare_and_swap): Legitimize operand.
7213         * config/s390/s390.md: Use the new nonsym_memory_operand
7214         with UNSPECV_CAS patterns.
7216 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7218         PR target/80080
7219         * config/s390/s390-passes.def: New file.
7220         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7221         declaration.
7222         (make_pass_s390_early_mach): Add declaration.
7223         * config/s390/s390.c (make_pass_s390_early_mach):
7224         (s390_option_override): Remove dynamic registration.
7225         * config/s390/t-s390: Add s390-passes.def.
7227 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7229         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7230         Remove UNSPEC_LTREL_BASE check.
7231         (annotate_constant_pool_refs): Likewise.
7232         (find_constant_pool_ref): Likewise.
7233         (find_ltrel_base): Removed.
7234         (replace_ltrel_base): Removed.
7235         (s390_mainpool_finish): Remove replace_ltrel_base call.
7236         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7237         (s390_chunkify_finish): Remove replace_ltrel_base call.
7238         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7240 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7242         PR target/86779
7243         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7244         to speculation_safe_value_not_needed.
7246 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7247             Bernd Schmidt  <bernds_cb1@t-online.de>
7249         * config/nvptx/nvptx-opts.h: New file.
7250         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7251         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7252         (ASM_SPEC): Define.
7253         (TARGET_SM35): New macro.
7254         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7255         correct predicate.
7256         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7257         values.
7258         (misa=): New option.
7259         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7261 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7263         * config/i386/i386.md (truncdfsf2): Remove expander.
7264         (truncdfsf2_with_temp): Ditto.
7265         (truncxf<mode>2): Ditto.
7266         (*truncdfsf_fast_mixed): Remove insn pattern.
7267         (*truncdfsf_fast_i387): Ditto.
7268         (*truncdfsf_mixed): Ditto.
7269         (*truncdfsf_i387): Ditto.
7270         (*truncdfsf2_i387_1): Ditto.
7271         (*truncxfsf2_mixed): Ditto.
7272         (*truncxfdf2_mixed): Ditto.
7273         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7274         to call gen_truncxf<mode>2 instead.
7275         (*truncxf<mode>2_i387): Remove.
7276         (reg->reg splitters): Remove splitter pattern.
7277         (reg->mem splitters): Ditto.
7279         (truncdfsf2): New insn pattern.
7280         (truncxf<mode>2): Ditto.
7282 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7284         * tree-ssa-math-opts.c (is_mult_by): New function.
7285         (is_square_of): Use the above.
7286         (optimize_recip_sqrt): New function.
7287         (pass_cse_reciprocals::execute): Use the above.
7289 2018-09-05  Richard Biener  <rguenther@suse.de>
7291         PR bootstrap/87134
7292         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7293         to zero-init the emplaced vec.
7295 2018-09-05  Martin Liska  <mliska@suse.cz>
7297         PR tree-optimization/87205
7298         * tree-switch-conversion.c (pass_lower_switch::execute):
7299         Group cases for switch statements.
7301 2018-09-05  Richard Biener  <rguenther@suse.de>
7303         PR tree-optimization/87217
7304         * tree-ssa-sccvn.c (vuse_valueize): New.
7305         (vn_reference_lookup_pieces): Use it.
7306         (vn_reference_lookup): Likewise.
7308 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7310         PR c++/87137
7311         * stor-layout.c (place_field): Scan forwards to check last
7312         bitfield when ms_bitfield_placement is in effect.
7314 2018-09-05  Richard Biener  <rguenther@suse.de>
7316         PR bootstrap/87225
7317         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7318         return.
7320 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7321             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7323         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7324         * config.gcc (extra_objs): Build it.
7325         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7326         Likewise.
7327         * config/aarch64/aarch64-passes.def
7328         (pass_tag_collision_avoidance): New pass.
7329         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7330         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7331         (aarch64_classify_address): Remove static qualifier.
7332         (aarch64_address_info, aarch64_address_type): Move to...
7333         * config/aarch64/aarch64-protos.h: ... here.
7334         (make_pass_tag_collision_avoidance): New function.
7335         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7336         New tuning flag.
7338 2018-09-05  Martin Liska  <mliska@suse.cz>
7340         * doc/gcov.texi: Update documentation of humar
7341         readable mode.
7342         * gcov.c (format_count): Print one decimal place, it provides
7343         more fine number of situations like '1G' vs. '1.4G'.
7345 2018-09-05  Martin Liska  <mliska@suse.cz>
7347         PR target/87164
7348         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7349         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7350         to generate a MASK value.
7352 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7354         PR debug/86593
7355         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7356         if frame pointer isn't used.
7357         (compute_frame_pointer_to_fb_displacement): Likewise.
7359 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7361         PR target/87198
7362         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7363         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7364         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7365         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7366         and OPTION_MASK_ISA_XSAVEC_UNSET.
7368 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7370         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7371         XOR operations in NAND case.
7373 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7375         * wide-int-range.cc (wide_int_range_convert): New.
7376         * wide-int-range.h (wide_int_range_convert): New.
7377         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7378         code into wide_int_range_convert.
7379         (extract_range_into_wide_ints): Do not munge anti range constants
7380         into the entire domain.  Just return the range back.
7382 2018-09-04  Martin Liska  <mliska@suse.cz>
7384         * genmatch.c (output_line_directive): Add new argument
7385         fnargs.
7386         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7388 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7390         * doc/invoke.texi (Option Summary): Add whitespace.
7392         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7394 2018-09-04  Richard Biener  <rguenther@suse.de>
7396         PR tree-optimization/87211
7397         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7398         backedge value we're supposed to treat as VARYING also number
7399         the PHI to VARYING in case it got a different value-number already.
7401 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7403         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7404         (extract_range_from_binary_expr_1): Do not call
7405         vrp_can_optimize_bit_op.
7406         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7407         static.
7408         (wide_int_range_get_mask_and_bounds): New.
7409         (wide_int_range_optimize_bit_op): New.
7410         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7411         (wide_int_range_bit_and): Same.
7412         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7413         (wide_int_range_optimize_bit_op): New.
7414         (wide_int_range_get_mask_and_bounds): New.
7416 2018-09-04  Richard Biener  <rguenther@suse.de>
7418         PR tree-optimization/87176
7419         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7420         variable.  When value-numbering a virtual PHI node make sure
7421         to not value-number to the backedge value.
7423 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7425         * doc/extend.texi (Long Long, Hex Floats): Document support for
7426         long long and hex floats in more recent versions of ISO C++.
7428 2018-09-03  Richard Biener  <rguenther@suse.de>
7430         PR tree-optimization/87177
7431         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7432         cleanup.
7434 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7436         * bb-reorder.c (edge_order): Convert to C-qsort-style
7437         tri-state comparator.
7438         (reorder_basic_blocks_simple): Change std::stable_sort to
7439         gcc_stablesort.
7441 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7443         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7444         tri-state comparator.
7445         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7447 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7449         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7450         (mergesort): ... here as maximum count for using netsort.
7451         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7452         (gcc_stablesort): New.
7453         * system.h (gcc_stablesort): Declare.
7455 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7457         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7458         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7459         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7461 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7463         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7464         lxsdx and stxsdx alternatives.
7465         (*mov<mode>_hardfloat64): Ditto.
7466         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7468 2018-09-03  Richard Biener  <rguenther@suse.de>
7470         PR tree-optimization/87200
7471         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7472         simplify result.
7474 2018-09-03  Martin Liska  <mliska@suse.cz>
7476         PR tree-optimization/87201
7477         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7478         Fix parenthesis in an expression.
7480 2018-09-03  Richard Biener  <rguenther@suse.de>
7482         PR tree-optimization/87197
7483         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7484         visited.  CSE the VN_INFO hashtable lookup.
7486         PR tree-optimization/87169
7487         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7488         iterating make sure there's no extra backedges from irreducible
7489         regions feeding the header.  Mark the destination block
7490         executable.
7492 2018-09-03  Martin Liska  <mliska@suse.cz>
7494         PR driver/83193
7495         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7496         * common/common-targhooks.c (default_get_valid_option_values):
7497         New function.
7498         * common/common-targhooks.h (default_get_valid_option_values):
7499         Likewise.
7500         * common/config/i386/i386-common.c: Move processor_target_table
7501         from i386.c.
7502         (ix86_get_valid_option_values): New function.
7503         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7504         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7505         (PTA_*): Move all defined masks into i386-common.c.
7506         (ix86_function_specific_restore): Use new processor_cost_table.
7507         * config/i386/i386.h (struct ptt): Moved from i386.c.
7508         (struct pta): Likewise.
7509         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7510         * doc/tm.texi.in: Likewise.
7511         * opt-suggestions.c (option_proposer::suggest_option):
7512         Pass prefix to build_option_suggestions.
7513         (option_proposer::get_completions): Likewise.
7514         (option_proposer::build_option_suggestions): Use the new target
7515         hook.
7516         * opts.c (struct option_help_tuple): New struct.
7517         (print_filtered_help): Use the new target hook.
7519 2018-09-03  Martin Liska  <mliska@suse.cz>
7521         PR middle-end/59521
7522         * predict.c (set_even_probabilities): Add likely_edges
7523         argument and handle cases where we have precisely one
7524         likely edge.
7525         (combine_predictions_for_bb): Catch also likely_edges.
7526         (tree_predict_by_opcode): Handle gswitch statements.
7527         * tree-cfg.h (find_case_label_for_value): New declaration.
7528         (find_taken_edge_switch_expr): Likewise.
7529         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7530         Find pivot in decision tree based on probabily, not by number of
7531         nodes.
7533 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7535         * doc/standards.texi (Standards): Update Objective-C reference.
7537 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7539         * doc/install.texi (Prerequisites): Update link for MPC.
7541 2018-09-01  Michael Matz  <matz@suse.de>
7543         PR tree-optimization/87074
7544         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7545         PHIs for outer-loop uses.
7547 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7549         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7550         * doc/invoke.texi (C Dialect Options): Ditto.
7552 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7554         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7556 2018-08-31  Richard Biener  <rguenther@suse.de>
7558         PR tree-optimization/87168
7559         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7560         (rpo_elim::eliminate_avail): When OP was not visited it must
7561         be available.
7563 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7565         * tree-vrp.c (copy_value_range): Convert param "from" from
7566         "value_range *" to "const value_range *".
7567         (range_is_null): Likewise for param "vr".
7568         (range_int_cst_p): Likewise.
7569         (range_int_cst_singleton_p): Likewise.
7570         (symbolic_range_p): Likewise.
7571         (value_ranges_intersect_p): Likewise for both params.
7572         (value_range_nonnegative_p): Likewise for param "vr".
7573         (value_range_constant_singleton): Likewise.
7574         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7575         (extract_range_into_wide_ints): Likewise for param "vr".
7576         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7577         and "vr1".
7578         (vrp_can_optimize_bit_op): Likewise.
7579         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7580         "vr1_".
7581         (extract_range_from_unary_expr): Likewise.
7582         (debug_value_range): Likewise for param "vr".
7583         (value_range::dump): Add "const" qualifier.
7584         (vrp_prop::check_array_ref): Convert local "vr" from
7585         "value_range *" to "const value_range *".
7586         (vrp_prop::check_mem_ref): Likewise.
7587         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7588         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7589         (vrp_intersect_ranges): Likewise.
7590         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7591         (vrp_prop::vrp_finalize): Likewise.
7592         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7593         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7594         (extract_range_from_unary_expr): Likewise.
7595         (value_range_constant_singleton): Likewise.
7596         (symbolic_range_p): Likewise.
7597         (copy_value_range): Likewise.
7598         (extract_range_from_binary_expr_1): Likewise.
7599         (range_int_cst_p): Likewise.
7600         (vrp_set_zero_nonzero_bits): Likewise.
7601         (range_int_cst_singleton_p): Likewise.
7603 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7605         * config/aarch64/arm_neon.h (vabsd_s64): New.
7606         (vnegd_s64): Likewise.
7608 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7610         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7612 2018-08-31  Martin Liska  <mliska@suse.cz>
7614         * ipa-icf.c (sem_item::add_type): Use
7615         sem_item::m_type_hash_cache.
7616         * ipa-icf.h: Move the cache from sem_item_optimizer
7617         to sem_item.
7619 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7621         * doc/extend.texi (Backwards Compatibility): Remove implicit
7622         extern C leeway of () being (...).
7624 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7626         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7628 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7630         PR target/86684
7631         PR target/87149
7632         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7634 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7636         PR middle-end/87138
7637         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7638         gen_int_mode.  Formatting fixes.
7640 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7642         * target.def (custom_function_descriptors): Improve documentation.
7643         * doc/tm.texi.in (Trampolines): Expand discussion of function
7644         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7645         beginning of the section.
7646         * doc/tm.texi: Regenerated.
7648 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7650         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7651         base class in the initializer list.  This is a workaround for
7652         building with older GCC.
7653         (class auto_bb_flag): Likewise.
7655 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7657         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7658         (altivec_vcmpequ<VI_char>_p): Remove star.
7659         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7660         vector load modes.
7661         (expand_strncmp_vec_sequence): New function.
7662         (emit_final_str_compare_vec): New function.
7663         (expand_strn_compare): Add support for vector strncmp.
7664         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7665         length specification to bytes.
7666         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7667         (vcmpnezb_p): New pattern.
7668         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7669         for option -mstring-compare-inline-limit.
7671 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7673         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7674         (PTA_SKYLAKE): Add PTA_AES.
7675         (PTA_GOLDMONT): Likewise.
7677 2018-08-29  Jan Hubicka  <jh@suse.cz>
7679         PR lto/86517
7680         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7681         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7683 2018-08-29  Jan Hubicka  <jh@suse.cz>
7685         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7686         TYPE_STUB_DECL.
7687         (hash_tree): Do not visit TYPE_STUB_DECL.
7688         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7689         stream TYPE_STUB_DECL.
7690         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7691         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7692         after free_lang_data.
7693         (type_in_anonymous_namespace_p): Likewise.
7695 2018-08-29  Jan Hubicka  <jh@suse.cz>
7697         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7698         comment that it has to be even number.
7699         (class sreal): Change m_sig type to int32_t.
7700         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7701         int64_t for temporary calculations.
7702         (sreal_verify_basics): Drop one bit from minimum and maximum.
7704 2018-08-30  Richard Biener  <rguenther@suse.de>
7706         PR tree-optimization/87147
7707         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7708         (visit_phi): When the degenerate result is from the backedge and
7709         we didn't visit its definition yet drop to VARYING.
7710         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7712 2018-08-29  Jan Hubicka  <jh@suse.cz>
7714         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7715         DECL_VINDEX.
7716         (hash_tree): Likewise.
7718 2018-08-29  Jan Hubicka  <jh@suse.cz>
7720         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
7721         and TYPE_NEXT_REF_TO.
7723 2018-08-29  Jan Hubicka  <jh@suse.cz>
7725         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7726         comment that it has to be even number.
7727         (class sreal): Change m_sig type to int32_t.
7728         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7729         int64_t for temporary calculations.
7730         (sreal_verify_basics): Drop one bit from minimum and maximum.
7732 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
7734         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
7736 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
7738         PR middle-end/86995
7739         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
7740         if to_add is negative.
7742 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7744         PR middle-end/87053
7745         * builtins.c (c_strlen): Improve range checks.
7747 2018-08-29  Martin Sebor  <msebor@redhat.com>
7748             Jeff Law  <law@redhat.com>
7750         PR tree-optimization/86714
7751         PR tree-optimization/86711
7752         * builtins.c (c_strlen): Add arguments to call to string_constant.
7753         * expr.c (string_constant): Add argument.  Detect missing nul
7754         terminator and outermost declaration it's missing in.
7755         * expr.h (string_constant): Add argument.
7756         * fold-const.c (read_from_constant_string): Add arguments to call to
7757         string_constant.
7758         (c_getstr): Likewise.
7759         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7760         to string_constant.
7761         * tree-ssa-strlen.c (get_stridx): Likewise.
7763 2018-08-29  Jan Hubicka  <jh@suse.cz>
7765         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
7766         Do not stream DECL_VINDEX.
7767         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
7768         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
7769         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
7771 2018-08-29  Richard Biener  <rguenther@suse.de>
7773         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
7774         virtual operands that are not default defs to honor region
7775         boundaries.
7776         (rpo_vn_valueize): Remove ineffective code here.
7778 2018-08-29  Richard Biener  <rguenther@suse.de>
7780         PR tree-optimization/87132
7781         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
7782         when skipping defs reachable over backedges.
7784 2018-08-29  Richard Biener  <rguenther@suse.de>
7786         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
7787         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
7788         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
7789         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
7790         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
7791         (vn_reference_lookup_3): Remove use of const_parms.
7792         (free_rpo_vn): Do not free const_parms.
7793         (do_rpo_vn): Do not call init_const_parms.
7794         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
7795         SSA_NAME_POINTS_TO_READONLY_MEMORY.
7796         (call_may_clobber_ref_p_1): Likewise.
7798 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
7800         PR other/86726
7801         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
7802         (-O): Ditto.
7803         (-ftree-scev-cprop): Document.
7805 2018-08-29  Jan Hubicka  <jh@suse.cz>
7807         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
7808         parameters.
7809         (sreal constructor): Update.
7810         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
7811         sreal:operator/): Update.
7813 2018-08-29  Martin Liska  <mliska@suse.cz>
7815         * tree-switch-conversion.c (switch_conversion::expand):
7816         Strenghten assumption about gswitch statements.
7818 2018-08-29  Richard Biener  <rguenther@suse.de>
7820         PR tree-optimization/87117
7821         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
7822         re-value-number released SSA VDEFs.
7824 2018-08-29  Richard Biener  <rguenther@suse.de>
7826         PR tree-optimization/87126
7827         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
7829 2018-08-28  Jim Wilson  <jimw@sifive.com>
7831         * config/riscv/pic.md: Rewrite.
7832         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
7833         invalid address.
7834         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
7835         (SOFTF, default_load, softload, softstore): New.
7837 2018-08-28  Jeff Law  <law@redhat.com>
7839         * fold-const.c (fold_binary_loc): Remove recently added assert.
7841 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
7843         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
7844         to OP parmeter of generated function.
7846 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
7848         PR tree-optimization/87009
7849         * match.pd: Add boolean optimizations.
7851 2018-08-28  Martin Sebor  <msebor@redhat.com>
7853         PR middle-end/86631
7854         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
7855         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
7856         (pass_walloca::gate): Use it.
7857         (alloca_call_type): Same.
7858         (pass_walloca::execute): Same.
7859         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
7861 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
7863         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
7864         GCC_VERSION for usage of "__gcc_dump_printf__" format from
7865         >= 3005 to >= 9000.
7867 2018-08-28  Richard Biener  <rguenther@suse.de>
7869         PR tree-optimization/87124
7870         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
7871         constants before looking up avail.
7873 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
7875         PR middle-end/87099
7876         * calls.c (maybe_warn_nonstring_arg): Punt early if
7877         warn_stringop_overflow is zero.  Don't call get_range_strlen
7878         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
7879         Swap comparison operands to have constants on rhs.  Only use
7880         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
7881         increment lenrng[0].
7883 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
7885         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
7886         use of tree_to_shwi.  Remove duplicated test for the size being
7887         a whole number of bytes.
7889 2018-08-28  Richard Biener  <rguenther@suse.de>
7891         PR tree-optimization/87117
7892         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
7893         Handle removed stmt without LHS (GIMPLE_NOP).
7895 2018-08-28  Richard Biener  <rguenther@suse.de>
7897         PR tree-optimization/87117
7898         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
7899         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
7901 2018-08-28  Richard Biener  <rguenther@suse.de>
7903         PR tree-optimization/87117
7904         * tree-ssa-pre.c (compute_avail): Do not make expressions
7905         with predicated values available.
7906         (get_expr_value_id): Assert we do not run into predicated value
7907         expressions.
7909 2018-08-28  Richard Biener  <rguenther@suse.de>
7911         PR tree-optimization/87117
7912         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
7913         get virtual operands.
7914         (get_expr_operands): Handle STRING_CST like other decls.
7916 2018-08-28  Martin Liska  <mliska@suse.cz>
7918         * tree.h: Update documentation of fndecl_built_in_p
7919         functions.
7922 2018-08-27  Jeff Law  <law@redhat.com>
7923         PR tree-optimization/87110
7924         * tree-ssa-dse.c (compute_trims): Handle non-constant
7925         TYPE_SIZE_UNIT.
7927 2018-08-27  Martin Sebor  <msebor@redhat.com>
7929         PR tree-optimization/86914
7930         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
7932 2018-08-27  Martin Sebor  <msebor@redhat.com>
7934         PR tree-optimization/87112
7935         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
7936         the type of the bound argument.
7938 2018-08-27  Jeff Law  <law@redhat.com>
7940         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
7941         type does not have a TYPE_SIZE_UNIT.
7943 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
7945         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
7946         with include of backend.h.
7948 2018-08-27  Richard Biener  <rguenther@suse.de>
7950         PR tree-optimization/86927
7951         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
7952         use const cond reduction code.
7954 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
7956         PR tree-optimization/85758
7957         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
7959 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7961         PR c++/87091
7962         * diagnostic-show-locus.c (class layout_range): Update for
7963         conversion of show_caret_p to a tri-state.
7964         (layout_range::layout_range): Likewise.
7965         (make_range): Likewise.
7966         (layout::maybe_add_location_range): Likewise.
7967         (layout::should_print_annotation_line_p): Don't show annotation
7968         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
7969         (layout::get_state_at_point): Update for conversion of
7970         show_caret_p to a tri-state.  Bail out early for
7971         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
7972         underlining or source colorization.
7973         (gcc_rich_location::add_location_if_nearby): Update for conversion
7974         of show_caret_p to a tri-state.
7975         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
7976         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
7977         Likewise.
7978         (selftest::test_one_liner_labels): Likewise.
7979         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
7980         conversion of show_caret_p to a tri-state.
7981         * pretty-print.c (text_info::set_location): Likewise.
7982         * pretty-print.h (text_info::set_location): Likewise.
7983         * substring-locations.c (format_warning_n_va): Likewise.
7984         * tree-diagnostic.c (default_tree_printer): Likewise.
7985         * tree-pretty-print.c (newline_and_indent): Likewise.
7987 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7989         PR c++/87091
7990         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
7991         line above for line-insertion fix-it hints.
7992         (selftest::test_fixit_insert_containing_newline): Update the
7993         expected results, and add a test with line-numbering enabled.
7995 2018-08-27  Martin Liska  <mliska@suse.cz>
7997         PR sanitizer/86962
7998         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
7999         params with DECL_HAS_VALUE_EXPR_P.
8001 2018-08-27  Martin Liska  <mliska@suse.cz>
8003         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8004         selected expansion strategy.
8006 2018-08-27  Martin Liska  <mliska@suse.cz>
8008         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8009         * builtins.c (is_builtin_fn): Likewise.
8010         * attribs.c (diag_attr_exclusions): Use new function
8011         fndecl_built_in_p and remove check for FUNCTION_DECL if
8012         possible.
8013         (builtin_mathfn_code): Likewise.
8014         (fold_builtin_expect): Likewise.
8015         (fold_call_expr): Likewise.
8016         (fold_builtin_call_array): Likewise.
8017         (fold_call_stmt): Likewise.
8018         (set_builtin_user_assembler_name): Likewise.
8019         (is_simple_builtin): Likewise.
8020         * calls.c (gimple_alloca_call_p): Likewise.
8021         (maybe_warn_nonstring_arg): Likewise.
8022         * cfgexpand.c (expand_call_stmt): Likewise.
8023         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8024         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8025         (cgraph_node::verify_node): Likewise.
8026         * cgraphclones.c (build_function_decl_skip_args): Likewise.
8027         (cgraph_node::create_clone): Likewise.
8028         * config/arm/arm.c (arm_insert_attributes): Likewise.
8029         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8030         * dse.c (scan_insn): Likewise.
8031         * expr.c (expand_expr_real_1): Likewise.
8032         * fold-const.c (operand_equal_p): Likewise.
8033         (fold_binary_loc): Likewise.
8034         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8035         * gimple-low.c (lower_stmt): Likewise.
8036         * gimple-pretty-print.c (dump_gimple_call): Likewise.
8037         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8038         Likewise.
8039         * gimple.c (gimple_build_call_from_tree): Likewise.
8040         (gimple_call_builtin_p): Likewise.
8041         (gimple_call_combined_fn): Likewise.
8042         * gimplify.c (gimplify_call_expr): Likewise.
8043         (gimple_boolify): Likewise.
8044         (gimplify_modify_expr): Likewise.
8045         (gimplify_addr_expr): Likewise.
8046         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8047         * ipa-cp.c (determine_versionability): Likewise.
8048         * ipa-fnsummary.c (compute_fn_summary): Likewise.
8049         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8050         * ipa-split.c (visit_bb): Likewise.
8051         (split_function): Likewise.
8052         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8053         * lto-cgraph.c (input_node): Likewise.
8054         * lto-streamer-out.c (write_symbol): Likewise.
8055         * omp-low.c (setjmp_or_longjmp_p): Likewise.
8056         (lower_omp_1): Likewise.
8057         * predict.c (strip_predict_hints): Likewise.
8058         * print-tree.c (print_node): Likewise.
8059         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8060         * trans-mem.c (is_tm_irrevocable): Likewise.
8061         (is_tm_load): Likewise.
8062         (is_tm_simple_load): Likewise.
8063         (is_tm_store): Likewise.
8064         (is_tm_simple_store): Likewise.
8065         (is_tm_abort): Likewise.
8066         (tm_region_init_1): Likewise.
8067         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8068         * tree-cfg.c (verify_gimple_call): Likewise.
8069         (move_stmt_r): Likewise.
8070         (stmt_can_terminate_bb_p): Likewise.
8071         * tree-eh.c (lower_eh_constructs_2): Likewise.
8072         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
8073         * tree-inline.c (remap_gimple_stmt): Likewise.
8074         (copy_bb): Likewise.
8075         (estimate_num_insns): Likewise.
8076         (fold_marked_statements): Likewise.
8077         * tree-sra.c (scan_function): Likewise.
8078         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
8079         (optimize_stack_restore): Likewise.
8080         (pass_fold_builtins::execute): Likewise.
8081         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
8082         (mark_all_reaching_defs_necessary_1): Likewise.
8083         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
8084         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8085         (pass_forwprop::execute): Likewise.
8086         * tree-ssa-loop-im.c (stmt_cost): Likewise.
8087         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
8088         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
8089         * tree-ssa-strlen.c (get_string_length): Likewise.
8090         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
8091         (find_func_aliases_for_call): Likewise.
8092         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
8093         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
8094         * tree-tailcall.c (find_tail_calls): Likewise.
8095         * tree.c (need_assembler_name_p): Likewise.
8096         (free_lang_data_in_decl): Likewise.
8097         (get_call_combined_fn): Likewise.
8098         * ubsan.c (is_ubsan_builtin_p): Likewise.
8099         * varasm.c (incorporeal_function_p): Likewise.
8100         * tree.h (DECL_BUILT_IN): Remove and replace with
8101         fndecl_built_in_p.
8102         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
8103         (fndecl_built_in_p): New.
8105 2018-08-27  Martin Liska  <mliska@suse.cz>
8107         PR tree-optimization/86847
8108         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
8109         Dump also subtree probability.
8110         (switch_decision_tree::do_jump_if_equal): New function.
8111         (switch_decision_tree::emit_case_nodes): Handle special
8112         situations in balanced tree that can be emitted much simpler.
8113         Fix calculation of probabilities that happen in tree expansion.
8114         * tree-switch-conversion.h (struct cluster): Add
8115         is_single_value_p.
8116         (struct simple_cluster): Likewise.
8117         (struct case_tree_node): Add new function has_child.
8118         (do_jump_if_equal): New.
8120 2018-08-27  Martin Liska  <mliska@suse.cz>
8122         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
8123         Add new argument to bit_test_cluster constructor.
8124         (bit_test_cluster::emit): Set bits really number of values
8125         handlel by a test.
8126         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
8127         probability argument.
8128         * tree-switch-conversion.h (struct bit_test_cluster):
8129         Add m_handles_entire_switch.
8131 2018-08-27  Martin Liska  <mliska@suse.cz>
8133         PR tree-optimization/86702
8134         * tree-switch-conversion.c (jump_table_cluster::emit):
8135         Make probabilities even for values in jump table
8136         according to number of cases handled.
8137         (switch_decision_tree::compute_cases_per_edge): Pass
8138         argument to reset_out_edges_aux function.
8139         (switch_decision_tree::analyze_switch_statement): Likewise.
8140         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8141         Make it static.
8143 2018-08-27  Martin Liska  <mliska@suse.cz>
8145         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8146         cfun argument explicitly.
8147         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8148         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8149         function gimple_switch_default_bb.
8150         (convert_switch_statements):
8151         (expand_builtins):
8152         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8153         * stmt.c (label_to_block_fn): Use label_to_block and pass
8154         cfun argument explicitly and use gimple_switch_label_bb.
8155         (expand_case): Likewise.
8156         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8157         cfun argument explicitly. Likewise.
8158         (make_edges_bb): Likewise.
8159         (make_cond_expr_edges): Likewise.
8160         (get_cases_for_edge): Likewise.
8161         (make_gimple_switch_edges): Likewise.
8162         (label_to_block_fn): Likewise.
8163         (label_to_block): Likewise.
8164         (make_goto_expr_edges): Likewise.
8165         (make_gimple_asm_edges): Likewise.
8166         (main_block_label): Likewise.
8167         (group_case_labels_stmt): Likewise.
8168         (find_taken_edge_computed_goto): Likewise.
8169         (find_taken_edge_switch_expr): Likewise.
8170         (gimple_verify_flow_info): Likewise.
8171         (gimple_redirect_edge_and_branch): Likewise.
8172         (gimple_switch_label_bb): New function.
8173         (gimple_switch_default_bb): Likewise.
8174         (gimple_switch_edge): Likewise.
8175         (gimple_switch_default_edge): Likewise.
8176         * tree-cfg.h (label_to_block_fn): Remove and replace ...
8177         (label_to_block): ... with this.
8178         (gimple_switch_label_bb): New.
8179         (gimple_switch_default_bb): Likewise.
8180         (gimple_switch_edge): Likewise.
8181         (gimple_switch_default_edge): Likewise.
8182         * tree-cfgcleanup.c (convert_single_case_switch): Use
8183         new gimple functions and pass new argument to label_to_block.
8184         (cleanup_control_flow_bb):
8185         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8186         cfun argument explicitly.
8187         (make_eh_edges): Likewise.
8188         (redirect_eh_dispatch_edge): Likewise.
8189         (lower_resx): Likewise.
8190         (lower_eh_dispatch): Likewise.
8191         (maybe_remove_unreachable_handlers): Likewise.
8192         (unsplit_eh): Likewise.
8193         (cleanup_empty_eh): Likewise.
8194         (verify_eh_edges): Likewise.
8195         (verify_eh_dispatch_edge): Likewise.
8196         * tree-ssa-dom.c (record_edge_info): Likewise.
8197         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8198         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8199         (thread_through_normal_block): Likewise.
8200         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8201         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8202         * tree-switch-conversion.c (switch_conversion::collect): Use new
8203         gimple functions.
8204         (switch_conversion::check_final_bb): Likewise.
8205         (switch_conversion::gather_default_values): Pass new argument
8206         to label_to_block.
8207         (switch_conversion::build_constructors): Likewise.
8208         (switch_decision_tree::compute_cases_per_edge): Use new
8209         gimple_switch_edge function.
8210         (switch_decision_tree::analyze_switch_statement): Pass new argument
8211         to label_to_block.
8212         (switch_decision_tree::try_switch_expansion): Use
8213         gimple_switch_default_edge.
8214         * tree-vrp.c (find_switch_asserts): Pass new argument
8215         to label_to_block.
8216         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8217         (vr_values::simplify_switch_using_ranges): Likewise.
8219 2018-08-27  Richard Biener  <rguenther@suse.de>
8221         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8222         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8224         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8225         (struct vn_nary_op_s): Add unwind_to member.  Add
8226         predicated_values flag and put result into a union together
8227         with a linked list of vn_pval.
8228         (struct vn_ssa_aux): Add name member to make maintaining
8229         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8230         needed info, dfsnum, low, visited, on_sccstack, use_processed
8231         and range_info_anti_range_p members.
8232         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8233         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8234         New functions.
8235         (vn_valueize): New global.
8236         (vn_context_bb): Likewise.
8237         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8238         VN_INFO_PTR_INFO): Remove.
8239         * tree-ssa-sccvn.c: ... (rewrite)
8240         (pass_fre::execute): For -O2+ initialize loops and run
8241         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8242         run RPO VN in non-optimistic mode.
8243         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8244         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8245         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8246         (rpo-vn-max-loop-depth): Document.
8247         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8248         when valuezing the VUSE signals we walked out of the region.
8249         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8250         (phi_translate): Set VN context block to use for availability
8251         lookup.
8252         (compute_avail): Likewise.
8253         (pre_valueize): New function.
8254         (pass_pre::execute): Adjust to the RPO VN API.
8256         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8257         (propagate_constants_for_unrolling): Remove.
8258         (tree_unroll_loops_completely): Perform value-numbering
8259         on the unrolled bodies loop parent.
8261 2018-08-27  Richard Biener  <rguenther@suse.de>
8263         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8264         for partial antic compute.
8266 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8268         PR rtl-optimization/87065
8269         * combine.c (simplify_if_then_else): Formatting fix.
8270         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8271         check.
8272         (known_cond): Don't return const_true_rtx for vector modes.  Use
8273         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8275 2018-08-27  Martin Liska  <mliska@suse.cz>
8277         PR gcov-profile/87069
8278         * gcov.c (process_file): Record files already processed
8279         and warn about a file being processed multiple times.
8281 2018-08-27  Martin Liska  <mliska@suse.cz>
8283         PR driver/83193
8284         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8285         Set default values for x_aarch64_*_string strings.
8286         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8287         prefix.  For -mabi do not print '=ABI' in help and use
8288         <option_value> format for -msve-vector-bits and -moverride
8289         options.
8291 2018-08-26  Jeff Law  <law@redhat.com>
8293         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8294         "cfg.h"
8296 2018-08-26  Marek Polacek  <polacek@redhat.com>
8298         PR c++/87029, Implement -Wredundant-move.
8299         * doc/invoke.texi: Document -Wredundant-move.
8301 2018-08-25  Martin Sebor  <msebor@redhat.com>
8303         PR tree-optimization/87059
8304         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8305         to the same type as the other.
8306         * fold-const.c (fold_binary_loc): Assert expectation.
8308 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8310         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8311         extraneous parentheses, dead code section and formatting.
8313 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8315         PR c++/87091
8316         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8317         wide enough for jumps in the line-numbering to be visible.
8318         (layout::print_gap_in_line_numbering): New member function.
8319         (layout::calculate_line_spans): When using line numbering, merge
8320         line spans that are only 1 line apart.
8321         (diagnostic_show_locus): When printing line numbers, show gaps in
8322         line numbering directly, rather than printing headers.
8323         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8324         line-numbering with multiple line spans.
8325         (selftest::test_fixit_insert_containing_newline_2): Add test of
8326         line-numbering, in which the spans are close enough to be merged.
8328 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8330         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8331         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8332         * tree-vrp.c (range_is_nonnull): Remove.
8333         (range_includes_zero_p): Accept value_range instead of min/max.
8334         (extract_range_from_binary_expr_1): Do not early bail on
8335         POINTER_PLUS_EXPR.
8336         Use range_includes_zero_p instead of range_is_nonnull.
8337         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8338         of range_is_nonnull.
8339         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8340         special case VR_ANTI_RANGE.
8341         (vrp_finalize): Same.
8342         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8343         instead of min/max.
8344         (range_is_nonnull): Remove.
8345         * vr-values.c (vrp_stmt_computes_nonzero): Use
8346         range_includes_zero_p instead of range_is_nonnull.
8347         (extract_range_basic): Pass value_range to range_includes_zero_p
8348         instead of range_is_nonnull.
8350 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8352         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8353         * emit-rtl.h (rtl_data): Remove return_bnd.
8354         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8355         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8356         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8357         (POINTER_BOUNDS_MODE): Remove definition.
8358         (make_pointer_bounds_mode): Remove.
8359         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8360         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8361         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8362         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8363         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8364         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8365         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8367         * config/i386/i386-modes.def (BND32, BND64): Remove.
8368         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8369         (dbx64_register_map): Ditto.
8370         (svr4_dbx_register_map): Ditto.
8371         (indirect_thunk_bnd_needed): Remove.
8372         (indirect_thunks_bnd_used): Ditto.
8373         (indirect_return_bnd_needed): Ditto.
8374         (indirect_return_via_cx_bnd): Ditto.
8375         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8376         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8377         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8378         (output_indirect_thunk_function): Remove handling of
8379         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8380         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8381         (ix86_save_reg): Remove handling of crtl->return_bnd.
8382         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8383         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8384         and UNSPEC_BNDLX_ADDR.
8385         (ix86_output_indirect_branch_via_reg): Remove handling of
8386         indirect_thunk_prefix_bnd.
8387         (ix86_output_indirect_branch_via_push): Ditto.
8388         (ix86_output_function_return): Ditto.
8389         (ix86_output_indirect_function_return): Ditto.
8390         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8391         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8392         (CALL_USED_REGISTERS): Ditto.
8393         (REG_ALLOC_ORDER): Update for removal of bound registers.
8394         (HI_REGISTER_NAMES): Ditto.
8395         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8396         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8397         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8398         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8399         (FIRST_PSEUDO_REG): Update.
8400         (BND): Remove mode iterator.
8401         * config/i386/predicates.md (bnd_mem_operator): Remove.
8403 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8405         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8406         vectors.
8408 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8410         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8411         the case in which the permute needs only a single element and
8412         repeats for every vector of the result.  Extend that case to
8413         handle variable-length vectors.
8414         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8416 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8418         PR debug/79342
8419         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8420         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8422 2018-08-24  Richard Biener  <rguenther@suse.de>
8424         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8425         bb_flags_allocated members.
8426         (auto_flag): New RAII class for allocating flags.
8427         (auto_edge_flag): New RAII class for allocating edge flags.
8428         (auto_bb_flag): New RAII class for allocating bb flags.
8429         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8430         flag dynamically.
8431         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8432         in favor of temporarily allocated BB flag.
8433         * hsa-brig.c: Re-order includes.
8434         * hsa-dump.c: Likewise.
8435         * hsa-regalloc.c: Likewise.
8436         * print-rtl.c: Likewise.
8437         * profile-count.c: Likewise.
8439 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8441         PR target/86989
8442         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8443         the TOC register.
8445 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8447         PR 87073/bootstrap
8448         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8449         from wide_int_range_multiplicative_op.
8451 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8453         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8454         "permutaion".
8456 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8458         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8459         to 'expanded'.
8461 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8463         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8464         full GENERIC expression used for replacement.
8466 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8468         * tree-vrp.c (abs_extent_range): Remove.
8469         (extract_range_into_wide_ints): Pass wide ints by reference.
8470         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8471         Pass wide ints by reference in all calls to
8472         extract_range_into_wide_ints.
8473         * wide-int-range.cc (wide_int_range_div): New.
8474         * wide-int-range.h (wide_int_range_div): New.
8475         (wide_int_range_includes_zero_p): New.
8476         (wide_int_range_zero_p): New.
8478 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8480         * config/aarch64/aarch64.md (arches): New enum.
8481         (arch): New enum attr.
8482         (arch_enabled): New attr.
8483         (enabled): Now uses arch_enabled only.
8484         (simd, sve, fp16): Removed attribute.
8485         (fp): Attr now defined in terms of 'arch'.
8486         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8487         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8488         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8489         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8490         attributes into 'arch'.
8491         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8492         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8493         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8494         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8495         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8496         'simd' attribute into 'arch'.
8497         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8498         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8499         Convert use of 'fp' attribute to 'arch'.
8500         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8501         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8502         into 'arch'.
8503         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8504         (different modes) Merge 'fp' and 'simd' into 'arch'.
8505         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8506         'simd' into 'arch'.
8508 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8510         PR rtl-optimization/87026
8511         * expmed.c (canonicalize_comparison): If we can no longer create
8512         pseudoregisters, don't.
8514 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8516         PR target/86951
8517         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8518         prototype.
8519         * config/arm/arm.c (speculation_barrier_libfunc): New static
8520         variable.
8521         (arm_init_libfuncs): Initialize it.
8522         (arm_emit_speculation_barrier): New function.
8523         * config/arm/arm.md (speculation_barrier): Call
8524         arm_emit_speculation_barrier for architectures that do not have
8525         DSB or ISB.
8526         (speculation_barrier_insn): Only match on Armv7 or later.
8528 2018-08-23  Richard Biener  <rguenther@suse.de>
8530         PR middle-end/87024
8531         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8532         calls.
8534 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8536         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8537         of single-vector TBLs.
8538         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8539         one input is given.
8541 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8543         PR target/85910
8544         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8545         aarch64_evpc_tbl guard.
8547 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8549         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8550         behaviour.
8552 2018-08-22  Martin Sebor  <msebor@redhat.com>
8554         PR middle-end/87052
8555         * tree-pretty-print.c (pretty_print_string): Add argument.
8556         (dump_generic_node): Call to pretty_print_string with string size.
8558 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8560         PR rtl-optimization/86771
8561         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8562         of two SETs into those two SETs, one to be placed at i2, if that SETs
8563         destination is modified between i2 and i3.
8565 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8567         PR tree-optimization/86725
8568         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8569         function.
8570         (vect_analyze_scalar_cycles_1): Check it.
8572 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8574         PR tree-optimization/86725
8575         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8576         an outer loop phi as a double reduction, make sure that the
8577         single user of the phi result is an inner loop phi.
8579 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8581         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8582         grouped stores with gaps to a strided group.
8584 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8586         * tree-vect-stmts.c (get_group_load_store_type)
8587         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8588         first statement in a group.
8590 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8592         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8593         the sequence used in gcc/gcc.c.
8595 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8597         PR other/704
8598         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8599         building it.
8601 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8603         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8604         Darwin10-specific unwinder-shim.
8605         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8606         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8607         New to cater for Darwin10 Rosetta.
8609 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8611         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8612         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8614 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8616         PR bootstrap/81033
8617         PR target/81733
8618         PR target/52795
8619         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8620         (dwarf2out_switch_text_section): Generate a local label for the
8621         second function sub-section and apply it as the second FDE start
8622         label.
8623         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8624         second sub-section start.
8626 2018-08-22  Richard Biener  <rguenther@suse.de>
8628         PR tree-optimization/86988
8629         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8631 2018-08-22  Richard Biener  <rguenther@suse.de>
8633         PR tree-optimization/86945
8634         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8636 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8638         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8639         a comment about how uses of r2 for .sdata2 come about.
8641 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8643         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8645 2018-08-21  Marek Polacek  <polacek@redhat.com>
8647         PR c++/86981, Implement -Wpessimizing-move.
8648         * doc/invoke.texi: Document -Wpessimizing-move.
8650 2018-08-21  Jan Hubicka  <jh@suse.cz>
8652         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8653         * tree.h (is_redundant_typedef): Remove.
8654         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8656 2018-08-21  Jan Hubicka  <jh@suse.cz>
8658         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8659         when possible.
8661 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8663         * expmed.c (extract_low_bits): Reject invalid subregs early.
8665 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8667         PR middle-end/86121
8668         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8669         behaviour.
8671 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8673         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8674         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8675         * config/vxworks.c: Likewise.
8677 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8679         * config/vxworks.c: Set targetm.have_ctors_dtors
8680         if HAVE_INITFINI_ARRAY_SUPPORT.
8681         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8682         if HAVE_INITFINI_ARRAY_SUPPORT.
8684 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8686         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8687         default search path for VxWorks < 7.
8689 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8691         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8692         (get_format_string): Refer to c_getstr.
8694 2018-08-21  Tom de Vries  <tdevries@suse.de>
8696         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8697         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8698         (debuginfo_early_stop): Declare.
8699         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8700         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8701         (debuginfo_early_stop): New function.
8702         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8703         and debuginfo_early_stop.
8704         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8705         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8706         (general_init): Call debuginfo_early_init.
8707         (finalize): Call debuginfo_fini.
8708         (do_compile): Call debuginfo_init.
8709         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8710         -fdump-early-debug.
8711         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8713 2018-08-21  Tom de Vries  <tdevries@suse.de>
8715         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8716         flag_dump_noaddr and flag_dump_unnumbered.
8718 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
8720         * wide-int-range.cc (wide_int_range_abs): New.
8721         (wide_int_range_order_set): Rename from wide_int_range_min_max.
8722         * wide-int-range.h (wide_int_range_abs): New.
8723         (wide_int_range_min_max): New.
8724         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
8725         case to call wide_int_range_abs.
8726         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
8727         (extract_range_from_abs_expr): Delete.
8729 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
8731         PR target/87033
8732         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
8733         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
8734         for indexed loads.
8736 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8737             Jeff Law <law@redhat.com>
8739         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
8740         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
8742 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
8744         PR other/84889
8745         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
8746         (decl_attributes): Likewise.
8747         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
8748         instance.
8749         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8750         * diagnostic-core.h (class auto_diagnostic_group): New class.
8751         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
8752         (diagnostic_report_diagnostic): Handle the first diagnostics within
8753         a group.
8754         (emit_diagnostic): Add auto_diagnostic_group instance.
8755         (inform): Likewise.
8756         (inform_n): Likewise.
8757         (warning): Likewise.
8758         (warning_at): Likewise.
8759         (warning_n): Likewise.
8760         (pedwarn): Likewise.
8761         (permerror): Likewise.
8762         (error): Likewise.
8763         (error_n): Likewise.
8764         (error_at): Likewise.
8765         (sorry): Likewise.
8766         (fatal_error): Likewise.
8767         (internal_error): Likewise.
8768         (internal_error_no_backtrace): Likewise.
8769         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
8770         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
8771         * diagnostic.h (struct diagnostic_context): Add fields
8772         "diagnostic_group_nesting_depth",
8773         "diagnostic_group_emission_count", "begin_group_cb",
8774         "end_group_cb".
8775         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
8776         Add auto_diagnostic_group instance(s).
8777         (find_explicit_erroneous_behavior): Likewise.
8778         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
8779         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
8780         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
8781         (gimplify_va_arg_expr): Likewise.
8782         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
8783         (HSA_SORRY_AT): Likewise.
8784         * ipa-devirt.c (compare_virtual_tables): Likewise.
8785         (warn_odr): Likewise.
8786         * multiple_target.c (expand_target_clones): Likewise.
8787         * opts-common.c (cmdline_handle_error): Likewise.
8788         * reginfo.c (globalize_reg): Likewise.
8789         * substring-locations.c (format_warning_n_va): Likewise.
8790         * tree-inline.c (expand_call_inline): Likewise.
8791         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
8792         * tree-ssa-loop-niter.c
8793         (do_warn_aggressive_loop_optimizations): Likewise.
8794         * tree-ssa-uninit.c (warn_uninit): Likewise.
8795         * tree.c (warn_deprecated_use): Likewise.
8797 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
8799         PR target/87014
8800         * config/i386/i386.md (eh_return): Always update EH return
8801         address in word_mode.
8803 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
8805         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
8806         TARGET_SPLIT_COMPLEX_ARG is defined.
8808 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8810         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
8812 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8814         PR target/86984
8815         * expr.c (expand_assignment): Assert that bitpos is positive.
8816         (store_field): Likewise
8817         (expand_expr_real_1): Make sure that bitpos is positive.
8818         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
8819         integer overflow.
8821 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8823         * Makefile.in (CPP_ID_DATA_H): Delete.
8824         (CPP_INTERNAL_H): Don't add it.
8825         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
8826         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
8828 2018-08-20  Richard Biener  <rguenther@suse.de>
8830         PR tree-optimization/78655
8831         * tree-vrp.c (extract_range_from_binary_expr_1): Make
8832         pointer + offset nonnull if either operand is nonnull work.
8834 2018-08-20  Tom de Vries  <tdevries@suse.de>
8836         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
8837         unless the referenced die describes the added property using
8838         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
8839         Otherwise, add a DW_AT_location to the referenced die.
8841 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
8843         PR target/86994
8844         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
8845         register_operand when calling ix86_set_reg_reg_cost.
8846         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
8847         Set *total to 0 for operands that satisfy x86_64_immediate_operand
8848         predicate and to 1 otherwise.
8850 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
8852         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
8853         emit a diagnostic that it is not supported and reset the option.
8854         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
8855         supported and consume the option.  (ASM_FINAL_SPEC): New.
8857 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8859         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
8860         a sentence.
8862 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
8864         C-SKY port: Documentation
8866         * doc/extend.texi (C-SKY Function Attributes): New section.
8867         * doc/invoke.texi (Option Summary): Add C-SKY options.
8868         (C-SKY Options): New section.
8869         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
8871 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8872             Huibin Wang  <huibin_wang@c-sky.com>
8873             Sandra Loosemore  <sandra@codesourcery.com>
8874             Chung-Lin Tang  <cltang@codesourcery.com>
8876         C-SKY port: Backend implementation
8878         * config/csky/*: New.
8879         * common/config/csky/*: New.
8881 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8882             Huibin Wang  <huibin_wang@c-sky.com>
8883             Sandra Loosemore  <sandra@codesourcery.com>
8884             Chung-Lin Tang  <cltang@codesourcery.com>
8885             Andrew Jenner  <andrew@codesourcery.com>
8887         C-SKY port: Configury
8889         * config.gcc (csky-*-*): New.
8890         * configure.ac: Add csky to targets for dwarf2 debug_line support.
8891         * configure: Regenerated.
8893 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
8895         * dump-context.h: Include "dumpfile.h".
8896         (dump_context::dump_printf_va): Convert final param from va_list
8897         to va_list *.  Convert from ATTRIBUTE_PRINTF to
8898         ATTRIBUTE_GCC_DUMP_PRINTF.
8899         (dump_context::dump_printf_loc_va): Likewise.
8900         * dumpfile.c: Include "stringpool.h".
8901         (make_item_for_dump_printf_va): Delete.
8902         (make_item_for_dump_printf): Delete.
8903         (class dump_pretty_printer): New class.
8904         (dump_pretty_printer::dump_pretty_printer): New ctor.
8905         (dump_pretty_printer::emit_items): New member function.
8906         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
8907         function.
8908         (dump_pretty_printer::emit_item): New member function.
8909         (dump_pretty_printer::stash_item): New member function.
8910         (dump_pretty_printer::format_decoder_cb): New member function.
8911         (dump_pretty_printer::decode_format): New member function.
8912         (dump_context::dump_printf_va): Reimplement in terms of
8913         dump_pretty_printer.
8914         (dump_context::dump_printf_loc_va): Convert final param from va_list
8915         to va_list *.
8916         (dump_context::begin_scope): Reimplement call to
8917         make_item_for_dump_printf.
8918         (dump_printf): Update for change to dump_printf_va.
8919         (dump_printf_loc): Likewise.
8920         (selftest::test_capture_of_dump_calls): Convert "stmt" from
8921         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
8922         with %T, %E, and %G.
8923         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
8924         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
8925         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
8926         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
8927         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
8928         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
8929         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
8930         within a dump_printf_loc call to "%wu".
8931         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
8932         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
8933         missing space after "=".
8934         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
8935         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
8936         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
8937         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
8938         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
8939         duplicate "vectorized" from message.
8941 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
8943         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
8944         polyNxK_t element's TYPE_STRING_FLAG.
8946 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8948         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
8949         (they were unnamed before).  Fix comments.
8951 2018-08-17  Nathan Sidwell  <nathan@acm.org>
8953         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
8955 2018-08-17  Richard Biener  <rguenther@suse.de>
8957         PR tree-optimization/86841
8958         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
8960 2018-08-17  Martin Liska  <mliska@suse.cz>
8962         * common.opt: Remove Warn, Init and Report for options with
8963         Ignore/Deprecated flag. Warning is done automatically for
8964         Deprecated flags.
8965         * config/i386/i386.opt: Likewise.
8966         * config/ia64/ia64.opt: Likewise.
8967         * config/rs6000/rs6000.opt: Likewise.
8968         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8969         Remove usage of flag_check_pointer_bounds.
8970         * lto-wrapper.c (merge_and_complain): Do not handle
8971         OPT_fcheck_pointer_bounds.
8972         (append_compiler_options): Likewise.
8973         * opt-functions.awk: Do not handle Deprecated.
8974         * optc-gen.awk: Check that Var, Report and Init are not
8975         used for an option with Ignore/Deprecated flag.
8976         * opts-common.c (decode_cmdline_option): Do not report
8977         CL_ERR_DEPRECATED.
8978         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
8979         options.
8980         * opts.h (struct cl_option): Remove cl_deprecated flag.
8981         (CL_ERR_DEPRECATED): Remove error enum value.
8983 2018-08-17  Richard Biener  <rguenther@suse.de>
8985         PR middle-end/86505
8986         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
8987         across a va-arg-pack using call adjust its return value accordingly.
8989 2018-08-16  Martin Sebor  <msebor@redhat.com>
8991         PR tree-optimization/86853
8992         * gimple-ssa-sprintf.c (struct format_result): Rename member.
8993         (struct fmtresult): Add member and initialize it in ctors.
8994         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
8995         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
8996         (format_directive): Set POSUNDER4K when MAYFAIL is set.
8997         (parse_directive): Handle %C same as %c.
8998         (sprintf_dom_walker::compute_format_length): Adjust.
8999         (is_call_safe): Adjust.
9001 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9003         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
9004         for determining how to count the elements.
9005         * builtins.h (c_strlen): Adjust prototype.
9006         * expr.c (string_constant): Add new parameter mem_size.
9007         Set *mem_size appropriately.
9008         * expr.h (string_constant): Adjust protoype.
9009         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9010         * gimple-fold.h (get_range_strlen): Adjust prototype.
9011         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9012         (format_string): Call get_string_length with eltsize.
9014 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9016         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9017         to emit the span, rather than setting it as the prefix.
9019 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9021         * diagnostic-show-locus.c (layout::start_annotation_line): Add
9022         "margin_char" parameter, defaulting to space.  Use it in place
9023         of pp_space for the initial part of the margin.
9024         (layout::print_leading_fixits): Use '+' when filling the margin
9025         of line-insertion fix-it hints.
9027 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9029         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9030         Delete.
9032 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9034         * config/rs6000/altivec.md: Don't set length attribute to the default
9035         value.
9036         * config/rs6000/darwin.md: Ditto.
9037         * config/rs6000/dfp.md: Ditto.
9038         * config/rs6000/htm.md: Ditto.
9039         * config/rs6000/rs6000.md: Ditto.
9040         * config/rs6000/sync.md: Ditto.
9041         * config/rs6000/vsx.md: Ditto.
9043 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9045         * config/rs6000/altivec.md: Don't set length attribute to the default
9046         value, for branch instructions.
9047         * config/rs6000/darwin.md: Ditto.
9048         * config/rs6000/rs6000.md: Ditto.
9050 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9052         * config/rs6000/rs6000.md (length): Always define as const_int 4.
9053         (unnamed conditional branch define_insn): Set length to 4 or 8
9054         depending on offset.
9055         (<bd>_<mode>): Similar, for alternative 0.
9056         (<bd>tf_<mode>): Ditto.
9058 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9060         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9062 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
9064         * doc/rtl.texi: Replace old RTX class names with new names.
9067 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
9069         * expmed.h (canonicalize_comparison): New declaration.
9070         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
9071         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
9072         * optabs.c (prepare_cmp_insn): Likewise.
9073         * rtl.h (unsigned_condition_p): New function which checks if a
9074         comparison operator is unsigned.
9076 2018-08-16  Nathan Sidwell  <nathan@acm.org>
9078         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
9079         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
9081 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9083         PR target/84711
9084         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
9085         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
9086         (mov<mov>): ..this and enable unconditionally.
9088 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9090         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
9092 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
9094         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
9095         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
9096         "Common" with "Target".
9098 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9100         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
9101         * doc/invoke.texi (mmitigate-rop): Remove.
9102         * config/i386/i386.c: Do not include "regrename.h".
9103         (ix86_rop_should_change_byte_p, reg_encoded_number)
9104         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
9105         Remove.
9106         (ix86_reorg): Remove call to ix86_mitigate_rop.
9107         * config/i386/i386.md (attr "modrm_class"): Remove.
9108         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
9109         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
9110         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
9112 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
9114         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
9115         allow folding of mergeh() and mergel() for the float and double types.
9116         (fold_mergehl_helper): Rework to handle building a permute tree
9117         for float vectors.
9119 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9121         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
9122         for TARGET_SSE.
9124 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
9126         * common.opt (fdiagnostics-show-labels): New option.
9127         * diagnostic-show-locus.c (class layout_range): Add field
9128         "m_label".
9129         (class layout): Add field "m_show_labels_p".
9130         (layout_range::layout_range): Add param "label" and use it to
9131         initialize m_label.
9132         (make_range): Pass in NULL for new "label" param of layout_range's
9133         ctor.
9134         (layout::layout): Initialize m_show_labels_p.
9135         (layout::maybe_add_location_range): Pass in loc_range->m_label
9136         when constructing layout_range instances.
9137         (struct line_label): New struct.
9138         (layout::print_any_labels): New member function.
9139         (layout::print_line): Call it if label-printing is enabled.
9140         (selftest::test_one_liner_labels): New test.
9141         (selftest::test_diagnostic_show_locus_one_liner): Call it.
9142         * diagnostic.c (diagnostic_initialize): Initialize
9143         context->show_labels_p.
9144         * diagnostic.h (struct diagnostic_context): Add field
9145         "show_labels_p".
9146         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9147         -fno-diagnostics-show-labels.
9148         * dwarf2out.c (gen_producer_string): Add
9149         OPT_fdiagnostics_show_labels to the ignored options.
9150         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9151         param.
9152         (gcc_rich_location::maybe_add_expr): Likewise.
9153         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9154         label" param, defaulting to NULL.
9155         (gcc_rich_location::add_expr): Add "label" param.
9156         (gcc_rich_location::maybe_add_expr): Likewise.
9157         (class text_range_label): New class.
9158         (class range_label_for_type_mismatch): New class.
9159         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9160         of format_warning_va.
9161         (fmtwarn_n): Likewise for new params of format_warning_n_va.
9162         * lto-wrapper.c (merge_and_complain): Add
9163         OPT_fdiagnostics_show_labels to the "pick one setting" options.
9164         (append_compiler_options): Likewise to the dropped options.
9165         (append_diag_options): Likewise to the passed-on options.
9166         * opts.c (common_handle_option): Handle the new option.
9167         * selftest-diagnostic.c
9168         (test_diagnostic_context::test_diagnostic_context): Enable
9169         show_labels_p.
9170         * substring-locations.c: Include "gcc-rich-location.h".
9171         (format_warning_n_va): Add "fmt_label" and "param_label" params
9172         and use them as appropriate.
9173         (format_warning_va): Add "fmt_label" and "param_label" params,
9174         passing them on to format_warning_n_va.
9175         (format_warning_at_substring): Likewise.
9176         (format_warning_at_substring_n): Likewise.
9177         * substring-locations.h (format_warning_va): Add "fmt_label" and
9178         "param_label" params.
9179         (format_warning_n_va): Likewise.
9180         (format_warning_at_substring): Likewise.
9181         (format_warning_at_substring_n): Likewise.
9182         * toplev.c (general_init): Initialize global_dc->show_labels_p.
9184 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
9186         PR testsuite/86519
9187         * builtins.c (expand_builtin_memcmp): Do not expand the call
9188         when overflow is detected.
9190 2018-08-15  Martin Sebor  <msebor@redhat.com>
9192         PR tree-optimization/71625
9193         * config/aarch64/aarch64-builtins.c
9194         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9196 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
9198         * config/s390/s390.c (s390_reorg): Remove loop.
9200 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9202         * config/darwin.c
9203          (darwin_function_switched_text_sections): Delete.
9204         * gcc/config/darwin.h
9205          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9207 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9209         PR target/81685
9210         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9211         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
9212         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9214 2018-08-15  Martin Liska  <mliska@suse.cz>
9216         PR tree-optimization/86925
9217         * predict.c (expr_expected_value_1): When taking
9218         later predictor, assign also probability.
9219         Use fold_build2_initializer_loc in order to fold
9220         the expression in -frounding-math.
9222 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9224         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9225         patterns.
9226         (expand_vec_perm_1): Try the new method.
9228 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9230         PR target/86547
9231         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9232         Check whether lra_live_max_point is 0 before dividing.
9234 2018-08-14  Martin Sebor  <msebor@redhat.com>
9236         PR tree-optimization/86650
9237         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9238         (vrp_prop::check_mem_ref): Same.
9240 2018-08-13  Liu Hao <lh_mouse@126.com>
9242         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9243         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9244         and clear it thereafter, as it only works for DBCS.
9246 2018-08-13  Liu Hao <lh_mouse@126.com>
9248         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9249         handle returned by _get_osf_handle().
9251 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9253         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9254         for folding vec_perm.
9256 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9258         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9259         Add support for gimple-folding of vec_pack() and vec_unpack()
9260         intrinsics.
9262 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9264         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9265         vec_xst variants to the list.
9266         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9267         vector loads and stores.
9269 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9271         * config.gcc (rs6000-ibm-aix4.x): Delete.
9272         (rs6000-ibm-aix5.1): Delete.
9273         (rs6000-ibm-aix5.2): Delete.
9274         (rs6000-ibm-aix5.3): Delete.
9275         * config/rs6000/aix43.h: Delete.
9276         * config/rs6000/aix51.h: Delete.
9277         * config/rs6000/aix52.h: Delete.
9278         * config/rs6000/t-aix43: Delete.
9280 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9282         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9283         New function.
9284         (s390_decompose_address): Factor out constant pool ref
9285         decomposition.
9287 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9289         * config/nds32/nds32-predicates.c
9290         (nds32_can_use_bclr_p): Change return type as bool.
9291         (nds32_can_use_bset_p): Ditto.
9292         (nds32_can_use_btgl_p): Ditto.
9293         (nds32_can_use_bitci_p): Ditto.
9294         * config/nds32/nds32-protos.h
9295         (nds32_can_use_bclr_p): Change declaration.
9296         (nds32_can_use_bset_p): Ditto.
9297         (nds32_can_use_btgl_p): Ditto.
9298         (nds32_can_use_bitci_p): Ditto.
9300 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9302         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9303         Support -msched-prolog-epilog option.
9304         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9306 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9308         * common/config/nds32/nds32-common.c
9309         (nds32_option_optimization_table): Enalbe -malways-align.
9311 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9313         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9314         extra_headers.
9315         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9316         OPT_misr_secure_ case.
9317         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9318         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9319         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9320         secure attribute.
9321         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9322         (nds32_isr_info): New field security_level.
9323         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9324         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9325         * config/nds32/nds32.opt (misr-secure): New option.
9326         * config/nds32/nds32_init.inc: New file.
9327         * config/nds32/nds32_isr.h: New file.
9329 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9331         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9332         Update comment for atomic instructions.
9333         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9334         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9335         Remove.
9336         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9337         load for __ATOMIC_SEQ_CST model.
9338         (atomic_loaddi_1): Remove float register target.
9339         (atomic_storedi): Handle CONST_INT values.
9340         (atomic_storedi_1): Remove float register source.  Add special case
9341         for zero value.
9342         (memory_barrier): New expander and insn.
9344 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9346         PR tree-optimization/86835
9347         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9348         new_stmt after def_gsi, make sure to insert new_square_stmt after
9349         that stmt, not 2 stmts before it.
9351 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9353         PR target/82418
9354         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9355         instead of SWI48.
9357 2018-08-10  Martin Liska  <mliska@suse.cz>
9359         PR target/83610
9360         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9361         function type.
9362         * builtins.c (expand_builtin_expect_with_probability):
9363         New function.
9364         (expand_builtin_expect_with_probability): New function.
9365         (build_builtin_expect_predicate): Add new argumnet probability
9366         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9367         (fold_builtin_expect):
9368         (fold_builtin_2):
9369         (fold_builtin_3):
9370         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9371         * builtins.h (fold_builtin_expect): Set new argument.
9372         * doc/extend.texi: Document __builtin_expect_with_probability.
9373         * doc/invoke.texi: Likewise.
9374         * gimple-fold.c (gimple_fold_call): Pass new argument.
9375         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9376         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9377         * predict.c (get_predictor_value): New function.
9378         (expr_expected_value): Add new argument probability. Assume
9379         that predictor and probability are always non-null.
9380         (expr_expected_value_1): Likewise.  For __builtin_expect and
9381         __builtin_expect_with_probability set probability.  Handle
9382         combination in binary expressions.
9383         (tree_predict_by_opcode): Simplify code by simply calling
9384         get_predictor_value.
9385         (pass_strip_predict_hints::execute): Add handling of
9386         BUILT_IN_EXPECT_WITH_PROBABILITY.
9387         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9388         new predictor.
9389         * tree.h (DECL_BUILT_IN_P): New function.
9391 2018-08-10  Martin Liska  <mliska@suse.cz>
9393         PR tree-optimization/85799
9394         * passes.def: Add argument for pass_strip_predict_hints.
9395         * predict.c (class pass_strip_predict_hints): Add new argument
9396         early_p.
9397         (strip_predictor_early): New function.
9398         (pass_strip_predict_hints::execute): Call the function to
9399         strip predictors.
9400         (strip_predict_hints): New function.
9401         * predict.def: Fix comment.
9403 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9405         * Makefile.in: Clarify which tm.texi to copy over to assert the
9406         right to grant a GFDL license for all.
9408 2018-08-09  Jeff Law  <law@redhat.com>
9410         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9411         unused variable.
9413 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9415         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9416         prototype.
9418 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9420         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9421         reductions for variable-length vectors.
9423 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9425         PR other/84889
9426         * common.opt (fdiagnostics-show-line-numbers): New option.
9427         * diagnostic-show-locus.c (class layout): Add fields
9428         "m_show_line_numbers_p" and "m_linenum_width";
9429         (num_digits): New function.
9430         (test_num_digits): New function.
9431         (layout::layout): Initialize new fields.  Update m_x_offset
9432         logic to handle any left margin.
9433         (layout::print_source_line): Print line number when requested.
9434         (layout::start_annotation_line): New member function.
9435         (layout::print_annotation_line): Call it.
9436         (layout::print_leading_fixits): Likewise.
9437         (layout::print_trailing_fixits): Likewise.  Update calls to
9438         move_to_column for new parameter.
9439         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9440         it to potentially call start_annotation_line.
9441         (layout::show_ruler): Call start_annotation_line.
9442         (selftest::test_line_numbers_multiline_range): New selftest.
9443         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9444         and selftest::test_line_numbers_multiline_range.
9445         * diagnostic.c (diagnostic_initialize): Initialize
9446         show_line_numbers_p.
9447         * diagnostic.h (struct diagnostic_context): Add field
9448         "show_line_numbers_p".
9449         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9450         -fno-diagnostics-show-line-numbers.
9451         * dwarf2out.c (gen_producer_string): Add
9452         OPT_fdiagnostics_show_line_numbers to the ignored options.
9453         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9454         one setting" options.
9455         (append_compiler_options): Likewise to the dropped options.
9456         (append_diag_options): Likewise to the passed-on options.
9457         * opts.c (common_handle_option): Handle the new option.
9458         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9460 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9462         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9463         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9464         third argument of type "const signed char" to descriptions of
9465         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9466         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9467         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9468         __builtin_bcdsub_ov functions.
9470 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9472         PR tree-optimization/86858
9473         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9474         flow_bb_inside_loop_p calls.
9476 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9478         PR tree-optimization/86871
9479         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9480         instead of gimple_assign_lhs.
9482 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9484         PR target/86887
9485         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9486         register constraint to operand 0.
9487         (add<mode>3_carryinC): Likewise.
9488         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9490 2018-08-09  Martin Liska  <mliska@suse.cz>
9492         PR c/86895
9493         * common.opt: Remove extra line.
9495 2018-08-09  Martin Liska  <mliska@suse.cz>
9497         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9498         at the end of a line, make first letter capital and end up
9499         a sentence with a dot.
9500         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9501         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9502         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9503         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9504         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9505         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9506         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9507         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9508         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9509         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9510         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9511         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9512         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9513         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9514         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9515         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9516         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9517         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9519 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9521         PR target/84332
9522         * config/s390/s390.c (s390_option_override_internal): Reduce the
9523         stack-clash-protection-probe-interval param if it would be too big
9524         for z900.
9526 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9528         PR target/46179
9529         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9530         * config/m68k/m68k.c (handle_move_double): Don't call
9531         m68k_final_prescan_insn.
9532         (m68k_adjust_decorated_operand): Renamed from
9533         m68k_final_prescan_insn, remove first and third operand and
9534         simplify.
9535         (print_operand): Call it.
9536         (print_operand_address): Call it.
9538 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9540         * diagnostic.c (diagnostic_report_current_module): Use
9541         linemap_included_from & linemap_included_from_linemap.
9543 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9545         * config/aarch64/aarch64-cores.def: Add phecda core.
9546         * config/aarch64/aarch64-tune.md: Regenerate.
9547         * doc/invoke.texi: Add phecda core.
9549 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9551         PR target/85295
9552         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9553         definitions.
9554         * config/s390/s390.md ("movti"): Add more alternatives for
9555         constant to GPR copies.
9557 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9559         * config/s390/s390.c: Fix whitespace damage throughout the file.
9560         * config/s390/s390.h: Likewise.
9561         * config/s390/tpf.h: Likewise.
9563 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9565         * config/s390/s390.c (s390_loadrelative_operand_p):
9566         Remove TARGET_CPU_ZARCH usages.
9567         (s390_rtx_costs): Likewise.
9568         (s390_legitimate_constant_p): Likewise.
9569         (s390_cannot_force_const_mem): Likewise.
9570         (legitimate_reload_constant_p): Likewise.
9571         (s390_preferred_reload_class): Likewise.
9572         (legitimize_pic_address): Likewise.
9573         (legitimize_tls_address): Likewise.
9574         (s390_split_branches): Removed.
9575         (s390_add_execute): Removed.
9576         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9577         (s390_mainpool_start): Likewise.
9578         (s390_mainpool_finish): Likewise.
9579         (s390_mainpool_cancel): Removed.
9580         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9581         (s390_chunkify_cancel): Likewise.
9582         (s390_return_addr_rtx): Likewise.
9583         (s390_register_info): Remove split_branches_pending_p uages.
9584         (s390_optimize_register_info): Likewise.
9585         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9586         split_branches_pending_p usages.
9587         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9588         (s390_load_got): Likewise.
9589         (s390_expand_split_stack_prologue): Likewise.
9590         (output_asm_nops): Likewise.
9591         (s390_function_profiler): Likewise.
9592         (s390_emit_call): Likewise.
9593         (s390_conditional_register_usage): Likewise.
9594         (s390_optimize_prologue): Likewise.
9595         (s390_reorg): Remove TARGET_CPU_ZARCH and
9596         split_branches_pending_p usages.
9597         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9598         usages.
9599         (s390_output_indirect_thunk_function): Likewise.
9600         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9601         (TARGET_CPU_ZARCH_P): Removed.
9602         (struct machine_function): Remove split_branches_pending_p.
9603         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9605 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9607         * common/config/s390/s390-common.c (processor_flags_table):
9608         Remove flags.
9609         * config.gcc: Remove with_arch/with_tune support.
9610         * config/s390/2064.md: Remove cpu attribute comparisons.
9611         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9612         Remove MTN.
9613         * config/s390/linux.h (ASM_SPEC):
9614         Remove -march support.
9615         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9616         Use a table to get an arch level.
9617         * config/s390/s390-opts.h (enum processor_type):
9618         Remove enum values.
9619         * config/s390/s390.c
9620         (processor_table): Remove entries, add arch_level values.
9621         (s390_issue_rate): Remove cases.
9622         (s390_option_override): Adjust
9623         s390_option_override_internal() call.
9624         (s390_option_override_internal): Remove deprecation warning.
9625         (s390_valid_target_attribute_tree): Adjust
9626         s390_option_override_internal() call.
9627         * config/s390/s390.h (struct s390_processor):
9628         Share with s390-c.c, add arch_level field.
9629         * config/s390/s390.md:
9630         Remove occurrences in cpu attribute.
9631         * config/s390/s390.opt: Remove -march/-mtune support.
9632         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9633         * doc/invoke.texi: Remove deprecation warning.
9635 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9637         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9638         global.
9639         (qdf24xx_tunings): Set vector cost structure to
9640         qdf24xx_vector_cost.
9642         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9643         <register_sextend>: Set to 3.
9645 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9647         PR target/86838
9648         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9649         * config/aarch64/aarch64-simd.md
9650         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9651         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9652         (aarch64_frecpx<mode>): ...this new pattern.
9653         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9654         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9656 2018-08-07  Martin Liska  <mliska@suse.cz>
9658         PR middle-end/83023
9659         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9660         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9661         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9662         * doc/extend.texi: Document that malloc attribute adds
9663         hit to compiler.
9665 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9667         PR target/86785
9668         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9669         Define to speculation_safe_value_not_needed.
9671 2018-08-06  Jeff Law  <law@redhat.com>
9673         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9674         the vr_values instance to cprop_into_stmt.
9675         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9676         (cprop_operand): Also query EVRP to determine if OP is a constant.
9678 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9680         * diagnostic.c (diagnostic_report_current_module): Reroll
9681         included-at loop.  Translate text.
9683 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9685         * function-tests.c (selftest::test_expansion_to_rtl): Call
9686         free_after_compilation.
9688 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9690         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9692 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9694         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9695         loops with memory block operations from getting unrolled.
9697 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9699         PR target/86807
9700         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9701         Define to speculation_safe_value_not_needed.
9703 2018-08-06  Jeff Law  <law@redhat.com>
9705         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9706         assert.
9708 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9710         PR target/86662
9711         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9712         with all enabled __intN types.
9714         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9716 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9718         * alias.c (record_set): Check for clobber high.
9719         * cfgexpand.c (expand_gimple_stmt): Likewise.
9720         * combine-stack-adj.c (single_set_for_csa): Likewise.
9721         * combine.c (find_single_use_1): Likewise.
9722         (set_nonzero_bits_and_sign_copies): Likewise.
9723         (get_combine_src_dest): Likewise.
9724         (is_parallel_of_n_reg_sets): Likewise.
9725         (try_combine): Likewise.
9726         (record_dead_and_set_regs_1): Likewise.
9727         (reg_dead_at_p_1): Likewise.
9728         (reg_dead_at_p): Likewise.
9729         * dce.c (deletable_insn_p): Likewise.
9730         (mark_nonreg_stores_1): Likewise.
9731         (mark_nonreg_stores_2): Likewise.
9732         * df-scan.c (df_find_hard_reg_defs): Likewise.
9733         (df_uses_record): Likewise.
9734         (df_get_call_refs): Likewise.
9735         * dwarf2out.c (mem_loc_descriptor): Likewise.
9736         * haifa-sched.c (haifa_classify_rtx): Likewise.
9737         * ira-build.c (create_insn_allocnos): Likewise.
9738         * ira-costs.c (scan_one_insn): Likewise.
9739         * ira.c (equiv_init_movable_p): Likewise.
9740         (rtx_moveable_p): Likewise.
9741         (interesting_dest_for_shprep): Likewise.
9742         * jump.c (mark_jump_label_1): Likewise.
9743         * postreload-gcse.c (record_opr_changes): Likewise.
9744         * postreload.c (reload_cse_simplify): Likewise.
9745         (struct reg_use): Add source expr.
9746         (reload_combine): Check for clobber high.
9747         (reload_combine_note_use): Likewise.
9748         (reload_cse_move2add): Likewise.
9749         (move2add_note_store): Likewise.
9750         * print-rtl.c (print_pattern): Likewise.
9751         * recog.c (decode_asm_operands): Likewise.
9752         (store_data_bypass_p): Likewise.
9753         (if_test_bypass_p): Likewise.
9754         * regcprop.c (kill_clobbered_value): Likewise.
9755         (kill_set_value): Likewise.
9756         * reginfo.c (reg_scan_mark_refs): Likewise.
9757         * reload1.c (maybe_fix_stack_asms): Likewise.
9758         (eliminate_regs_1): Likewise.
9759         (elimination_effects): Likewise.
9760         (mark_not_eliminable): Likewise.
9761         (scan_paradoxical_subregs): Likewise.
9762         (forget_old_reloads_1): Likewise.
9763         * reorg.c (find_end_label): Likewise.
9764         (try_merge_delay_insns): Likewise.
9765         (redundant_insn): Likewise.
9766         (own_thread_p): Likewise.
9767         (fill_simple_delay_slots): Likewise.
9768         (fill_slots_from_thread): Likewise.
9769         (dbr_schedule): Likewise.
9770         * resource.c (update_live_status): Likewise.
9771         (mark_referenced_resources): Likewise.
9772         (mark_set_resources): Likewise.
9773         * rtl.c (copy_rtx): Likewise.
9774         * rtlanal.c (reg_referenced_p): Likewise.
9775         (single_set_2): Likewise.
9776         (noop_move_p): Likewise.
9777         (note_stores): Likewise.
9778         * sched-deps.c (sched_analyze_reg): Likewise.
9779         (sched_analyze_insn): Likewise.
9781 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9783         * cse.c (invalidate_reg): New function extracted from...
9784         (invalidate): ...here.
9785         (canonicalize_insn): Check for clobber high.
9786         (invalidate_from_clobbers): invalidate clobber highs.
9787         (invalidate_from_sets_and_clobbers): Likewise.
9788         (count_reg_usage): Check for clobber high.
9789         (insn_live_p): Likewise.
9790         * cselib.c (cselib_expand_value_rtx_1):Likewise.
9791         (cselib_invalidate_regno): Check for clobber in setter.
9792         (cselib_invalidate_rtx): Pass through setter.
9793         (cselib_invalidate_rtx_note_stores):
9794         (cselib_process_insn): Check for clobber high.
9795         * cselib.h (cselib_invalidate_rtx): Add operand.
9797 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9799         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
9800         (mark_not_eliminable): Likewise.
9801         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
9802         * lra-lives.c (process_bb_lives): Check for clobber high.
9803         * lra.c (new_insn_reg): Remember clobber highs.
9804         (collect_non_operand_hard_regs): Check for clobber high.
9805         (lra_set_insn_recog_data): Likewise.
9806         (add_regs_to_insn_regno_info): Likewise.
9807         (lra_update_insn_regno_info): Likewise.
9809 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9811         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
9812         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
9814 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9816         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
9817         (copy_insn_1): Likewise.
9818         (gen_hard_reg_clobber_high): New gen function.
9819         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
9820         * genemit.c (gen_exp): Likewise.
9821         (gen_emit_seq): Pass through info.
9822         (gen_insn): Check for CLOBBER_HIGH.
9823         (gen_expand): Pass through info.
9824         (gen_split): Likewise.
9825         (output_add_clobbers): Likewise.
9826         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
9827         (remove_clobbers): Likewise.
9828         * rtl.h (gen_hard_reg_clobber_high): New declaration.
9830 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9832         * doc/rtl.texi (clobber_high): Add.
9833         (parallel): Add in clobber high
9834         * rtl.c (rtl_check_failed_code3): Add function.
9835         * rtl.def (CLOBBER_HIGH): Add expression.
9836         * rtl.h (RTL_CHECKC3): Add macro.
9837         (rtl_check_failed_code3): Add declaration.
9838         (XC3EXP): Add macro.
9840 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
9842         PR target/86386
9843         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
9844         cfun->machine->max_used_stack_alignment if needed.
9846 2018-08-04  Martin Sebor  <msebor@redhat.com>
9848         PR tree-optimization/86571
9849         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
9850         NaN output to 4.
9852 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
9854         PR target/86799
9855         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9856         Define.
9858 2018-08-03  Jeff Law  <law@redhat.com>
9860         PR target/86795
9861         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9862         Define to speculation_safe_value_not_needed.
9864 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
9866         * doc/gcov.texi (-x): Remove duplicate "to".
9867         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
9868         (-Wif-not-aligned): Remove duplicate "is".
9869         (-flto): Remove duplicate "the".
9870         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
9871         duplicate "v5.00.b".
9872         (MSP430 Options): Remove duplicate "and" from the description
9873         of "-mgprel-sec=regexp".
9874         (x86 Options): Remove duplicate copies of "vmldLog102" and
9875         vmlsLog104 from description of "-mveclibabi=type".
9877 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
9879         * internal-fn.h (first_commutative_argument): Declare.
9880         * internal-fn.c (first_commutative_argument): New function.
9881         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
9882         restrictions for pattern statements.  Use first_commutative_argument
9883         to look for commutative operands in calls to internal functions.
9885 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
9887         * Makefile.in (wide-int-range.o): New.
9888         * tree-vrp.c: Move all the wide_int_* functions to...
9889         * wide-int-range.cc: ...here.
9890         * tree-vrp.h: Move all the wide_int_* prototypes to...
9891         * wide-int-range.h: ...here.
9893 2018-08-03  Tom de Vries  <tdevries@suse.de>
9895         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
9896         UI_NONE.
9897         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
9898         * except.c (output_function_exception_table): Do early exit if
9899         targetm_common.except_unwind_info (&global_options) == UI_NONE.
9901 2018-08-03  Martin Liska  <mliska@suse.cz>
9903         * predict.c (dump_prediction): Change to 2 digits
9904         in fraction part.
9906 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9908         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
9909         neon_dup_q to...
9910         (falkor_am_1_gtov_gtov): ... a new insn reservation.
9912 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
9914         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
9915         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
9916         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
9917         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
9918         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
9920 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9922         * diagnostic-show-locus.c (diagnostic_show_locus): Use
9923         pp_take_prefix when saving the existing prefix.
9924         * diagnostic.c (diagnostic_append_note): Likewise.
9925         * langhooks.c (lhd_print_error_function): Likewise.
9926         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
9927         param's type.  Free the existing prefix.
9928         (pp_take_prefix): New function.
9929         (pretty_printer::pretty_printer): Drop the prefix parameter.
9930         Rename the length parameter to match the comment.
9931         (pretty_printer::~pretty_printer): Free the prefix.
9932         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
9933         parameter.
9934         (struct pretty_printer): Drop the "const" from "prefix" field's
9935         type and clarify memory management.
9936         (pp_set_prefix): Drop the "const" from the 2nd param.
9937         (pp_take_prefix): New decl.
9939 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
9941         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
9942         for word_mode_ok here instead of passing as argument.
9943         (expand_block_compare): Change select_block_compare_mode() call.
9944         (expand_strncmp_gpr_sequence): New function.
9945         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
9947 2018-08-02  Jeff Law  <law@redhat.com>
9949         PR target/86790
9950         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9951         Define to speculation_safe_value_not_needed.
9953         PR target/86784
9954         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9955         Define to speculation_safe_value_not_needed.
9957 2018-08-02  Tom de Vries  <tdevries@suse.de>
9959         PR target/86660
9960         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
9961         function.  Return UI_TARGET unconditionally.
9962         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
9963         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
9965 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9967         * genemit.c (print_overload_test): Fix typo.
9969 2018-08-02  Richard Biener  <rguenther@suse.de>
9971         PR tree-optimization/86816
9972         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
9973         which checks for value availability before querying it.
9974         (gvn_uses_equal): Use it.
9975         (same_succ_hash): Likewise.
9976         (gimple_equal_p): Likewise.
9978 2018-08-02  Nick Clifton  <nickc@redhat.com>
9980         PR target/86813
9981         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9982         Define to speculation_safe_value_not_needed.
9984         PR target/86810
9985         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9986         Define to speculation_safe_value_not_needed.
9988         PR target/86810
9989         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9990         Define to speculation_safe_value_not_needed.
9992         PR target/86803
9993         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9994         Define to speculation_safe_value_not_needed.
9996         PR target/86797
9997         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9998         Define to speculation_safe_value_not_needed.
10000         PR target/86791
10001         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10002         Define to speculation_safe_value_not_needed.
10004         PR target/86789
10005         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10006         Define to speculation_safe_value_not_needed.
10008         PR target/86787
10009         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10010         Define to speculation_safe_value_not_needed.
10012         PR target/86782
10013         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10014         speculation_safe_value_not_needed.
10016         PR target/86781
10017         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10018         to speculation_safe_value_not_needed.
10020 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10022         * doc/md.texi: Expand the documentation of instruction names
10023         to mention port-local uses.  Document '@' in pattern names.
10024         * read-md.h (overloaded_instance, overloaded_name): New structs.
10025         (mapping): Declare.
10026         (md_reader::handle_overloaded_name): New member function.
10027         (md_reader::get_overloads): Likewise.
10028         (md_reader::m_first_overload): New member variable.
10029         (md_reader::m_next_overload_ptr): Likewise.
10030         (md_reader::m_overloads_htab): Likewise.
10031         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10032         m_next_overload_ptr and m_overloads_htab.
10033         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10034         (get_mode_token, get_code_token, get_int_token): New functions.
10035         (map_attr_string): Add an optional argument that passes back
10036         the associated iterator.
10037         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10038         (md_reader::handle_overloaded_name, add_overload_instance): New
10039         functions.
10040         (apply_iterators): Handle '@' names.  Report an error if '@'
10041         is used without iterators.
10042         (initialize_iterators): Initialize the new iterator_group fields.
10043         * genopinit.c (handle_overloaded_code_for)
10044         (handle_overloaded_gen): New functions.
10045         (main): Use them to print declarations of maybe_code_for_* and
10046         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10047         * genemit.c (print_overload_arguments, print_overload_test)
10048         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10049         (main): Use it to print definitions of maybe_code_for_* and
10050         maybe_gen_* functions.
10051         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10052         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10053         instead of explicit mode checks.
10054         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10055         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10056         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10057         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10058         (aarch64_expand_compare_and_swap): Likewise
10059         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10060         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10061         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10062         (aarch64_constant_pool_reload_icode): Delete.
10063         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10064         instead of aarch64_constant_pool_reload_icode.  Use
10065         code_for_aarch64_reload_mov instead of explicit mode checks.
10066         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10067         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10068         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10069         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
10070         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
10071         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
10072         (aarch64_atomic_load_op_code): Delete.
10073         (aarch64_emit_atomic_load_op): Likewise.
10074         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
10075         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
10076         instead of aarch64_emit_atomic_load_op.
10077         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
10078         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
10079         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
10080         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
10081         character before the pattern name.
10082         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
10083         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
10084         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
10085         (aarch64_frecps<mode>): Likewise.
10086         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
10087         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
10088         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
10089         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
10090         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
10092 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10094         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
10095         Allow HFmode constants if TARGET_FP_F16INST.
10097 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
10099         PR target/86014
10100         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
10101         No longer check last store for clobber of address register.
10103 2018-08-02  Martin Liska  <mliska@suse.cz>
10105         PR gcov-profile/86817
10106         * gcov.c (process_all_functions): New function.
10107         (main): Call it.
10108         (process_file): Move functions processing to
10109         process_all_functions.
10111 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10113         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
10114         "const" to the "gimple *" and "rtx_insn *" parameters.
10115         * dumpfile.h (dump_user_location_t::dump_user_location_t):
10116         Likewise.
10117         (dump_location_t::dump_location_t): Likewise.
10119 2018-08-01  Martin Sebor  <msebor@redhat.com>
10121         PR tree-optimization/86650
10122         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
10123         rather than a "gcall *".  Directly pass the data of interest
10124         to percent_K_format, rather than building a temporary CALL_EXPR
10125         to hold it.
10126         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
10127         (gimple_fold_builtin_strncat): Adjust.
10128         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
10129         gcall* argument with gimple*.
10130         * gimple-ssa-warn-restrict.c (check_call): Same.
10131         (wrestrict_dom_walker::before_dom_children): Same.
10132         (builtin_access::builtin_access): Same.
10133         (check_bounds_or_overlap): Same
10134         (maybe_diag_overlap): Same.
10135         (maybe_diag_offset_bounds): Same.
10136         * tree-diagnostic.c (default_tree_printer): Move usage of
10137         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10138         to this callsite.
10139         * tree-pretty-print.c (percent_K_format): Add argument.
10140         * tree-pretty-print.h: Add argument.
10141         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10142         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10143         (maybe_diag_stxncpy_trunc): Same.
10144         (handle_builtin_stxncpy): Same.
10145         (handle_builtin_strcat): Same.
10147 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10149         * match.pd: Optimise pointer range checks.
10151 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10153         PR tree-optimization/86758
10154         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10155         to remove pattern statements.
10157 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10159         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10160         result of dfs_enumerate_from when constructing stmt_vec_infos,
10161         instead of additionally calling get_loop_body.
10163 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10165         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10166         parameter.
10167         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10168         When creating an iv, assert that the step is not known to be zero.
10169         (vect_setup_realignment): Update call accordingly.
10170         * tree-vect-stmts.c (vectorizable_store): Likewise.
10171         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
10173 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10175         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10176         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10177         (vectorizable_reduction): Likewise.
10178         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10179         (vect_detect_hybrid_slp_stmts): Likewise.
10180         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10182 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
10184         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10185         (wide_int_set_zero_nonzero_bits): ...this.
10186         (zero_nonzero_bits_from_vr): Rename to...
10187         (vrp_set_zero_nonzero_bits): ...this.
10188         (extract_range_from_multiplicative_op_1): Abstract wide int
10189         code...
10190         (wide_int_range_multiplicative_op): ...here.
10191         (extract_range_from_binary_expr_1): Extract wide int binary
10192         operations into their own functions.
10193         (wide_int_range_lshift): New.
10194         (wide_int_range_can_optimize_bit_op): New.
10195         (wide_int_range_shift_undefined_p): New.
10196         (wide_int_range_bit_xor): New.
10197         (wide_int_range_bit_ior): New.
10198         (wide_int_range_bit_and): New.
10199         (wide_int_range_trunc_mod): New.
10200         (extract_range_into_wide_ints): New.
10201         (vrp_shift_undefined_p): New.
10202         (extract_range_from_multiplicative_op): New.
10203         (vrp_can_optimize_bit_op): New.
10204         * tree-vrp.h (value_range::dump): New.
10205         (wide_int_range_multiplicative_op): New.
10206         (wide_int_range_lshift):New.
10207         (wide_int_range_shift_undefined_p): New.
10208         (wide_int_range_bit_xor): New.
10209         (wide_int_range_bit_ior): New.
10210         (wide_int_range_bit_and): New.
10211         (wide_int_range_trunc_mod): New.
10212         (zero_nonzero_bits_from_bounds): Rename to...
10213         (wide_int_set_zero_nonzero_bits): ...this.
10214         (zero_nonzero_bits_from_vr): Rename to...
10215         (vrp_set_zero_nonzero_bits): ...this.
10216         (range_easy_mask_min_max): Rename to...
10217         (wide_int_range_can_optimize_bit_op): this.
10218         * vr-values.c (simplify_bit_ops_using_ranges): Rename
10219         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10221 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10223         * tree-vectorizer.h (vect_orig_stmt): New function.
10224         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10225         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10226         (vect_create_epilog_for_reduction): Likewise.
10227         (vectorizable_live_operation): Likewise.
10228         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10229         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10230         * tree-vect-stmts.c (vectorizable_call): Likewise.
10231         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10233 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10235         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10236         argument.
10237         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10238         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10239         (vect_transform_loop): Likewise.
10240         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10242 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10244         * tree-vectorizer.h (vect_schedule_slp): Return void.
10245         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10246         (vect_schedule_slp): Likewise.
10248 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10250         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10251         argument.
10252         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10253         instances before the main loop, if any exist.
10255 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10257         PR tree-optimization/86749
10258         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10259         If the lhs is used in a COND_EXPR, check that it is being used
10260         as the "then" or "else" value.
10262 2018-08-01  Tom de Vries  <tdevries@suse.de>
10264         PR target/86800
10265         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10266         speculation_safe_value_not_needed.
10268 2018-08-01  Richard Biener  <rguenther@suse.de>
10270         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10271         as base and offset.
10273 2018-08-01  Martin Liska  <mliska@suse.cz>
10275         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10276         format how successful transformation is dumped.
10277         (gimple_mod_pow2_value_transform): Likewise.
10278         (gimple_mod_subtract_transform): Likewise.
10279         (gimple_stringops_transform): Likewise.
10281 2018-08-01  Martin Liska  <mliska@suse.cz>
10283         PR value-prof/35543
10284         * value-prof.c (interesting_stringop_to_profile_p):
10285         Simplify the code and add BUILT_IN_MEMMOVE.
10286         (gimple_stringops_transform): Likewise.
10288 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10290         * config/aarch64/aarch64-simd.md
10291         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10292         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10293         use GPI iterator instead of SI mode.
10295 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10297         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10298         rs6000_speculation_barrier.
10299         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10300         new barrier pattern name.
10302 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10304         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10305         (speculation_barrier): New insn.
10307 2018-08-01  Richard Biener  <rguenther@suse.de>
10309         PR bootstrap/86724
10310         * graphite.h: Include isl/id.h and isl/space.h to allow build
10311         with ISL 0.20.
10313 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10315         PR target/86651
10316         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10317         mode for COFF targets.
10318         * defaults.h (TARGET_COFF): Define.
10319         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10320         TARGET_COFF): Define.
10321         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10322         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10323         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10325 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10327         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10328         * gimple-streamer-out.c (output_bb): Save it.
10329         * lto-streamer-in.c (input_struct_function_base): Restore
10330         instance discriminator if available.  Create map on demand.
10331         * lto-streamer-out.c (output_struct_function_base): Save it if
10332         available.
10333         * final.c (decl_to_instance_map): Document LTO strategy.
10335 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10336             Olivier Hainque  <hainque@adacore.com>
10338         * debug.h (decl_to_instance_map_t): New type.
10339         (decl_to_instance_map): Declare.
10340         (maybe_create_decl_to_instance_map): New inline function.
10341         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10342         to track basic block discriminators.
10343         (final_start_function_1): Initialize them.
10344         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10345         bb_discriminator.
10346         (decl_to_instance_map): New variable.
10347         (map_decl_to_instance, maybe_set_discriminator): New functions.
10348         (notice_source_line): Set discriminator.
10350 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10352         * targhooks.c (default_have_speculation_safe_value): Add
10353         ATTRIBUTE_UNUSED.
10355 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10357         * dump-context.h: Include "pretty-print.h".
10358         (dump_context::refresh_dumps_are_enabled): New decl.
10359         (dump_context::emit_item): New decl.
10360         (class dump_context): Add fields "m_test_pp" and
10361         "m_test_pp_flags".
10362         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10363         (temp_dump_context::get_dumped_text): New decl.
10364         (class temp_dump_context): Add field "m_pp".
10365         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10366         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10367         m_test_pp.
10368         (set_dump_file): Update for above change.
10369         (set_alt_dump_file): Likewise.
10370         (dump_loc): New overload, taking a pretty_printer *.
10371         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10372         to any test pretty-printer.
10373         (make_item_for_dump_gimple_stmt): New function, adapted from
10374         optinfo::add_gimple_stmt.
10375         (dump_context::dump_gimple_stmt): Call it, and use the result,
10376         eliminating the direct usage of dump_file and alt_dump_file in
10377         favor of indirectly using them via emit_item.
10378         (make_item_for_dump_gimple_expr): New function, adapted from
10379         optinfo::add_gimple_expr.
10380         (dump_context::dump_gimple_expr): Call it, and use the result,
10381         eliminating the direct usage of dump_file and alt_dump_file in
10382         favor of indirectly using them via emit_item.
10383         (make_item_for_dump_generic_expr): New function, adapted from
10384         optinfo::add_tree.
10385         (dump_context::dump_generic_expr): Call it, and use the result,
10386         eliminating the direct usage of dump_file and alt_dump_file in
10387         favor of indirectly using them via emit_item.
10388         (make_item_for_dump_printf_va): New function, adapted from
10389         optinfo::add_printf_va.
10390         (make_item_for_dump_printf): New function.
10391         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10392         and use the result, eliminating the direct usage of dump_file and
10393         alt_dump_file in favor of indirectly using them via emit_item.
10394         (make_item_for_dump_dec): New function.
10395         (dump_context::dump_dec): Call it, and use the result,
10396         eliminating the direct usage of dump_file and alt_dump_file in
10397         favor of indirectly using them via emit_item.
10398         (make_item_for_dump_symtab_node): New function, adapted from
10399         optinfo::add_symtab_node.
10400         (dump_context::dump_symtab_node): Call it, and use the result,
10401         eliminating the direct usage of dump_file and alt_dump_file in
10402         favor of indirectly using them via emit_item.
10403         (dump_context::begin_scope): Reimplement, avoiding direct usage
10404         of dump_file and alt_dump_file in favor of indirectly using them
10405         via emit_item.
10406         (dump_context::emit_item): New member function.
10407         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10408         Set up test pretty-printer on the underlying context.  Call
10409         refresh_dumps_are_enabled.
10410         (temp_dump_context::~temp_dump_context): Call
10411         refresh_dumps_are_enabled.
10412         (temp_dump_context::get_dumped_text): New member function.
10413         (selftest::verify_dumped_text): New function.
10414         (ASSERT_DUMPED_TEXT_EQ): New macro.
10415         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10416         and then without optinfo enabled.  Add uses of
10417         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10418         * dumpfile.h: Update comment for the dump_* API.
10419         * optinfo-emit-json.cc
10420         (selftest::test_building_json_from_dump_calls): Update for new
10421         param for temp_dump_context ctor.
10422         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10423         and "m_owned" field.
10424         (optinfo_item::~optinfo_item): Likewise.
10425         (optinfo::add_item): New member function.
10426         (optinfo::emit): Update comment.
10427         (optinfo::add_string): Delete.
10428         (optinfo::add_printf): Delete.
10429         (optinfo::add_printf_va): Delete.
10430         (optinfo::add_gimple_stmt): Delete.
10431         (optinfo::add_gimple_expr): Delete.
10432         (optinfo::add_tree): Delete.
10433         (optinfo::add_symtab_node): Delete.
10434         (optinfo::add_dec): Delete.
10435         * optinfo.h (class dump_context): New forward decl.
10436         (optinfo::add_item): New decl.
10437         (optinfo::add_string): Delete.
10438         (optinfo::add_printf): Delete.
10439         (optinfo::add_printf_va): Delete.
10440         (optinfo::add_gimple_stmt): Delete.
10441         (optinfo::add_gimple_expr): Delete.
10442         (optinfo::add_tree): Delete.
10443         (optinfo::add_symtab_node): Delete.
10444         (optinfo::add_dec): Delete.
10445         (optinfo::add_poly_int): Delete.
10446         (optinfo_item::optinfo_item): Remove "owned" param.
10447         (class optinfo_item): Remove field "m_owned".
10449 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10451         PR middle-end/86705
10452         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10453         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10454         requested variable alignment.
10455         (expand_one_ssa_partition): Likewise.
10456         (expand_one_var): Likewise.
10458 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10460         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10461         to speculation_safe_value_not_needed.
10463 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10465         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10466         * targhooks.c (speculation_safe_value_not_needed): New function.
10467         * target.def (have_speculation_safe_value): Update documentation.
10468         * doc/tm.texi: Regenerated.
10470 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10472         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10473         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10474         expand.
10475         (despeculate_copy<ALLI:mode>_insn): New insn.
10476         (despeculate_copyti_insn): New insn.
10477         (despeculate_simple<ALLI:mode>): New insn
10478         (despeculate_simpleti): New insn.
10479         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10480         function.
10481         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10482         aarch64_speculation_safe_value.
10483         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10485 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10487         * config/aarch64/aarch64-speculation.cc: New file.
10488         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10489         before pass_reorder_blocks.
10490         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10491         prototype.
10492         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10493         X14 and X15 when tracking speculation.
10494         * config/aarch64/aarch64.md (register name constants): Add
10495         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10496         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10497         (speculation_barrier): New insn attribute.
10498         (cmp<mode>): Allow SP in comparisons.
10499         (speculation_tracker): New insn.
10500         (speculation_barrier): Add speculation_barrier attribute.
10501         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10502         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10503         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10505 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10507         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10508         aarch64_track_speculation is true.
10509         (tb<optab><mode>1): Likewise.
10510         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10511         generate CB[N]Z when tracking speculation.
10512         (aarch64_split_compare_and_swap): Likewise.
10513         (aarch64_split_atomic_op): Likewise.
10515 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10517         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10519 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10521         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10522         (speculation_barrier): New insn.
10524 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10526         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10527         * config/arm/arm.md (speculation_barrier): New expand.
10528         (speculation_barrier_insn): New pattern.
10530 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10532         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10533         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10534         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10535         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10536         list.
10537         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10538         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10539         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10540         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10541         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10542         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10543         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10544         * builtins.c (expand_speculation_safe_value): New function.
10545         (expand_builtin): Call it.
10546         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10547         * doc/extend.texi: Document __builtin_speculation_safe_value.
10548         * doc/md.texi: Document "speculation_barrier" pattern.
10549         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10550         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10551         * doc/tm.texi: Regenerated.
10552         * target.def (have_speculation_safe_value, speculation_safe_value): New
10553         hooks.
10554         * targhooks.c (default_have_speculation_safe_value): New function.
10555         (default_speculation_safe_value): New function.
10556         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10557         (default_speculation_safe_value): Add prototype.
10559 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10561         * dump-context.h (dump_context::dump_loc): New decl.
10562         * dumpfile.c (dump_context::dump_loc): New member function.
10563         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10564         and dump_gimple_stmt.
10565         (dump_context::dump_gimple_expr_loc): Likewise, using
10566         dump_gimple_expr.
10567         (dump_context::dump_generic_expr_loc): Likewise, using
10568         dump_generic_expr.
10569         (dump_context::dump_printf_loc_va): Likewise, using
10570         dump_printf_va.
10571         (dump_context::begin_scope): Explicitly using the global function
10572         "dump_loc", rather than the member function.
10574 2018-07-31  Martin Sebor  <msebor@redhat.com>
10576         PR tree-optimization/86741
10577         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10579 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10581         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10582         depend on whether prefetch instructions will be emitted or not.
10583         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10584         will be emitted or not.
10585         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10586         (TARGET_SETMEM_PFD): New macros.
10588 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10590         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10591         (NULL_STMT_VEC_INFO): Delete.
10592         (stmt_vec_info::operator*): Likewise.
10593         (stmt_vec_info::operator gimple *): Likewise.
10594         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10595         of NULL_STMT_VEC_INFO.
10596         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10597         (vect_reassociating_reduction_p): Likewise.
10598         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10599         (vectorizable_store): Likewise.
10600         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10601         (vec_info::free_stmt_vec_infos): Likewise.
10603 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10605         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10606         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10608 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10610         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10611         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10612         (vec_info::free_stmt_vec_info): New private member functions.
10613         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10614         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10615         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10616         set_stmt_vec_info_vec and free_stmt_vec_infos.
10617         * tree-vect-loop.c (_loop_vec_info): Remove call to
10618         set_stmt_vec_info_vec.
10619         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10620         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10621         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10622         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10623         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10624         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10625         with the clearing in the xcalloc.
10626         (stmt_vec_info_vec): Delete.
10627         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10628         (vectorize_loops): Likewise.
10629         (vec_info::~vec_info): Remove argument from call to
10630         free_stmt_vec_infos.
10631         (vec_info::add_stmt): Remove vinfo argument from call to
10632         new_stmt_vec_info.
10634 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10636         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10637         rather than a gimple stmt.
10638         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10639         information for pattern statements when passed the original
10640         statement; instead wait to be passed the pattern statement itself.
10641         Don't call set_vinfo_for_stmt here.
10642         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10643         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10644         stmt_vec_infos here.
10645         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10646         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10647         stmt_vec_infos entry.
10649 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10651         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10652         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10653         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10654         * tree-vect-stmts.c (vectorizable_call): Likewise.
10655         (vectorizable_simd_clone_call): Likewise.
10657 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10659         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10660         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10661         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10662         * tree-vect-loop.c (vect_transform_loop): Likewise.
10663         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10664         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10666 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10668         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10669         (vect_dr_stmt): Delete.
10670         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10671         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10672         of DR_VECT_AUX.
10673         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10674         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10675         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10676         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10677         (vect_analyze_data_refs): Likewise.
10678         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10679         argument.
10680         (vect_find_same_alignment_drs): Likewise.
10681         (vect_slp_analyze_node_dependences): Update calls accordingly.
10682         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10683         instead of DR_VECT_AUX.
10684         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10685         of a vector data references.  Use vec_info::lookup_dr instead of
10686         DR_VECT_AUX.
10687         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10688         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10689         instead of DR_VECT_AUX.
10691 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10693         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10694         dr_vec_info.
10695         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10696         accordingly.
10697         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10698         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10699         (vect_gen_prolog_loop_niters): Likewise.
10701 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10703         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10704         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10705         (vect_known_alignment_in_bytes, vect_dr_behavior)
10706         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10707         instead of data_references.  Update calls to other routines for
10708         which the same change has been made.
10709         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10710         dr_vec_infos instead of stmt_vec_infos.
10711         (vect_analyze_data_ref_dependence): Update call accordingly.
10712         (vect_slp_analyze_data_ref_dependence)
10713         (vect_record_base_alignments): Use DR_VECT_AUX.
10714         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10715         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10716         (vector_alignment_reachable_p, vect_get_data_access_cost)
10717         (vect_peeling_supportable, vect_analyze_group_access_1)
10718         (vect_analyze_group_access, vect_analyze_data_ref_access)
10719         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10720         (vect_compile_time_alias, vect_small_gap_p)
10721         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
10722         (vect_supportable_dr_alignment): Take references as dr_vec_infos
10723         instead of data_references.  Update calls to other routines for
10724         which the same change has been made.
10725         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
10726         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10727         (vect_slp_analyze_and_verify_node_alignment)
10728         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10729         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10730         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
10731         above changes.
10732         (_vect_peel_info::dr): Replace with...
10733         (_vect_peel_info::dr_info): ...this new field.
10734         (vect_peeling_hash_get_most_frequent)
10735         (vect_peeling_hash_choose_best_peeling): Update accordingly.
10736         (vect_peeling_hash_get_lowest_cost):
10737         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
10738         routines for which the same change has been made.
10739         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
10740         data_reference.
10741         * tree-vect-loop-manip.c (get_misalign_in_elems)
10742         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
10743         above changes.
10744         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10745         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
10746         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
10747         (get_group_load_store_type, get_negative_load_store_type)
10748         (vect_get_data_ptr_increment, vectorizable_store)
10749         (vectorizable_load): Likewise.
10750         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
10751         Update calls to other routines for which the same change has been made.
10753 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10755         * tree-vectorizer.h (vec_info::move_dr): New member function.
10756         (dataref_aux): Rename to...
10757         (dr_vec_info): ...this and add "dr" and "stmt" fields.
10758         (_stmt_vec_info::dr_aux): Update accordingly.
10759         (_stmt_vec_info::data_ref_info): Delete.
10760         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
10761         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
10762         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
10763         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
10764         of data_ref.
10765         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
10766         (STMT_VINFO_DR_INFO): New macro.
10767         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
10768         (set_dr_misalignment): Update after rename of dataref_aux.
10769         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
10770         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
10771         initialization of STMT_VINFO_DATA_REF.
10772         * tree-vectorizer.c (vec_info::move_dr): New function.
10773         * tree-vect-patterns.c (vect_recog_bool_pattern)
10774         (vect_recog_mask_conversion_pattern)
10775         (vect_recog_gather_scatter_pattern): Use it.
10776         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
10777         the "dr" and "stmt" fields of dr_vec_info instead of
10778         STMT_VINFO_DATA_REF.
10780 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10782         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
10783         (is_pattern_stmt_p): Use it.
10784         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
10785         on pattern statements.
10787 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10789         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
10790         original stmt as a stmt_vec_info rather than a gimple stmt.
10791         (vect_pattern_recog_1): Take the statement directly as a
10792         stmt_vec_info, rather than via a gimple_stmt_iterator.
10793         Update call to vect_mark_pattern_stmts.
10794         (vect_pattern_recog): Update calls accordingly.
10796 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10798         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
10799         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
10800         a vect_def_type for the first argument.
10801         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
10802         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
10803         operand if it isn't defined by a vectorized statement.
10804         (vect_build_gather_load_calls): Remove the mask_dt argument and
10805         update calls to vect_get_vec_def_for_stmt_copy.
10806         (vectorizable_bswap): Likewise the dt argument.
10807         (vectorizable_call): Update calls to vectorizable_bswap and
10808         vect_get_vec_def_for_stmt_copy.
10809         (vectorizable_simd_clone_call, vectorizable_assignment)
10810         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
10811         (vectorizable_comparison): Update calls to
10812         vect_get_vec_def_for_stmt_copy.
10813         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
10814         vect_is_simple_use.
10815         (vect_get_loop_based_defs): Remove dt argument and update call
10816         to vect_get_vec_def_for_stmt_copy.
10817         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
10818         and vect_get_vec_def_for_stmt_copy.
10819         (vectorizable_load): Update calls to vect_build_gather_load_calls
10820         and vect_get_vec_def_for_stmt_copy.
10821         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10822         (vectorizable_reduction, vectorizable_live_operation): Update calls
10823         to vect_get_vec_def_for_stmt_copy.
10825 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10827         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
10828         and gimple stmt arguments with a stmt_vec_info.
10829         (vect_record_base_alignments): Update calls accordingly.
10830         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
10831         and gimple stmt arguments with a stmt_vec_info.
10832         (vect_build_slp_tree_1): Remove vinfo argument and update call
10833         to vect_record_max_nunits.
10834         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
10835         and vect_record_max_nunits.
10837 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10839         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
10840         file and take a stmt_vec_info instead of a gimple stmt.
10841         (supportable_widening_operation, vect_finish_replace_stmt)
10842         (vect_finish_stmt_generation, vect_get_store_rhs)
10843         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
10844         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
10845         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
10846         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
10847         (vect_create_data_ref_ptr, bump_vector_ptr)
10848         (vect_permute_store_chain, vect_setup_realignment)
10849         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
10850         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
10851         (vectorizable_reduction, vectorizable_induction)
10852         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
10853         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
10854         than gimple stmts as arguments.
10855         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
10856         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
10857         (can_group_stmts_p, vect_check_gather_scatter)
10858         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10859         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
10860         (vect_permute_load_chain, vect_shift_permute_load_chain)
10861         (vect_transform_grouped_load)
10862         (vect_record_grouped_load_vectors): Likewise.
10863         * tree-vect-loop.c (vect_fixup_reduc_chain)
10864         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
10865         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
10866         (vectorizable_reduction, vectorizable_induction)
10867         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
10868         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
10869         (vect_get_load_store_mask): Likewise.
10870         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
10871         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
10872         * tree-vect-stmts.c (vect_mark_relevant)
10873         (is_simple_and_all_uses_invariant)
10874         (exist_non_indexing_operands_for_use_p, process_use)
10875         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
10876         (vect_get_vec_def_for_operand, vect_get_vec_defs)
10877         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
10878         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
10879         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
10880         (get_negative_load_store_type, get_load_store_type)
10881         (vect_check_load_store_mask, vect_check_store_rhs)
10882         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
10883         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10884         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10885         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10886         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
10887         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
10888         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
10889         (supportable_widening_operation): Likewise.
10891 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10893         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
10894         a stmt_vec_info instead of a gcall.
10895         (vect_check_gather_scatter): Update call accordingly.
10896         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
10897         of a gphi.
10898         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
10899         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
10900         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
10901         instead of a gimple stmt.
10902         (vect_transform_loop): Update calls accordingly.
10903         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
10904         stmt_vec_infos instead of gimple stmts.
10905         (vect_analyze_slp_instance): Update use accordingly.
10906         * tree-vect-stmts.c (read_vector_array, write_vector_array)
10907         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
10908         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
10909         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
10910         (vect_gen_widened_results_half, vect_get_loop_based_defs)
10911         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
10912         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
10913         down to subroutines.
10915 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10917         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
10918         of the worklist from a vector of gimple stmts to a vector of
10919         stmt_vec_infos.
10920         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10921         (vect_mark_stmts_to_be_vectorized): Likewise
10923 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10925         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
10926         statement before passing it to vect_analyze_stmt.
10927         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
10928         the chain of phi vector definitions.  Track the exit phi via its
10929         stmt_vec_info.
10930         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
10931         STMT_VINFO_REDUC_DEF.
10932         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
10933         stmt_vec_infos to handle the statement chains.
10934         (vect_get_slp_defs): Record the first statement in the node
10935         using a stmt_vec_info.
10936         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
10937         statements here and pass their stmt_vec_info down to subroutines.
10938         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
10939         down to vect_finish_stmt_generation.
10940         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
10941         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
10942         stmt_vec_infos to subroutines.
10943         (vect_remove_stores): Use stmt_vec_infos to handle the statement
10944         chains.
10946 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10948         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
10949         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
10950         (vect_permute_store_chain, vect_setup_realignment)
10951         (vect_permute_load_chain, vect_shift_permute_load_chain)
10952         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
10953         stmts internally, and when passing values to other vectorizer routines.
10954         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10955         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
10956         (vect_analyze_loop_operations, get_initial_def_for_reduction)
10957         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
10958         (vectorizable_reduction, vectorizable_induction)
10959         (vectorizable_live_operation, vect_transform_loop_stmt)
10960         (vect_transform_loop): Likewise.
10961         * tree-vect-patterns.c (vect_reassociating_reduction_p)
10962         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
10963         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
10964         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10965         (vect_slp_analyze_node_operations_1): Likewise.
10966         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10967         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
10968         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
10969         (vect_finish_stmt_generation_1, get_group_load_store_type)
10970         (get_load_store_type, vect_build_gather_load_calls)
10971         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10972         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10973         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10974         (vectorizable_store, vectorizable_load, vectorizable_condition)
10975         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
10976         (supportable_widening_operation): Likewise.
10977         (vect_get_vector_types_for_stmt): Likewise.
10978         * tree-vectorizer.h (vect_dr_behavior): Likewise.
10980 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10982         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10983         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
10984         (vect_permute_store_chain, vect_permute_load_chain)
10985         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
10986         repeated stmt_vec_info lookups.
10987         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10988         (vect_update_ivs_after_vectorizer): Likewise.
10989         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
10990         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10991         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
10992         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10993         (vect_bb_slp_scalar_cost): Likewise.
10994         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
10996 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10998         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
10999         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11000         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11001         to get gassigns and gcalls, rather than operating on generc gimple
11002         stmts.
11003         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11004         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11005         (vectorizable_load, vect_analyze_stmt): Likewise.
11006         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11008 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11010         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11011         return stmt_vec_infos rather than gimple stmts.  Do not accept
11012         null arguments.
11013         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11014         of a gimple stmt.
11015         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11016         Update use of get_later_stmt.
11017         (vect_get_constant_vectors): Update call accordingly.
11018         (vect_schedule_slp_instance): Likewise
11019         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11020         (vect_slp_analyze_instance_dependence): Likewise.
11021         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11023 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11025         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11026         (stmt_info_for_cost::stmt_info): ...this new field.
11027         (add_stmt_costs): Update accordingly.
11028         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11029         (vect_get_known_peeling_cost): Likewise.
11030         (vect_estimate_min_profitable_iters): Likewise.
11031         * tree-vect-stmts.c (record_stmt_cost): Likewise.
11033 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11035         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11036         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11037         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11038         accordingly.
11039         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11041 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11043         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11044         a gimple stmt to a stmt_vec_info.
11045         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11047 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11049         * tree-vectorizer.h (vec_info::grouped_stores): Change from
11050         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11051         (_loop_vec_info::reduction_chains): Likewise.
11052         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11053         accordingly.
11054         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11056 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11058         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11059         a gimple stmt to a stmt_vec_info.
11060         (_stmt_vec_info::next_element): Likewise.
11061         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11062         (vect_slp_analyze_and_verify_node_alignment)
11063         (vect_analyze_group_access_1, vect_analyze_group_access)
11064         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11065         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11066         (vect_supportable_dr_alignment): Update accordingly.
11067         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11068         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11069         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
11070         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11071         * tree-vect-slp.c (vect_build_slp_tree_1)
11072         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
11073         (vect_split_slp_store_group, vect_analyze_slp_instance)
11074         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
11075         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
11076         (get_group_load_store_type, get_load_store_type)
11077         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
11078         (vect_transform_stmt, vect_remove_stores): Likewise.
11080 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11082         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
11083         than a gimple stmt.
11084         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11085         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
11086         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
11087         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
11088         (vector_alignment_reachable_p, vect_get_data_access_cost)
11089         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11090         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
11091         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11092         (vect_analyze_group_access_1, vect_analyze_group_access)
11093         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
11094         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
11095         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
11096         result of vect_dr_stmt and use the stmt_vec_info instead of
11097         the associated gimple stmt.
11098         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11099         (vect_gen_prolog_loop_niters): Likewise.
11100         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11102 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11104         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
11105         to a vec<stmt_vec_info>.
11106         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
11107         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
11108         vec<stmt_vec_info>.
11109         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
11110         to a vec<stmt_vec_info>.
11111         (bst_traits::value_type, bst_traits::value_type): Likewise.
11112         (bst_traits::hash): Update accordingly.
11113         (vect_get_and_check_slp_defs): Change the stmts parameter from
11114         a vec<gimple *> to a vec<stmt_vec_info>.
11115         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
11116         (vect_build_slp_tree): Likewise.
11117         (vect_build_slp_tree_2): Likewise.  Update uses of
11118         SLP_TREE_SCALAR_STMTS.
11119         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
11120         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
11121         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
11122         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
11123         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
11124         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
11125         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
11126         (vect_get_constant_vectors, vect_get_slp_defs)
11127         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
11128         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
11129         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
11130         instead of gimple stmts.
11131         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11132         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11133         (vect_slp_analyze_instance_dependence): Update uses of
11134         SLP_TREE_SCALAR_STMTS.
11135         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11136         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11137         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11138         (get_initial_defs_for_reduction): Likewise.
11139         (vect_create_epilog_for_reduction): Likewise.
11140         (vectorize_fold_left_reduction): Likewise.
11141         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11142         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11143         (can_vectorize_live_stmts): Likewise.
11145 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11147         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11148         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11149         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11150         than gimple stmts.
11151         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11152         of a gimple stmt.
11153         (gather_scalar_reductions): Update after above interface changes.
11154         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11155         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11156         than gimple stmts.
11157         (vect_force_simple_reduction): Likewise.
11158         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11159         LOOP_VINFO_REDUCTIONS.
11160         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11162 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11164         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11165         a gimple stmt to a stmt_vec_info.
11166         * tree-vect-loop.c (vect_active_double_reduction_p)
11167         (vect_force_simple_reduction, vectorizable_reduction): Update
11168         accordingly.
11170 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11172         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11173         vec<gimple *> to a vec<stmt_vec_info>.
11174         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11175         the reduction_phis argument from a vec<gimple *> to a
11176         vec<stmt_vec_info>.
11177         (vectorizable_reduction): Likewise the phis local variable that
11178         is passed to vect_create_epilog_for_reduction.  Update for new type
11179         of SLP_TREE_VEC_STMTS.
11180         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11181         (vectorizable_live_operation): Likewise.
11182         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11183         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11185 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11187         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11188         a gimple stmt to a stmt_vec_info.
11189         (vectorizable_condition, vectorizable_live_operation)
11190         (vectorizable_reduction, vectorizable_induction): Pass back the
11191         vectorized statement as a stmt_vec_info.
11192         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11193         use of STMT_VINFO_VEC_STMT.
11194         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11195         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11196         as stmt_vec_infos rather than gimple stmts.
11197         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11198         to a stmt_vec_info.
11199         (vectorizable_live_operation): Likewise.
11200         (vectorizable_reduction, vectorizable_induction): Likewise,
11201         updating use of STMT_VINFO_VEC_STMT.
11202         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11203         of STMT_VINFO_VEC_STMT.
11204         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11205         (vectorizable_simd_clone_call, vectorizable_conversion)
11206         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11207         (vectorizable_store, vectorizable_load, vectorizable_condition)
11208         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11209         from a gimple stmt to a stmt_vec_info.
11210         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
11211         pointer to a stmt_vec_info to the vectorizable_* routines.
11213 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11215         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11216         a gimple stmt to a stmt_vec_info.
11217         (is_pattern_stmt_p): Update accordingly.
11218         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11219         (vect_record_grouped_load_vectors): Likewise.
11220         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11221         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11222         (vect_model_reduction_cost): Likewise.
11223         (vect_create_epilog_for_reduction): Likewise.
11224         (vectorizable_reduction, vectorizable_induction): Likewise.
11225         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11226         Return the stmt_vec_info for the pattern statement.
11227         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11228         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11229         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11230         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11231         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11232         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11233         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11234         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11236 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11238         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11239         (vect_finish_stmt_generation): Likewise.
11240         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11241         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11242         (vect_build_gather_load_calls): Use the return value of the above
11243         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11244         scopes for the input gimple stmt and wider scopes for the associated
11245         stmt_vec_info.  Use vec_info::lookup_def when setting these
11246         stmt_vec_infos from an SSA_NAME definition.
11247         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11248         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11249         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11250         (vectorizable_store, vectorizable_load, vectorizable_condition)
11251         (vectorizable_comparison): Likewise.
11252         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11253         (vectorizable_reduction): Likewise.
11255 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11257         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11258         stmt_vec_info * parameter before the optional gimple **.
11259         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11260         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11261         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11262         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11263         (vectorizable_live_operation): Likewise.
11264         * tree-vect-patterns.c (type_conversion_p): Likewise.
11265         (vect_look_through_possible_promotion): Likewise.
11266         (vect_recog_rotate_pattern): Likewise.
11267         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11269 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11271         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11272         a typedef to a wrapper class.
11273         (NULL_STMT_VEC_INFO): New macro.
11274         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11275         (stmt_vec_info::operator*): New function.
11276         (stmt_vec_info::operator gimple *): Likewise.
11277         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11278         (add_stmt_costs): Likewise.
11279         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11280         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11281         (vect_get_known_peeling_cost): Likewise.
11282         (vect_estimate_min_profitable_iters): Likewise.
11283         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11284         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11285         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11286         (vectorizable_store, free_stmt_vec_infos): Likewise.
11287         (new_stmt_vec_info): Change return type of xcalloc to
11288         _stmt_vec_info *.
11290 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11292         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11293         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11294         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11295         a single_imm_use-based sequence.
11296         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11298 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11300         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11301         * tree-vectorizer.c (vec_info::lookup_def): New function.
11302         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11303         (vect_widened_op_tree): Likewise.
11304         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11305         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11306         (vectorizable_reduction): Likewise.
11307         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11308         of a gimple *.
11309         (vect_is_slp_reduction): Update calls accordingly.  Use
11310         vec_info::lookup_def.
11311         (vect_is_simple_reduction): Likewise
11312         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11314 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11316         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11317         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11318         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11319         of vinfo_for_stmt.
11320         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11321         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11322         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11323         (vect_is_slp_reduction, vectorizable_induction)
11324         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11325         * tree-vect-patterns.c (vect_init_pattern_stmt):
11326         (vect_determine_min_output_precision_1, vect_determine_precisions)
11327         (vect_pattern_recog): Likewise.
11328         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11329         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11330         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11331         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11332         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11333         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11334         info field from a loop to a loop_vec_info.
11336 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11338         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11339         (vec_info::add_stmt): Declare.
11340         * tree-vectorizer.c (vec_info::add_stmt): New function.
11341         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11342         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11343         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11344         (vectorizable_induction): Likewise.
11345         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11346         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11347         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11348         (vectorizable_load): Likewise.
11349         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11350         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11351         (vect_recog_gather_scatter_pattern): Likewise.
11352         (append_pattern_def_seq): Likewise.  Remove a check that is
11353         performed by add_stmt itself.
11355 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11357         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11358         which make_ssa_name was called with new_stmt before new_stmt
11359         had been created.
11361 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11363         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11364         split out from...
11365         (vect_is_slp_reduction): ...here...
11366         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11367         that are already known to be false.
11369 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11371         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11372         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11373         STMT_VINFO_NUM_SLP_USES when it's true.
11374         (vect_free_slp_instance): Add a final_p parameter and pass it to
11375         vect_free_slp_tree.
11376         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11377         (vect_analyze_slp_instance): Likewise.
11378         (vect_slp_analyze_operations): Likewise.
11379         (vect_slp_analyze_bb_1): Likewise.
11380         * tree-vectorizer.c (vec_info): Likewise.
11381         * tree-vect-loop.c (vect_transform_loop): Likewise.
11383 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11385         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11386         function is not called for second and subsequent members of
11387         a reduction group.
11389 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11391         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11392         cases for nested loops from here to ...
11393         (vect_create_epilog_for_reduction): ...here.  Only call
11394         vect_is_simple_use for inner-loop reductions.
11396 2018-07-31  Martin Liska  <mliska@suse.cz>
11398         PR gcov-profile/85338
11399         PR gcov-profile/85350
11400         PR gcov-profile/85372
11401         * profile.c (struct location_triplet): New.
11402         (struct location_triplet_hash): Likewise.
11403         (output_location): Do not output a BB that
11404         is already recorded for a line.
11405         (branch_prob): Use streamed_locations.
11407 2018-07-31  Martin Liska  <mliska@suse.cz>
11409         PR gcov-profile/85370
11410         * coverage.c (coverage_begin_function): Do not mark target
11411         clones as artificial functions.
11413 2018-07-31  Martin Liska  <mliska@suse.cz>
11415         PR gcov-profile/83813
11416         PR gcov-profile/84758
11417         PR gcov-profile/85217
11418         PR gcov-profile/85332
11419         * profile.c (branch_prob): Do not record GOTO expressions
11420         for GIMPLE statements which locations are already streamed.
11422 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11424         * gcc.c (handle_spec_function): Accept a soft_matched_part
11425         argument, as do_spec_1.  Pass it down to ...
11426         (eval_spec_function): Accept a soft_matched_part argument,
11427         and pass it down to ...
11428         (do_spec_2): Accept a soft_matched_part argument, and pass
11429         it down to do_spec_1.
11430         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11431         (handle_braces): Update call to handle_spec_function.
11432         (driver::set_up_specs): Update calls to do_spec_2.
11433         (compare_debug_dump_opt_spec_function): Likewise.
11434         (compare_debug_self_opt_spec_function): Likewise.
11436 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11438         * common.opt (nolibc): New option.
11439         * doc/invoke.texi (Link Options): Document it.
11440         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11441         * config/alpha/linux.h: Likewise.
11442         * config/arc/elf.h: Likewise.
11443         * config/arm/uclinux-elf.h: Likewise.
11444         * config/arm/unknown-elf.h: Likewise.
11445         * config/avr/avrlibc.h: Likewise.
11446         * config/bfin/bfin.h: Likewise.
11447         * config/bfin/linux.h: Likewise.
11448         * config/bfin/uclinux.h: Likewise.
11449         * config/darwin.h: Likewise.
11450         * config/darwin10.h: Likewise.
11451         * config/darwin12.h: Likewise.
11452         * config/gnu-user.h: Likewise.
11453         * config/lm32/uclinux-elf.h: Likewise.
11454         * config/pa/pa-hpux11.h: Likewise.
11455         * config/pa/pa64-hpux.h: Likewise.
11456         * config/sparc/sparc.h: Likewise.
11458 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11460         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11461         undefined variables.
11463 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11465         PR target/86640
11466         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11467         instead of GEN_INT.
11469 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11471         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11472         terminated string literal.
11474 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11476         PR rtl-optimization/85160
11477         * combine.c (is_just_move): New function.
11478         (try_combine): Allow combining two instructions into two if neither of
11479         the original instructions was a move.
11481 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11483         PR target/86673
11484         * doc/extend.texi (Global Register Variables): Discourage use of type
11485         qualifiers.
11486         (Local Register Variables): Likewise.
11488 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11490         PR tree-optimization/86506
11491         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11493 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11495         PR target/86547
11496         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11497         hard_regno, make sure no insn between `from` and `to` clobbers it.
11499 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11500             Tom de Vries  <tdevries@suse.de>
11502         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11503         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11504         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11505         PTX_DEFAULT_RUNTIME_DIM.
11506         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11508 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11510         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11511         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11512         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11514 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11516         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11517         used.
11518         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11519         predicate to reflect toc fusion has been deleted.
11520         (toc_fusion_mem_raw): Delete, no longer used.
11521         (toc_fusion_mem_wrapped): Likewise.
11522         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11523         fusion mask bit.
11524         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11525         Delete, no longer used.
11526         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11527         meant to be used for toc fusion.
11528         (rs6000_debug_print_mode): Delete toc fusion debugging.
11529         (rs6000_debug_reg_global): Likewise.
11530         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11531         fusion and secondary reload support that were never used.
11532         (rs6000_option_override_internal): Delete TOC fusion, that was only
11533         partially defined, and it did not work unless you also used the
11534         -mcmodel= switch.
11535         (rs6000_legitimate_address_p): Delete TOC fusion support.
11536         (rs6000_opt_masks): Likewise.
11537         (fusion_wrap_memory_address): Delete function, no longer used.
11538         (fusion_split_address); Delete TOC fusion support.
11539         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11540         longer used with toc fusion being deleted.
11541         (TARGET_TOC_FUSION_FP): Likewise.
11542         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11543         UNSPEC.
11544         (toc fusion spliter): Delete TOC fusion support.
11545         (toc_fusionload_<mode>): Likewise.
11546         (toc_fusionload_di): Likewise.
11547         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11548         longer needs to be named.  Rename predicate to delete TOC fusion.
11549         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11550         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11551         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11552         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11553         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11555 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11557         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11558         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11559         __int128 in built-in function prototypes.
11560         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11561         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11563 2018-07-27  Martin Sebor  <msebor@redhat.com>
11565         PR tree-optimization/86696
11566         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11567         types, including enums.
11568         (handle_char_store): Be prepared for the above function to fail.
11570 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11572         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11573         when optimization level is lower than 2 or optimize for size.
11575 2018-07-26  Martin Sebor  <msebor@redhat.com>
11577         PR tree-optimization/86043
11578         PR tree-optimization/86042
11579         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11580         (get_string_cst_length): Rename...
11581         (get_min_string_length): ...to this.  Add argument.
11582         (handle_char_store): Extend to handle multi-character stores by
11583         MEM_REF.
11584         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11585         * tree.h (initializer_zerop): Add argument.
11587 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11589         PR middle-end/86660
11590         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11591         declare target to variables if they have always,{to,from,tofrom} map
11592         kinds.
11594 2018-07-26  Martin Liska  <mliska@suse.cz>
11596         PR lto/86548
11597         * lto-wrapper.c: Add linker_output as prefix
11598         for ltrans_output_file.
11600 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11602         PR rtl-optimization/85805
11603         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11604         value for hard registers if that was written in the same mode.
11606 2018-07-26  Martin Liska  <mliska@suse.cz>
11608         PR gcov-profile/86536
11609         * gcov.c (format_gcov): Use printf format %.*f directly
11610         and do not handle special values.
11612 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11614         * common/config/arc/arc-common.c (arc_option_optimization_table):
11615         Update default optimizations for size.
11617 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11619         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11620         * config/arc/constraints.md (CfZ): New constraint.
11621         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11622         (subsf3_fpu): Likewise.
11623         (cmpsf_fpu): Likewise.
11624         (cmpsf_fpu_uneq): Likewise.
11626 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11628         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11629         accesses as well.
11630         (arc_is_uncached_mem_p): uncached applies to both the variable and
11631         the pointer.
11633 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11635         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11636         register names.
11638 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11640         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11641         field "m_scopes" from vec to auto_vec.
11643 2018-07-25  Martin Liska  <mliska@suse.cz>
11645         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11646         return type.
11648 2018-07-25  Richard Biener  <rguenther@suse.de>
11650         PR debug/86654
11651         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11652         special wrt context_die late.
11653         (gen_subprogram_die): Re-use DIEs in local scope.
11655 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11657         PR tree-optimization/86644
11658         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11660 2018-07-25  Martin Liska  <mliska@suse.cz>
11662         PR middle-end/86645
11663         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11664         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11666 2018-07-25  Martin Liska  <mliska@suse.cz>
11668         PR sanitizer/79635
11669         * params.def: Explain ASan abbreviation and provide
11670         a documentation link.
11672 2018-07-24  Martin Sebor  <msebor@redhat.com>
11674         PR tree-optimization/86622
11675         PR tree-optimization/86532
11676         * builtins.h (string_length): Declare.
11677         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11678         (check_access): Be prepared for non-constant length ranges.
11679         (string_length): Make extern.
11680         * expr.c (string_constant): Only handle the minor non-constant
11681         array index.  Use string_constant to compute the length of
11682         a generic string constant.
11684 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11686         PR tree-optimization/86618
11687         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11688         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11690 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11692         PR tree-optimization/86636
11693         * json.cc (json::object::set): Fix comment.  Add assertions.
11694         (json::array::append): Move here from json.h.  Add comment and an
11695         assertion.
11696         (json::string::string): Likewise.
11697         * json.h (json::array::append): Move to json.cc.
11698         (json::string::string): Likewise.
11699         * optinfo-emit-json.cc
11700         (optrecord_json_writer::impl_location_to_json): Assert that we
11701         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11702         wrapper around it.  Expand the location once, rather than three
11703         times.
11704         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11705         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11706         wrappers.
11707         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11708         places.  Fix some overlong lines.
11710 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11712         * config/aarch64/aarch64-simd.md
11713         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11714         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11715         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11716         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11717         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11718         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11719         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11720         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
11721         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
11723 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
11725         PR middle-end/86627
11726         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
11727         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
11728         and abs_d == d, do the power of two handling if profitable.
11730 2018-07-24  Richard Biener  <rguenther@suse.de>
11732         * match.pd: Add BIT_FIELD_REF canonicalizations.
11734 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11736         PR c/86617
11737         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
11739 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11741         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
11742         terminated STRING_CST object.
11744 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11746         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
11748 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11750         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
11751         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
11752         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
11753         the elements into a register.
11754         (rs6000_split_v4si_init_di_reg): Delete.
11755         (rs6000_split_v4si_init): Delete.
11756         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
11757         (vsx_init_v4si): Rewrite as a define_expand.
11759 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11761         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
11762         zero_extend argument from memory): New.
11764 2018-07-22  Martin Sebor  <msebor@redhat.com>
11766         PR bootstrap/86621
11767         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
11768         diagnosing calls with unknown arguments unless -Walloca-larger-than
11769         is restricted to less than PTRDIFF_MAX bytes.
11771 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
11773         * doc/gcov.texi (Invoking Gcov): Editorial changes.
11775 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11777         * pretty-print.c (text_info::set_location): Remove redundant
11778         "line_table" parameter from call to rich_location::set_range.
11780 2018-07-20  Martin Sebor  <msebor@redhat.com>
11782         PR middle-end/82063
11783         * builtins.c (expand_builtin_alloca): Adjust.
11784         * calls.c (alloc_max_size): Simplify.
11785         * cgraphunit.c (cgraph_node::expand): Adjust.
11786         * common.opt (larger_than_size, warn_frame_larger_than): Remove
11787         variables.
11788         (frame_larger_than_size): Same.
11789         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
11790         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
11791         Initialize.
11792         * doc/invoke.texi (GCC Command Options): Document option arguments.
11793         Explain byte-size arguments and suffixes.
11794         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
11795         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
11796         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
11797         * doc/options.texi (UInteger): Expand.
11798         (Host_Wide_Int, ByteSize): Document new properties.
11799         * final.c (final_start_function_1): Include sizes in an error message.
11800         * function.c (frame_offset_overflow): Same.
11801         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
11802         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
11803         Diagnose unbounded alloca calls only for limits of less than
11804         PTRDIFF_MAX.
11805         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
11806         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
11807         for alloca(0).
11808         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
11809         only for limits of less than PTRDIFF_MAX.
11810         * langhooks-def.h (lhd_handle_option): Change function argument
11811         to HOST_WIDE_INT.
11812         * langhooks.c (lhd_handle_option): Same.
11813         * langhooks.h (handle_option): Same.
11814         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
11815         ByteSize flags.
11816         (var_type, var_type_struct): Same.
11817         (var_set): Handle ByteSize flag.
11818         * optc-gen.awk: Add comments to output to ease debugging.  Make
11819         use of HOST_WIDE_INT where appropriate.
11820         * opts-gen-save.awk:  Use %lx to format unsigned long.
11821         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
11822         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
11823         arguments.  Parse bytes-size suffixes.
11824         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
11825         (enum_value_to_arg): Same.
11826         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
11827         (handle_option): Adjust.
11828         (generate_option): Change function argument to HOST_WIDE_INT.
11829         (cmdline_handle_error): Adjust.
11830         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
11831         (set_option): Change function argument to HOST_WIDE_INT.
11832         (option_enabled): Handle cl_host_wide_int.
11833         (get_option_state): Handle CLVC_SIZE.
11834         (control_warning_option): Same.
11835         * opts.c (common_handle_option): Change function argument to
11836         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
11837         OPT_Wvla_larger_than_.
11838         * opts.h (enum cl_var_type): Add an enumerator.
11839         * stor-layout.c (layout_decl): Print a more meaningful warning.
11840         * toplev.c (output_stack_usage): Adjust.
11842 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
11844         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
11845         call to inline_expand_builtin_string_cmp.
11846         (expand_builtin_strcmp): Likewise.
11847         (expand_builtin_strncmp): Likewise.
11848         (inline_string_cmp): Delete the last parameter, change char_type_node
11849         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
11850         two operands.
11851         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
11852         the inlining expansion on target where the type of the call has same or
11853         narrower precision than unsigned char.
11855 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11857         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
11858         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
11859         * common.opt (fsave-optimization-record): New option.
11860         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
11861         * doc/invoke.texi (-fsave-optimization-record): New option.
11862         * dumpfile.c: Include "optinfo-emit-json.h".
11863         (struct kv_pair): Move to coretypes.h.
11864         (optgroup_options): Make non-static.
11865         (dump_context::end_scope): Call
11866         optimization_records_maybe_pop_dump_scope.
11867         * dumpfile.h (optgroup_options): New decl.
11868         * json.cc: New file.
11869         * json.h: New file.
11870         * optinfo-emit-json.cc: New file.
11871         * optinfo-emit-json.h: New file.
11872         * optinfo.cc: Include "optinfo-emit-json.h".
11873         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
11874         (optinfo_enabled_p): Check optimization_records_enabled_p.
11875         (optinfo_wants_inlining_info_p): Likewise.
11876         * optinfo.h: Update comment.
11877         * profile-count.c (profile_quality_as_string): New function.
11878         * profile-count.h (profile_quality_as_string): New decl.
11879         (profile_count::quality): New accessor.
11880         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
11881         and optinfo_emit_json_cc_tests.
11882         * selftest.h (selftest::json_cc_tests): New decl.
11883         (selftest::optinfo_emit_json_cc_tests): New decl.
11884         * toplev.c: Include "optinfo-emit-json.h".
11885         (compile_file): Call optimization_records_finish.
11886         (do_compile): Call optimization_records_start.
11887         * tree-ssa-live.c: Include optinfo.h.
11888         (remove_unused_scope_block_p): Retain inlining information if
11889         optinfo_wants_inlining_info_p returns true.
11891 2018-07-20  Richard Biener  <rguenther@suse.de>
11893         PR debug/86585
11894         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
11895         to cover -flto-partition=none.
11897 2018-07-20  Martin Liska  <mliska@suse.cz>
11899         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
11900         (get_decl_source_range): Remove unused function.
11902 2018-07-20  Richard Biener  <rguenther@suse.de>
11904         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
11905         (struct vn_phi_s): Likewise.
11906         (struct vn_reference_s): Likewise.
11907         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
11908         for searching the slot of an entry known to be in the hash itself.
11909         (vn_phi_hasher::equal): Likewise.
11910         (vn_reference_hasher::equal): Likewise.
11911         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
11912         globals.
11913         (optimistic_info, current_info): Remove, keeping only valid_info.
11914         (vn_reference_lookup_1): Remove fallback lookup.
11915         (vn_reference_lookup_2): Likewise.
11916         (vn_nary_op_lookup_1): Likewise.
11917         (vn_phi_lookup): Likewise.
11918         (vn_nary_build_or_lookup_1): Make sure to not chain the built
11919         hash element.
11920         (vn_reference_insert): Adjust, chain the inserted hash element
11921         at last_inserted_ref.
11922         (vn_reference_insert_pieces): Likewise.
11923         (visit_reference_op_call): Likewise.
11924         (vn_nary_op_insert_into): Chain the inserted hash element at
11925         last_inserted_nary.
11926         (vn_nary_op_insert_pieces): Adjust.
11927         (vn_nary_op_insert): Likewise.
11928         (vn_nary_op_insert_stmt): Likewise.
11929         (vn_phi_insert): Adjust, chain the inserted hash element at
11930         last_inserted_phi.
11931         (process_scc): Remove clearing and copying the optimistic
11932         table.  Instead remove elements inserted during an optimistic
11933         iteration from the single table we maintain.
11934         (init_scc_vn): Adjust.
11935         (free_scc_vn): Likewise.
11936         (sccvn_dom_walker::record_cond): Likewise.
11937         (sccvn_dom_walker::after_dom_children): Likewise.
11939 2018-07-19  Martin Sebor  <msebor@redhat.com>
11941         PR tree-optimization/84047
11942         PR tree-optimization/83776
11943         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
11944         (check_array_bounds): Call it.
11946 2018-07-19  Martin Sebor  <msebor@redhat.com>
11948         * align.h (align_flags): Use member initialization.
11950 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
11952         * Makefile.in (OBJS): Add optinfo.o.
11953         * coretypes.h (class symtab_node): New forward decl.
11954         (struct cgraph_node): New forward decl.
11955         (class varpool_node): New forward decl.
11956         * dump-context.h: New file.
11957         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
11958         "tree-pass.h".
11959         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
11960         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
11961         (set_alt_dump_file): Likewise.
11962         (dump_context::~dump_context): New dtor.
11963         (dump_gimple_stmt): Move implementation to...
11964         (dump_context::dump_gimple_stmt): ...this new member function.
11965         Add the stmt to any pending optinfo, creating one if need be.
11966         (dump_gimple_stmt_loc): Move implementation to...
11967         (dump_context::dump_gimple_stmt_loc): ...this new member function.
11968         Start a new optinfo and add the stmt to it.
11969         (dump_gimple_expr): Move implementation to...
11970         (dump_context::dump_gimple_expr): ...this new member function.
11971         Add the stmt to any pending optinfo, creating one if need be.
11972         (dump_gimple_expr_loc): Move implementation to...
11973         (dump_context::dump_gimple_expr_loc): ...this new member function.
11974         Start a new optinfo and add the stmt to it.
11975         (dump_generic_expr): Move implementation to...
11976         (dump_context::dump_generic_expr): ...this new member function.
11977         Add the tree to any pending optinfo, creating one if need be.
11978         (dump_generic_expr_loc): Move implementation to...
11979         (dump_context::dump_generic_expr_loc): ...this new member
11980         function.  Add the tree to any pending optinfo, creating one if
11981         need be.
11982         (dump_printf): Move implementation to...
11983         (dump_context::dump_printf_va): ...this new member function.  Add
11984         the text to any pending optinfo, creating one if need be.
11985         (dump_printf_loc): Move implementation to...
11986         (dump_context::dump_printf_loc_va): ...this new member function.
11987         Start a new optinfo and add the stmt to it.
11988         (dump_dec): Move implementation to...
11989         (dump_context::dump_dec): ...this new member function.  Add the
11990         value to any pending optinfo, creating one if need be.
11991         (dump_context::dump_symtab_node): New member function.
11992         (dump_context::get_scope_depth): New member function.
11993         (dump_context::begin_scope): New member function.
11994         (dump_context::end_scope): New member function.
11995         (dump_context::ensure_pending_optinfo): New member function.
11996         (dump_context::begin_next_optinfo): New member function.
11997         (dump_context::end_any_optinfo): New member function.
11998         (dump_context::s_current): New global.
11999         (dump_context::s_default): New global.
12000         (dump_scope_depth): Delete global.
12001         (dumpfile_ensure_any_optinfo_are_flushed): New function.
12002         (dump_symtab_node): New function.
12003         (get_dump_scope_depth): Reimplement in terms of dump_context.
12004         (dump_begin_scope): Likewise.
12005         (dump_end_scope): Likewise.
12006         (selftest::temp_dump_context::temp_dump_context): New ctor.
12007         (selftest::temp_dump_context::~temp_dump_context): New dtor.
12008         (selftest::verify_item): New function.
12009         (ASSERT_IS_TEXT): New macro.
12010         (ASSERT_IS_TREE): New macro.
12011         (ASSERT_IS_GIMPLE): New macro.
12012         (selftest::test_capture_of_dump_calls): New test.
12013         (selftest::dumpfile_c_tests): Call it.
12014         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12015         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12016         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12017         descriptive comment.
12018         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12019         (dump_node, dump_bb): Move these unrelated decls.
12020         (class dump_manager): Add leading comment.
12021         * optinfo.cc: New file.
12022         * optinfo.h: New file.
12024 2018-07-19  Michael Collison  <michael.collison@arm.com>
12025             Richard Henderson <rth@redhat.com>
12027         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12028         (subti): Handle op1 zero.
12029         (subvti4, usub4ti4): New.
12030         (*sub<GPI>3_compare1_imm): New.
12031         (sub<GPI>3_carryinCV): New.
12032         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12033         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12035 2018-07-19  Michael Collison  <michael.collison@arm.com>
12036             Richard Henderson <rth@redhat.com>
12038         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12039         (addti3): Create simpler code if low part is already known to be 0.
12040         (addvti4, uaddvti4): New.
12041         (*add<GPI>3_compareC_cconly_imm): New.
12042         (*add<GPI>3_compareC_cconly): New.
12043         (*add<GPI>3_compareC_imm): New.
12044         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12045         handle constants within this pattern..
12046         (*add<GPI>3_compareV_cconly_imm): New.
12047         (*add<GPI>3_compareV_cconly): New.
12048         (*add<GPI>3_compareV_imm): New.
12049         (add<GPI>3_compareV): New.
12050         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12051         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12052         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12053         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12054         with 'comparison' operator.
12055         (*add<GPI>3_compareV_cconly_imm): Ditto.
12056         (*add<GPI>3_compareV_cconly): Ditto.
12057         (*add<GPI>3_compareV_imm): Ditto.
12058         (add<GPI>3_compareV): Ditto.
12059         (add<mode>3_carryinC): Ditto.
12060         (*add<mode>3_carryinC_zero): Ditto.
12061         (*add<mode>3_carryinC): Ditto.
12062         (add<mode>3_carryinV): Ditto.
12063         (*add<mode>3_carryinV_zero): Ditto.
12064         (*add<mode>3_carryinV): Ditto.
12066 2018-07-19  Michael Collison  <michael.collison@arm.com>
12067             Richard Henderson <rth@redhat.com>
12069         * config/aarch64/aarch64-modes.def (CC_V): New.
12070         * config/aarch64/aarch64-protos.h
12071         (aarch64_addti_scratch_regs): Declare
12072         (aarch64_subvti_scratch_regs): Declare.
12073         (aarch64_expand_subvti): Declare.
12074         (aarch64_gen_unlikely_cbranch): Declare
12075         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
12076         for signed overflow using CC_Vmode.
12077         (aarch64_get_condition_code_1): Handle CC_Vmode.
12078         (aarch64_gen_unlikely_cbranch): New function.
12079         (aarch64_addti_scratch_regs): New function.
12080         (aarch64_subvti_scratch_regs): New function.
12081         (aarch64_expand_subvti): New function.
12083 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12085         * config/aarch64/aarch64-option-extensions.def: New entry for profile
12086         extension.
12087         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
12088         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
12089         extension.
12091 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12093         PR target/83009
12094         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12095         address check not strict.
12097 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12099         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
12100         Umq with Umn.
12101         (store_pair_lanes<mode>): Likewise.
12102         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
12103         enum value 'ADDR_QUERY_LDP_STP_N'.
12104         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
12105         (aarch64_print_address_internal): Add declaration.
12106         (aarch64_print_ldpstp_address): Remove.
12107         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
12108         (aarch64_print_operand): Change printing of 'y'.
12109         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
12110         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
12111         'true' rather than '1'.
12112         * config/aarch64/constraints.md (Uml): Likewise.
12113         (Uml): Rename to Umn.
12114         (Umq): Remove.
12116 2018-07-19  Richard Biener  <rguenther@suse.de>
12118         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
12119         a trailing array.
12120         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
12121         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
12122         (vn_reference_hasher): Likewise.
12123         (struct vn_tables_s): Remove obstack and alloc-pool members.
12124         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
12125         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
12126         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
12127         (vn_reference_insert_pieces): Likewise.
12128         (alloc_vn_nary_op_noinit): Adjust.
12129         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
12130         (vn_phi_eq): Adjust.
12131         (shared_lookup_phiargs): Remove.
12132         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12133         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12134         (visit_reference_op_call): Likewise.
12135         (copy_nary, copy_phi, copy_reference): Remove.
12136         (process_scc): Rewind the obstack when iterating.  Do not
12137         copy the elements to valid_info but just move them from one
12138         hashtable to the other.
12139         (allocate_vn_table): Adjust.
12140         (free_vn_table): Likewise.
12141         (init_scc_vn): Likewise.
12142         (free_scc_vn): Likewise.
12144 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
12146         PR target/86560
12147         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12148         indirect_return as function type attribute.
12149         (ix86_attribute_table): Change indirect_return to function
12150         type attribute.
12151         * doc/extend.texi: Update indirect_return attribute.
12153 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
12155         * wide-int.h (widest2_int): New.
12156         * gimple-fold.c (arith_overflowed_p): Use it.
12157         * tree.h (widest2_int_cst): New.
12158         * tree-vrp.c (wide_int_binop_overflow): Rename from
12159         vrp_int_const_binop.
12160         Rewrite to work on trees.
12161         (extract_range_from_multiplicative_op_1): Abstract code to...
12162         (wide_int_range_min_max): ...here.
12163         (wide_int_range_cross_product): ...and here.
12164         (extract_range_from_binary_expr_1): Abstract overflow code to...
12165         (wide_int_range_mult_wrapping): ...here.
12166         * tree-vrp.h (wide_int_range_cross_product): New.
12167         (wide_int_range_mult_wrapping): New.
12169 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
12170             Julia Koval  <julia.koval@intel.com>
12172         * config/i386/x86-tune-costs.h (skylake_memcpy,
12173         skylake_memset): Replace rep_prefix with unrolling for size 512.
12175 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
12177         PR middle-end/86544
12178         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12179         comparision with EQ_EXPR in last stmt.
12181 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12183         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12184         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12185         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12186         previously known as "PowerPC AltiVec Built-in Functions".  Move
12187         some material to new subsubsections "PowerPC AltiVec Built-in
12188         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12189         ISA 2.07".
12190         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12191         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12192         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12193         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12195 2018-07-18  Richard Biener  <rguenther@suse.de>
12197         PR tree-optimization/86557
12198         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12199         EXACT_DIV_EXPR.
12201 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
12203         * config/s390/s390.c (s390_function_profiler): Generate CFI.
12205 2018-07-17  Jeff Law  <law@redhat.com>
12207         * config/arm/arm.c (get_label_padding): Update for recent
12208         changes to label_to_alignment.
12210         PR tree-optimization/86010
12211         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12213         * config/mips/mips.c (vr4130_align_insns): Update for recent
12214         changes to label_to_alignment.
12216         * config/frv/frv.c (frv_label_align): Update for recent changes
12217         to label_to_alignment.
12219         * config/nios2/nios2.c (nios2_label_align): Update for recent
12220         changes which dropped ALIGN_LABELS_LOG.
12222 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12224         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12225         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12227 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12229         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12230         deprecated align_labels_log.
12232 2018-07-17  Richard Biener  <rguenther@suse.de>
12234         PR lto/86456
12235         * dwarf2out.c (init_sections_and_labels): Always generate
12236         a debug_line_str_section for early LTO debug.
12237         (dwarf2out_finish): Reset debug_line_str_hash output early.
12238         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12239         with fat LTO part.
12240         (dwarf2out_early_finish): Output debug_line_str.
12242 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12244         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12245         index register on z196 or later.
12247 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12249         * config/s390/s390.c (s390_default_align): Set default function
12250         alignment to 16.
12251         (s390_override_options_after_change): Call s390_default align.
12252         (s390_option_override_internal): Call s390_default_align.
12253         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12255 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12257         PR middle-end/86542
12258         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12259         to _looptemp_ clauses, other than the first two.
12261 2018-07-17  Martin Liska  <mliska@suse.cz>
12263         * opts.c: Do not enable OPT_falign_* for -Os.
12265 2018-07-17  Martin Liska  <mliska@suse.cz>
12267         * align.h (MAX_CODE_ALIGN): New.
12268         (MAX_CODE_ALIGN_VALUE): New.
12269         * common/config/i386/i386-common.c (ix86_handle_option):
12270         (MAX_CODE_ALIGN): Moved to align.h.
12271         * final.c (MAX_CODE_ALIGN): Likewise.
12272         * opts.c (parse_and_check_align_values):
12273         (MAX_CODE_ALIGN): Likewise.
12274         (MAX_CODE_ALIGN_VALUE): Likewise.
12276 2018-07-17  Martin Liska  <mliska@suse.cz>
12278         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12279         in order to fulfil coding style.
12280         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12281         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12282         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12283         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12284         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12285         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12286         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12287         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12288         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12290 2018-07-17  Martin Liska  <mliska@suse.cz>
12292         * align.h: New file.
12293         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12294         directly.
12295         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12296         align_flags of label_to_alignment.
12297         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12298         align_flags class.
12299         * config/m68k/m68k.c: Do not use removed align_labels_value and
12300         align_loops_value.
12301         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12302         (LOOP_ALIGN): Likewise.
12303         (LABEL_ALIGN): Likewise.
12304         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12305         Remove not used macro.
12306         (rs6000_loop_align): Change return type to align_flags.
12307         (rs6000_loop_align_max_skip): Remove.
12308         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12309         Change return type to align_flags.
12310         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12311         Remove not used macro.
12312         (rs6000_loop_align):  Change return type to align_flags.
12313         (rs6000_loop_align_max_skip): Remove.
12314         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12315         * config/rx/rx-protos.h (rx_align_for_label): Make it
12316         static function.
12317         * config/rx/rx.c (rx_align_for_label): Change return type
12318         to align_flags.
12319         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12320         macro definitions.
12321         into align_flags class.
12322         (LABEL_ALIGN): Likewise.
12323         (LOOP_ALIGN): Likewise.
12324         * config/s390/s390.c (s390_label_align): Use align_flags
12325         class member.
12326         (s390_asm_output_function_label): Likewise.
12327         * config/sh/sh.c (sh_override_options_after_change):
12328         Use align_flags class directly without macros.
12329         (find_barrier): Likewise.
12330         (barrier_align): Likewise.
12331         (sh_loop_align): Likewise.
12332         * config/spu/spu.c (spu_option_override):
12333         Use align_flags_tuple::get_value instead of removed macros.
12334         (spu_sched_init): Likewise.
12335         * config/spu/spu.h (GTY): Likewise.
12336         * config/visium/visium.c (visium_option_override):
12337         Set "8" as default secondary alignment.
12338         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12339         in order to guarantee secondary alignment of 8.
12340         * coretypes.h: Include align.h header file.
12341         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12342         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12343         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12344         * doc/tm.texi.in: Likewise.
12345         * final.c (struct label_alignment): Remove not used structure.
12346         (LABEL_ALIGN): Change type to align_flags.
12347         (LOOP_ALIGN): Likewise.
12348         (JUMP_ALIGN): Likewise.
12349         (default_loop_align_max_skip): Remove.
12350         (default_label_align_max_skip): Likewise.
12351         (default_jump_align_max_skip): Likewise.
12352         (default_label_align_after_barrier_max_skip):
12353         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12354         (LABEL_TO_MAX_SKIP): Remove.
12355         (label_to_alignment): Return align_flags type instead of integer.
12356         (label_to_max_skip): Remove.
12357         (align_fuzz): Use align_flags type.
12358         (compute_alignments): Use align_flags type and use align_flags::max
12359         to combine multiple alignments.
12360         (grow_label_align): Grow vec instead of C array.
12361         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12362         (shorten_branches):  Use align_flags type and use align_flags::max
12363         to combine multiple alignments.
12364         (final_scan_insn_1): Remove usage of secondary alignment that comes
12365         from label alignment, but instead use proper secondary alignment
12366         which is computed in grow_label_align.
12367         * flags.h (struct align_flags_tuple): Move to align.h.
12368         (struct align_flags): Likewise.
12369         (state_align_loops): Rename to align_loops.
12370         (state_align_jumps): Rename to align_jumps.
12371         (state_align_labels): Rename to align_labels.
12372         (state_align_functions): Rename to align_functions.
12373         (align_loops_log): Remove.
12374         (align_jumps_log): Remove.
12375         (align_labels_log): Remove.
12376         (align_functions_log): Remove.
12377         (align_loops_max_skip): Remove.
12378         (align_jumps_max_skip): Remove.
12379         (align_labels_max_skip): Remove.
12380         (align_functions_max_skip): Remove.
12381         (align_loops_value): Remove.
12382         (align_jumps_value): Remove.
12383         (align_labels_value): Remove.
12384         (align_functions_value): Remove.
12385         * output.h (label_to_alignment): Change return type to align_flags.
12386         (label_to_max_skip): Remove.
12387         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12388         jump_align_max_skip macros.
12389         * targhooks.h (default_loop_align_max_skip): Remove.
12390         (default_label_align_max_skip): Likewise.
12391         (default_jump_align_max_skip): Likewise.
12392         (default_label_align_after_barrier_max_skip): Remove.
12393         * toplev.c (read_log_maxskip): Use ::normalize function.
12394         (parse_N_M): Remove not used argument and also call ::normalize.
12395         (parse_alignment_opts): Do not pass unused arguments.
12396         * varasm.c (assemble_start_function): Use directly align_functions
12397         instead of removed macros.
12398         * system.h: Do not poison removed macros.
12400 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12402         PR middle-end/86539
12403         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12404         and cond temporaries don't have reference type if iterator has
12405         pointer type.  For init use &for_pre_body instead of pre_p if
12406         for_pre_body is non-empty.
12408 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12410         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12411         double-double modes to SFmode directly directly.
12412         (trunc<mode>sf2_fprs): Delete.
12414 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12416         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12417         for conversions between IFmode and the decimal floating point modes.
12418         (init_float128_ieee): Use the correct names for conversions between
12419         KFmode and the decimal floating point modes.
12421 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12423         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12424         for the conversions between TDmode and IFmode.
12425         (init_float128_ieee): Use more correct names for the conversions
12426         between TDmode and KFmode.
12428 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12430         PR tree-optimization/86526
12431         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12432         (inline_expand_builtin_string_cmp): Likewise.
12433         (inline_string_cmp): Likewise.  Use c_readstr instead of
12434         builtin_memcpy_read_str.  Add unit_mode temporary.
12436 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12438         PR middle-end/86528
12439         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12440         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12442 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12444         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12445         Alphabetize prototypes of built-in functions, separating out
12446         built-in functions that are listed in this section but should be
12447         described elsewhere.
12449 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12451         PR target/86511
12452         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12453         conditional move when trapping comparison was split to a
12454         non-trapping one (and vice versa).
12456 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12458         * config/s390/s390.c (s390_function_profiler): Generate nops
12459         instead of profiler call sequences.
12460         * config/s390/s390.opt: Add the new option.
12462 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12464         * config/s390/s390.c (s390_function_profiler): Generate
12465         __mcount_loc section.
12466         * config/s390/s390.opt: Add the new option.
12468 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12470         * common.opt: Add the new warning.
12471         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12472         %r0,__fentry__" when -mfentry is specified.
12473         (s390_option_override_internal): Disallow -mfentry for 31-bit
12474         CPUs.
12475         * config/s390/s390.opt: Add the new option.
12477 2018-07-16  Richard Biener  <rguenther@suse.de>
12479         PR lto/86523
12480         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12481         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12483 2018-07-16  Martin Liska  <mliska@suse.cz>
12485         PR ipa/86529
12486         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12487         to ::get_create.
12489 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12491         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12493 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12495         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12496         for ARCHS4x.
12497         * config/arc/arc-cpus.def (hs4x): New cpu.
12498         (hs4xd): Likewise.
12499         * config/arc/arc-tables.opt: Regenerate.
12500         * config/arc/arc.c (arc_sched_issue_rate): New function.
12501         (TARGET_SCHED_ISSUE_RATE): Define.
12502         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12503         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12504         fpu_cvt.
12505         (attr tune): Add ARCHS4x tune values.
12506         (attr tune_dspmpy): Define.
12507         (*tst): Correct instruction type.
12508         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12509         * config/arc/arcHS4x.md: New file.
12510         * config/arc/fpu.md: Update instruction type attributes.
12511         * config/arc/t-multilib: Regenerate.
12513 2018-07-16  Tom de Vries  <tdevries@suse.de>
12515         PR debug/86455
12516         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12518 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12520         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12521         early clobber.
12523 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12525         PR tree-optimization/86514
12526         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12527         conversion to a boolean type from a type with greater precision.
12529 2018-07-16  Tom de Vries  <tdevries@suse.de>
12531         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12532         dump_flags request TDF_SLIM.
12534 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12536         * fold-const.c (int_const_binop_1): Abstract...
12537         (wide_int_binop): ...wide int code here.
12538         (poly_int_binop): ...poly int code here.
12539         Abstract the rest of int_const_binop_1 into int_const_binop.
12540         * fold-const.h (wide_int_binop): New.
12541         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12542         Remove useless PLUS/MINUS_EXPR case.
12543         (zero_nonzero_bits_from_vr): Move wide int code...
12544         (zero_nonzero_bits_from_bounds): ...here.
12545         (extract_range_from_binary_expr_1): Move mask optimization code...
12546         (range_easy_mask_min_max): ...here.
12547         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12548         (range_easy_mask_min_max): New.
12550 2018-07-15  Jeff Law  <law@redhat.com>
12552         PR target/85993
12553         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12554         block.
12556 2018-07-14  Jim Wilson  <jimw@sifive.com>
12558         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12560 2018-07-14  Paul Koning  <ni1d@arrl.net>
12562         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12564 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12566         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12567         copying section.
12569 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12570             Steve Munroe  <munroesj52@gmail.com>
12572         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12573         (_mm_andnot_si128): Likewise.
12574         (_mm_or_si128): Likewise.
12575         (_mm_xor_si128): Likewise.
12577 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12579         PR middle-end/78809
12580         * builtins.c (expand_builtin_memcmp): Inline the calls first
12581         when result_eq is false.
12582         (expand_builtin_strcmp): Inline the calls first.
12583         (expand_builtin_strncmp): Likewise.
12584         (inline_string_cmp): New routine. Expand a string compare
12585         call by using a sequence of char comparison.
12586         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12587         a call to str(n)cmp/memcmp.
12588         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12589         New option.
12590         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12592 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12594         * config/arm/driver-arm.c: Include arm-native.h.
12595         (host_detect_local_cpu): Use auto-generated data tables.
12596         (vendors, arm_cpu_table): Delete.  Move part information to ...
12597         * config/arm/arm-cpus.in: ... here.
12598         * config/arm/parsecpu.awk (gen_native): New function.
12599         (vendor, part): New CPU fields.
12600         (END): Add support for building the native CPU detection tables.
12601         * config/arm/t-arm (arm-native.h): Add build rule.
12602         (driver-arm.o): Add dependency on arm-native.h.
12604 2018-07-13  Richard Biener  <rguenther@suse.de>
12606         PR middle-end/85974
12607         * match.pd (addr1 - addr2): Allow either of the operand to
12608         have a conversion.
12610 2018-07-13  Tom de Vries  <tdevries@suse.de>
12612         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12613         in remap_ssa_name.
12615 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12617         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12618         arrays instead of numbered variables.
12620 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12622         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12623         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12625 2018-07-13  Richard Biener  <rguenther@suse.de>
12627         PR debug/86452
12628         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12629         instead of get_context_die.
12631 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12632             Richard Biener  <rguenther@suse.de>
12634         PR middle-end/86489
12635         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12636         that the loop latch destination where phi is defined.
12638 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12640         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12641         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12642         (riscv_set_current_function): Extract getting interrupt type to new
12643         function.
12644         (riscv_get_interrupt_type): New function.
12645         (riscv_merge_decl_attributes): New function, checking interrupt type is
12646         same.
12647         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12649 2018-07-12  Paul Koning  <ni1d@arrl.net>
12651         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12652         directive.
12654 2018-07-12  Paul Koning  <ni1d@arrl.net>
12656         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12657         zero reference, add doloop_end instead.
12658         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12659         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12660         detail for doloop_end.
12662 2018-07-12  Martin Sebor  <msebor@redhat.com>
12664         PR c/86453
12665         * attribs.c (decl_attributes): Reject conflicting attributes before
12666         calling attribute handlers.
12668 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12670         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12671          parameter.
12672         (gcc::dump_manager::get_dump_file_name): likewise.
12673         (dump_begin): Likewise.
12674         * dumpfile.h (dump_begin): Update prototype.
12675         (gcc::dump_manager::get_dump_file_name,
12676         gcc::dump_manager::get_dump_file_name): Update prototype.
12678 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12680         * internal-fn.h (vectorizable_internal_fn_p): New function.
12681         * tree-vect-slp.c (compatible_calls_p): Likewise.
12682         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12683         to internal functions.
12684         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12686 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12688         * fold-const.h (inverse_conditions_p): Declare.
12689         * fold-const.c (inverse_conditions_p): New function.
12690         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12691         that test the inverse condition of a conditional internal function.
12692         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12693         * internal-fn.c (internal_fn_mask_index): Handle conditional
12694         internal functions.
12695         (vectorized_internal_fn_supported_p): New function.
12696         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12697         (any_pred_load_store): Replace with...
12698         (need_to_predicate): ...this new variable.
12699         (redundant_ssa_names): New variable.
12700         (ifcvt_can_use_mask_load_store): Move initial checks to...
12701         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12702         for which a conditional internal function exists.
12703         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12704         instead of ifcvt_can_use_mask_load_store.  Update after variable
12705         name change.
12706         (predicate_load_or_store): New function, split out from
12707         predicate_mem_writes.
12708         (check_redundant_cond_expr): New function.
12709         (value_available_p): Likewise.
12710         (predicate_rhs_code): Likewise.
12711         (predicate_mem_writes): Rename to...
12712         (predicate_statements): ...this.  Use predicate_load_or_store
12713         and predicate_rhs_code.
12714         (combine_blocks, tree_if_conversion): Update after above name changes.
12715         (ifcvt_local_dce): Handle redundant_ssa_names.
12716         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12717         general conditional functions.
12718         * tree-vect-stmts.c (vectorizable_call): Likewise.
12720 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12721             Alan Hayward  <alan.hayward@arm.com>
12722             David Sherwood  <david.sherwood@arm.com>
12724         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
12725         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
12726         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
12727         plus and minus and convert them into IFN_COND_FMA-based sequences.
12728         (convert_mult_to_fma): Handle conditional plus and minus.
12730 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12732         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
12733         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
12734         (cond_fnms_optab): New optabs.
12735         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
12736         internal functions.
12737         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
12738         * internal-fn.h (get_conditional_internal_fn): Declare.
12739         (get_unconditional_internal_fn): Likewise.
12740         * internal-fn.c (cond_ternary_direct): New macro.
12741         (expand_cond_ternary_optab_fn): Likewise.
12742         (direct_cond_ternary_optab_supported_p): Likewise.
12743         (FOR_EACH_COND_FN_PAIR): Likewise.
12744         (get_conditional_internal_fn): New function.
12745         (get_unconditional_internal_fn): Likewise.
12746         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
12747         (gimple_match_op::gimple_match_op): Add a new overload for 5
12748         operands.
12749         (gimple_match_op::set_op): Likewise.
12750         (gimple_resimplify5): Declare.
12751         * genmatch.c (decision_tree::gen): Generate simplifications for
12752         5 operands.
12753         * gimple-match-head.c (gimple_simplify): Define an overload for
12754         5 operands.  Handle calls with 5 arguments in the top-level overload.
12755         (convert_conditional_op): Handle conversions from unconditional
12756         internal functions to conditional ones.
12757         (gimple_resimplify5): New function.
12758         (build_call_internal): Pass a fifth operand.
12759         (maybe_push_res_to_seq): Likewise.
12760         (try_conditional_simplification): Try converting conditional
12761         internal functions to unconditional internal functions.
12762         Handle 3-operand unconditional forms.
12763         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
12764         Define ternary equivalents of the current rules for binary conditional
12765         internal functions.
12766         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
12767         ternary operations.
12768         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
12769         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
12770         (optab): Handle them.
12771         (SVE_COND_FP_TERNARY): New int iterator.
12772         (sve_fmla_op, sve_fmad_op): New int attributes.
12773         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
12774         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
12775         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
12777 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12779         * target.def (preferred_else_value): New target hook.
12780         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
12781         * doc/tm.texi: Regenerate.
12782         * targhooks.h (default_preferred_else_value): Declare.
12783         * targhooks.c (default_preferred_else_value): New function.
12784         * internal-fn.h (conditional_internal_fn_code): Declare.
12785         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
12786         (get_conditional_internal_fn): Use it.
12787         (conditional_internal_fn_code): New function.
12788         * gimple-match.h (gimple_match_cond): New struct.
12789         (gimple_match_op): Add a cond member function.
12790         (gimple_match_op::gimple_match_op): Update all forms to take a
12791         gimple_match_cond.
12792         * genmatch.c (expr::gen_transform): Use the same condition as res_op
12793         for the suboperation, but don't specify a particular else_value.
12794         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
12795         (visit_nary_op, visit_reference_op_load): Pass
12796         gimple_match_cond::UNCOND to the gimple_match_op constructor.
12797         * gimple-match-head.c: Include tree-eh.h
12798         (convert_conditional_op): New function.
12799         (maybe_resimplify_conditional_op): Likewise.
12800         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
12801         (gimple_resimplify2): Likewise.
12802         (gimple_resimplify3): Likewise.
12803         (gimple_resimplify4): Likewise.
12804         (maybe_push_res_to_seq): Return null for conditional operations.
12805         (try_conditional_simplification): New function.
12806         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
12807         constructor.
12808         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
12809         IFN_COND_* call.
12810         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
12811         function.
12812         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
12814 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12816         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12817         DECL_FCONTEXT
12818         (hash_tree): Do not hash DECL_FCONTEXT
12819         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
12820         Do not stream DECL_FCONTEXT.
12821         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
12822         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
12824 2018-07-12  Richard Biener  <rguenther@suse.de>
12826         PR debug/86462
12827         * dwarf2out.c (gen_block_die): Only output blocks when they have
12828         at least one !DECL_IGNORED_P variable.
12830 2018-07-12  Richard Biener  <rguenther@suse.de>
12832         PR target/84829
12833         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
12834         Remove -mieee-fp handling.
12836 2018-07-12  Richard Biener  <rguenther@suse.de>
12838         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
12839         left-over from last patch.
12841 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
12843         PR tree-optimization/86492
12844         * gimple-ssa-store-merging.c
12845         (imm_store_chain_info::coalesce_immediate_stores): Call
12846         check_no_overlap even for the merge_overlapping case.  Formatting fix.
12848 2018-07-12  Richard Biener  <rguenther@suse.de>
12850         PR middle-end/86479
12851         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
12852         move possibly trapping operations into the conditional.
12854 2018-07-12  Richard Biener  <rguenther@suse.de>
12856         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
12857         (vn_lookup_simplify_result): Remove recursion limit applied
12858         here.
12859         (vn_nary_build_or_lookup_1): Adjust.
12860         (try_to_simplify): Likewise.
12861         * gimple-match-head.c (gimple_resimplify1): Instead apply one
12862         here.
12863         (gimple_resimplify2): Likewise.
12864         (gimple_resimplify3): Likewise.
12865         (gimple_resimplify4): Likewise.
12867 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
12869         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
12870         Use __mmask64 type instead of __mmask8 for __M argument.
12871         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
12872         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
12873         __U argument.
12874         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
12875         __mmask16 for __M argument.
12876         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
12877         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12878         to __mmask16 instead of __mmask8.
12879         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
12880         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
12881         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
12882         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
12883         instead of __mmask16 for __U argument.
12884         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
12885         __mmask16 instead of __mmask8 for __U argument.
12886         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
12887         __U argument.
12888         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
12889         __mmask16.
12890         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
12891         argument.
12892         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
12893         __U argument.
12894         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
12895         __mmask16.
12896         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
12897         of __mmask16.
12898         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
12899         __U argument.
12900         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
12901         __U argument.
12902         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
12903         __U argument.
12904         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
12905         __U argument.
12906         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
12907         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
12908         return type as well as __M argument type and all casts from __mmask8
12909         to __mmask32.
12910         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
12911         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
12912         return type as well as __M argument type and all casts from __mmask8
12913         to __mmask16.
12914         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
12915         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
12916         return type as well as __M argument type and all casts from __mmask8
12917         to __mmask32.
12918         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
12919         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
12920         return type as well as __M argument type and all casts from __mmask8
12921         to __mmask16.
12922         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
12923         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
12924         __mmask16.
12926 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
12928         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
12929         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
12930         for __U argument.
12932 2018-07-11  Paul Koning  <ni1d@arrl.net>
12934         * doc/md.texi (define_subst): Document how multiple occurrences of
12935         the same argument in the replacement pattern are handled.
12937 2018-07-11  Paul Koning  <ni1d@arrl.net>
12939         * doc/extend.texi (Common Variable Attributes): Move "mode" into
12940         alphabetical order.
12941         (Common Type Attributes): Add "mode" attribute.
12943 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
12945         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
12946         stream DECL_ORIGINAL_TYPE.
12947         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
12948         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
12949         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
12950         Do not walk original type.
12951         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
12952         external decls.
12953         (write_ts_decl_non_common_tree_pointers): Do not stream
12954         DECL_ORIGINAL_TYPE
12955         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
12956         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
12958 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12960         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
12961         thread twice from the same starting edge.
12963 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12965         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
12966         * gimple.c (gimple_call_nonnull_result_p): ...here...
12967         (gimple_call_nonnull_arg): ...and here.
12968         * gimple.h (gimple_call_nonnull_result_p): New.
12969         (gimple_call_nonnull_arg): New.
12971 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
12973         * config/arm/arm-cpus.in: Move information from fpu field of each
12974         cpu definition to the isa field.
12975         * config/arm/parsecpu.awk (fpu): Delete match rule.
12976         (gen_comm_data): Don't add bits from the CPU's FPU entry.
12978 2018-07-11  Richard Biener  <rguenther@suse.de>
12980         PR debug/86457
12981         * dwarf2out.c (init_sections_and_labels): Use
12982         output_asm_line_debug_info consistently.
12983         (dwarf2out_early_finish): Likewise.
12984         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
12985         type units.
12987 2018-07-11  Richard Biener  <rguenther@suse.de>
12989         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
12990         Rework father_bb setting in a way to avoid propagating constants
12991         multiple times on a loop body.
12993 2018-07-10  Mark Wielaard  <mark@klomp.org>
12995         PR debug/86459
12996         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
12998 2018-07-10  Richard Biener  <rguenther@suse.de>
13000         * hash-map.h (hash_map::iterator::operator*): Return
13001         references to key and value.
13003 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
13005         PR c++/86443
13006         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13007         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13008         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13009         (gimplify_omp_for): For composite loops, move outer
13010         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13011         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
13012         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
13013         TREE_LIST for both the original class iterator and the "last" helper
13014         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
13015         loop, remember has_decl_expr from outer composite loops for the
13016         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13018 2018-07-09  Martin Sebor  <msebor@redhat.com>
13020         PR middle-end/77357
13021         PR middle-end/86428
13022         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13023         accessing implicitly initialized array elements.
13024         * expr.c (string_constant): Handle string initializers of
13025         character arrays within aggregates.
13026         * gimple-fold.c (fold_array_ctor_reference): Add argument.
13027         Store element offset.  As a special case, handle zero size.
13028         (fold_nonarray_ctor_reference): Same.
13029         (fold_ctor_reference): Add argument.  Store subobject offset.
13030         * gimple-fold.h (fold_ctor_reference): Add argument.
13032 2018-07-09  Paul Koning  <ni1d@arrl.net>
13034         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13035         (pdp11_insn_cost): New function.
13036         (pdp11_md_asm_adjust): New function.
13037         (TARGET_INVALID_WITHIN_DOLOOP): Define.
13038         (pdp11_rtx_costs): Update to match machine better.
13039         (output_addr_const_pdp11): Correct format mismatch warnings.
13040         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13041         * config/pdp11/pdp11.md: General change to add base_cost and/or
13042         length attributes for use by new pdp11_insn_cost function.
13043         (MIN_BRANCH): Correct definition.
13044         (MIN_SOB): Ditto.
13045         (doloop_end): Use standard pattern name for looping pattern.
13046         (doloop_end_nocc): New.
13047         (movsf): Add another constraint alternative.
13048         (zero_extendqihi2): Add constraint alternatives for not in place
13049         extend.
13050         (zero_extendhisi2): Remove.
13051         (shift patterns): Add CC handling variants.
13052         (bswaphi2): New.
13053         (bswapsi2): New.
13054         (rothi3): New.
13055         (define_peephole2): New peephole to recognize mov that sets CC for
13056         subsequent test.
13058 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13060         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13061         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13062         wi::add.
13064 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
13066         PR c/86420
13067         * real.c (real_nextafter): Return true if result is denormal.
13069 2018-07-09  Martin Liska  <mliska@suse.cz>
13071         * common.opt: Add back wrongly removed attribute.
13073 2018-07-09  Richard Biener  <rguenther@suse.de>
13075         PR debug/86413
13076         * dwarf2out.c (gen_block_die): For an early generated DIE
13077         always output high/low PC attributes.
13079 2018-07-09  Tom de Vries  <tdevries@suse.de>
13081         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
13082         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
13083         onto VAR_DECL with abstract origin.
13085 2018-07-07  Jim Wilson  <jimw@sifive.com>
13087         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
13089 2018-07-07  Tom de Vries  <tdevries@suse.de>
13091         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
13093 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
13095         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
13096         overflow_type.
13097         (combine_bound): Use wide-int overflow calculation instead of
13098         rolling our own.
13099         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
13100         overflow_type.
13101         * fold-const.c (int_const_binop_2): Same.
13102         (extract_muldiv_1): Same.
13103         (fold_div_compare): Same.
13104         (fold_abs_const): Same.
13105         * match.pd: Same.
13106         * poly-int.h (add): Same.
13107         (sub): Same.
13108         (neg): Same.
13109         (mul): Same.
13110         * predict.c (predict_iv_comparison): Same.
13111         * profile-count.c (slow_safe_scale_64bit): Same.
13112         * simplify-rtx.c (simplify_const_binary_operation): Same.
13113         * tree-chrec.c (tree_fold_binomial): Same.
13114         * tree-data-ref.c (split_constant_offset_1): Same.
13115         * tree-if-conv.c (idx_within_array_bound): Same.
13116         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13117         * tree-ssa-phiopt.c (minmax_replacement): Same.
13118         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
13119         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
13120         * vr-values.c (vr_values::adjust_range_with_scev): Same.
13121         * wide-int.cc (wi::add_large): Same.
13122         (wi::mul_internal): Same.
13123         (wi::sub_large): Same.
13124         (wi::divmod_internal): Same.
13125         * wide-int.h: Change overflow type to overflow_type for neg, add,
13126         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
13127         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
13128         mul_internal, divmod_internal.
13129         (overflow_type): New enum.
13130         (accumulate_overflow): New.
13132 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
13134         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13135         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13137 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13139         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13140         argument is checked for zero before entering loop, avoid checking again.
13142 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13144         * gimplify.h (generic_expr_could_trap_p): Set as global function.
13145         * gimplify.h (generic_expr_could_trap_p): Likwise.
13146         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13148 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
13150         PR tree-optimization/86401
13151         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13152         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13153         (fold_bit_and_mask): ... here.  New helper function for match.pd.
13154         * fold-const.h (fold_bit_and_mask): Declare.
13155         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13157 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
13159         PR target/86324
13160         * target.def (translate_mode_attribute): New hook.
13161         * targhooks.h (default_translate_mode_attribute): Declare.
13162         * targhooks.c (default_translate_mode_attribute): New function.
13163         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13164         * doc/tm.texi: Regenerate.
13165         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13166         (rs6000_translate_mode_attribute): New function.
13168 2018-07-06  Paul Koning  <ni1d@arrl.net>
13170         * doc/md.texi (define_split): Document DONE and FAIL.
13171         (define_peephole2): Ditto.
13173 2018-07-05  Jeff Law  <law@redhat.com>
13175         PR tree-optimization/86010
13176         * tree-ssa-dse.c (compute_trims): More aggressively trim at
13177         both the head and tail of mem* and str* calls.
13179 2018-07-05  Jim Wilson  <jimw@sifive.com>
13181         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13183 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
13185         * config/aarch64/aarch64-simd.md: correct flags text for
13186         MIN_EXPR replacement.
13188 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
13190         * configure: Regenerated.
13192 2018-07-05  Carl Love  <cel@us.ibm.com>
13194         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13195         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13196         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13197         VSX_BUILTIN_DOUBLEL_V4SF.
13199 2018-07-05  Martin Sebor  <msebor@redhat.com>
13201         PR c++/86400
13202         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13203         than its domain to compute its the upper bound of a char array.
13205 2018-07-05  Nathan Sidwell  <nathan@acm.org>
13207         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13208         * doc/cpp.texi: Update comment.
13209         * doc/tm.texi: Rebuilt.
13210         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13211         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13212         * doc/extend.texi (Backwards Compatibility): Clarify it is system
13213         headers affected by extern "C".
13214         * system.h: Poison NO_IMPLICIT_EXTERN_C.
13215         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13216         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13217         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13218         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13219         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13220         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13221         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13222         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13223         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13224         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13225         config/sparc/sp64-elf.h, config/spu/spu.h,
13226         config/stormy16/stormy16.h, config/v850/v850.h,
13227         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13228         define NO_IMPLICIT_EXTERN_C.
13229         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13231 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13233         PR target/84711
13234         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13235         instead of GET_MODE_SIZE when comparing Units.
13237 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13239         PR target/84711
13240         * rtlanal.c (set_noop_p): Constrain on mode change,
13241         include hard-reg-set.h
13243 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13245         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13247 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13249         Revert
13250         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13252         PR sanitizer/84250
13253         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13254         libasan.
13255         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13257 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13259         PR sanitizer/84250
13260         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13261         libasan.
13262         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13264 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13266         PR middle-end/86380
13267         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13269 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13271         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13272         neg_*_op* variables.
13274 2018-07-04  Martin Liska  <mliska@suse.cz>
13276         * tree-switch-conversion.c: Define
13277         max_ratio_for_speed and max_ratio_for_size constants.
13279 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13280             Martin Liska  <mliska@suse.cz>
13282         PR middle-end/66240
13283         PR target/45996
13284         PR c/84100
13285         * common.opt: Rename align options with 'str_' prefix.
13286         * common/config/i386/i386-common.c (set_malign_value): New
13287         function.
13288         (ix86_handle_option): Use it to set -falign-* options/
13289         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13290         type from int to string.
13291         * config/aarch64/aarch64.c: Update default values from int
13292         to string.
13293         * config/alpha/alpha.c (alpha_override_options_after_change):
13294         Likewise.
13295         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13296         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13297         max skip conditionally.
13298         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13299         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13300         max skip conditionally.
13301         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13302         max skip conditionally.
13303         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13304         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13305         max skip conditionally.
13306         * config/i386/i386.c (struct ptt): Change type from int to
13307         string.
13308         (ix86_default_align): Set default values.
13309         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13310         max skip conditionally.
13311         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13312         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13313         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13314         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13315         max skip conditionally.
13316         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13317         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13318         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13319         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13320         max skip conditionally.
13321         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13322         * config/ia64/ia64.c (ia64_option_override): Set default values
13323         for alignment options.
13324         * config/m68k/m68k.c: Handle new str_align_* options.
13325         * config/mips/mips.c (mips_set_compression_mode): Change
13326         type of constants.
13327         (mips_option_override): Set default values for options.
13328         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13329         Likewise.
13330         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13331         Likewise.
13332         * config/rx/rx.c (rx_option_override): Likewise.
13333         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13334         (LABEL_ALIGN): Use align_labels_log.
13335         (LOOP_ALIGN): Use align_loops_align.
13336         * config/s390/s390.c (s390_asm_output_function_label): Use new
13337         macros.
13338         * config/sh/sh.c (sh_override_options_after_change):
13339         Change type of constants.
13340         * config/spu/spu.c (spu_sched_init): Likewise.
13341         * config/sparc/sparc.c (sparc_option_override): Set default
13342         values for options.
13343         * config/visium/visium.c (visium_option_override): Likewise.
13344         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13345         emit p2align format with last argument if it's not needed.
13346         * doc/invoke.texi: Document extended format of -falign-*.
13347         * final.c: Use align_labels alignment.
13348         * flags.h (struct target_flag_state): Change type to use
13349         align_flags.
13350         (struct align_flags_tuple): New.
13351         (struct align_flags): Likewise.
13352         (align_loops_log): Redefine macro to use new types.
13353         (align_loops_max_skip): Redefine macro to use new types.
13354         (align_jumps_log): Redefine macro to use new types.
13355         (align_jumps_max_skip): Redefine macro to use new types.
13356         (align_labels_log): Redefine macro to use new types.
13357         (align_labels_max_skip): Redefine macro to use new types.
13358         (align_functions_log): Redefine macro to use new types.
13359         (align_loops): Redefine macro to use new types.
13360         (align_jumps): Redefine macro to use new types.
13361         (align_labels): Redefine macro to use new types.
13362         (align_functions): Redefine macro to use new types.
13363         (align_functions_max_skip): Redefine macro to use new types.
13364         (align_loops_value): New macro.
13365         (align_jumps_value): New macro.
13366         (align_labels_value): New macro.
13367         (align_functions_value): New macro.
13368         * function.c (invoke_set_current_function_hook): Propagate
13369         alignment values from flags to global variables default in
13370         topleev.h.
13371         * ipa-icf.c (sem_function::equals_wpa): Use
13372         cl_optimization_option_eq instead of memcmp.
13373         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13374         of string types.
13375         (cl_optimization_stream_in): Likewise.
13376         * optc-save-gen.awk: Support strings in cl_optimization.
13377         * opth-gen.awk: Likewise.
13378         * opts.c (finish_options): Remove error checking of invalid
13379         value ranges.
13380         (MAX_CODE_ALIGN): Remove.
13381         (MAX_CODE_ALIGN_VALUE): Likewise.
13382         (parse_and_check_align_values): New function.
13383         (check_alignment_argument): Likewise.
13384         (common_handle_option): Use check_alignment_argument.
13385         * opts.h (parse_and_check_align_values): Declare.
13386         * toplev.c (init_alignments): Remove.
13387         (read_log_maxskip): New.
13388         (parse_N_M): Likewise.
13389         (parse_alignment_opts): Likewise.
13390         (backend_init_target): Remove usage of init_alignments.
13391         * toplev.h (parse_alignment_opts): Declare.
13392         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13393         argument.
13394         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13395         * tree.c (cl_option_hasher::equal): New.
13396         * varasm.c: Use new global macros.
13398 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13400         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13401         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13402         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13403         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13404         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13405         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13406         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13407         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13408         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13409         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13410         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13411         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13412         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13414 2018-07-04  Martin Liska  <mliska@suse.cz>
13415             Jonathan Wakely  <jwakely@redhat.com>
13417         * coverage.c: Use correct type.
13418         * doc/invoke.texi: Language correction.
13420 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13422         PR target/85620
13423         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13424         ENDBRANCH for non-tail call which may return via indirect branch.
13425         * doc/extend.texi: Document indirect_return attribute.
13427 2018-07-03  Martin Sebor  <msebor@redhat.com>
13429         PR tree-optimization/86274
13430         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13431         precondition.
13432         (format_floating): Correct handling of infinities and NaNs.
13434 2018-07-03  Martin Sebor  <msebor@redhat.com>
13436         * print-tree.c (print_real_cst): New function.
13437         (print_node_brief): Call it.
13438         (print_node): Ditto.
13440 2018-07-03  Jeff Law  <law@redhat.com>
13442         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13443         into a single pattern.
13445         * config/h8300/h8300.md (ors code_iterator): New.
13446         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13447         a single pattern and single splitter.
13448         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13449         (iorqi3_1, xorqi3_1): Likewise.
13450         (iorqi3, xorqi3 expanders): Similarly.
13452         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13453         (movmd_internal) into a single pattern using the P mode iterator.
13454         (movmd splitters): Similarly.
13455         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13456         (movsd splitters): Similarly.
13458         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13459         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13460         right table based on the size of the operand.
13461         * config/h8300/h8300.md (length_table): Corresponding changes. All
13462         references to "addb", "addw" and "addl" changed to "add".
13463         (btst patterns): Merge two variants into a single pattern.
13464         (tstqi, tsthi): Likewise.
13465         (addhi3_incdec, addsi3_incdec): Likewise.
13466         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13467         (mulhi3, mulsi3): Likewise.
13468         (udivhi3, udivsi3): Likewise.
13469         (divhi3, divsi3): Likewise.
13470         (andorqi3, andorhi3, andorsi3): Likewise.
13472 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13474         PR target/85694
13475         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13476         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13478 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13480         PR tree-optimization/85694
13481         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13482         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13483         UNSPEC_URHADD.
13484         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13485         (<u>avg<mode>3_ceil): New patterns.
13487 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13489         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13490         scan-tree-dump directive.
13491         * gcc.dg/vect/slp-perm-2.c: Likewise.
13492         * gcc.dg/vect/slp-perm-3.c: Likewise.
13493         * gcc.dg/vect/slp-perm-5.c: Likewise.
13494         * gcc.dg/vect/slp-perm-6.c: Likewise.
13495         * gcc.dg/vect/slp-perm-7.c: Likewise.
13496         * gcc.dg/vect/slp-perm-8.c: Likewise.
13498 2018-07-03  Marek Polacek  <polacek@redhat.com>
13500         PR middle-end/86202
13501         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13503 2018-07-03  Richard Biener  <rguenther@suse.de>
13505         PR ipa/86389
13506         * tree-ssa-structalias.c (find_func_clobbers): Properly
13507         handle indirect calls.
13509 2018-07-03  Jeff Law  <law@redhat.com>
13511         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13512         (shifts): New code iterator.
13513         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13514         expander.  Fix HImode handling on H8/SX.
13515         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13516         (subqi3, subhi3, subsi3 expanders): Likewise.
13517         (andqi3, andhi3, andsi3 expanders): Likewise.
13518         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13519         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13520         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13521         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13522         (zero_extendqihi2, zero_extendqisi2): Likewise.
13523         (extendqihi2, extendqisi2): Likewise.
13524         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13525         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13526         (rotlqi3_1, rotlhi3_1): Likewise.
13527         (logicalhi3_sn, logicalsi3_sn): Likewise.
13528         (logicalhi3, logicalsi3): Likewise.
13530 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13532         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13533         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13534         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13535         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13536         type to append_pattern_def_seq instead of creating a stmt_vec_info
13537         directly.
13538         (build_mask_conversion): Likewise.  Remove vinfo argument.
13539         (vect_add_conversion_to_patterm): Likewise, renaming to...
13540         (vect_add_conversion_to_pattern): ...this.
13541         (vect_recog_mask_conversion_pattern): Update call to
13542         build_mask_conversion.  Pass the vector type to
13543         append_pattern_def_seq here too.
13544         (vect_recog_gather_scatter_pattern): Update call to
13545         vect_add_conversion_to_pattern.
13547 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13549         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13550         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13551         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13552         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13553         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13554         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13555         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13556         append_pattern_def_seq instead of new_pattern_def_seq.
13557         (vect_recog_divmod_pattern): Do both of the above.
13558         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13559         is null.
13561 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13563         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13564         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13565         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13566         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13567         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13568         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13569         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13570         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13571         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13572         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13573         parameter with a single stmt_vec_info.
13574         (vect_recog_func_ptr): Likewise.
13575         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13576         (vect_try_gather_scatter_pattern): ...this.
13577         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13578         the stmt_vec_info of the statement to be matched.  Don't clear
13579         STMT_VINFO_RELATED_STMT.
13580         (vect_pattern_recog): Update call accordingly.
13582 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13584         PR tree-optimization/85694
13585         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13586         (uavgM3_ceil): Document new optabs.
13587         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13588         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13589         functions.
13590         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13591         (savg_ceil_optab): New optabs.
13592         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13593         (vect_vect_recog_func_ptrs): Add it.
13594         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13595         constant directly from the associated lhs.
13597 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13599         * tree-vect-patterns.c (vect_split_statement): New function.
13600         (vect_convert_input): Use it to try to split an existing cast.
13602 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13604         * poly-int.h (print_hex): New function.
13605         * dumpfile.h (dump_dec, dump_hex): Declare.
13606         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13607         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13608         min_input_precision, operation_precision and operation_sign.
13609         * tree-vect-patterns.c (vect_get_range_info): New function.
13610         (vect_same_loop_or_bb_p, vect_single_imm_use)
13611         (vect_operation_fits_smaller_type): Delete.
13612         (vect_look_through_possible_promotion): Add an optional
13613         single_use_p parameter.
13614         (vect_recog_over_widening_pattern): Rewrite to use new
13615         stmt_vec_info infomration.  Handle one operation at a time.
13616         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13617         (vect_truncatable_operation_p, vect_set_operation_type)
13618         (vect_set_min_input_precision): New functions.
13619         (vect_determine_min_output_precision_1): Likewise.
13620         (vect_determine_min_output_precision): Likewise.
13621         (vect_determine_precisions_from_range): Likewise.
13622         (vect_determine_precisions_from_users): Likewise.
13623         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13624         (vect_vect_recog_func_ptrs): Put over_widening first.
13625         Add cast_forwprop.
13626         (vect_pattern_recog): Call vect_determine_precisions.
13628 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13630         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13631         statements that have been replaced by further pattern statements.
13632         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13634 2018-07-03  Richard Biener  <rguenther@suse.de>
13636         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13637         always set *dt.  Dump vectype in vectype overload.
13638         * dumpfile.h (dump_gimple_expr): New function.
13639         (dump_gimple_expr_loc): Likewise.
13640         * dumpfile.c (dump_gimple_expr): New function.
13641         (dump_gimple_expr_loc): Likewise.
13643 2018-07-02  Jeff Law  <law@redhat.com>
13645         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13646         the H8/300, H8/300H and H8/S variants into a single pattern.
13647         (movhi_h8300, movqi_h8300hs): Similarly.
13648         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13649         (QHI mode iterator): New.
13651         * config/h8300/h8300.md: Remove trailing whitespace.
13653 2018-07-02  Jim Wilson  <jimw@sifive.com>
13655         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13656         instead of emit_insn for interrupt returns.
13657         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13658         (riscv_sret, riscv_uret): Likewise.
13660 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13662         * pretty-print.c (selftest::test_pp_format): Move save and restore
13663         of quotes to class auto_fix_quotes, and add an instance.
13664         * selftest.c: Include "intl.h".
13665         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13666         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13667         * selftest.h (selftest::auto_fix_quotes): New class.
13669 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13671         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13672         (aarch64_sve_prepare_conditional_op): Remove.
13673         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13674         Allow aarch64_simd_reg_or_zero as select operand; remove
13675         the aarch64_sve_prepare_conditional_op call.
13676         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13677         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13678         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13679         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13680         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13681         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13682         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13683         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13684         and a splitters to match all of the *_any patterns.
13685         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13687         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13688         (SVE_COND_FP_BINARY_REV): Remove.
13689         (sve_int_op_rev, sve_fp_op_rev): New.
13690         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13691         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13692         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13693         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13694         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13695         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13696         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13697         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13698         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13700         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13701         Remove match_dup 1 from the inner unspec.
13702         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13704         * config/aarch64/aarch64.md (movprfx): New attr.
13705         (length): Default movprfx to 8.
13706         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13707         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13708         (*<su>mul<SVE_I>3_highpart): Likewise.
13709         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13710         (*v<ASHIFT><SVE_I>3): Likewise.
13711         (*<su><MAXMIN><SVE_I>3): Likewise.
13712         (*<su><MAXMIN><SVE_F>3): Likewise.
13713         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13714         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13715         (*div<SVE_F>4): Likewise.
13717 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
13719         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13720         in dump string.
13722 2018-07-02  Richard Biener  <rguenther@suse.de>
13724         PR tree-optimization/86363
13725         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
13726         memset argument refers to a non-variable address.
13728 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13730         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
13731         {PLUS,MINUS}_EXPR code to...
13732         (adjust_symbolic_bound): ...here,
13733         (combine_bound): ...here,
13734         (set_value_range_with_overflow): ...and here.
13736 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13738         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
13739         code...
13740         (extract_range_from_abs_expr): ...here.
13742 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
13744         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
13745         -fno-omit-frame-pointer when not optimizing.
13747 2018-07-02  Martin Liska  <mliska@suse.cz>
13749         PR ipa/86279
13750         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
13751         (propagate_nothrow): Likewise.
13753 2018-07-02  Martin Liska  <mliska@suse.cz>
13755         PR ipa/86323
13756         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
13758 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13760         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
13761         function in r262149, changing "loc" param from source_location to
13762         const dump_location_t &.
13763         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
13764         declaration, as above.
13766 2018-07-01  Paul Koning  <ni1d@arrl.net>
13768         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13769         -munit-asm, -mgnu-asm, -mdec-asm.
13770         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
13771         (pdp11_output_labelref): New.
13772         (pdp11_output_def): New.
13773         (pdp11_output_addr_vec_elt): New.
13774         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
13775         %# and %@ format codes.
13776         (pdp11_option_override): New.
13777         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
13778         (pdp11_output_ident): New.
13779         (pdp11_asm_named_section): New.
13780         (pdp11_asm_init_sections): New.
13781         (pdp11_file_start): New.
13782         (pdp11_file_end): New.
13783         (output_ascii): Use .ascii/.asciz for -mdec-asm.
13784         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
13785         %o, like %c but octal.
13786         (pdp11_option_override): New.
13787         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
13788         -mdec-asm.
13789         (DATA_SECTION_ASM_OP): Ditto.
13790         (READONLY_DATA_SECTION_ASM_OP): New.
13791         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
13792         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
13793         (ASM_OUTPUT_LABELREF): Ditto.
13794         (ASM_OUTPUT_DEF): Ditto.
13795         (ASM_OUTPUT_EXTERNAL): New.
13796         (ASM_OUTPUT_SOURCE_FILENAME): New.
13797         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
13798         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
13799         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
13800         %# and %@ format codes.
13801         * config/pdp11/pdp11.opt (mgnu-asm): New.
13802         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
13803         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
13804         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
13806 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
13808         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
13809         dereferencing path[] beyond its length.
13810         (debug_path): New.
13811         (debug_all_paths): New.
13812         (rewire_first_differing_edge): New.
13813         (adjust_paths_after_duplication): New.
13814         (duplicate_thread_path): Call adjust_paths_after_duplication.
13815         Add new argument.
13816         (thread_through_all_blocks): Add new argument to
13817         duplicate_thread_path.
13819 2018-06-30  Jim Wilson  <jimw@sifive.com>
13821         * config/riscv/predicates.md (p2m1_shift_operand): New.
13822         (high_mask_shift_operand): New.
13823         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
13824         pattern using p2m1_shift_operand.
13825         (lshsi3_zero_extend_3+2): New combiner pattern using
13826         high_mask_shift_operand.
13828 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13830         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
13831         split out from...
13832         (vect_recog_rotate_pattern): ...here.
13833         (vect_convert_input): Try to insert casts of invariants in the
13834         preheader.
13835         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
13836         preheader to be empty.
13838 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13840         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
13841         vector type.  If given, install it in the new statement's
13842         STMT_VINFO_VECTYPE.
13843         (vect_element_precision): New function.
13844         (vect_unpromoted_value): New struct.
13845         (vect_unpromoted_value::vect_unpromoted_value): New function.
13846         (vect_unpromoted_value::set_op): Likewise.
13847         (vect_look_through_possible_promotion): Likewise.
13848         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
13849         (vect_widened_op_tree, vect_convert_input): Likewise.
13850         (vect_convert_inputs, vect_convert_output): Likewise.
13851         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
13852         to handle the optional cast of the multiplication result and
13853         vect_widened_op_tree to detect the widened multiplication itself.
13854         Do not require the input and output of promotion casts to have
13855         the same sign, but base the signedness of the operation on the
13856         input rather than the result.  If the pattern includes two
13857         promotions, check that those promotions have the same sign.
13858         Do not restrict the MULT_EXPR handling to a double-width result;
13859         handle quadruple-width results and wider.  Use vect_convert_inputs
13860         to convert the inputs to the common type.
13861         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
13862         to handle the optional cast of the ABS result.  Also allow a sign
13863         change or a sign extension between the ABS and MINUS.
13864         Use vect_widened_op_tree to detect the widened subtraction and use
13865         vect_convert_inputs to convert the inputs to the common type.
13866         (vect_handle_widen_op_by_const): Delete.
13867         (vect_recog_widen_op_pattern): New function.
13868         (vect_recog_widen_mult_pattern): Use it.
13869         (vect_recog_widen_shift_pattern): Likewise.
13870         (vect_recog_widen_sum_pattern): Use
13871         vect_look_through_possible_promotion to handle the promoted
13872         PLUS_EXPR operand.
13874 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13876         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
13877         the containing gimple_seq *.
13878         * gimple-iterator.h (gsi_for_stmt): Declare it.
13879         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
13880         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
13881         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13882         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13883         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
13884         checks.
13885         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
13886         split out from...
13887         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
13888         statement being replaced is part of an existing pattern
13889         definition sequence, inserting the new pattern statements before
13890         the original one.
13891         (vect_pattern_recog_1): Don't return a bool.  If the statement
13892         is already part of a pattern, instead apply pattern matching
13893         to the pattern definition statements.  Don't clear the
13894         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
13895         (vect_pattern_recog): Don't break after the first match;
13896         continue processing the pattern definition statements instead.
13897         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
13899 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13901         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13902         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13903         (vect_recog_widen_sum_pattern): Use it.
13905 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13907         * tree-vect-loop.c (vectorizable_reduction): Assert that the
13908         phi is not a pattern statement and has not been replaced by
13909         a pattern statement.
13910         * tree-vect-patterns.c (type_conversion_p): Don't check
13911         STMT_VINFO_IN_PATTERN_P.
13912         (vect_recog_vector_vector_shift_pattern): Likewise.
13913         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
13914         the pattern statement rather than the original statement; check
13915         directly for a WIDEN_MULT_EXPR here.
13916         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
13917         vect_is_simple_use to return the pattern statement rather
13918         than the original statement; use is_pattern_stmt_p to check
13919         for such a pattern statement.
13920         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
13921         to return the pattern statement rather than the original statement;
13922         don't do the same transformation here.
13923         (vect_is_simple_use): If the defining statement has been replaced
13924         by a pattern statement, return the pattern statement instead.
13925         Remove the corresponding (local) transformation from the vectype
13926         overload.
13928 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13930         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
13931         end and default to null.
13932         * tree-vect-loop.c (vect_create_epilog_for_reduction)
13933         (vectorizable_reduction): Update calls accordingly, dropping the
13934         gimple ** argument if the passed-back statement isn't needed.
13935         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
13936         (vect_recog_rotate_pattern): Likewise.
13937         (vect_recog_mask_conversion_pattern): Likewise.
13938         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13939         (vect_mask_constant_operand_p): Likewise.
13940         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
13941         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
13942         (get_group_load_store_type, get_load_store_type): Likewise.
13943         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
13944         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
13945         (vectorizable_conversion, vectorizable_assignment): Likewise.
13946         (vectorizable_shift, vectorizable_operation): Likewise.
13947         (vectorizable_store, vect_is_simple_cond): Likewise.
13948         (vectorizable_condition, vectorizable_comparison): Likewise.
13949         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
13950         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
13951         and move it to the end.  Cope with null def_stmt_outs.
13953 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13955         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
13957 2018-06-29  Jeff Law  <law@redhat.com>
13959         * config/v850/v850.c (v850_legitimate_address_p): Handle large
13960         displacements for TARGET_V850E2V3 and newer.
13961         (TARGET_LRA_P): Remove.  Defaults to LRA now.
13962         * config/v850/v850.md (sign23byte_load): Remove.
13963         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
13964         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
13966 2018-06-29  Martin Liska  <mliska@suse.cz>
13968         PR lto/85759
13969         * coverage.c (coverage_init): Mangle full path name.
13970         * doc/invoke.texi: Document the change.
13971         * gcov-io.c (mangle_path): New.
13972         * gcov-io.h (mangle_path): Likewise.
13973         * gcov.c (mangle_name): Use mangle_path for path mangling.
13975 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13977         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
13978         if starting source register is not even.
13980 2018-06-29  Martin Liska  <mliska@suse.cz>
13982         PR tree-optimization/86263
13983         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
13984         Make edge redirection.
13986 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
13988         * dumpfile.c (dump_loc): Add indentation based on scope depth.
13989         (dump_scope_depth): New variable.
13990         (get_dump_scope_depth): New function.
13991         (dump_begin_scope): New function.
13992         (dump_end_scope): New function.
13993         * dumpfile.h (get_dump_scope_depth): New declaration.
13994         (dump_begin_scope): New declaration.
13995         (dump_end_scope): New declaration.
13996         (class auto_dump_scope): New class.
13997         (AUTO_DUMP_SCOPE): New macro.
13998         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
13999         AUTO_DUMP_SCOPE.
14001 2018-06-29  Richard Biener  <rguenther@suse.de>
14003         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14004         compute_all_dependences succeeds.
14005         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14006         exceed --param loop-max-datarefs-for-datadeps.
14008 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
14010         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14012 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
14014         PR target/86348
14015         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14016         alternative 0 in preferred_for_speed attribute.
14018 2018-06-28  Paul Koning  <ni1d@arrl.net>
14020         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14021         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14022         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14023         * config/pdp11/pdp11.md: Correct "length" attribute calculation
14024         for shift insn patterns.
14026 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
14028         * cgraph.c (cgraph_node::get_body): Replace assignments to
14029         "dump_file" with calls to set_dump_file.
14030         * dumpfile.c (alt_dump_file): Make static, and group with...
14031         (alt_flags): ...this definition.
14032         (dumps_are_enabled): New variable.
14033         (refresh_dumps_are_enabled): New function.
14034         (set_dump_file): New function.
14035         (set_alt_dump_file): New function.
14036         (gcc::dump_manager::dump_start): Replace assignments to
14037         "dump_file" and "alt_dump_file" with calls to set_dump_file and
14038         set_alt_dump_file.
14039         (gcc::dump_manager::dump_finish): Likewise.
14040         * dumpfile.h (alt_dump_file): Delete decl.
14041         (dumps_are_enabled): New variable decl.
14042         (set_dump_file): New function decl.
14043         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14044         global.
14045         * tree-nested.c (lower_nested_functions): Replace assignments to
14046         "dump_file" with calls to set_dump_file.
14048 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14050         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14051         goto_locus of each outgoing edge of each basic block.
14053 2018-06-28  Richard Biener  <rguenther@suse.de>
14055         * dwarf2out.c (decl_scope_table): Remove.
14056         (push_decl_scope): Likewise.
14057         (pop_decl_scope): Likewise.
14058         (gen_type_die_for_member): Do not call push/pop_decl_scope.
14059         (gen_struct_or_union_type_die): Likewise.
14060         (gen_tagged_type_die): Likewise.
14061         (dwarf2out_init): Do not initialize decl_scope_table.
14062         (dwarf2out_c_finalize): Do not free it.
14064 2018-06-28  Richard Biener  <rguenther@suse.de>
14066         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14067         deciding whether to not re-use a DIE.
14069 2018-06-28  Richard Biener  <rguenther@suse.de>
14071         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
14072         DW_AT_abstract_origin attribute.
14074 2018-06-28  Martin Liska  <mliska@suse.cz>
14076         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14077         Use newly introduced constants.
14078         * tree-switch-conversion.h (struct jump_table_cluster):
14079         Define max_ratio_for_size and max_ratio_for_speed.
14081 2018-06-28  Martin Liska  <mliska@suse.cz>
14083         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14084         Add new checking assert to catch invalid state.
14085         (jump_table_cluster::can_be_handled): Handle single case
14086         clusters.
14087         (jump_table_cluster::is_beneficial): Bail out for such case.
14088         (bit_test_cluster::find_bit_tests):
14089         Add new checking assert to catch invalid state.
14090         (bit_test_cluster::can_be_handled): Handle single case
14091         clusters.
14092         (bit_test_cluster::is_beneficial): Bail out for such case.
14093         (switch_decision_tree::analyze_switch_statement):
14094         Fix comment.
14096 2018-06-28  Martin Liska  <mliska@suse.cz>
14098         * common.opt: Introduce -completion option.
14099         * gcc.c (driver_handle_option): Handle it.
14100         (driver::main): Print completions if completion
14101         is set.
14102         * opt-suggestions.c (option_proposer::get_completions):
14103         New function.
14104         (option_proposer::suggest_completion): Likewise.
14105         (option_proposer::find_param_completions): Likewise.
14106         (verify_autocompletions): Likewise.
14107         (test_completion_valid_options): Likewise.
14108         (test_completion_valid_params): Likewise.
14109         (in_completion_p): Likewise.
14110         (empty_completion_p): Likewise.
14111         (test_completion_partial_match): Likewise.
14112         (test_completion_garbage): Likewise.
14113         (opt_proposer_c_tests): Likewise.
14114         * opt-suggestions.h: Declare new functions.
14115         * opts.c (common_handle_option): Handle OPT__completion_.
14116         * selftest-run-tests.c (selftest::run_tests): Add
14117         opt_proposer_c_tests.
14118         * selftest.c (assert_str_startswith): New.
14119         * selftest.h (assert_str_startswith): Likewise.
14120         (opt_proposer_c_tests): New.
14121         (ASSERT_STR_STARTSWITH): Likewise.
14123 2018-06-28  Martin Liska  <mliska@suse.cz>
14125         * Makefile.in: Add opt-suggestions.o.
14126         * gcc-main.c: Include opt-suggestions.h.
14127         * gcc.c (driver::driver): Likewise.
14128         (driver::~driver): Remove m_option_suggestions.
14129         (driver::build_option_suggestions): Moved to option_proposer.
14130         (driver::suggest_option): Likewise.
14131         (driver::handle_unrecognized_options): Use option_proposer.
14132         * gcc.h (class driver): Add new memver m_option_proposer.
14133         * opt-suggestions.c: New file.
14134         * opt-suggestions.h: New file.
14136 2018-06-28  Martin Liska  <mliska@suse.cz>
14138         * vec.h (class auto_string_vec): New (moved from auto_argvec).
14139         (auto_string_vec::~auto_string_vec): Likewise.
14141 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14143         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14144         prevent_decl_creation_for_types fields up and add reset_location field.
14145         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14146         statement if id->reset_location is true.
14147         (copy_edges_for_bb): Do not set goto_locus on the new edges if
14148         id->reset_location is true.
14149         (copy_phis_for_bb): Force input_location on the arguments if
14150         id->reset_location is true.
14151         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14152         is set on the function to be inlined.
14154 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
14156         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14158 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
14160         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14161         registers for Pmode.
14162         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14163         hard registers for the clobbered pseudo.
14165 2018-06-27  Paul Koning  <ni1d@arrl.net>
14167         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14168         mutually exclusive options.
14169         * config/pdp11/constraints.md (h): New constraint.
14170         (O): Update definition to match shift code generation.
14171         (D): New constraint.
14172         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14173         (CCFP): Remove.
14174         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14175         function.
14176         (output_jump): Change arguments.
14177         (pdp11_fixed_cc_regs): New function.
14178         (pdp11_cc_mode): Ditto.
14179         (pdp11_expand_shift): Ditto.
14180         (pdp11_assemble_shift): Ditto.
14181         (pdp11_small_shift): Ditto.
14182         (pdp11_branch_cost): Remove.
14183         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14184         from output.
14185         (pdp11_register_move_cost): Update for CC registers.
14186         (pdp11_rtx_costs): Add case for LSHIFTRT.
14187         (pdp11_output_jump): Add CCNZ mode conditional branches.
14188         (notice_update_cc_on_set): Remove.
14189         (pdp11_cc_mode): New function.
14190         (simple_memory_operand): Correct pre/post decrement case.
14191         (no_side_effect_operand): New function.
14192         (pdp11_regno_reg_class): Add CC_REGS class.
14193         (pdp11_fixed_cc_regs): New function.
14194         (pdp11_small_shift): New function.
14195         (pdp11_expand_shift): New function to expand shift insns.
14196         (pdp11_assemble_shift): New function to output shifts.
14197         (pdp11_branch_cost): Remove.
14198         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14199         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14200         (WCHAR_TYPE): Ditto.
14201         (PTRDIFF_TYPE): Ditto.
14202         (ADJUST_INSN_LENGTH): New macro.
14203         (FIXED_REGISTERS): Add CC registers.
14204         (CALL_USED_REGISTERS): Ditto.
14205         (reg_class): Ditto.
14206         (REG_CLASS_NAMES): Ditto.
14207         (REG_CLASS_CONTENTS): Ditto.
14208         (SELECT_CC_MODE): Use new function.
14209         (TARGET_FLAGS_REGNUM): New macro.
14210         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14211         (cc0_reg_rtx): Remove.
14212         (CC_STATUS_MDEP): Remove.
14213         (CC_STATUS_MDEFP_INIT): Remove.
14214         (CC_IN_FPU): Remove.
14215         (NOTICE_UPDATE_CC): Remove.
14216         (REGISTER_NAMES): Add CC registers.
14217         (BRANCH_COST): Change to constant 1.
14218         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14219         handling.
14220         * config/pdp11/pdp11.opt (mbcopy): Remove.
14221         (mbcopy-builtin): Remove.
14222         (mbranch-cheap): Remove.
14223         (mbranch-expensive): Remove.
14224         * config/pdp11/predicates.md (expand_shift_operand): Update to
14225         match shift code generation.
14226         (ccnz_operator): New predicate.
14227         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14228         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14229         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14230         exclusive options.
14231         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14232         description of O constraint.
14234 2018-06-27  Jeff Law  <law@redhat.com>
14235             Austin Law  <austinklaw@gmail.com>
14237         * config/v850/v850.md (addsi3_set_flags): New pattern.
14238         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14239         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14240         (zero_extendhisi2_v850_set_flags): Likewise.
14241         (zero_extendqisi2_v850_set_flags): Likewise.
14242         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14243         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14244         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14246         * config/v850/v850-protos.h (notice_update_cc): Remove.
14247         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14248         (v850_print_operand): Handle 'D' and "d".
14249         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14250         Add handling of arithmetic/logical operations compared against zero.
14251         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14252         Do not look at v850_compare_op, instead get mode from last argument.
14253         (v850_gen_compare): Remove
14254         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14255         after reload for prologue insns.
14256         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14257         patterns.
14258         (construct_save_jarl): Likewise.
14259         (TARGET_FLAGS_REGNUM): Define.
14260         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14261         (NOTICE_UPDATE_CC): Remove.
14262         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14263         than cc0.  Conditionalize on reload_completed.
14264         (cmpsi_insn, setfcc_insn): Likewise.
14265         (tst1 splitter): Turn into define_and_split which sets the flags
14266         after reload.
14267         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14268         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14269         (cstoresf4, cstoredf4): Clobber the flags.
14270         (cmpsi, cmpsf, cmpdf): Remove expanders.
14271         (setf_insn): Remove pattern.
14272         (addsi3): Turn into define_and_split which clobbers the flags after
14273         reload and a suitable pattern (addsi3_clobber_flags) for use after
14274         reload.
14275         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14276         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14277         (ashrsi3, ashrsi3_v850e2): Likewise.
14278         (bins): Clobber the flags.
14279         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14280         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14281         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14282         (call_value_internal_short, call_value_internal_long): Likewise.
14283         (callt_save_interrupt, callt_return_interrupt): Likewise.
14284         (save_interrupt, return_interrupt): Likewise.
14285         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14286         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14287         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14288         (All FP comparisons): Only allow after reload has completed.
14289         (trfsr): Likewise.
14290         (divh, divhu): Tweak output template.
14291         (branch_z_normal, branch_z_invert): Remove
14292         (branch_nz_normal, branch_nz_invert): Likewise.
14293         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14295         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14296         * config/v850/v850.c (notice_update_cc): Remove.
14297         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14298         (CC_NO_CARRY): Likewise.
14299         (NOTICE_UPDATE_CC): Define to nothing.
14300         * config/v850/v850.md: Remove block comment on cc0 handling
14301         Remove "cc" attribute from all patterns.  Remove cc_status handling
14302         from all patterns.  Minor formatting fixes.
14304 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14306         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14307         (cortex-a76.cortex-a55): Likewise.
14308         * config/aarch64/aarch64-tune.md: Regenerate.
14309         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14310         cortex-a76.cortex-a55.
14312 2018-06-27  Jeff Law  <law@redhat.com>
14314         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14315         (MULTILIB_DIRNAMES): Similarly.
14317 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14319         * gimple.h (gimple_return_retbnd): Delete.
14320         (gimple_return_set_retbnd): Likewise.
14321         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14322         gimple_return_set_retbnd.
14323         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14324         gimple_return_retbnd and adjust.
14325         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14326         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14327         Explicitly return NULL in a couple more cases.  Move assertion
14328         on debug statements and remove unreachable code.
14329         (reset_debug_binding): Do not test id->retbnd.
14330         (expand_call_inline): Do not set it.
14332 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14334         * configure.ac: Add --disable-gcov option.
14335         * configure: Regenerate.
14336         * Makefile.in: Honour @enable_gcov@.
14337         * doc/install.texi: Document --disable-gcov.
14339 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14341         * config/arm/arm-cpus.in (cortex-a76): New entry.
14342         (cortex-a76.cortex-a55): Likewise.
14343         * config/arm/arm-tables.opt: Regenerate.
14344         * config/arm/arm-tune.md: Likewise.
14345         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14346         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14347         cortex-a76.cortex-a55.
14349 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14351         PR target/85769
14352         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14354 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14356         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14357         comment.
14358         (EPILOGUE_USES): Likewise.
14360 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14362         * tree-inline.c (remap_location): New function extracted from...
14363         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14364         (copy_phis_for_bb): ...here.  Call remap_location.
14365         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14367 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14369         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14370         unaligned vsx for 16B memset.
14372 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14374         PR target/86285
14375         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14376         ieee128_float_type_node to long_double_type_node unless
14377         TARGET_LONG_DOUBLE_128 is set.
14379 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14381         * cfgloop.c (get_loop_location): Convert return type from
14382         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14383         by implicit construction from rtx_insn *, and using
14384         dump_user_location_t::from_function_decl for the fallback case.
14385         * cfgloop.h (get_loop_location): Convert return type from
14386         location_t to dump_user_location_t.
14387         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14388         dump_printf_loc to pass in a dump_location_t rather than a
14389         location_t, via the gimple stmt.
14390         * coverage.c (get_coverage_counts): Update calls to
14391         dump_printf_loc to pass in dump_location_t rather than a
14392         location_t.
14393         * doc/optinfo.texi (Dump types): Convert example of
14394         dump_printf_loc from taking "locus" to taking "insn".  Update
14395         description of the "_loc" calls to cover dump_location_t.
14396         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14397         "selftest.h".
14398         (dump_user_location_t::dump_user_location_t): New constructors,
14399         from gimple *stmt and rtx_insn *.
14400         (dump_user_location_t::from_function_decl): New function.
14401         (dump_loc): Make static.
14402         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14403         const dump_location_t &.
14404         (dump_generic_expr_loc): Delete.
14405         (dump_printf_loc): Convert param "loc" from location_t to
14406         const dump_location_t &.
14407         (selftest::test_impl_location): New function.
14408         (selftest::dumpfile_c_tests): New function.
14409         * dumpfile.h: Include "profile-count.h".
14410         (class dump_user_location_t): New class.
14411         (struct dump_impl_location_t): New struct.
14412         (class dump_location_t): New class.
14413         (dump_printf_loc): Convert 2nd param from source_location to
14414         const dump_location_t &.
14415         (dump_generic_expr_loc): Delete.
14416         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14417         const dump_location_t &.
14418         * gimple-fold.c (fold_gimple_assign): Update call to
14419         dump_printf_loc to pass in a dump_location_t rather than a
14420         location_t, via the gimple stmt.
14421         (gimple_fold_call): Likewise.
14422         * gimple-loop-interchange.cc
14423         (loop_cand::analyze_iloop_reduction_var): Update for change to
14424         check_reduction_path.
14425         (tree_loop_interchange::interchange): Update for change to
14426         find_loop_location.
14427         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14428         change in return-type of find_loop_location.
14429         (graphite_regenerate_ast_isl): Likewise.
14430         * graphite-optimize-isl.c (optimize_isl): Likewise.
14431         * graphite.c (graphite_transform_loops): Likewise.
14432         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14433         pass in a dump_location_t rather than a location_t, via the
14434         gimple stmt.
14435         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14436         * ipa.c (walk_polymorphic_call_targets): Likewise.
14437         * loop-unroll.c (report_unroll): Convert "locus" param from
14438         location_t to dump_location_t.
14439         (decide_unrolling): Update for change to get_loop_location's
14440         return type.
14441         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14442         location_t to dump_user_location_t.
14443         (grid_find_single_omp_among_assignments_1): Updates calls to
14444         dump_printf_loc to pass in a dump_location_t rather than a
14445         location_t, via the gimple stmt.
14446         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14447         location_t to dump_location_t.  Updates calls to dump_printf_loc
14448         to pass in a dump_location_t rather than a location_t, via the
14449         gimple stmt.
14450         (grid_inner_loop_gridifiable_p): Likewise.
14451         (grid_dist_follows_simple_pattern): Likewise.
14452         (grid_gfor_follows_tiling_pattern): Likewise.
14453         (grid_target_follows_gridifiable_pattern): Likewise.
14454         (grid_attempt_target_gridification): Convert initialization
14455         of local "grid" from memset to zero-initialization; FIXME: does
14456         this require C++11?  Update call to dump_printf_loc to pass in a
14457         optinfo_location rather than a location_t, via the gimple stmt.
14458         * profile.c (read_profile_edge_counts): Updates call to
14459         dump_printf_loc to pass in a dump_location_t rather than a
14460         location_t
14461         (compute_branch_probabilities): Likewise.
14462         * selftest-run-tests.c (selftest::run_tests): Call
14463         dumpfile_c_tests.
14464         * selftest.h (dumpfile_c_tests): New decl.
14465         * tree-loop-distribution.c (pass_loop_distribution::execute):
14466         Update for change in return type of find_loop_location.
14467         * tree-parloops.c (parallelize_loops): Likewise.
14468         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14469         "locus" from location_t to dump_user_location_t.
14470         (canonicalize_loop_induction_variables): Likewise.
14471         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14472         for change in return type of find_loop_location.
14473         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14474         to dump_printf_loc to pass in a dump_location_t rather than a
14475         location_t, via the stmt.
14476         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14477         Likewise.
14478         * tree-vect-loop-manip.c (find_loop_location): Convert return
14479         type from source_location to dump_user_location_t.
14480         (vect_do_peeling): Update for above change.
14481         (vect_loop_versioning): Update for change in type of
14482         vect_location.
14483         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14484         from location_t to dump_user_location_t.
14485         (vect_estimate_min_profitable_iters): Update for change in type
14486         of vect_location.
14487         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14488         location_t to dump_location_t.
14489         (vect_slp_bb): Update for change in type of vect_location.
14490         * tree-vectorizer.c (vect_location): Convert from source_location
14491         to dump_user_location_t.
14492         (try_vectorize_loop_1): Update for change in vect_location's type.
14493         (vectorize_loops): Likewise.
14494         (increase_alignment): Likewise.
14495         * tree-vectorizer.h (vect_location): Convert from source_location
14496         to dump_user_location_t.
14497         (find_loop_location): Convert return type from source_location to
14498         dump_user_location_t.
14499         (check_reduction_path): Convert 1st param from location_t to
14500         dump_user_location_t.
14501         * value-prof.c (check_counter): Update call to dump_printf_loc to
14502         pass in a dump_user_location_t rather than a location_t; update
14503         call to error_at for change in type of "locus".
14504         (check_ic_target): Update call to dump_printf_loc to
14505         pass in a dump_user_location_t rather than a location_t, via the
14506         call_stmt.
14508 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14510         * config/s390/s390.h (enum processor_flags): Do not use
14511         default tune parameter when -march was specified.
14513 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14515         PR target/86314
14516         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14517         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14518         operands.
14520 2018-06-26  Richard Biener  <rguenther@suse.de>
14522         PR tree-optimization/86287
14523         PR bootstrap/86316
14524         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14525         (vect_analyze_loop): Initialize n_stmts.
14527 2018-06-26  Richard Biener  <rguenther@suse.de>
14529         PR middle-end/86271
14530         * fold-const.c (fold_convertible_p): Pointer extension
14531         isn't valid.
14533 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14535         PR debug/86064
14536         * dwarf2out.c (loc_list_has_views): Adjust comments.
14537         (dw_loc_list): Split single cross-partition range with
14538         nonzero locview.
14540 2018-06-25  Jeff Law  <law@redhat.com>
14542         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14543         on -mbig-switch by default.
14545         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14546         test.
14547         (const_float_0_operand): Remove unused predicate.
14548         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14549         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14550         (recipsf2): New expander.  Original pattern now called
14551         (recipsf2_insn).
14552         (recipdf2, recipdf2_insn): Similarly.
14553         (rsqrtsf2, rsqrtsf2_insn): Similarly
14554         (rsqrtdf2, rsqrtdf2_insn): Similarly
14556 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14558         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14559         Simplify logic for FreeBSD (twice).
14561 2018-06-25  Martin Sebor  <msebor@redhat.com>
14563         PR tree-optimization/86204
14564         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14565         a strnlen result if it's less than the length of the string.
14567 2018-06-25  Martin Sebor  <msebor@redhat.com>
14569         PR tree-optimization/85700
14570         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14571         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14572         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14574 2018-06-25  Martin Sebor  <msebor@redhat.com>
14576         * doc/extend.texi (Zero-length arrays): Update and clarify.
14578 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14580         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14581         added IEEE/IBM long double multilib support on PowerPC little
14582         endian Linux systems.
14583         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14584         (MULTILIB_DEFAULTS): Likewise.
14585         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14586         Likewise.
14587         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14588         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14589         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14591 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14593         PR middle-end/86311
14594         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14595         (REORDER_45): Likewise.
14597 2018-06-25  Jeff Law  <law@redhat.com>
14599         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14600         dividend to 32 bits.  Adjust length.
14601         (udivmodhi4): Cleanup output template.  Fix length.
14603 2018-06-25  Carl Love  <cel@us.ibm.com>
14605         * config/rs6000/vsx.md: Change word selector to prefered location.
14607 2018-06-25  Richard Biener  <rguenther@suse.de>
14609         PR tree-optimization/86304
14610         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14611         epilogue-if-converted loops as well.
14613 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14615         * lto-section-out.c (lto_begin_section): Do not print section
14616         name for noaddr and unnumbered dumps.
14618 2018-06-25  Richard Biener  <rguenther@suse.de>
14620         * tree-vectorizer.h (struct vec_info_shared): New structure
14621         with parts split out from struct vec_info and loop_nest from
14622         struct _loop_vec_info.
14623         (struct vec_info): Adjust accordingly.
14624         (struct _loop_vec_info): Likewise.
14625         (LOOP_VINFO_LOOP_NEST): Adjust.
14626         (LOOP_VINFO_DATAREFS): Likewise.
14627         (LOOP_VINFO_DDRS): Likewise.
14628         (struct _bb_vec_info): Likewise.
14629         (BB_VINFO_DATAREFS): Likewise.
14630         (BB_VINFO_DDRS): Likewise.
14631         (struct _stmt_vec_info): Add dr_aux member.
14632         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14633         (DR_MISALIGNMENT_UNINITIALIZED): New.
14634         (set_dr_misalignment): Adjust.
14635         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14636         (vect_analyze_loop): Adjust prototype.
14637         (vect_analyze_loop_form): Likewise.
14638         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14639         Compute dependences lazily.
14640         (vect_record_base_alignments): Use shared datarefs/ddrs.
14641         (vect_verify_datarefs_alignment): Likewise.
14642         (vect_analyze_data_refs_alignment): Likewise.
14643         (vect_analyze_data_ref_accesses): Likewise.
14644         (vect_analyze_data_refs): Likewise.
14645         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14646         constructor parameter for shared part.
14647         (vect_analyze_loop_form): Pass in shared part and adjust.
14648         (vect_analyze_loop_2): Pass in storage for the number of
14649         stmts.  Move loop nest finding to the caller.  Compute
14650         datarefs lazily.
14651         (vect_analyze_loop): Pass in shared part.
14652         (vect_transform_loop): Verify shared datarefs are unchanged.
14653         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14654         constructor parameter for shared part.
14655         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14656         (vect_slp_bb): Verify shared datarefs are unchanged before
14657         transform.
14658         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14659         change.
14660         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14661         DR_MISALIGNMENT_UNINITIALIZED.
14662         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14663         parameter for shared part.
14664         (vec_info::~vec_info): Adjust.
14665         (vec_info_shared::vec_info_shared): New.
14666         (vec_info_shared::~vec_info_shared): Likewise.
14667         (vec_info_shared::save_datarefs): Likewise.
14668         (vec_info_shared::check_datarefs): Likewise.
14669         (try_vectorize_loop_1): Construct shared part live for analyses
14670         of a single loop for multiple vector sizes.
14671         * tree-parloops.c (gather_scalar_reductions): Adjust.
14673 2018-06-25  Richard Biener  <rguenther@suse.de>
14675         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14676         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14677         (vect_analyze_data_refs): Remove similar code from here and
14678         simplify accordingly.
14680 2018-06-25  Richard Biener  <rguenther@suse.de>
14682         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14683         for reverse storage order accesses rather than asserting
14684         they cannot happen here.
14686 2018-06-25  Tom de Vries  <tdevries@suse.de>
14688         PR debug/86257
14689         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14690         Use data16 instead of .byte for insn prefix.
14692 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14694         PR C++/86082
14695         * parser.c (make_char_string_pack): Pass this literal chars
14696         through cpp_interpret_string.
14697         (cp_parser_userdef_numeric_literal): Check the result of
14698         make_char_string_pack.
14700 2018-06-24  Maya Rashish  <coypu@sdf.org>
14702         * ginclude/stddef.h: Simplify conditions around avoiding
14703         re-definition of __size_t.
14705 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14707         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14708         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14710 2018-06-22  Maya Rashish  <coypu@sdf.org>
14712         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14713         made to OpenBSD and NetBSD through the years.
14715 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14717         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14718         behavior of vec_pack (vector double, vector double) to match
14719         behavior of vec_float2 (vector double, vector double).
14721 2018-06-22  Olivier Hainque  <hainque@adacore.com>
14723         * gimplify.c (gimplify_function_tree): Prevent creation
14724         of a trampoline for the address of the current function
14725         passed to entry/exit instrumentation hooks.
14727 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
14729         PR target/86222
14730         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
14731         correctly.
14733 2018-06-22  Martin Liska  <mliska@suse.cz>
14735         PR tree-optimization/86263
14736         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14737         Bail out if is_enabled is false.
14738         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
14739         New declaration.
14740         (jump_table_cluster::is_enabled): New function.
14742 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14744         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14745         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
14746         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
14747         (lto_input_ts_binfo_tree_pointers): Likewise.
14748         * tree-streamer-out.c (streamer_write_tree_bitfields,
14749         write_ts_binfo_tree_pointers): Likewise.
14750         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
14752 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14754         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
14756 2018-06-22  Martin Liska  <mliska@suse.cz>
14758         * symbol-summary.h (get): Make it pure and inline move
14759         functionality from ::get function.
14760         (get): Remove and inline into ::get and ::get_create.
14761         (get_create): Move code from ::get function.
14763 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14765         PR target/85994
14766         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
14767         -x assembler-with-cpp.
14769 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14771         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
14772         _FILE_OFFSET_BITS=64 for C++.
14774 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
14776         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
14777         conversion insn that shows up when pr85657-3.c is compiled using
14778         IEEE 128-bit long double.
14779         (neg<mode>2_internal): Use the correct mode to check whether the
14780         mode is IBM extended.
14781         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
14782         multiply and divide external functions from being created more
14783         than once.
14785 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14787         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
14788         functions.
14789         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
14790         the edge can be forwarded.
14791         (cfg_layout_merge_blocks): Likewise.
14793 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14795         * except.c (finish_eh_generation): Commit edge insertions only after
14796         the EH edges have been redirected from post-landing to landing pads.
14798 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14800         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
14801         create_tmp_var_for to create the FRAME decl.
14802         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
14804 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14806         * tree-inline.c (copy_edges_for_bb): Minor tweak.
14807         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
14808         debug statement when resetting its value.
14809         (expand_call_inline): Copy the locus of the call onto the assignment
14810         of the return value, if any.  Use local variable in more cases.
14812 2018-06-21  Martin Liska  <mliska@suse.cz>
14814         * ipa-pure-const.c (propagate_nothrow): Use
14815         funct_state_summaries->get.
14816         (dump_malloc_lattice): Likewise.
14817         (propagate_malloc): Likewise.
14819 2018-06-21  Richard Biener  <rguenther@suse.de>
14821         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
14822         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
14823         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
14824         comment.
14825         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
14826         BLOCK_ABSTRACT_ORIGIN unconditionally.
14828 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
14830         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
14831         deleting it.
14832         * ipa-reference.c (ipa_reference_c_finalize): Delete
14833         ipa_ref_opt_sum_summaries and set it to NULL.
14835 2018-06-21  Tom de Vries  <tdevries@suse.de>
14837         PR tree-optimization/85859
14838         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
14839         test with comment from bb_no_side_effects_p.
14841 2018-06-21  Richard Biener  <rguenther@suse.de>
14843         PR tree-optimization/86232
14844         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
14845         max for constant niter.
14847 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14849         * config/aarch64/aarch64-simd.md
14850         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
14852 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14854         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
14855         Make opernads of the unspec commutative.
14857 2018-06-21  Richard Biener  <rguenther@suse.de>
14859         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
14860         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14861         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
14862         (vect_analyze_data_ref_dependence): Re-order checks to deal with
14863         NULL DR_STEP.
14864         (vect_record_base_alignments): Do not record base alignment
14865         for gathers or scatters.
14866         (vect_compute_data_ref_alignment): Drop return value that is always
14867         true.  Bail out early for gathers or scatters.
14868         (vect_enhance_data_refs_alignment): Bail out early for gathers
14869         or scatters.
14870         (vect_find_same_alignment_drs): Likewise.
14871         (vect_analyze_data_refs_alignment): Remove dead code.
14872         (vect_slp_analyze_and_verify_node_alignment): Likewise.
14873         (vect_analyze_data_refs): For possible gathers or scatters do
14874         not create an alternate DR, just check their possible validity
14875         and mark them.  Adjust DECL_NONALIASED handling to not rely
14876         on DR_BASE_ADDRESS.
14877         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
14878         update inits of gathers or scatters.
14879         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
14880         Also copy gather/scatter flag to pattern vinfo.
14882 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14884         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14885         behavior of vec_packsu (vector unsigned long long, vector unsigned
14886         long long) to match behavior of vec_packs with same signature.
14888 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
14889             Thomas Schwinge <thomas@codesourcery.com>
14890             Cesar Philippidis  <cesar@codesourcery.com>
14892         * gimplify.c (gimplify_scan_omp_clauses): Add support for
14893         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14894         (gimplify_adjust_omp_clauses): Likewise.
14895         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
14896         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
14897         (gimplify_omp_target_update): Update handling of acc update and
14898         enter/exit data.
14899         * omp-low.c (install_var_field): Remove unused parameter
14900         base_pointers_restrict.
14901         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
14902         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
14903         FINALIZE}
14904         (omp_target_base_pointers_restrict_p): Delete.
14905         (scan_omp_target): Update call to scan_sharing_clauses.
14906         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
14907         FINALIZE}.
14908         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
14909         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14910         (convert_local_omp_clauses): Likewise.
14911         * tree-pretty-print.c (dump_omp_clause): Likewise.
14912         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
14913         FINALIZE}.
14914         (omp_clause_code_name): Likewise.
14916 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14918         PR debug/86194
14919         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
14920         be narrowed.
14922         PR tree-optimization/86231
14923         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
14924         anti-range don't overwrite *vr0min before using it to compute *vr0max.
14926 2018-06-20  Tom de Vries  <tdevries@suse.de>
14928         PR tree-optimization/86097
14929         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
14930         iv type if signedness of iv type is not the same as that of *nit.
14932 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14934         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
14935         EDGE_EH edges, verify they are all EDGE_EH.
14937 2018-06-20  Maya Rashish  <coypu@sdf.org>
14939         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
14941 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14943         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
14944         * config/aarch64/aarch64.c (xgene1_tunings): Add
14945         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
14946         (aarch64_mode_valid_for_sched_fusion_p):
14947         Allow 16-byte modes.
14948         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
14949         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
14950         128-bit modes.
14951         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
14952         New pattern.
14953         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
14954         * config/aarch64/iterators.md (VQ2): New mode iterator.
14956 2018-06-20  Martin Liska  <mliska@suse.cz>
14958         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14959         Change default ratio from 10 to 8.
14961 2018-06-20  Martin Liska  <mliska@suse.cz>
14963         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14964         New.
14965         (bit_test_cluster::find_bit_tests): Likewise.
14966         (switch_decision_tree::analyze_switch_statement): Find clusters.
14967         * tree-switch-conversion.h (struct jump_table_cluster): Document
14968         hierarchy.
14970 2018-06-20  Martin Liska  <mliska@suse.cz>
14972         * tree-switch-conversion.c (switch_conversion::collect):
14973         Record m_uniq property.
14974         (switch_conversion::expand): Bail out for special conditions.
14975         (group_cluster::~group_cluster): New.
14976         (group_cluster::group_cluster): Likewise.
14977         (group_cluster::dump): Likewise.
14978         (jump_table_cluster::emit): New.
14979         (switch_decision_tree::fix_phi_operands_for_edges): New.
14980         (struct case_node): Remove struct.
14981         (jump_table_cluster::can_be_handled): New.
14982         (case_values_threshold): Moved to header.
14983         (reset_out_edges_aux): Likewise.
14984         (jump_table_cluster::is_beneficial): New.
14985         (bit_test_cluster::can_be_handled): Likewise.
14986         (add_case_node): Remove.
14987         (bit_test_cluster::is_beneficial): New.
14988         (case_bit_test::cmp): New.
14989         (bit_test_cluster::emit): New.
14990         (expand_switch_as_decision_tree_p): Remove.
14991         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
14992         (fix_phi_operands_for_edge): Likewise.
14993         (switch_decision_tree::analyze_switch_statement): New.
14994         (compute_cases_per_edge): Move ...
14995         (switch_decision_tree::compute_cases_per_edge): ... here.
14996         (try_switch_expansion): Likewise.
14997         (switch_decision_tree::try_switch_expansion): Likewise.
14998         (record_phi_operand_mapping): Likewise.
14999         (switch_decision_tree::record_phi_operand_mapping): Likewise.
15000         (emit_case_decision_tree): Likewise.
15001         (switch_decision_tree::emit): Likewise.
15002         (balance_case_nodes): Likewise.
15003         (switch_decision_tree::balance_case_nodes): Likewise.
15004         (dump_case_nodes): Likewise.
15005         (switch_decision_tree::dump_case_nodes): Likewise.
15006         (emit_jump): Likewise.
15007         (switch_decision_tree::emit_jump): Likewise.
15008         (emit_cmp_and_jump_insns): Likewise.
15009         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15010         (emit_case_nodes): Likewise.
15011         (switch_decision_tree::emit_case_nodes): Likewise.
15012         (conditional_probability): Remove.
15013         * tree-switch-conversion.h (enum cluster_type): New.
15014         (PRINT_CASE): New.
15015         (struct cluster): Likewise.
15016         (cluster::cluster): Likewise.
15017         (struct simple_cluster): Likewise.
15018         (simple_cluster::simple_cluster): Likewise.
15019         (struct group_cluster): Likewise.
15020         (struct jump_table_cluster): Likewise.
15021         (struct bit_test_cluster): Likewise.
15022         (struct min_cluster_item): Likewise.
15023         (struct case_tree_node): Likewise.
15024         (case_tree_node::case_tree_node): Likewise.
15025         (jump_table_cluster::case_values_threshold): Likewise.
15026         (struct case_bit_test): Likewise.
15027         (struct switch_decision_tree): Likewise.
15028         (struct switch_conversion): Likewise.
15029         (switch_decision_tree::reset_out_edges_aux): Likewise.
15031 2018-06-20  Martin Liska  <mliska@suse.cz>
15033         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15034         (hoist_edge_and_branch_if_true): Likewise.
15035         (expand_switch_using_bit_tests_p): Likewise.
15036         (struct case_bit_test): Likewise.
15037         (case_bit_test_cmp): Likewise.
15038         (emit_case_bit_tests): Likewise.
15039         (switch_conversion::switch_conversion): New class.
15040         (struct switch_conv_info): Remove old struct.
15041         (collect_switch_conv_info): More to ...
15042         (switch_conversion::collect): ... this.
15043         (check_range): Likewise.
15044         (switch_conversion::check_range): Likewise.
15045         (check_all_empty_except_final): Likewise.
15046         (switch_conversion::check_all_empty_except_final): Likewise.
15047         (check_final_bb): Likewise.
15048         (switch_conversion::check_final_bb): Likewise.
15049         (create_temp_arrays): Likewise.
15050         (switch_conversion::create_temp_arrays): Likewise.
15051         (free_temp_arrays): Likewise.
15052         (gather_default_values): Likewise.
15053         (switch_conversion::gather_default_values): Likewise.
15054         (build_constructors): Likewise.
15055         (switch_conversion::build_constructors): Likewise.
15056         (constructor_contains_same_values_p): Likewise.
15057         (switch_conversion::contains_same_values_p): Likewise.
15058         (array_value_type): Likewise.
15059         (switch_conversion::array_value_type): Likewise.
15060         (build_one_array): Likewise.
15061         (switch_conversion::build_one_array): Likewise.
15062         (build_arrays): Likewise.
15063         (switch_conversion::build_arrays): Likewise.
15064         (gen_def_assigns): Likewise.
15065         (switch_conversion::gen_def_assigns): Likewise.
15066         (prune_bbs): Likewise.
15067         (switch_conversion::prune_bbs): Likewise.
15068         (fix_phi_nodes): Likewise.
15069         (switch_conversion::fix_phi_nodes): Likewise.
15070         (gen_inbound_check): Likewise.
15071         (switch_conversion::gen_inbound_check): Likewise.
15072         (process_switch): Use the newly created class.
15073         (switch_conversion::expand): New.
15074         (switch_conversion::~switch_conversion): New.
15075         * tree-switch-conversion.h: New file.
15077 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15079         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
15080         tree-vect-patterns.c.
15081         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
15082         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
15083         (vect_recog_sad_pattern): Likewise.
15084         (vect_recog_widen_sum_pattern): Likewise.
15085         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
15086         (vect_recog_widen_shift_pattern): Remove the type_in argument.
15087         (vect_recog_rotate_pattern): Likewise.
15088         (vect_recog_mult_pattern): Likewise.
15089         (vect_recog_vector_vector_shift_pattern): Likewise.
15090         (vect_recog_divmod_pattern): Likewise.
15091         (vect_recog_mixed_size_cond_pattern): Likewise.
15092         (vect_recog_bool_pattern): Likewise.
15093         (vect_recog_mask_conversion_pattern): Likewise.
15094         (vect_try_gather_scatter_pattern): Likewise.
15095         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
15096         (vect_recog_over_widening_pattern): Likewise.
15097         (vect_recog_gather_scatter_pattern): Likewise.
15098         (vect_recog_func_ptr): Move from tree-vectorizer.h
15099         (vect_vect_recog_func_ptrs): Move further down the file.
15100         (vect_recog_func): Likewise.  Remove the third argument.
15101         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
15102         (vect_pattern_recog_1): Expect the pattern function to do any
15103         necessary target tests.  Also expect it to provide a vector type.
15104         Remove the type_in handling.
15106 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15108         * tree-vect-patterns.c (vect_pattern_detected): New function.
15109         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
15110         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
15111         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
15112         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
15113         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15114         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15115         (vect_recog_mask_conversion_pattern)
15116         (vect_try_gather_scatter_pattern): Likewise.
15118 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15120         * tree-vect-patterns.c (vect_get_internal_def): New function.
15121         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15122         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
15123         (search_type_for_mask_1): Use it.
15125 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15127         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15128         redundant WIDEN_SUM_EXPR handling.
15129         (vect_recog_sad_pattern): Likewise.
15131 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15133         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15134         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15135         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15136         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15137         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15139 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15141         * tree-vect-stmts.c (vectorizable_call): Make sure that we
15142         use the stmt_vec_info of the original bb statement for the
15143         new zero assignment, even if the call is part of a pattern.
15145 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15147         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15148         that the sequence is attached to the original statement rather
15149         than the pattern statement.
15150         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15151         PATTERN_DEF_SEQ from the original statement rather than
15152         the main pattern statement.
15153         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15154         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15155         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15157 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15159         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15160         definition statements before the early exit for statements that aren't
15161         live or relevant.
15162         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15163         split out from...
15164         (vect_transform_loop): ...here.  Process pattern definition
15165         statements without first checking whether the main pattern
15166         statement is live or relevant.
15168 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15170         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15171         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15173 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
15175         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15176         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15177         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15178         (expand_block_compare): Change select_block_compare_mode call.
15179         (expand_strncmp_align_check): Use new functions, fix comment.
15180         (emit_final_str_compare_gpr): New function.
15181         (expand_strn_compare): Refactor and clean up code.
15182         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15184 2018-06-19  Tony Reix  <tony.reix@atos.com>
15185             Damien Bergamini  <damien.bergamini@atos.com>
15186             David Edelsohn  <dje.gcc@gmail.com>
15188         * collect2.c (static_obj): New variable.
15189         (static_libs): New variable.
15190         (is_in_list): Uncomment declaration.
15191         (main): Track AIX libraries linked statically.
15192         (is_in_list): Uncomment definition.
15193         (scan_prog_file): Don't add AIX shared libraries initializer
15194         to constructor list if linking statically.
15196 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
15198         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15199         constant.
15200         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15202 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
15204         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15205         blocks.
15207 2018-06-19  Martin Liska  <mliska@suse.cz>
15209         * config/i386/i386.c (ix86_can_inline_p): Do not use
15210         ipa_fn_summaries::get_create.
15211         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15212         get.
15213         (devirtualization_time_bonus): Likewise.
15214         (ipcp_propagate_stage): Likewise.
15215         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15216         (edge_set_predicate): Likewise.
15217         (evaluate_conditions_for_known_args): Likewise.
15218         (evaluate_properties_for_edge): Likewise.
15219         (ipa_call_summary::reset): Tranform to ...
15220         (ipa_call_summary::~ipa_call_summary): ... this.
15221         (ipa_fn_summary::reset): Transform to ...
15222         (ipa_fn_summary::~ipa_fn_summary): ... this.
15223         (ipa_fn_summary_t::remove): Rename to ...
15224         (ipa_fn_summary_t::remove_callees): ... this.
15225         (ipa_fn_summary_t::duplicate): Use placement new
15226         instead of memory copy.
15227         (ipa_call_summary_t::duplicate): Likewise.
15228         (ipa_call_summary_t::remove): Remove.
15229         (dump_ipa_call_summary): Change get_create to get.
15230         (ipa_dump_fn_summary): Dump only when summary exists.
15231         (analyze_function_body): Use symbol_summary::get instead
15232         of get_create.
15233         (compute_fn_summary): Likewise.
15234         (estimate_edge_devirt_benefit): Likewise.
15235         (estimate_edge_size_and_time): Likewise.
15236         (inline_update_callee_summaries): Likewise.
15237         (remap_edge_change_prob): Likewise.
15238         (remap_edge_summaries): Likewise.
15239         (ipa_merge_fn_summary_after_inlining): Likewise.
15240         (write_ipa_call_summary): Likewise.
15241         (ipa_fn_summary_write): Likewise.
15242         (ipa_free_fn_summary): Likewise.
15243         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15244         (struct ipa_call_summary): Likewise.
15245         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15246         of get_create.
15247         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15248         (estimate_size_after_inlining): Likewise.
15249         (estimate_growth): Likewise.
15250         (growth_likely_positive): Likewise.
15251         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15252         (inline_call): Likewise.
15253         * ipa-inline.c (caller_growth_limits): Likewise.
15254         (can_inline_edge_p): Likewise.
15255         (can_inline_edge_by_limits_p): Likewise.
15256         (compute_uninlined_call_time): Likewise.
15257         (compute_inlined_call_time): Likewise.
15258         (want_inline_small_function_p): Likewise.
15259         (edge_badness): Likewise.
15260         (update_caller_keys): Likewise.
15261         (update_callee_keys): Likewise.
15262         (inline_small_functions): Likewise.
15263         (inline_to_all_callers_1): Likewise.
15264         (dump_overall_stats): Likewise.
15265         (early_inline_small_functions): Likewise.
15266         (early_inliner): Likewise.
15267         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15268         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15269         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15270         * ipa-split.c (execute_split_functions): Likewise.
15271         * symbol-summary.h: Likewise.
15272         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15274 2018-06-19  Richard Biener  <rguenther@suse.de>
15276         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15277         (vectorize_loops): ... here.  Fix dbgcnt handling.
15278         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15280 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15282         PR target/86197
15283         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15284         ieee128 argument takes up only one (vector) register, not two (floating
15285         point) registers.
15287 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15289         * gimplify.c (gimplify_init_constructor): Really never clear for an
15290         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15292 2018-06-19  Richard Biener  <rguenther@suse.de>
15294         PR tree-optimization/86179
15295         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15296         after failed recognition.
15298 2018-06-18  Martin Sebor  <msebor@redhat.com>
15300         PR middle-end/85602
15301         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15302         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15303         Handle integer subtraction.
15304         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15305         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15307 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15309         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15310         param from rtx to rtx_insn *.
15311         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15312         param.
15313         (frv_ifcvt_modify_insn): Likwise.
15314         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15315         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15316         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15317         as_a <rtx_insn *> cast to local "unprotected_region" once
15318         it's been established that it's not NULL or pc_rtx.
15319         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15320         param "sethi" from rtx to rtx_insn *.
15321         (nds32_group_float_insns): Likewise for param "insn".
15322         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15323         param.
15324         (vax_output_int_subtract): Likewise.
15325         * config/vax/vax.c (vax_output_int_add): Likewise for param
15326         "insn".
15327         (vax_output_int_subtract): Likewise.
15328         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15329         (emit_pattern_after): Likewise for param "after".
15330         (emit_insn_after): Likewise.
15331         (emit_jump_insn_after): Likewise.
15332         (emit_call_insn_after): Likewise.
15333         (emit_debug_insn_after): Likewise.
15334         (emit_pattern_before): Likewise for param "before".
15335         (emit_insn_before): Likewise.
15336         (emit_jump_insn_before): Likewise.
15337         * final.c (get_insn_template): Likewise for param "insn", removing
15338         a cast.
15339         * output.h (get_insn_template): Likewise for 2nd param.
15340         * rtl.h (emit_insn_before): Likewise.
15341         (emit_jump_insn_before): Likewise.
15342         (emit_debug_insn_before_noloc): Likewise.
15343         (emit_insn_after): Likewise.
15344         (emit_jump_insn_after): Likewise.
15345         (emit_call_insn_after): Likewise.
15346         (emit_debug_insn_after): Likewise.
15347         (set_insn_deleted): Likewise for param.
15349 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15351         PR target/85358
15352         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15353         floating point modes, so that IFmode is numerically greater than
15354         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15355         to declare the ordering.  This prevents IFmode from being
15356         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15357         machine.  Include rs6000-modes.h to share the fractional values
15358         between genmodes* and the rest of the compiler.
15359         (IFmode): Likewise.
15360         (KFmode): Likewise.
15361         (TFmode): Likewise.
15362         * config/rs6000/rs6000-modes.h: New file.
15363         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15364         meaning of rs6000_long_double_size so that 126..128 selects an
15365         appropriate 128-bit floating point type.
15366         (rs6000_option_override_internal): Likewise.
15367         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15368         (TARGET_LONG_DOUBLE_128): Change the meaning of
15369         rs6000_long_double_size so that 126..128 selects an appropriate
15370         128-bit floating point type.
15371         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15372         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15373         source and destination to match the standard usage.
15374         (truncifkf2): Likewise.
15375         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15376         ISA 2.07 to use an explicit clobber, instead of passing in a
15377         temporary.
15378         (copysign<mode>3_soft): Likewise.
15380 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15382         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15383         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15384         (vect_slp_analyze_instance_dependence): Likewise.
15385         (vect_enhance_data_refs_alignment): Likewise.
15386         (vect_analyze_data_refs_alignment): Likewise.
15387         (vect_slp_analyze_and_verify_instance_alignment
15388         (vect_analyze_data_ref_accesses): Likewise.
15389         (vect_prune_runtime_alias_test_list): Likewise.
15390         (vect_analyze_data_refs): Likewise.
15391         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15392         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15393         (vect_analyze_scalar_cycles_1): Likewise.
15394         (vect_get_loop_niters): Likewise.
15395         (vect_analyze_loop_form_1): Likewise.
15396         (vect_update_vf_for_slp): Likewise.
15397         (vect_analyze_loop_operations): Likewise.
15398         (vect_analyze_loop): Likewise.
15399         (vectorizable_induction): Likewise.
15400         (vect_transform_loop): Likewise.
15401         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15402         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15403         (vect_make_slp_decision): Likewise.
15404         (vect_detect_hybrid_slp): Likewise.
15405         (vect_slp_analyze_operations): Likewise.
15406         (vect_slp_bb): Likewise.
15407         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15408         (vectorizable_bswap): Likewise.
15409         (vectorizable_call): Likewise.
15410         (vectorizable_simd_clone_call): Likewise.
15411         (vectorizable_conversion): Likewise.
15412         (vectorizable_assignment): Likewise.
15413         (vectorizable_shift): Likewise.
15414         (vectorizable_operation): Likewise.
15415         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15417 2018-06-18  Martin Sebor  <msebor@redhat.com>
15419         PR tree-optimization/81384
15420         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15421         * builtins.c (expand_builtin_strnlen): New function.
15422         (expand_builtin): Call it.
15423         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15424         * builtins.def (BUILT_IN_STRNLEN): New.
15425         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15426         Warn for bounds in excess of maximum object size.
15427         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15428         single-value ranges.  Handle strnlen.
15429         (handle_builtin_strlen): Handle strnlen.
15430         (strlen_check_and_optimize_stmt): Same.
15431         * doc/extend.texi (Other Builtins): Document strnlen.
15433 2018-06-18  Maya Rashish  <coypu@sdf.org>
15435         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15436         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15437         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15439         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15440         here to ...
15441         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15443 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15445         * tree.c (escaped_string::escape): Replace cast to char * by
15446         const_cast<char *> (unescaped).
15448 2018-06-18  Nick Clifton  <nickc@redhat.com>
15450         PR 84195
15451         * tree.c (escaped_string): New class.  Converts an unescaped
15452         string into its escaped equivalent.
15453         (warn_deprecated_use): Use the new class to convert the
15454         deprecation message, if present.
15455         (test_escaped_strings): New self test.
15456         (test_c_tests): Add test_escaped_strings.
15457         * doc/extend.texi (deprecated): Add a note that the
15458         deprecation message is affected by the -fmessage-length
15459         option, and that control characters will be escaped.
15460         (#pragma GCC error): Document this pragma.
15461         (#pragma GCC warning): Likewise.
15462         * doc/invoke.texi (-fmessage-length): Document this option's
15463         effect on the #warning and #error preprocessor directives and
15464         the deprecated attribute.
15466 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15468         * tree.c (decl_value_expr_lookup): Revert latest change.
15469         (decl_value_expr_insert): Likewise.
15471 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15473         * gimplify.c (nonlocal_vlas): Delete.
15474         (nonlocal_vla_vars): Likewise.
15475         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15476         referenced VLAs.
15477         (gimplify_body): Do not create and destroy nonlocal_vlas.
15478         * tree-nested.c: Include diagnostic.h.
15479         (use_pointer_in_frame): Tweak.
15480         (lookup_field_for_decl): Add assertion and declare the transformation.
15481         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15482         internal error when the reference is in a wrong context.  Do not
15483         create a debug decl by default.
15484         (note_nonlocal_block_vlas): Delete.
15485         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15486         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15487         create a debug decl by default.
15488         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15489         call is in a wrong context.
15490         (fixup_vla_decls): New function.
15491         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15492         debug variables were created.
15493         * tree.c (decl_value_expr_lookup): Add checking assertion.
15494         (decl_value_expr_insert): Likewise.
15496 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15498         PR middle-end/82479
15499         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15500         * tree-scalar-evolution.c (interpret_expr): Likewise.
15501         (expression_expensive_p): Likewise.
15502         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15503         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15504         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15505         (ssa_defined_by_minus_one_stmt_p): New.
15507 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15509         PR middle-end/64946
15510         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15511         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15512         * dojump.c (do_jump): Likewise.
15513         * expr.c (expand_expr_real_2): Check operand type's sign.
15514         * fold-const.c (const_unop): Handle ABSU_EXPR.
15515         (fold_abs_const): Likewise.
15516         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15517         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15518         (strip_sign_op_1): Likesise.
15519         * match.pd: Add new pattern to generate ABSU_EXPR.
15520         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15521         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15522         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15523         * tree-inline.c (estimate_operator_cost): Likewise.
15524         * tree-pretty-print.c (dump_generic_node): Likewise.
15525         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15526         * tree.def (ABSU_EXPR): New.
15528 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15530         PR middle-end/86095
15531         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15532         documented as preserved for backward compatibility only.
15533         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15535         PR rtl-optimization/86108
15536         * bb-reorder.c (create_forwarder_block): Renamed to ...
15537         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15538         jump from new landing pad to the second part.
15539         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15540         Adjust callers.
15542 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15544         PR middle-end/85878
15545         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15546         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15547         Only call store_expr for halves if the mode is the same.
15549         PR middle-end/86123
15550         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15551         Fix up comment formatting.
15553 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15555         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15556         (typed_splay_tree::closure,
15557         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15558         (typed_splay_tree::typed_splay_tree,
15559         typed_splay_tree::operator =): Declared private.
15560         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15561         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15562         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15563         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15564         typed_splay_tree::splay_tree_splay,
15565         typed_splay_tree::splay_tree_foreach_helper,
15566         typed_splay_tree::splay_tree_insert,
15567         typed_splay_tree::splay_tree_remove,
15568         typed_splay_tree::splay_tree_lookup,
15569         typed_splay_tree::splay_tree_predecessor,
15570         typed_splay_tree::splay_tree_successor,
15571         typed_splay_tree::splay_tree_min,
15572         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15573         (typed_splay_tree::root, typed_splay_tree::comp,
15574         typed_splay_tree::delete_key,
15575         typed_splay_tree::delete_value): New data members.
15576         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15577         typed_splay_tree::remove.
15579 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15581         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15582         -mginv and -mno-ginv to the assembler.
15583         * config/mips/mips.opt (-mcrc): New option.
15584         (-mginv): Likewise.
15585         * doc/invoke.text (-mcrc): Document.
15586         (-mginv): Likewise.
15588 2018-06-15  Nick Clifton  <nickc@redhat.com>
15590         PR 84195
15591         * tree.c (escaped_string): New class.  Converts an unescaped
15592         string into its escaped equivalent.
15593         (warn_deprecated_use): Use the new class to convert the
15594         deprecation message, if present.
15595         (test_escaped_strings): New self test.
15596         (test_c_tests): Add test_escaped_strings.
15597         * doc/extend.texi (deprecated): Add a note that the
15598         deprecation message is affected by the -fmessage-length
15599         option, and that control characters will be escaped.
15600         (#pragma GCC error): Document this pragma.
15601         (#pragma GCC warning): Likewise.
15602         * doc/invoke.texi (-fmessage-length): Document this option's
15603         effect on the #warning and #error preprocessor directives and
15604         the deprecated attribute.
15606 2018-06-15  Richard Biener  <rguenther@suse.de>
15608         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15609         here, also noting vector size used.
15610         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15611         size used in MSG_OPTIMIZED_LOCATIONS dump.
15612         (pass_slp_vectorize::execute): Adjust.
15614 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15616         PR target/85968
15617         * config/arc/arc.c (arc_return_address_register): Fix
15618         if-condition.
15620 2018-06-15  Richard Biener  <rguenther@suse.de>
15622         PR middle-end/86159
15623         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15624         leave useless conversion stripping to force_gimple_operand_gsi.
15625         (gimplify_build2): Likewise.
15626         (gimplify_build1): Likewise.
15628 2018-06-15  Richard Biener  <rguenther@suse.de>
15630         PR middle-end/86076
15631         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15632         before adjusting their block.
15634 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15636         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15637         multilibs for *-*-rtems*.
15638         * config/riscv/t-rtems: New file.
15640 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15642         PR middle-end/86122
15643         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15644         unsigned_type_for returns NULL.
15646         PR target/85945
15647         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15648         subregs of multi-word pseudos unless the float mode has word size.
15650 2018-06-14  Richard Biener  <rguenther@suse.de>
15652         PR middle-end/86139
15653         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15654         duplicate and harmful type_hash_canon.
15655         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15657 2018-06-14  Richard Biener  <rguenther@suse.de>
15659         PR ipa/86124
15660         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15661         NULL cgraph_node.
15663 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15665         * config/rtems.h (STDINT_LONG32): Define.
15667 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15668             Prachi Godbole  <prachi.godbole@imgtec.com>
15670         * config/mips/mips-cpus.def: Define P6600.
15671         * config/mips/mips-tables.opt: Regenerate.
15672         * config/mips/mips.c (mips_ucbranch_type): New enum.
15673         (mips_rtx_cost_data): Add support for P6600.
15674         (mips_issue_rate): Likewise.
15675         (mips_multipass_dfa_lookahead): Likewise.
15676         (mips_avoid_hazard): Likewise.
15677         (mips_reorg_process_insns): Likewise.
15678         (mips_classify_branch_p6600): New function.
15679         * config/mips/mips.h (TUNE_P6600): New define.
15680         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15681         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15682         * config/mips/mips.md: Include p6600.md.
15683         (processor): Add p6600.
15684         * config/mips/p6600.md: New file.
15685         * doc/invoke.texi: Add p6600 to supported architectures.
15687 2018-06-13  Martin Sebor  <msebor@redhat.com>
15689         PR tree-optimization/86114
15690         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15691         of integer types.
15692         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15694 2018-06-13  Richard Biener  <rguenther@suse.de>
15696         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15697         Properly set vector type of the intermediate stmt.
15698         * tree-vect-stmts.c (vectorizable_operation): The destination
15699         var always has vectype_out type.
15701 2018-06-13  Jeff Law  <law@redhat.com>
15703         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15704         integer 0 for argument to print_rtl_with_bb.
15705         (rl78_reorg): Likewise.
15707 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15709         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15710         from rtx to rtx_insn *.
15711         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15712         "label".
15713         (add_sched_insns_for_speculation): Likewise for local "target",
15714         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15715         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15716         from rtx_insn ** to rtx_code_label **.
15717         (reorg_emit_nops): Likewise.
15718         (c6x_reorg): Likewise for local "call_labels".
15719         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15720         rtx to rtx_insn *.
15721         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
15722         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
15723         the loops over LABEL_REFS.
15724         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
15725         braf_label.
15726         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15727         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
15728         (split_branches): Strengthen local "olabel" from rtx to
15729         rtx_insn *, adding a safe_as_a cast.
15730         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
15731         to "rtx_insn *".
15732         (add_insn_after): Likewise for first two params.
15733         (add_insn_before): Likewise.
15734         (remove_insn): Likewise for param.
15735         (emit_pattern_before_noloc): Likewise for second and third params.
15736         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
15737         (emit_call_insn_before_noloc): Likewise.
15738         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
15739         to "rtx_insn *".
15740         (emit_barrier_before): Likewise.
15741         (emit_label_before): Strengthen "label" param from "rtx" to
15742         "rtx_code_label *".  Strengthen "before" param from "rtx" to
15743         "rtx_insn *".
15744         (emit_insn_after_1): Strengthen "after" param from "rtx" to
15745         "rtx_insn *".
15746         (emit_pattern_after_noloc): Likewise.
15747         (emit_insn_after_noloc): Likewise.
15748         (emit_jump_insn_after_noloc): Likewise.
15749         (emit_call_insn_after_noloc): Likewise.
15750         (emit_debug_insn_after_noloc): Likewise.
15751         (emit_barrier_after): Likewise.
15752         (emit_label_after): Likewise for both params.
15753         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
15754         "loc" param from "int" to "location_t".
15755         (emit_insn_after_setloc): Likewise.
15756         (emit_jump_insn_after_setloc): Likewise.
15757         (emit_call_insn_after_setloc): Likewise.
15758         (emit_debug_insn_after_setloc): Likewise.
15759         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
15760         "loc" param from "int" to "location_t".
15761         (emit_pattern_before): Convert NULL_RTX to NULL.
15762         (emit_insn_before_setloc): Convert "loc" param from "int" to
15763         "location_t".
15764         (emit_jump_insn_before_setloc): Likewise.
15765         (emit_call_insn_before_setloc): Likewise.
15766         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
15767         rtx_insn *.  Convert "loc" param from "int" to "location_t".
15768         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
15769         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
15770         Convert 3rd param from "int" to "location_t".
15771         (emit_barrier_before, emit_barrier_after, next_real_insn):
15772         Strengthen param from rtx to rtx_insn *.
15773         (emit_label_before): Strengthen 1st param from "rtx" to
15774         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
15775         "rtx_insn *".
15776         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
15777         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
15778         Strengthen 2nd param from "rtx" to "rtx_insn *".
15779         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
15780         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
15781         Likewise. Convert 3rd param from "int" to "location_t".
15782         (emit_label_after): Strengthen 1st param from "rtx" to
15783         "rtx_code_label *".
15784         (next_real_insn, remove_insn): Strengthen param from "rtx" to
15785         "rtx_insn *".
15786         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
15787         from "rtx" to "rtx_insn *".
15789 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
15791         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
15792         bodies streamed in with -Q.
15793         * dumpfile.c (dump_files): Add lto-stream-out dump file.
15794         * dumpfile.h (tree_dump_index): Add lto_stream_out.
15795         * gimple-streamer-out.c: Include gimple-pretty-print.h
15796         (output_bb): Dump stmts streamed.
15797         * lto-section-out.c: Include print-tree.h
15798         (lto_begin_section): Dump sections created.
15799         (lto_output_decl_index): Dump decl encoded.
15800         * lto-streamer-out.c: Include print-tree.h
15801         (create_output_block): Dump output block created.
15802         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
15803         (output_function): Dump function output.
15804         (output_constructor): Dump constructor streamed.
15805         (write_global_stream): Output indexes encoded.
15806         (produce_asm_for_decls): Dump streams encoded.
15807         * lto-streamer.c (streamer_dump_file): New global var.
15808         * lto-streamer.h (streamer_dump_file): Declare.
15809         * passes.c (ipa_write_summaries): Initialize streamer dump.
15810         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
15811         in.
15813 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
15815         PR target/86048
15816         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
15817         offsets for register save directives.  Emit a second batch of save
15818         directives, if need be, when the function accesses prior frames.
15820 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15822         * config/arc/fpu.md (fmasf4): Force operand to register.
15823         (fnmasf4): Likewise.
15825 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15827         * config/arc/arc-protos.h (arc_pad_return): Remove.
15828         * config/arc/arc.c (machine_function): Remove force_short_suffix
15829         and size_reason.
15830         (arc_print_operand): Adjust printing of '&'.
15831         (arc_verify_short): Remove conditional printing of short suffix.
15832         (arc_final_prescan_insn): Remove reference to size_reason.
15833         (pad_return): New function.
15834         (arc_reorg): Call pad_return.
15835         (arc_pad_return): Remove.
15836         (arc_init_machine_status): Remove reference to force_short_suffix.
15837         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
15838         (attr length): When attribute iscompact is true force to 2
15839         regardless; in the case of maybe check if we want to force the
15840         instruction to have 4 bytes length.
15841         (nopv): Change it to generate 4 byte long nop as well.
15842         (blockage): New pattern.
15843         (simple_return): Remove call to arc_pad_return.
15844         (p_return_i): Likewise.
15846 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15848         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
15850 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15852         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
15853         ARC cores.
15855 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15857         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
15858         for ARC700 and ARCv2.
15860 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
15862         PR target/86076
15863         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
15864         operands[2] instead of operands[1].
15867 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15869         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
15870         case, check whether the outer register overlaps an unallocatable
15871         register, not just whether it fits the required class.
15873 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15875         * poly-int.h (can_div_trunc_p): Add new overload in which all values
15876         are poly_ints.
15877         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
15878         (memrefs_conflict_p): Likewise.
15879         (init_alias_analysis): Likewise.
15880         * cfgexpand.c (expand_debug_expr): Likewise.
15881         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
15882         * cse.c (fold_rtx): Likewise.
15883         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
15884         * expr.c (emit_block_move_hints): Likewise.
15885         (clear_storage_hints, push_block, emit_push_insn): Likewise.
15886         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
15887         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
15888         (emit_group_store): Likewise.
15889         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
15890         to read the PRE/POST_MODIFY increment.
15891         * calls.c (store_one_arg): Use strip_offset.
15892         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
15893         poly_int_rtx_p.
15894         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
15895         by a VEC_SELECT.
15896         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
15897         (simplify_binary_operation_1): Extend CONST_INT handling to
15898         poly_int_rtx_p.
15899         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
15900         than a HOST_WIDE_INT.
15901         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
15902         poly_int64.
15903         (adjust_mems, add_stores): Update accodingly.
15904         (vt_canonicalize_addr): Track polynomial offsets.
15905         (emit_note_insn_var_location): Likewise.
15906         (vt_add_function_parameter): Likewise.
15907         (vt_initialize): Likewise.
15909 2018-06-12  Jeff Law  <law@redhat.com>
15911         * config.gcc (alpha*-*-freebsd*): Remove.
15912         * config/alpha/freebsd.h: Remove.
15914 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
15916         PR other/69968
15917         * spellcheck-tree.c (levenshtein_distance): Rename to...
15918         (get_edit_distance): ...this, and update for underlying renaming.
15919         * spellcheck-tree.h (levenshtein_distance): Rename to...
15920         (get_edit_distance): ...this.
15921         * spellcheck.c (levenshtein_distance): Rename to...
15922         (get_edit_distance): ...this.  Convert from Levenshtein distance
15923         to Damerau-Levenshtein distance by supporting transpositions of
15924         adjacent characters.  Rename "v1" to "v_next" and "v0" to
15925         "v_one_ago".
15926         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
15927         (selftest::test_edit_distance_unit_test_oneway): ...this, and
15928         update for underlying renaming.
15929         (selftest::levenshtein_distance_unit_test): Rename to...
15930         (selftest::test_get_edit_distance_unit): ...this, and update for
15931         underlying renaming.
15932         (selftest::test_find_closest_string): Add example from PR 69968
15933         where transposition helps
15934         (selftest::test_metric_conditions): Update for renaming.
15935         (selftest::test_metric_conditions): Likewise.
15936         (selftest::spellcheck_c_tests): Likewise.
15937         * spellcheck.h (levenshtein_distance): Rename both overloads to...
15938         (get_edit_distance): ...this.
15939         (best_match::consider): Update for renaming.
15941 2018-06-12  Martin Sebor  <msebor@redhat.com>
15943         PR tree-optimization/85259
15944         * builtins.c (compute_objsize): Handle constant offsets.
15945         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
15946         true iff a warning has been issued.
15947         * gimple.h (gimple_nonartificial_location): New function.
15948         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
15949         gimple_nonartificial_location and handle -Wno-system-headers.
15950         (handle_builtin_stxncpy): Same.
15952 2018-06-12  Martin Sebor  <msebor@redhat.com>
15954         PR c/85931
15955         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
15957 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
15959         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
15960         BUILTIN_VEC_XST entries for pointer to double and long long.
15962 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
15964         PR target/85990
15965         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15966         Update comments.
15967         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15968         Likewise.
15970 2018-06-12  Martin Liska  <mliska@suse.cz>
15972         * doc/options.texi: Document IntegerRange.
15974 2018-06-12  Martin Liska  <mliska@suse.cz>
15976         * config/i386/i386.opt: Make MPX-related options as Deprecated.
15977         * opt-functions.awk: Handle Deprecated flag.
15978         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
15979         and report error.
15980         (read_cmdline_option): Report warning for a deprecated option.
15981         * opts.h (struct cl_option): Add new field cl_deprecated.
15982         (CL_ERR_DEPRECATED): New.
15984 2018-06-12  Martin Liska  <mliska@suse.cz>
15986         * doc/options.texi: Document Deprecated option flag.
15988 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15990         * config/arc/arc-arch.h (arc_extras): New enum.
15991         (arc_cpu_t):Add field extra.
15992         (arc_cpu_types): Consider the extras.
15993         * config/arc/arc-cpus.def: Add extras info.
15994         * config/arc/arc-opts.h (processor_type): Consider extra field.
15995         * config/arc/arc.c (arc_override_options): Handle extra field.
15997 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15999         * config/arc/arc-arch.h: Update ARC_OPTX macro.
16000         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16001         field.
16002         * config/arc/arc.c (arc_init): Update pic warning.
16003         (irq_range): Update irq range parsing warnings.
16004         (arc_override_options): Update various warning messages.
16005         (arc_handle_aux_attribute): Likewise.
16007 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
16009         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16011 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16013         * doc/sourcebuild.texi: Document usage of line number 0 in verify
16014         compiler messages directives.
16016 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
16018         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16019         * config/mips/mips-tables.opt: Regenerate.
16020         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16021         mips64r6.
16022         * doc/invoke.texi: Document -march=i6500.
16024 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
16026         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16027         (i6400_gpmul): Add cpu_unit.
16028         (i6400_gpdiv): Likewise.
16029         (i6400_msa_add_d): Update reservations.
16030         (i6400_msa_int_add) Likewise.
16031         (i6400_msa_short_logic3) Likewise.
16032         (i6400_msa_short_logic2) Likewise.
16033         (i6400_msa_short_logic) Likewise.
16034         (i6400_msa_move) Likewise.
16035         (i6400_msa_cmp) Likewise.
16036         (i6400_msa_short_float2) Likewise.
16037         (i6400_msa_div_d) Likewise.
16038         (i6400_msa_long_logic1) Likewise.
16039         (i6400_msa_long_logic2) Likewise.
16040         (i6400_msa_mult) Likewise.
16041         (i6400_msa_long_float2) Likewise.
16042         (i6400_msa_long_float4) Likewise.
16043         (i6400_msa_long_float5) Likewise.
16044         (i6400_msa_long_float8) Likewise.
16045         (i6400_fpu_fadd): Include frint type.
16046         (i6400_fpu_store): New define_insn_reservation.
16047         (i6400_fpu_load): Likewise.
16048         (i6400_fpu_move): Likewise.
16049         (i6400_fpu_fcmp): Likewise.
16050         (i6400_fpu_fmadd): Likewise.
16051         (i6400_int_mult): Include imul3nc type and update reservation.
16052         (i6400_int_div): Include idiv3 type and update reservation.
16053         (i6400_int_load): Update to check type not move_type.
16054         (i6400_int_store): Likewise.
16055         (i6400_int_prefetch): Set zero latency.
16057 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
16059         * gcc.c: Document new %@{...} sequence.
16060         (LINK_COMMAND_SPEC): Use it for the -L switches.
16061         (cpp_unique_options): Use it for the -I switches.
16062         (at_file_argbuf): New global variable.
16063         (in_at_file): Likewise.
16064         (alloc_args): Create at_file_argbuf.
16065         (clear_args): Truncate at_file_argbuf.
16066         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16067         (open_at_file): New function.
16068         (close_at_file): Likewise.
16069         (create_at_file): Delete.
16070         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
16071         <'o'>: Likewise.
16072         <'@'>: New case.
16073         (validate_switches_from_spec): Deal with %@{...} sequence.
16074         (validate_switches): Likewise.
16075         (driver::finalize): Call clear_args.
16077 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
16079         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
16081 2018-06-11  Martin Sebor  <msebor@redhat.com>
16083         * doc/invoke.texi (-Wall): List -Wc++17-compat.
16084         (Wno-class-memaccess): Add @opindex.
16085         (Wno-templates, Wno-multiple-inheritance): Same.
16086         (Wno-virtual-inheritance, Wno-namespaces): Same.
16087         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
16088         (Wno-format-overflow, Wno-format-truncation): Same.
16089         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
16090         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
16091         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
16092         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
16093         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
16094         (Wno-misspelled-isr): Same.
16096 2018-06-11  Martin Sebor  <msebor@redhat.com>
16098         * PR tree-optimization/86083
16099         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
16101 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
16103         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
16105 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
16107         PR target/85755
16108         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
16109         on the correct operand.
16110         (*movdi_internal64): Ditto.
16112 2018-06-11  Martin Liska  <mliska@suse.cz>
16114         PR tree-optimization/86089
16115         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
16117 2018-06-11  Julia Koval  <julia.koval@intel.com>
16119         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
16120         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
16121         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
16123 2018-06-11  Olivier Hainque  <hainque@adacore.com>
16125         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
16126         for Ada with strict dwarf2.
16128 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
16130         PR target/85755
16131         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16132         addresses.
16134 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
16136         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16138 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
16140         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16141         TARGET_ELF.
16143 2018-06-08  Martin Liska  <mliska@suse.cz>
16145         * tree-cfg.h (debug_function): Fix argument type to match
16146         implementation.
16148 2018-06-08  Martin Liska  <mliska@suse.cz>
16150         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16151         Remove usage of MPX-related (and removed) fields.
16152         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16154 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
16156         * cfg.c (debug): Use TDF_NONE rather than 0.
16157         * cfghooks.c (debug): Likewise.
16158         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16159         (struct dump_option_value_info): Convert to...
16160         (struct kv_pair): ...this template type.
16161         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16162         rather than 0.
16163         (optinfo_verbosity_options): Likewise.
16164         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16165         OPTGROUP_NONE.
16166         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16167         than int for "optgroup_flags" param.
16168         (dump_generic_expr_loc): Use dump_flags_t rather than int for
16169         "dump_kind" param.
16170         (dump_dec): Likewise.
16171         (dump_finish): Use TDF_NONE rather than 0.
16172         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16173         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
16174         than 0.  Update for change to option_ptr.
16175         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16176         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
16177         0.  Update for changes to optinfo_verbosity_options and
16178         optgroup_options.
16179         (opt_info_switch_p): Convert optgroup_flags from int to
16180         optgroup_flags_t.
16181         (dump_basic_block): Use dump_flags_t rather than int
16182         for "dump_kind" param.
16183         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16184         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16185         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16186         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16187         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16188         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16189         TDF_NONE): Convert from macros to...
16190         (enum dump_flag): ...this new enum.
16191         (dump_flags_t): Update to use enum.
16192         (operator|, operator&, operator~, operator|=, operator&=):
16193         Implement for dump_flags_t.
16194         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16195         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16196         Convert from macros to...
16197         (enum optgroup_flag): ...this new enum.
16198         (optgroup_flags_t): New typedef.
16199         (operator|, operator|=): Implement for optgroup_flags_t.
16200         (struct dump_file_info): Convert field "alt_flags" to
16201         dump_flags_t.  Convert field "optgroup_flags" to
16202         optgroup_flags_t.
16203         (dump_basic_block): Use dump_flags_t rather than int for param.
16204         (dump_generic_expr_loc): Likewise.
16205         (dump_dec): Likewise.
16206         (dump_register): Convert param "optgroup_flags" to
16207         optgroup_flags_t.
16208         (opt_info_enable_passes): Likewise.
16209         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16210         than 0.
16211         * gimple-pretty-print.c (debug): Likewise.
16212         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16213         (merged_store_group::apply_stores): Likewise.
16214         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16215         * gimple.c (verify_gimple_pp): Likewise.
16216         * graphite-poly.c (print_pbb_body): Likewise.
16217         * passes.c (pass_manager::register_one_dump_file): Convert
16218         local "optgroup_flags" to optgroup_flags_t.
16219         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16220         (debug): Likewise.
16221         (debug_body): Likewise.
16222         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16223         to optgroup_flags_t.
16224         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16225         than 0.
16226         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16227         (convert_mult_to_fma): Likewise.
16228         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16229         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16230         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16231         "dump_kind" to dump_flags_t.
16233 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16235         * config/rs6000/rs6000.c (min, max): Delete.
16237 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16239         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16240         -mabi=spe and -mabi=no-spe.
16242 2018-06-08  Martin Liska  <mliska@suse.cz>
16244         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16245         where we expect an existing summary.
16247 2018-06-08  Martin Liska  <mliska@suse.cz>
16249         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16250         * ipa-inline.h (estimate_edge_growth): Likewise.
16252 2018-06-08  Martin Liska  <mliska@suse.cz>
16254         * cgraph.c (function_version_hasher::hash): Use
16255         cgraph_node::get_uid ().
16256         (function_version_hasher::equal):
16257         * cgraph.h (cgraph_node::get_uid): New method.
16258         * ipa-inline.c (update_caller_keys): Use
16259         cgraph_node::get_uid ().
16260         (update_callee_keys): Likewise.
16261         * ipa-utils.c (searchc): Likewise.
16262         (ipa_reduced_postorder): Likewise.
16263         * lto-cgraph.c (input_node): Likewise.
16264         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16265         * symbol-summary.h (symtab_insertion): Likewise.
16266         (symtab_removal): Likewise.
16267         (symtab_duplication): Likewise.
16268         * tree-pretty-print.c (dump_function_header): Likewise.
16269         * tree-sra.c (convert_callers_for_node): Likewise.
16271 2018-06-08  Martin Liska  <mliska@suse.cz>
16273         * cgraph.c (symbol_table::create_edge): Always assign a new
16274         unique number.
16275         (symbol_table::free_edge): Do not recycle numbers.
16276         * cgraph.h (cgraph_edge::get): New method.
16277         * symbol-summary.h (symtab_removal): Use it.
16278         (symtab_duplication): Likewise.
16279         (call_summary::hashable_uid): Remove.
16281 2018-06-08  Martin Liska  <mliska@suse.cz>
16283         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16284         (initialize_growth_caches): Remove.
16285         (free_growth_caches): Likewise.
16286         (do_estimate_edge_time): Use edge_growth_cache.
16287         (do_estimate_edge_size): Likewise.
16288         (do_estimate_edge_hints): Likewise.
16289         * ipa-inline.c (reset_edge_caches): Likewise.
16290         (recursive_inlining): Likewise.
16291         (inline_small_functions): Likewise.
16292         * ipa-inline.h (initialize_growth_caches): Remove.
16293         (estimate_edge_size): Likewise.
16294         (estimate_edge_time): Likewise.
16295         (estimate_edge_hints): Likewise.
16296         (reset_edge_growth_cache): Likewise.
16297         * symbol-summary.h (call_summary::remove): New method.
16299 2018-06-08  Martin Liska  <mliska@suse.cz>
16301         * ipa-cp.c (class edge_clone_summary): New summary.
16302         (grow_edge_clone_vectors): Remove.
16303         (ipcp_edge_duplication_hook): Remove.
16304         (class edge_clone_summary_t): New call_summary class.
16305         (ipcp_edge_removal_hook): Remove.
16306         (edge_clone_summary_t::duplicate): New function.
16307         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16308         (create_specialized_node): Likewise.
16309         (ipcp_driver): Initialize edge_clone_summaries and do not
16310         register hooks.
16312 2018-06-08  Martin Liska  <mliska@suse.cz>
16314         * symbol-summary.h (get): New function.
16315         (call_summary::m_initialize_when_cloning): New class member.
16317 2018-06-08  Martin Liska  <mliska@suse.cz>
16319         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16320         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16321         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16322         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16323         (remove_cgraph_node_from_order): Use the removed_nodes set.
16324         (do_per_function_toporder): Likwise.
16325         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16326         instead of summary_uid.
16327         (symtab_removal): Likewise.
16328         (symtab_duplication): Likewise.
16330 2018-06-08  Martin Liska  <mliska@suse.cz>
16332         * ipa-cp.c (ipcp_store_bits_results): Use
16333         ipcp_transformation_sum.
16334         (ipcp_store_vr_results): Likewise.
16335         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16336         to ...
16337         (ipcp_transformation_initialize): ... this.
16338         (ipa_set_node_agg_value_chain):
16339         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16340         (write_ipcp_transformation_info): Likewise.
16341         (read_ipcp_transformation_info): Likewise.
16342         (ipcp_update_bits): Likewise.
16343         (ipcp_update_vr): Likewise.
16344         (ipcp_transform_function): Likewise.
16345         * ipa-prop.h: Rename ipcp_transformation_summary to
16346         ipcp_transformation.
16347         (class ipcp_transformation_t): New function summary.
16348         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16349         (ipa_get_agg_replacements_for_node): Likewise.
16351 2018-06-08  Martin Liska  <mliska@suse.cz>
16353         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16354         of struct.
16355         (class funct_state_summary_t): New function_summary class.
16356         (has_function_state): Remove.
16357         (get_function_state): Likewise.
16358         (set_function_state): Likewise.
16359         (add_new_function): Likewise.
16360         (funct_state_summary_t::insert): New function.
16361         (duplicate_node_data): Remove.
16362         (remove_node_data): Remove.
16363         (funct_state_summary_t::duplicate): New function.
16364         (register_hooks): Create new funct_state_summaries.
16365         (pure_const_generate_summary): Use it.
16366         (pure_const_write_summary): Likewise.
16367         (pure_const_read_summary): Likewise.
16368         (propagate_pure_const): Likewise.
16369         (propagate_nothrow): Likewise.
16370         (dump_malloc_lattice): Likewise.
16371         (propagate_malloc): Likewise.
16372         (execute): Do not register hooks, just remove summary
16373         instead.
16374         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16375         constructor.
16377 2018-06-08  Martin Liska  <mliska@suse.cz>
16379         * ipa-reference.c (remove_node_data): Remove.
16380         (duplicate_node_data): Likewise.
16381         (class ipa_ref_var_info_summary_t): New class.
16382         (class ipa_ref_opt_summary_t): Likewise.
16383         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16384         (get_reference_optimization_summary): Use
16385         ipa_ref_opt_sum_summaries.
16386         (set_reference_vars_info): Remove.
16387         (set_reference_optimization_summary): Likewise.
16388         (ipa_init): Create summaries.
16389         (init_function_info): Use function summary.
16390         (ipa_ref_opt_summary_t::duplicate): New function.
16391         (ipa_ref_opt_summary_t::remove): New function.
16392         (get_read_write_all_from_node): Fix GNU coding style.
16393         (propagate): Use function summary.
16394         (write_node_summary_p): Fix GNU coding style.
16395         (stream_out_bitmap): Likewise.
16396         (ipa_reference_read_optimization_summary): Use function summary.
16397         (ipa_reference_c_finalize): Do not release hooks.
16399 2018-06-08  Martin Liska  <mliska@suse.cz>
16401         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16402         (analyze_function_body): Extract multiple calls of get_create.
16403         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16404         * ipa-inline.c (recursive_inlining): Use ::get method.
16405         * ipa-inline.h (estimate_edge_growth): Likewise.
16407 2018-06-08  Martin Liska  <mliska@suse.cz>
16409         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16410         HSA_INVALID.
16411         (hsa_function_summary::hsa_function_summary): Use the new enum
16412         value.
16413         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16414         * hsa-gen.c (hsa_get_host_function): Likewise.
16415         (get_brig_function_name): Likewise.
16416         * ipa-hsa.c (process_hsa_functions): Likewise.
16417         (ipa_hsa_write_summary): Likewise.
16418         * symbol-summary.h (symtab_duplication): Use ::get function/
16419         (get): New function.
16421 2018-06-08  Martin Liska  <mliska@suse.cz>
16423         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16424         of get.
16425         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16426         (hsa_register_kernel): Likewise.
16427         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16428         * hsa-gen.c (hsa_get_host_function): Likewise.
16429         (get_brig_function_name): Likewise.
16430         (generate_hsa): Likewise.
16431         (pass_gen_hsail::execute): Likewise.
16432         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16433         (devirtualization_time_bonus): Likewise.
16434         (ipcp_propagate_stage): Likewise.
16435         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16436         (edge_set_predicate): Likewise.
16437         (evaluate_conditions_for_known_args): Likewise.
16438         (evaluate_properties_for_edge): Likewise.
16439         (ipa_fn_summary::reset): Likewise.
16440         (ipa_fn_summary_t::duplicate): Likewise.
16441         (dump_ipa_call_summary): Likewise.
16442         (ipa_dump_fn_summary): Likewise.
16443         (analyze_function_body): Likewise.
16444         (compute_fn_summary): Likewise.
16445         (estimate_edge_devirt_benefit): Likewise.
16446         (estimate_edge_size_and_time): Likewise.
16447         (estimate_calls_size_and_time): Likewise.
16448         (estimate_node_size_and_time): Likewise.
16449         (inline_update_callee_summaries): Likewise.
16450         (remap_edge_change_prob): Likewise.
16451         (remap_edge_summaries): Likewise.
16452         (ipa_merge_fn_summary_after_inlining): Likewise.
16453         (ipa_update_overall_fn_summary): Likewise.
16454         (read_ipa_call_summary): Likewise.
16455         (inline_read_section): Likewise.
16456         (write_ipa_call_summary): Likewise.
16457         (ipa_fn_summary_write): Likewise.
16458         (ipa_free_fn_summary): Likewise.
16459         * ipa-hsa.c (process_hsa_functions): Likewise.
16460         (ipa_hsa_write_summary): Likewise.
16461         (ipa_hsa_read_section): Likewise.
16462         * ipa-icf.c (sem_function::merge): Likewise.
16463         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16464         (do_estimate_edge_time): Likewise.
16465         (estimate_size_after_inlining): Likewise.
16466         (estimate_growth): Likewise.
16467         (growth_likely_positive): Likewise.
16468         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16469         (inline_call): Likewise.
16470         * ipa-inline.c (caller_growth_limits): Likewise.
16471         (can_inline_edge_p): Likewise.
16472         (can_inline_edge_by_limits_p): Likewise.
16473         (compute_uninlined_call_time): Likewise.
16474         (compute_inlined_call_time): Likewise.
16475         (want_inline_small_function_p): Likewise.
16476         (edge_badness): Likewise.
16477         (update_caller_keys): Likewise.
16478         (update_callee_keys): Likewise.
16479         (recursive_inlining): Likewise.
16480         (inline_small_functions): Likewise.
16481         (inline_to_all_callers_1): Likewise.
16482         (dump_overall_stats): Likewise.
16483         (early_inline_small_functions): Likewise.
16484         (early_inliner): Likewise.
16485         * ipa-inline.h (estimate_edge_growth): Likewise.
16486         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16487         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16488         * ipa-prop.h (IPA_NODE_REF): Likewise.
16489         (IPA_EDGE_REF): Likewise.
16490         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16491         (propagate_malloc): Likewise.
16492         * ipa-split.c (execute_split_functions): Likewise.
16493         * symbol-summary.h: Rename get to get_create.
16494         (get): Likewise.
16495         (get_create): Likewise.
16496         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16498 2018-06-08  Martin Liska  <mliska@suse.cz>
16500         * symbol-summary.h (release): Move definition out of class
16501         declaration.
16502         (symtab_removal): Likewise.
16503         (symtab_duplication): Likewise.
16505 2018-06-08  Martin Liska  <mliska@suse.cz>
16507         * symbol-summary.h (function_summary): Move constructor
16508         implementation out of class declaration.
16509         (release): Likewise.
16510         (symtab_insertion): Likewise.
16511         (symtab_removal): Likewise.
16512         (symtab_duplication): Likewise.
16513         (get): Likewise.
16515 2018-06-08  Martin Liska  <mliska@suse.cz>
16517         * Makefile.in: Remove support for MPX (macros, related functions,
16518         fields in cgraph_node, ...).
16519         * builtin-types.def (BT_BND): Likewise.
16520         (BT_FN_BND_CONST_PTR): Likewise.
16521         (BT_FN_CONST_PTR_BND): Likewise.
16522         (BT_FN_VOID_PTR_BND): Likewise.
16523         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16524         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16525         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16526         (expand_builtin_mempcpy_with_bounds): Likewise.
16527         (expand_builtin_memset_with_bounds): Likewise.
16528         (expand_builtin_memset_args): Likewise.
16529         (std_expand_builtin_va_start): Likewise.
16530         (expand_builtin): Likewise.
16531         (expand_builtin_with_bounds): Likewise.
16532         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16533         (DEF_LIB_BUILTIN_CHKP): Likewise.
16534         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16535         (DEF_CHKP_BUILTIN): Likewise.
16536         (BUILT_IN_MEMCPY): Likewise.
16537         (BUILT_IN_MEMMOVE): Likewise.
16538         (BUILT_IN_MEMPCPY): Likewise.
16539         (BUILT_IN_MEMSET): Likewise.
16540         (BUILT_IN_STPCPY): Likewise.
16541         (BUILT_IN_STRCAT): Likewise.
16542         (BUILT_IN_STRCHR): Likewise.
16543         (BUILT_IN_STRCPY): Likewise.
16544         (BUILT_IN_STRLEN): Likewise.
16545         (BUILT_IN_MEMCPY_CHK): Likewise.
16546         (BUILT_IN_MEMMOVE_CHK): Likewise.
16547         (BUILT_IN_MEMPCPY_CHK): Likewise.
16548         (BUILT_IN_MEMSET_CHK): Likewise.
16549         (BUILT_IN_STPCPY_CHK): Likewise.
16550         (BUILT_IN_STRCAT_CHK): Likewise.
16551         (BUILT_IN_STRCPY_CHK): Likewise.
16552         * calls.c (store_bounds): Likewise.
16553         (emit_call_1): Likewise.
16554         (special_function_p): Likewise.
16555         (maybe_warn_nonstring_arg): Likewise.
16556         (initialize_argument_information): Likewise.
16557         (finalize_must_preallocate): Likewise.
16558         (compute_argument_addresses): Likewise.
16559         (expand_call): Likewise.
16560         * cfgexpand.c (expand_call_stmt): Likewise.
16561         (expand_return): Likewise.
16562         (expand_gimple_stmt_1): Likewise.
16563         (pass_expand::execute): Likewise.
16564         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16565         (cgraph_node::remove): Likewise.
16566         (cgraph_node::dump): Likewise.
16567         (cgraph_node::verify_node): Likewise.
16568         * cgraph.h (chkp_function_instrumented_p): Likewise.
16569         (symtab_node::get_alias_target): Likewise.
16570         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16571         (cgraph_local_p): Likewise.
16572         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16573         (cgraph_edge::rebuild_references): Likewise.
16574         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16575         (walk_polymorphic_call_targets): Likewise.
16576         (cgraph_node::expand_thunk): Likewise.
16577         (symbol_table::output_weakrefs): Likewise.
16578         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16579         (ix86_handle_option): Likewise.
16580         * config/i386/constraints.md: Likewise.
16581         * config/i386/i386-builtin-types.def (BND): Likewise.
16582         (VOID): Likewise.
16583         (PVOID): Likewise.
16584         (ULONG): Likewise.
16585         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16586         (BDESC_FIRST): Likewise.
16587         (BDESC): Likewise.
16588         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16589         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16590         * config/i386/i386.c (enum reg_class): Likewise.
16591         (ix86_target_string): Likewise.
16592         (ix86_option_override_internal): Likewise.
16593         (ix86_conditional_register_usage): Likewise.
16594         (ix86_valid_target_attribute_inner_p): Likewise.
16595         (ix86_set_indirect_branch_type): Likewise.
16596         (ix86_set_current_function): Likewise.
16597         (ix86_function_arg_regno_p): Likewise.
16598         (init_cumulative_args): Likewise.
16599         (ix86_function_arg_advance): Likewise.
16600         (ix86_function_arg): Likewise.
16601         (ix86_pass_by_reference): Likewise.
16602         (ix86_function_value_regno_p): Likewise.
16603         (ix86_function_value_1): Likewise.
16604         (ix86_function_value_bounds): Likewise.
16605         (ix86_return_in_memory): Likewise.
16606         (ix86_setup_incoming_vararg_bounds): Likewise.
16607         (ix86_va_start): Likewise.
16608         (indirect_thunk_need_prefix): Likewise.
16609         (print_reg): Likewise.
16610         (ix86_print_operand): Likewise.
16611         (ix86_expand_call): Likewise.
16612         (ix86_output_function_return): Likewise.
16613         (reg_encoded_number): Likewise.
16614         (BDESC_VERIFYS): Likewise.
16615         (ix86_init_mpx_builtins): Likewise.
16616         (ix86_init_builtins): Likewise.
16617         (ix86_emit_cmove): Likewise.
16618         (ix86_emit_move_max): Likewise.
16619         (ix86_expand_builtin): Likewise.
16620         (ix86_builtin_mpx_function): Likewise.
16621         (ix86_get_arg_address_for_bt): Likewise.
16622         (ix86_load_bounds): Likewise.
16623         (ix86_store_bounds): Likewise.
16624         (ix86_load_returned_bounds): Likewise.
16625         (ix86_store_returned_bounds): Likewise.
16626         (ix86_class_likely_spilled_p): Likewise.
16627         (ix86_hard_regno_mode_ok): Likewise.
16628         (x86_order_regs_for_local_alloc): Likewise.
16629         (ix86_mitigate_rop): Likewise.
16630         (ix86_bnd_prefixed_insn_p): Likewise.
16631         (ix86_mpx_bound_mode): Likewise.
16632         (ix86_make_bounds_constant): Likewise.
16633         (ix86_initialize_bounds): Likewise.
16634         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16635         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16636         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16637         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16638         (TARGET_CHKP_BOUND_MODE): Likewise.
16639         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16640         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16641         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16642         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16643         * config/i386/i386.h (TARGET_MPX): Likewise.
16644         (TARGET_MPX_P): Likewise.
16645         (VALID_BND_REG_MODE): Likewise.
16646         (FIRST_BND_REG): Likewise.
16647         (LAST_BND_REG): Likewise.
16648         (enum reg_class): Likewise.
16649         (BND_REG_P): Likewise.
16650         (BND_REGNO_P): Likewise.
16651         (BNDmode): Likewise.
16652         (ADJUST_INSN_LENGTH): Likewise.
16653         * config/i386/i386.md: Likewise.
16654         * config/i386/i386.opt: Likewise.
16655         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16656         (defined): Likewise.
16657         (LINK_MPX): Likewise.
16658         (MPX_SPEC): Likewise.
16659         (LIBMPX_SPEC): Likewise.
16660         (LIBMPXWRAPPERS_SPEC): Likewise.
16661         (CHKP_SPEC): Likewise.
16662         * config/i386/predicates.md: Likewise.
16663         * dbxout.c (dbxout_type): Likewise.
16664         * doc/extend.texi: Likewise.
16665         * doc/invoke.texi: Likewise.
16666         * doc/md.texi: Likewise.
16667         * doc/tm.texi: Likewise.
16668         * doc/tm.texi.in: Likewise.
16669         * dwarf2out.c (is_base_type): Likewise.
16670         (gen_formal_types_die): Likewise.
16671         (gen_subprogram_die): Likewise.
16672         (gen_type_die_with_usage): Likewise.
16673         (gen_decl_die): Likewise.
16674         (dwarf2out_late_global_decl): Likewise.
16675         * expr.c (expand_assignment): Likewise.
16676         (emit_storent_insn): Likewise.
16677         (store_expr_with_bounds): Likewise.
16678         (store_expr): Likewise.
16679         (expand_expr_real_1): Likewise.
16680         * expr.h (store_expr_with_bounds): Likewise.
16681         * function.c (use_register_for_decl): Likewise.
16682         (struct bounds_parm_data): Likewise.
16683         (assign_parms_augmented_arg_list): Likewise.
16684         (assign_parm_find_entry_rtl): Likewise.
16685         (assign_parm_is_stack_parm): Likewise.
16686         (assign_parm_load_bounds): Likewise.
16687         (assign_bounds): Likewise.
16688         (assign_parms): Likewise.
16689         (expand_function_start): Likewise.
16690         * gcc.c (CHKP_SPEC): Likewise.
16691         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16692         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16693         (wrestrict_dom_walker::check_call): Likewise.
16694         * gimple.c (gimple_build_call_from_tree): Likewise.
16695         * gimple.h (enum gf_mask): Likewise.
16696         (gimple_call_with_bounds_p): Likewise.
16697         (gimple_call_set_with_bounds): Likewise.
16698         * gimplify.c (gimplify_init_constructor): Likewise.
16699         * ipa-cp.c (initialize_node_lattices): Likewise.
16700         (propagate_constants_across_call): Likewise.
16701         (find_more_scalar_values_for_callers_subset): Likewise.
16702         * ipa-hsa.c (process_hsa_functions): Likewise.
16703         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16704         * ipa-icf.c (sem_function::merge): Likewise.
16705         * ipa-inline.c (early_inliner): Likewise.
16706         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16707         (warn_function_cold): Likewise.
16708         (propagate_pure_const): Likewise.
16709         * ipa-ref.h (enum GTY): Likewise.
16710         * ipa-split.c (find_retbnd): Likewise.
16711         (consider_split): Likewise.
16712         (split_function): Likewise.
16713         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16714         * ipa.c (walk_polymorphic_call_targets): Likewise.
16715         (symbol_table::remove_unreachable_nodes): Likewise.
16716         (process_references): Likewise.
16717         (cgraph_build_static_cdtor_1): Likewise.
16718         * lto-cgraph.c (lto_output_node): Likewise.
16719         (output_refs): Likewise.
16720         (compute_ltrans_boundary): Likewise.
16721         (input_overwrite_node): Likewise.
16722         (input_node): Likewise.
16723         (input_cgraph_1): Likewise.
16724         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
16725         * passes.c (pass_manager::execute_early_local_passes): Likewise.
16726         (class pass_chkp_instrumentation_passes): Likewise.
16727         (make_pass_chkp_instrumentation_passes): Likewise.
16728         * passes.def: Likewise.
16729         * rtl.h (struct GTY): Likewise.
16730         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
16731         * stor-layout.c (layout_type): Likewise.
16732         * symtab.c: Likewise.
16733         * target.def: Likewise.
16734         * targhooks.c (default_chkp_bound_type): Likewise.
16735         (default_chkp_bound_mode): Likewise.
16736         (default_builtin_chkp_function): Likewise.
16737         (default_chkp_function_value_bounds): Likewise.
16738         (default_chkp_make_bounds_constant): Likewise.
16739         (default_chkp_initialize_bounds): Likewise.
16740         * targhooks.h (default_chkp_bound_type): Likewise.
16741         (default_chkp_bound_mode): Likewise.
16742         (default_builtin_chkp_function): Likewise.
16743         (default_chkp_function_value_bounds): Likewise.
16744         (default_chkp_make_bounds_constant): Likewise.
16745         (default_chkp_initialize_bounds): Likewise.
16746         * toplev.c (compile_file): Likewise.
16747         (process_options): Likewise.
16748         * tree-core.h (DEF_BUILTIN): Likewise.
16749         (DEF_BUILTIN_CHKP): Likewise.
16750         * tree-inline.c (declare_return_variable): Likewise.
16751         (remap_gimple_stmt): Likewise.
16752         (copy_bb): Likewise.
16753         (initialize_inlined_parameters): Likewise.
16754         (expand_call_inline): Likewise.
16755         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
16756         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
16757         (make_pass_ipa_chkp_produce_thunks): Likewise.
16758         (make_pass_chkp): Likewise.
16759         (make_pass_chkp_opt): Likewise.
16760         (make_pass_chkp_instrumentation_passes): Likewise.
16761         * tree-pretty-print.c (dump_generic_node): Likewise.
16762         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
16763         * tree-ssa-dce.c (propagate_necessity): Likewise.
16764         (eliminate_unnecessary_stmts): Likewise.
16765         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
16766         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
16767         * tree-ssa-sccvn.h: Likewise.
16768         * tree-ssa-strlen.c (get_string_length): Likewise.
16769         (valid_builtin_call): Likewise.
16770         (adjust_last_stmt): Likewise.
16771         (handle_builtin_strchr): Likewise.
16772         (handle_builtin_strcpy): Likewise.
16773         (handle_builtin_stxncpy): Likewise.
16774         (handle_builtin_memcpy): Likewise.
16775         (handle_builtin_strcat): Likewise.
16776         (strlen_check_and_optimize_stmt): Likewise.
16777         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
16778         * tree-streamer-in.c: Likewise.
16779         * tree-streamer.c (record_common_node): Likewise.
16780         * tree.c (tree_code_size): Likewise.
16781         (wide_int_to_tree_1): Likewise.
16782         (type_contains_placeholder_1): Likewise.
16783         (build_common_tree_nodes): Likewise.
16784         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
16785         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
16786         (POINTER_BOUNDS_P): Likewise.
16787         (BOUNDED_TYPE_P): Likewise.
16788         (BOUNDED_P): Likewise.
16789         (CALL_WITH_BOUNDS_P): Likewise.
16790         (pointer_bounds_type_node): Likewise.
16791         * value-prof.c (gimple_ic): Likewise.
16792         * var-tracking.c (vt_add_function_parameters): Likewise.
16793         * varasm.c (make_decl_rtl): Likewise.
16794         (assemble_start_function): Likewise.
16795         (output_constant): Likewise.
16796         (maybe_assemble_visibility): Likewise.
16797         * varpool.c (ctor_for_folding): Likewise.
16798         * chkp-builtins.def: Remove.
16799         * ipa-chkp.c: Remove.
16800         * ipa-chkp.h: Remove.
16801         * rtl-chkp.c: Remove.
16802         * rtl-chkp.h: Remove.
16803         * tree-chkp-opt.c: Remove.
16804         * tree-chkp.c: Remove.
16805         * tree-chkp.h: Remove.
16807 2018-06-07  Carl Love  <cel@us.ibm.com>
16809         * config/rs6000/vsx.md (vextract_fp_from_shorth,
16810         vextract_fp_from_shortl): Add BE support.
16812 2018-06-07  Paul Koning  <ni1d@arrl.net>
16814         * compare-elim.c (try_merge_compare): Don't merge compare if
16815         address contains a side effect.
16816         (try_eliminate_compare): Likewise.
16818 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
16820         * config.gcc: Support "tremont".
16821         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
16822         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16823         PROCESSOR_TREMONT.
16824         * config/i386/i386.c (m_TREMONT): Define.
16825         (processor_target_table): Add "tremont".
16826         (PTA_TREMONT): Define.
16827         (ix86_lea_outperforms): Add TARGET_TREMONT.
16828         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
16829         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
16830         and M_INTEL_GOLDMONT_PLUS.
16831         (fold_builtin_cpu): Add "tremont".
16832         (ix86_add_stmt_cost): Add TARGET_TREMONT.
16833         (ix86_option_override_internal): Add "tremont".
16834         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
16835         (processor_type): Add PROCESSOR_TREMONT.
16836         * config/i386/x86-tune.def: Add m_TREMONT.
16837         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
16839 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16841         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
16842         symbol defined for msp430i* devices to be lower case.
16844 2018-06-07  Richard Biener  <rguenther@suse.de>
16846         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
16847         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
16848         Properly wrap signed arithmetic if overflow wraps.
16850 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
16852         PR tree-optimization/69615
16853         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
16854         of a cast from a same precision integral SSA_NAME in a bb dominated
16855         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
16856         cast to utype if rhs2 has already a compatible type.
16858 2018-06-07  Richard Biener  <rguenther@suse.de>
16860         PR tree-optimization/85935
16861         * graphite-scop-detection.c (find_params_in_bb): Analyze
16862         condition operands with respect to the correct loop.  Assert
16863         the analysis doesn't fail.
16865 2018-06-04  Carl Love  <cel@us.ibm.com>
16867         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
16868         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
16869         as it is slightly cheaper.
16870         (first_match_or_eos_index_<mode>):
16871         Calculate index using natural element order.
16872         (first_match_index_<mode>):
16873         Calculate index using natural element order.
16874         (first_match_or_eos_index_<mode>):
16875         Calculate index using natural order.
16876         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
16877         for BE and LE modes.
16878         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
16879         P9V_BUILTIN_VCLZLSBB_V16QI.
16880         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
16881         specific.
16883 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16885         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
16886         indentation and line wrap for many prototypes.  Add missing
16887         @smallexample directives around block of prototypes for vec_xl and
16888         vec_xst.
16890 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
16892         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
16893         track if we pass or return IEEE 128-bit floating point.
16894         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
16895         C++ mangling that is compatible with GCC 8.1.
16896         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
16897         (init_cumulative_args): Note if we pass or return IEEE 128-bit
16898         floating point types.
16899         (rs6000_function_arg_advance_1): Likewise.
16900         (rs6000_mangle_type): Optionally generate mangled names that match
16901         what GCC 8.1 generated for IEEE 128-bit floating point types.
16902         (rs6000_globalize_decl_name): If we have an external function that
16903         passes or returns IEEE 128-bit types, generate a weak reference
16904         from the mangled name used in GCC 8.1 to the current mangled
16905         name.
16906         (rs6000_init_builtins): Make __ibm128 use the long double type if
16907         long double is IBM extended double.  Make __float128 use the long
16908         double type if long double is IEEE 128-bit.
16910         PR target/85657
16911         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
16912         macro for __ibm128 built-in functions.
16913         (PACK_IF): Add __ibm128 pack/unpack functions.
16914         (UNPACK_IF): Likewise.
16915         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
16916         enable long double built-in functions if long double is IEEE
16917         128-bit floating point.
16918         (rs6000_invalid_builtin): Update long double built-in function
16919         error message.
16920         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
16921         functions, adjust the built-in function to use the long double
16922         built-in function if __ibm128 and long double are the same type.
16923         * doc/extend.texi (PowerPC builtins): Update documention for
16924         __builtin_{,un}pack_longdouble.  Add documentation for
16925         __builtin_{,un}pack_ibm128.
16927 2018-06-06  Jim Wilson  <jimw@sifive.com>
16929         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
16930         (struct machine_function): New field interrupt_mode.
16931         (riscv_handle_type_attribute): New function.  Add forward declaration.
16932         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
16933         (riscv_expand_epilogue): Check interrupt_mode field.
16934         (riscv_set_current_function): Check interrupt attribute args and
16935         set interrupt_mode field.
16936         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
16937         (riscv_sret, riscv_uret): New.
16938         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
16939         new arguments to interrupt attribute.
16941 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
16943         PR target/63177
16944         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
16945         Don't handle -mcpu=power8 if -mpower9-vector is also used.
16947 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16949         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
16950         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
16951         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16952         several redundant entries.
16954 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
16956         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
16957         type from "rtx" to "rtx_insn *".
16958         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
16959         for local "call_insn", removing cast.
16960         (ix86_expand_call): Likewise, introducing a "call_insn" local.
16962 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
16964         PR tree-optimization/86066
16965         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
16966         for BIT_INSERT_EXPR stores.
16968 2018-06-06  Richard Biener  <rguenther@suse.de>
16970         PR tree-optimization/86062
16971         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16972         component refs ontop
16973         of to be offsetted base.
16975 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16977         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
16978         to be static and remove check on interrupt attribute name.
16980 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16982         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16983         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
16985 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
16987         PR target/79924
16988         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
16989         second argument.
16990         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
16991         Remove second argument, change how error is called.
16992         (aarch64_layout_arg): Remove second argument from
16993         aarch64_err_no_fpadvsimd call.
16994         (aarch64_init_cumulative_args): Ditto.
16995         (aarch64_gimplify_va_arg_expr): Ditto.
16996         * config/aarch64/aarch64.md (mov<mode>): Ditto.
16998 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
17000         * config/i386/i386.md (simple_return_indirect_internal): New expander.
17001         (*simple_return_indirect_internal<mode>): Rename from
17002         simple_return_indirect_internal.  Use W mode iterator.
17003         (rstorssp): New expander.
17004         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
17005         (clrssbsy): New expander.
17006         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
17008 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17010         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17011         __typeof__.
17012         (cmse_check_pointed_object): Likewise.
17014 2018-06-05  Martin Liska  <mliska@suse.cz>
17016         PR gcov-profile/47618
17017         * doc/invoke.texi: Document how -fprofile-dir format
17018         is extended.
17020 2018-06-05  Richard Biener  <rguenther@suse.de>
17022         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17023         removal pretend DOM info isn't available so we do not update
17024         it and only remove edges, not dominated blocks.  Actually free
17025         DOM info in case we removed something.  Remove unreachable blocks.
17026         (mfb_keep_latches): Work with either DOM info or marked backedges.
17027         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17028         first.  Mark backedges if DOM info isn't available.
17029         (Re-)compute DOM info after cleanup_control_flow_pre.
17031 2018-06-05  Richard Biener  <rguenther@suse.de>
17033         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17034         (locus_discrim_hasher::hash): Adjust.
17035         (locus_discrim_hasher::equal): Likewise.
17036         (next_discriminator_for_locus): Work on line directly.
17037         (same_line_p): Pass in expanded locus1 as well.
17038         (assign_discriminators): Avoid redundant location expansions.
17040 2018-06-05  Richard Biener  <rguenther@suse.de>
17042         PR tree-optimization/86046
17043         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17044         if required after clearing TREE_ADDRESSABLE.
17046 2018-06-05  Richard Biener  <rguenther@suse.de>
17048         PR tree-optimization/86047
17049         * tree-ssa-loop.c (for_each_index): Glob handling of all
17050         decls and constants and really handle all of them.
17052 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17054         PR target/81497
17055         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17056         qualifier_void_pointer and qualifier_const_void_pointer.
17057         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17058         (arm_init_builtins): Handle the above.
17059         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17060         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17061         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17062         void intrinsics.
17064 2018-06-05  Martin Liska  <mliska@suse.cz>
17066         * auto-profile.c (read_autofdo_file): Do not use
17067         gcov_ctr_summary struct.
17068         (afdo_callsite_hot_enough_for_early_inline): Likewise.
17069         * coverage.c (struct counts_entry): Likewise.
17070         (read_counts_file): Read just single summary entry.
17071         (get_coverage_counts): Use gcov_summary struct.
17072         * coverage.h (get_coverage_counts): Likewise.
17073         * gcov-dump.c (dump_working_sets): Likewise.
17074         (tag_summary): Dump just single summary.
17075         * gcov-io.c (gcov_write_summary): Write just histogram
17076         summary.
17077         (gcov_read_summary): Read just single summary.
17078         (compute_working_sets): Use gcov_summary struct.
17079         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
17080         of GCOV_COUNTERS_SUMMABLE.
17081         (GCOV_COUNTERS_SUMMABLE): Remove.
17082         (GCOV_FIRST_VALUE_COUNTER): Replace with
17083         GCOV_COUNTER_V_INTERVAL.
17084         (struct gcov_ctr_summary): Remove.
17085         (struct gcov_summary): Directly use fields of former
17086         gcov_ctr_summary.
17087         (compute_working_sets): Use gcov_summary struct.
17088         * gcov.c (read_count_file): Do not use ctrs fields.
17089         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
17090         struct.
17091         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
17092         struct.
17093         * profile.c: Likewise.
17094         * profile.h: Likewise.
17096 2018-06-05  Martin Liska  <mliska@suse.cz>
17098         PR gcov-profile/84846
17099         * gcov.c (output_lines): Print working directory only
17100         in intermediate format.
17102 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
17104         * config/s390/s390-builtin-types.def: Add void function type.
17105         * config/s390/s390-builtins.def: Use the function type for the
17106         tbeginc builtin.
17108 2018-06-04  Jim Wilson  <jimw@sifive.com>
17110         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
17111         to int.
17112         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
17113         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
17114         handle EH_RETURN_DATA_REGNO registers properly.
17115         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
17116         (riscv_expand_epilogue): Update comment.  Change argument name and
17117         type.  Update code to use new name and type.  Pass new args to
17118         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
17119         EXCEPTION_RETURN.
17120         * config/riscv/riscv.md (NORMAL_RETURN): New.
17121         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
17122         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
17123         (eh_return): Call gen_eh_return_internal and emit barrier.
17124         (eh_return_internal): Call riscv_expand_epilogue.
17126 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
17128         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
17129         bit_insertion field and declare can_be_merged_into method.
17130         (merged_store_group::can_be_merged_into): New method.
17131         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17132         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
17133         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17135 2018-06-04  Richard Biener  <rguenther@suse.de>
17137         PR tree-optimization/85955
17138         * builtins.c (fold_builtin_sincos): Convert pointers to
17139         destination to appropriate type before dereferencing.
17141 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
17143         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17145 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
17147         * expr.c (expand_expr_real_1): Force the operand into memory if
17148         its TYPE_MODE is BLKmode and if there is no integer mode for
17149         the number of bits being extracted.
17151 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17153         PR target/85832
17154         PR target/86036
17155         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17156         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17158 2018-06-04  Richard Biener  <rguenther@suse.de>
17160         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17161         (cleanup_tree_cfg_noloop): ... single caller.  Do
17162         start_recording_case_labels later.
17164 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
17166         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17167         to _IMMINTRIN_H_INCLUDED.
17168         * config/i386/pconfigintrin.h: Ditto.
17169         * config/i386/waitpkgintrin.h: Ditto.
17170         * config/i386/immintrin.h: Add includes for sgxintrin.h,
17171         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17172         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17173         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17174         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17175         waitpkgintrin.h and cldemoteintrin.h.
17177 2018-06-04  Richard Biener  <rguenther@suse.de>
17179         PR tree-optimization/86038
17180         * tracer.c (find_best_successor): Check probability for
17181         being initialized, bail out if not.
17183 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
17185         PR target/86003
17186         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17187         of bits to ignore when comparing architectures.
17189 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17191         PR tree-optimization/69615
17192         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17193         maximum or minimum of the type, try to merge it also as if
17194         range1 is + [-, x - 1] or + [x + 1, -].
17196         PR c++/86025
17197         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17199 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
17201         PR tree-optimization/86034
17202         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17203         the unsigned bitfield type in a bit insertion sequence if it does not
17204         have a larger precision than the bitfield size.
17205         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17207 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
17209         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17211 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17213         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17214         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17215         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17216         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17218 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17220         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17221         Disable -fdelete-null-pointer-checks for ELF toolchain.
17223 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17224             Kito Cheng  <kito.cheng@gmail.com>
17226         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17227         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17228         (nds32*-*-*): Add glibc and uclibc conditions.
17229         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17230         (TARGET_EXCEPT_UNWIND_INFO): Define.
17231         * config/nds32/elf.h: New file.
17232         * config/nds32/linux.h: New file.
17233         * config/nds32/nds32-elf.opt: New file.
17234         * config/nds32/nds32-linux.opt: New file.
17235         * config/nds32/nds32-fp-as-gp.c
17236         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17237         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17238         TARGET_LINUX_ABI.
17239         (nds32_asm_file_end): Ditto.
17240         (nds32_print_operand): Ditto.
17241         (nds32_insert_attributes): Ditto.
17242         (nds32_init_libfuncs): New function.
17243         (TARGET_HAVE_TLS): Define.
17244         (TARGET_INIT_LIBFUNCS): Define.
17245         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17246         spec content.
17247         (TARGET_ELF): Apply different mcmodel setting.
17248         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17249         been migrated into elf.h and linux.h files.
17250         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17251         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17252         (mcmodel): The content has been migrated into nds32-elf.opt and
17253         nds32-linux.opt files.
17254         * config/nds32/t-elf: New file.
17255         * config/nds32/t-linux: New file.
17257 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17258             Shiva Chen  <shiva0217@gmail.com>
17260         * config/nds32/constants.md (unspec_volatile_element): Add
17261         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17262         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17263         optimization.
17264         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17265         (make_pass_nds32_fp_as_gp): Declare.
17266         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17267         optmization pass.
17268         (nds32_asm_function_end_prologue): Remove unused asm output.
17269         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17270         (nds32_asm_file_start): Output necessary fp_as_gp information.
17271         (nds32_option_override): Adjust register usage.
17272         (nds32_expand_prologue): Consider fp_as_gp situation.
17273         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17274         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17275         (epilogue): Ditto.
17276         (return): Ditto.
17277         (simple_return): Ditto.
17278         (omit_fp_begin): Output special directive for fp_as_gp.
17279         (omit_fp_end): Output special directive for fp_as_gp.
17280         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17281         mforbid-fp-as-gp): New options.
17283 2018-06-01  Mark Wielaard  <mark@klomp.org>
17285         * dwarf2out.c (dwarf2out_finish): Remove generation of
17286         DW_AT_loclists_base.
17288 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17290         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17291         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17292         (struct merged_store_group): Add bit_insertion field.
17293         (dump_char_array): Use standard hexadecimal format.
17294         (merged_store_group::merged_store_group): Set bit_insertion to false.
17295         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17296         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17297         also print the mask in the dump file.
17298         (pass_store_merging::gate): Minor tweak.
17299         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17300         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17301         stores with INTEGER_CST stores.
17302         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17303         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17304         and use it throughout.  Generate bit insertion sequences if need be.
17305         (pass_store_merging::process_store): Remove redundant condition.
17306         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17308 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17310         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17311         the 128-bit floating point types.  Fix function comment.
17313 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17315         * config/aarch64/aarch64-simd.md
17316         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17317         mnemonics.
17318         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17319         mnemonics.
17321 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17323         PR tree-optimization/85989
17324         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17325         variable.
17326         (backprop::intersect_uses): Check it when deciding whether this
17327         is a backedge reference.
17328         (backprop::process_block): Add each phi to m_visited_phis
17329         after visiting it, then clear it at the end.
17331 2018-06-01  Richard Biener  <rguenther@suse.de>
17333         * tree-vectorizer.h (vect_dr_stmt): New function.
17334         (vect_get_load_cost): Adjust.
17335         (vect_get_store_cost): Likewise.
17336         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17337         Use vect_dr_stmt instead of DR_SMTT.
17338         (vect_record_base_alignments): Likewise.
17339         (vect_calculate_target_alignment): Likewise.
17340         (vect_compute_data_ref_alignment): Likewise and make static.
17341         (vect_update_misalignment_for_peel): Likewise.
17342         (vect_verify_datarefs_alignment): Likewise.
17343         (vector_alignment_reachable_p): Likewise.
17344         (vect_get_data_access_cost): Likewise.  Pass down
17345         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17346         (vect_get_peeling_costs_all_drs): Likewise.
17347         (vect_peeling_hash_get_lowest_cost): Likewise.
17348         (vect_enhance_data_refs_alignment): Likewise.
17349         (vect_find_same_alignment_drs): Likewise.
17350         (vect_analyze_data_refs_alignment): Likewise.
17351         (vect_analyze_group_access_1): Likewise.
17352         (vect_analyze_group_access): Likewise.
17353         (vect_analyze_data_ref_access): Likewise.
17354         (vect_analyze_data_ref_accesses): Likewise.
17355         (vect_vfa_segment_size): Likewise.
17356         (vect_small_gap_p): Likewise.
17357         (vectorizable_with_step_bound_p): Likewise.
17358         (vect_prune_runtime_alias_test_list): Likewise.
17359         (vect_analyze_data_refs): Likewise.
17360         (vect_supportable_dr_alignment): Likewise.
17361         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17362         (vect_gen_prolog_loop_niters): Likewise.
17363         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17364         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17365         modify DR_STMT.
17366         (vect_recog_mask_conversion_pattern): Likewise.
17367         (vect_try_gather_scatter_pattern): Likewise.
17368         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17369         to vect_get_store_cost.
17370         (vect_get_store_cost): Get stmt_info instead of DR.
17371         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17372         (vect_get_load_cost): Get stmt_info instead of DR.
17374 2018-06-01  Richard Biener  <rguenther@suse.de>
17376         PR middle-end/86017
17377         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17378         vars wrapped in MEM_REFs.
17380 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17382         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17383         Fix subreg tests so that we only return a choice between
17384         GENERAL_REGS and FP_REGS if the original classes included both.
17386 2018-06-01  Richard Biener  <rguenther@suse.de>
17388         PR ipa/85960
17389         * tree-ssa-structalias.c (get_function_part_constraint):
17390         Handle NULL fi->decl.
17391         (find_func_aliases_for_call): Properly handle indirect
17392         fi from direct call.
17393         (find_func_clobbers): Likewise.
17394         (ipa_pta_execute): Likewise.
17395         (create_variable_info_for): For functions that are ifunc_resolver
17396         resolve to a varinfo that contains the result of the resolver call.
17397         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17398         aliases.
17400 2018-05-31  Michael Collison  <michael.collison@arm.com>
17402         * config/aarch64/aarch64.md:
17403         (*fix_to_zero_extenddfdi2): New pattern.
17404         * gcc.target/aarch64/fix_extend1.c: New testcase.
17406 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17408         PR middle-end/78809
17409         PR middle-end/83026
17410         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17411         and BUILT_IN_STRNCMP_EQ.
17412         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17413         BUILT_IN_STRNCMP_EQ.
17414         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17415         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17416         (gimple_fold_builtin): Likewise.
17417         * tree-ssa-strlen.c (compute_string_length): New function.
17418         (determine_min_obsize): New function.
17419         (handle_builtin_string_cmp): New function to handle calls to
17420         string compare functions.
17421         (strlen_optimize_stmt): Add handling to builtin string compare
17422         calls.
17423         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17424         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17425         * tree.c (build_common_builtin_nodes): Add new defines of
17426         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17428 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17430         PR target/85984
17431         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17432         functions with naked attribute.
17434 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17436         * config/i386/sse.md (avx_vec_concat<mode>):
17437         Substitute concat_tg_mode mode attribute with xtg_mode.
17438         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17439         (concat_tg_mode): Remove mode attribute.
17441 2018-05-31  Martin Sebor  <msebor@redhat.com>
17443         PR c/82063
17444         * calls.c (alloc_max_size): Correct a logic error/typo.
17445         Treat excessive arguments as infinite.  Warn for invalid arguments.
17446         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17448 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17450         PR target/85829
17451         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17452         and movx for Haswell.
17454 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17455             Cesar Philippidis  <cesar@codesourcery.com>
17457         PR middle-end/85879
17458         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17459         when emitting error on private/firstprivate reductions.
17460         * omp-low.c (lower_omp_target): Avoid reference-type processing
17461         on pointers for firstprivate clause.
17463 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17465         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17466         (st1x2): Likewise.
17467         (st1x3): Likewise.
17468         * config/aarch64/aarch64-simd.md
17469         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17470         (aarch64_ld1_x3_<mode>): Likewise
17471         (aarch64_st1x2<VALLDIF:mode>): Likewise
17472         (aarch64_st1_x2_<mode>): Likewise
17473         (aarch64_st1x3<VALLDIF:mode>): Likewise
17474         (aarch64_st1_x3_<mode>): Likewise
17475         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17476         (vld1_s8_x3): Likewise.
17477         (vld1_u16_x3): Likewise.
17478         (vld1_s16_x3): Likewise.
17479         (vld1_u32_x3): Likewise.
17480         (vld1_s32_x3): Likewise.
17481         (vld1_u64_x3): Likewise.
17482         (vld1_s64_x3): Likewise.
17483         (vld1_f16_x3): Likewise.
17484         (vld1_f32_x3): Likewise.
17485         (vld1_f64_x3): Likewise.
17486         (vld1_p8_x3): Likewise.
17487         (vld1_p16_x3): Likewise.
17488         (vld1_p64_x3): Likewise.
17489         (vld1q_u8_x3): Likewise.
17490         (vld1q_s8_x3): Likewise.
17491         (vld1q_u16_x3): Likewise.
17492         (vld1q_s16_x3): Likewise.
17493         (vld1q_u32_x3): Likewise.
17494         (vld1q_s32_x3): Likewise.
17495         (vld1q_u64_x3): Likewise.
17496         (vld1q_s64_x3): Likewise.
17497         (vld1q_f16_x3): Likewise.
17498         (vld1q_f32_x3): Likewise.
17499         (vld1q_f64_x3): Likewise.
17500         (vld1q_p8_x3): Likewise.
17501         (vld1q_p16_x3): Likewise.
17502         (vld1q_p64_x3): Likewise.
17503         (vst1_s64_x2): Likewise.
17504         (vst1_u64_x2): Likewise.
17505         (vst1_f64_x2): Likewise.
17506         (vst1_s8_x2): Likewise.
17507         (vst1_p8_x2): Likewise.
17508         (vst1_s16_x2): Likewise.
17509         (vst1_p16_x2): Likewise.
17510         (vst1_s32_x2): Likewise.
17511         (vst1_u8_x2): Likewise.
17512         (vst1_u16_x2): Likewise.
17513         (vst1_u32_x2): Likewise.
17514         (vst1_f16_x2): Likewise.
17515         (vst1_f32_x2): Likewise.
17516         (vst1_p64_x2): Likewise.
17517         (vst1q_s8_x2): Likewise.
17518         (vst1q_p8_x2): Likewise.
17519         (vst1q_s16_x2): Likewise.
17520         (vst1q_p16_x2): Likewise.
17521         (vst1q_s32_x2): Likewise.
17522         (vst1q_s64_x2): Likewise.
17523         (vst1q_u8_x2): Likewise.
17524         (vst1q_u16_x2): Likewise.
17525         (vst1q_u32_x2): Likewise.
17526         (vst1q_u64_x2): Likewise.
17527         (vst1q_f16_x2): Likewise.
17528         (vst1q_f32_x2): Likewise.
17529         (vst1q_f64_x2): Likewise.
17530         (vst1q_p64_x2): Likewise.
17531         (vst1_s64_x3): Likewise.
17532         (vst1_u64_x3): Likewise.
17533         (vst1_f64_x3): Likewise.
17534         (vst1_s8_x3): Likewise.
17535         (vst1_p8_x3): Likewise.
17536         (vst1_s16_x3): Likewise.
17537         (vst1_p16_x3): Likewise.
17538         (vst1_s32_x3): Likewise.
17539         (vst1_u8_x3): Likewise.
17540         (vst1_u16_x3): Likewise.
17541         (vst1_u32_x3): Likewise.
17542         (vst1_f16_x3): Likewise.
17543         (vst1_f32_x3): Likewise.
17544         (vst1_p64_x3): Likewise.
17545         (vst1q_s8_x3): Likewise.
17546         (vst1q_p8_x3): Likewise.
17547         (vst1q_s16_x3): Likewise.
17548         (vst1q_p16_x3): Likewise.
17549         (vst1q_s32_x3): Likewise.
17550         (vst1q_s64_x3): Likewise.
17551         (vst1q_u8_x3): Likewise.
17552         (vst1q_u16_x3): Likewise.
17553         (vst1q_u32_x3): Likewise.
17554         (vst1q_u64_x3): Likewise.
17555         (vst1q_f16_x3): Likewise.
17556         (vst1q_f32_x3): Likewise.
17557         (vst1q_f64_x3): Likewise.
17558         (vst1q_p64_x3): Likewise.
17560 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17562         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17563         user_label_prefix to name.
17565         * tree-core.h: Update comment about the format of NAME string
17566         passed to handler in attribute_spec.
17568         * config/msp430/msp430.md: Remove erroneous subreg expression from
17569         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17570         zero_extend{q,h}isi2.
17572 2018-05-30  Borislav Petkov  <bp@suse.de>
17574         * doc/extend.texi: Document some architecture specific
17575         constraints and sort entries.
17577 2018-05-30  Martin Sebor  <msebor@redhat.com>
17579         PR middle-end/85369
17580         * builtins.c (expand_builtin_stpcpy_1): New function.
17581         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17582         only if the former succeeds.
17584 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17586         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17587         in saphira.
17589 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17591         * doc/invoke.texi (-flinker-output): Document
17593 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17595         * passes.c (ipa_write_summaries): Only modify statements if body
17596         is in memory.
17597         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17598         incrementally linking.
17599         (ipa_passes): Likewise.
17600         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17601         pass down resolution info.
17602         * common.opt (flag_incremental_link): Update info.
17603         * gcc.c (plugin specs): Turn flinker-output=* to
17604         -plugin-opt=-linker-output-known
17605         * toplev.c (compile_file): Also cut compilation when doing incremental
17606         link.
17607         * flag-types. (enum lto_partition_model): Add
17608         LTO_LINKER_OUTPUT_NOLTOREL.
17609         (invoke.texi): Add -flinker-output docs.
17610         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17611         link same way as WPA; do not stream in dead initializers.
17613         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17614         darf2out_register_external_decl): Support incremental link.
17616 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17618         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17620 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17622         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17623         it down to simple_object_copy_lto_debug_sections.
17624         (run_gcc): Determine incremental LTO link time and configure
17625         lto1 into non-wpa mode, disable renaming of debug sections.
17627 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17629         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17630         descriptions of various incorrectly documented functions.
17632 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17634         Revert:
17635         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17636         address check not strict.
17638 2018-05-30  Richard Biener  <rguenther@suse.de>
17640         PR tree-optimization/85964
17641         * tracer.c (better_p): Drop initialized count check, we only
17642         call the function with initialized counts now.
17643         (find_best_successor): Do find a best edge if one
17644         has uninitialized count.
17645         (find_best_predecessor): Likewise.  Do BB frequency check only
17646         if count is initialized.
17648 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17650         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17651         (aarch64_ldrstr_offset_compare): New.
17652         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17653         load/store orderings.
17654         (aarch64_gen_adjusted_ldpstp): Likewise.
17656 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17658         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17659         Check for subset of GENERAL_REGS and FP_REGS.
17660         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17661         r=w alternative.
17663 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17665         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17666         and wi::to_poly_offset.  Add the current offset and then check
17667         whether the sum fits, rather than using an unchecked addition of
17668         a checked term.  Check for a shwi rather than a uhwi.
17669         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17670         (store_constructor): Use poly_int_tree_p.
17671         (expand_expr_real_1): Likewise.
17672         * function.c (assign_temp): Likewise.
17673         * fold-const.c (const_binop): Use poly_int_tree_p and
17674         wi::to_poly_offset.
17675         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17676         division.
17677         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17678         to_poly_offset for MEM offsets.
17679         * ipa-icf.c (sem_variable::equals): Likewise.
17680         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17681         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17682         wi::to_poly_offset for BIT_FIELD_REF offsets.
17683         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17684         wi::to_poly_offset.
17685         * var-tracking.c (emit_note_insn_var_location): Use
17686         tree_to_poly_uint64.
17688 2018-05-29  Jim Wilson  <jimw@sifive.com>
17690         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17692 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17694         PR target/85950
17695         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17696         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17697         sequence.
17698         (sse4_1_round<mode>2): Use nonimmediate_operand
17699         for operand 1 predicate.
17701 2018-05-29  Martin Sebor  <msebor@redhat.com>
17702             Richard Biener  <rguenther@suse.de>
17704         PR testsuite/85888
17705         * calls.c (get_size_range): Call determine_value_range instead
17706         of get_value_range..
17707         * tree-vrp.h (determine_value_range): Declared new function.
17708         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17710 2018-05-29  Richard Biener  <rguenther@suse.de>
17712         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17713         sure to use non-pattern stmts for get_earlier_stmt arguments.
17714         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17715         called on pattern stmts.
17716         (get_later_stmt): Likewise.
17718 2018-05-29  Martin Liska  <mliska@suse.cz>
17720         PR gcov-profile/85759
17721         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
17722         env variables.
17724 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
17726         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
17727         VEC_UNPACK_*_EXPR.
17728         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
17729         VEC_PACK_*_EXPR.
17731         PR target/85918
17732         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17733         VEC_PACK_FLOAT_EXPR): New tree codes.
17734         * tree-pretty-print.c (op_code_prio): Handle
17735         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17736         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17737         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17738         * tree-inline.c (estimate_operator_cost): Likewise.
17739         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
17740         * fold-const.c (const_binop): Likewise.
17741         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
17742         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17743         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17744         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
17745         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17746         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17747         * expr.c (expand_expr_real_2): Likewise.
17748         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
17749         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
17750         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
17751         optabs.
17752         * optabs.c (expand_widen_pattern_expr): For
17753         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
17754         sign from result type rather than operand's type.
17755         (expand_binop_directly): For vec_packu_float_optab and
17756         vec_packs_float_optab allow result type to be different from operand's
17757         type.
17758         * optabs-tree.c (optab_for_tree_code): Handle
17759         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17760         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
17761         * tree-vect-generic.c (expand_vector_operations_1):  Handle
17762         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17763         VEC_PACK_FLOAT_EXPR.
17764         * tree-vect-stmts.c (supportable_widening_operation): Handle
17765         FIX_TRUNC_EXPR.
17766         (supportable_narrowing_operation): Handle FLOAT_EXPR.
17767         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
17768         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
17769         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
17770         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
17771         mode attributes.
17772         (vec_pack<floatprefix>_float_<mode>): New expander.
17773         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
17774         attributes.
17775         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
17776         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
17777         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
17778         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
17779         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
17780         Document.
17781         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
17782         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
17783         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17784         VEC_PACK_FLOAT_EXPR): Document.
17786 2018-05-29  Richard Biener  <rguenther@suse.de>
17788         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
17789         member.
17790         (stmt_vec_info_vec): Make pointer.
17791         (init_stmt_vec_info_vec): Remove.
17792         (free_stmt_vec_info_vec): Likewise.
17793         (set_stmt_vec_info_vec): New function.
17794         (free_stmt_vec_infos): Likewise.
17795         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
17796         (set_vinfo_for_stmt): Likewise.
17797         (get_earlier_stmt): Likewise.
17798         (get_later_stmt): Likewise.
17799         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
17800         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
17801         (vec_info::~vec_info): Free stmt_vec_infos.
17802         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
17803         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
17804         (pass_slp_vectorize::execute): Likewise.
17805         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
17806         (free_stmt_vec_info_vec): Likewise.
17807         (set_stmt_vec_info_vec): New function.
17808         (free_stmt_vec_infos): Likewise.
17809         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
17810         the global stmt_vec_info_vec.
17811         * tree-parloops.c (gather_scalar_reductions): Use
17812         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
17813         vector.
17815 2018-05-29  Richard Biener  <rguenther@suse.de>
17817         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
17819 2018-05-29  Martin Liska  <mliska@suse.cz>
17820             David Malcolm  <dmalcolm@redhat.com>
17822         * vec.c (test_reverse): New.
17823         (vec_c_tests): Add new test.
17824         * vec.h (vl_ptr>::reverse): New function.
17826 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
17828         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
17830         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
17831         and later.
17833 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17835         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
17837 2018-05-28  Richard Biener  <rguenther@suse.de>
17839         PR tree-optimization/85933
17840         * tree-vect-data-refs.c (vect_record_base_alignments): Only
17841         look at stmts marked as vectorizable.
17843 2018-05-28  Richard Biener  <rguenther@suse.de>
17845         PR tree-optimization/85934
17846         * tree-vect-generic.c (expand_vector_operations_1): Hoist
17847         vector boolean check before scalar optimization.
17849 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
17851         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
17852         for armv5te.
17854 2018-05-28  Mark Wielaard  <mark@klomp.org>
17856         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
17857         if it is an expression containing a minus sign.
17859 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
17861         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
17863 2018-05-27  Paul Koning  <ni1d@arrl.net>
17865         * config/pdp11/pdp11.md (truncsihi2): Remove.
17867 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
17868             Chung-Ju Wu  <jasonwucj@gmail.com>
17870         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
17871         implementation.
17872         (unaligned_store_dw): Ditto.
17873         * config/nds32/nds32-memory-manipulation.c
17874         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17875         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
17876         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
17877         (emit_setmem_word_loop): Rename to ...
17878         (emit_setmem_doubleword_loop): ... this.
17879         (nds32_gen_dup_4_byte_to_word_value): New function.
17880         (nds32_gen_dup_8_byte_to_double_word_value): New function.
17881         (nds32_expand_setmem_loop): Refine implementation.
17882         (nds32_expand_setmem_loop_v3m): Ditto.
17883         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
17884         pattern.
17886 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17888         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
17890 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17892         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
17893         (nds32_init_machine_status): Initialize machine->attr_naked_p and
17894         machine->attr_no_prologue_p.
17895         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
17896         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
17897         (nds32_expand_epilogue): Consider attr_naked_p.
17898         (nds32_expand_epilogue_v3pop): Likewise.
17899         (nds32_can_use_return_insn): Likewise.
17900         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
17901         attr_no_prologue_p fields.
17902         * config/nds32/nds32.opt (mret-in-naked-func): New option.
17904 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
17906         PR target/85918
17907         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
17908         attributes.
17909         * config/i386/sse.md
17910         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
17911         Rename to ...
17912         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
17913         ... this.
17914         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
17915         Rename to ...
17916         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
17917         ... this.
17918         (*<floatsuffix>floatv2div2sf2): Rename to ...
17919         (*float<floatunssuffix>v2div2sf2): ... this.
17920         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
17921         (float<floatunssuffix>v2div2sf2_mask): ... this.
17922         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
17923         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
17924         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
17925         to ...
17926         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
17927         ... this.
17928         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17929         Rename to ...
17930         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17931         ... this.
17932         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17933         Rename to ...
17934         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17935         ... this.
17936         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
17937         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
17938         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
17939         gen_ufix_truncv8dfv8si2.
17940         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
17941         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
17942         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
17943         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
17944         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
17945         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
17946         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
17947         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
17949 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17951         PR target/85900
17952         PR target/85345
17953         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
17955 2018-05-25  Jim Wilson  <jimw@sifive.com>
17957         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
17958         * config/riscv/riscv.c (struct machine_function): Add
17959         interrupt_handler_p and attribute_checked_p fields.
17960         (riscv_attribute_table): Add interrupt.
17961         (riscv_interrupt_type_p): New.
17962         (riscv_save_reg_p): Save extra regs for interrupt handler.
17963         (riscv_use_save_libcall): Return false  for interrupt handler.
17964         (riscv_first_stack_step): Add forward declaration.
17965         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
17966         for interrupt handler with large frame.  Use it for saved reg list.
17967         (riscv_expand_prologue): Move flag_stack_usage_info support to
17968         eliminate duplication.
17969         (riscv_expand_epilogue): Generate mret for interrupt handler.
17970         (riscv_epilogue_uses): New.
17971         (riscv_can_use_return_insn): Return false for interrupt handler.
17972         (riscv_function_ok_for_sibcall): Likewise.
17973         (riscv_set_current_function): Add interrupt handler support.
17974         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
17975         * config/riscv/riscv.md (UNSPECV_MRET): New.
17976         (GP_REGNUM): New.
17977         (riscv_frflags, riscv_fsflags): Use tab after opcode.
17978         (riscv_mret): New.
17979         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
17981 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
17983         PR tree-optimization/85712
17984         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
17985         this candidate has already been replaced in-situ by a copy.
17987 2018-05-25  Jason Merrill  <jason@redhat.com>
17989         PR c++/80485 - inline function non-zero address.
17990         * symtab.c (nonzero_address): Check DECL_COMDAT.
17992 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
17994         PR target/83628
17995         * config/alpha/alpha.md (ashlsi3): New insn pattern.
17996         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
17997         extension of SImode operation.  Use const123_operand predicate.
17998         (*saddsi_1): Remove.
17999         (*saddl_se_1): Ditto.
18000         (*ssubsi_1): Ditto.
18001         (*ssubl_se_1): Ditto.
18002         * config/alpha/predicates.md (const123_operand): New predicate.
18003         * config/alpha/constraints.md (P): Use IN_RANGE.
18005 2018-05-25  Richard Biener  <rguenther@suse.de>
18007         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18008         defaulted to true.
18009         (ref_maybe_used_by_stmt_p): Likewise.
18010         (stmt_may_clobber_ref_p): Likewise.
18011         (stmt_may_clobber_ref_p_1): Likewise.
18012         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18013         and pass it along.
18014         (ref_maybe_used_by_stmt_p): Likewise.
18015         (stmt_may_clobber_ref_p): Likewise.
18016         (stmt_may_clobber_ref_p_1): Likewise.
18017         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18018         the alias oracle to disambiguate DRs with stmts DR analysis
18019         couldn't handle.
18020         (vect_analyze_data_refs): Do not give up on not analyzable
18021         DRs for BB vectorization.  Remove code truncating the dataref
18022         vector.
18024 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
18026         PR target/85832
18027         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18028         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
18029         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18031 2018-05-25  Richard Biener  <rguenther@suse.de>
18033         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18034         function, combining stmt data ref gathering and fatal analysis
18035         parts.
18036         (vect_analyze_data_refs): Remove now redudnant code and simplify.
18037         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18038         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18039         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18040         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18042 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18044         PR tree-optimization/85720
18045         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18046         SCC if all partitions are builtins.
18047         (version_loop_by_alias_check): New parameter.  Generate cancelable
18048         runtime alias check if all partitions are builtins.
18049         (distribute_loop): Update call to above function.
18051 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18053         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18054         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18055         (parm_default_def_partition_arg): Ditto.
18056         (set_parm_default_def_partition): Ditto.
18057         (get_parm_default_def_partitions): Ditto and make it static.
18058         (get_undefined_value_partitions): Ditto and make it static.
18059         (remove_ssa_form): Refactor call to init_var_map here.
18060         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18061         computation for loop region.
18062         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18063         (register_default_def): Delete.
18064         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18065         (parm_default_def_partition_arg): Ditto.
18066         (set_parm_default_def_partition): Ditto.
18067         (get_parm_default_def_partitions): Ditto and make it static.
18068         (get_undefined_value_partitions): Ditto and make it static.
18069         (coalesce_with_default, coalesce_with_default): Update comment.
18070         (create_coalesce_list_for_region): New func factored out from
18071         create_outofssa_var_map.
18072         (populate_coalesce_list_for_outofssa): New func factored out from
18073         create_outofssa_var_map and coalesce_ssa_name.
18074         (create_outofssa_var_map): Delete.
18075         (coalesce_ssa_name): Refactor to support live range computation.
18076         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
18077         (get_parm_default_def_partitions): Delete.
18078         (get_undefined_value_partitions): Ditto.
18079         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
18080         computation for loop region.
18081         (new_tree_live_info, loe_visit_block): Ditto.
18082         (live_worklist, set_var_live_on_entry): Ditto.
18083         (calculate_live_on_exit, verify_live_on_entry): Ditto.
18084         * tree-ssa-live.h (struct _var_map): New fields.
18085         (init_var_map): Change decl.
18086         (region_contains_p): New.
18088 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18090         * tree-ssa-live.h (live_merge_and_clear): Delete.
18092 2018-05-25  Richard Biener  <rguenther@suse.de>
18094         PR c++/85912
18095         * tree-dump.c (dequeue_and_dump): Remove access to removed
18096         operand 2 of a SWITCH_EXPR.
18098 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18100         * doc/sourcebuild.texi (vect_double_cond_arith): Include
18101         multiplication and division.
18102         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
18103         (cond_udiv@var{m}, cond_umod@var{m}): Document.
18104         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
18105         (cond_udiv_optab, cond_umod_optab): New optabs.
18106         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
18107         (IFN_COND_RDIV): New internal functions.
18108         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
18109         TRUNC_MOD_EXPR and RDIV_EXPR.
18110         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
18111         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
18112         New unspecs.
18113         (SVE_INT_BINARY): Include mult.
18114         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18115         (optab, sve_int_op): Handle mult.
18116         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
18117         UNSPEC_COND_DIV.
18118         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
18119         for SVE_INT_BINARY_SD.
18121 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18123         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
18124         (optab, sve_int_op): Handle div and udiv.
18125         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
18126         for SVE_INT_BINARY_SD.
18127         (*<optab><mode>3): New insn for the same.
18129 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18131         * tree-vect-patterns.c: Include predict.h.
18132         (vect_recog_divmod_pattern): Restrict check for division support
18133         to when optimizing for size.
18135 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18137         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18138         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18139         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18140         (gimple_match_op::set_op): Likewise.
18141         (gimple_resimplify4): Declare.
18142         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18143         (expr::gen_transform): Likewise.
18144         (decision_tree::gen): Generate a simplification routine for 4 operands.
18145         * gimple-match-head.c (gimple_simplify): Add an overload for
18146         4 operands.  In the top-level function, handle up to 4 call
18147         arguments and call gimple_resimplify4.
18148         (gimple_resimplify4): New function.
18149         (build_call_internal): Pass a fourth operand.
18150         (maybe_push_to_seq): Likewise.
18151         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18152         Fold VEC_COND_EXPRs of an operation and a default value into
18153         an IFN_COND_* function if possible.
18154         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18155         New unspecs.
18156         (SVE_COND_FP_BINARY): Include them.
18157         (optab, sve_fp_op): Handle them.
18158         (SVE_INT_BINARY_REV): New code iterator.
18159         (SVE_COND_FP_BINARY_REV): New int iterator.
18160         (commutative): New int attribute.
18161         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18162         Declare.
18163         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18164         function.
18165         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18166         (*cond_<optab><mode>): New patterns for reversed operands.
18168 2018-05-25  Richard Biener  <rguenther@suse.de>
18170         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18171         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18172         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18173         (STMT_VINFO_GROUPED_ACCESS): Adjust.
18174         * tree-vect-data-refs.c (everywhere): Adjust users.
18175         * tree-vect-loop.c (everywhere): Likewise.
18176         * tree-vect-slp.c (everywhere): Likewise.
18177         * tree-vect-stmts.c (everywhere): Likewise.
18178         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18180 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18182         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18183         Rename to...
18184         (gcc_cv_as_section_exclude): ... this.
18185         Try Solaris as #exclude syntax.
18186         * configure: Regenerate.
18187         * config.in: Regenerate.
18188         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18189         SECTION_EXCLUDE.
18190         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18191         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18193         * varasm.c (default_elf_asm_named_section): Don't check if
18194         HAVE_GAS_SECTION_EXCLUDE is defined.
18196 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18198         * doc/md.texi: Update the documentation of the cond_* optabs
18199         to mention the new final operand.  Fix GET_MODE_NUNITS call.
18200         Describe the scalar case too.
18201         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18202         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18203         instead of 2.
18204         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18205         (get_conditional_internal_fn): Update comment.
18206         * tree-vect-loop.c (vectorizable_reduction): Pass the original
18207         accumulator value as a final argument to conditional functions.
18208         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18209         a define_expand and add an "else" operand.  Assert for now that
18210         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
18211         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18212         (*cond_<optab><mode>): New patterns.
18213         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18214         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18215         (UNSPEC_COND_EOR): Delete.
18216         (optab): Remove associated mappings.
18217         (SVE_INT_BINARY): New code iterator.
18218         (sve_int_op): Remove int attribute and add "minus" to the code
18219         attribute.
18220         (SVE_COND_INT_OP): Delete.
18221         (SVE_COND_FP_OP): Rename to...
18222         (SVE_COND_FP_BINARY): ...this.
18224 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18226         * optabs.c (can_reuse_operands_p): New function.
18227         (maybe_legitimize_operands): Try to reuse the results for
18228         earlier operands.
18230 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18232         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18233         Add {q} suffix to insn mnemonic.
18235 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18237         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18238         (msp430_warn_func_return): New.
18240 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18242         * fold-const.c (tree_nonzero_bits): New function.
18243         * fold-const.h (tree_nonzero_bits): Likewise.
18244         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18245         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18247 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18249         PR target/85900
18250         PR target/85345
18251         * varasm.c (assemble_alias): Check ifunc_resolver only on
18252         FUNCTION_DECL.
18254 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18256         PR target/85903
18257         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18258         when memory input operand is handled.
18260 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18262         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18263         global.
18264         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18266 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18268         * match.pd: Delay FMA folds until after vectorization.
18270 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18272         PR target/83009
18273         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18274         address check not strict.
18276 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18278         * gimple-match.h (gimple_match_op): New class.
18279         (mprts_hook): Replace parameters with a gimple_match_op *.
18280         (maybe_build_generic_op): Likewise.
18281         (gimple_simplified_result_is_gimple_val): Replace parameters with
18282         a const gimple_match_op *.
18283         (gimple_simplify): Replace code_helper * and tree * parameters with
18284         a gimple_match_op * parameter.
18285         (gimple_resimplify1): Replace code_helper *, tree and tree *
18286         parameters with a gimple_match_op * parameter.
18287         (gimple_resimplify2): Likewise.
18288         (gimple_resimplify3): Likewise.
18289         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18290         parameters with a gimple_match_op * parameter.
18291         * gimple-match-head.c (gimple_simplify): Change prototypes of
18292         auto-generated functions to take a gimple_match_op * instead of
18293         separate code_helper * and tree * parameters.  Make the same
18294         change in the top-level overload and update calls to the
18295         gimple_resimplify routines.  Update calls to the auto-generated
18296         functions and to maybe_push_res_to_seq in the publicly-facing
18297         operation-specific gimple_simplify overloads.
18298         (gimple_match_op::MAX_NUM_OPS): Define.
18299         (gimple_resimplify1): Replace rcode and ops with a single res_op
18300         parameter.  Update call to gimple_simplify.
18301         (gimple_resimplify2): Likewise.
18302         (gimple_resimplify3): Likewise.
18303         (mprts_hook): Replace parameters with a gimple_match_op *.
18304         (maybe_build_generic_op): Likewise.
18305         (build_call_internal): Replace type, nargs and ops with
18306         a gimple_match_op *.
18307         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18308         with a single gimple_match_op *.  Update calls to mprts_hook,
18309         build_call_internal and gimple_simplified_result_is_gimple_val.
18310         Factor out code that is common to the tree_code and combined_fn cases.
18311         * genmatch.c (expr::gen_transform): Replace tem_code and
18312         tem_ops with a gimple_match_op called tem_op.  Update calls
18313         to the gimple_resimplify functions and maybe_push_res_to_seq.
18314         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18315         res_ops.  Update call to the gimple_resimplify functions.
18316         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18317         (decision_tree::gen): Make the functions take a gimple_match_op *
18318         called res_op instead of separate res_code and res_ops parameters.
18319         Update call accordingly.
18320         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18321         and ops with a single res_op parameter.  Update calls to
18322         maybe_build_generic_op and maybe_push_res_to_seq.
18323         (fold_stmt_1): Update calls to gimple_simplify and
18324         replace_stmt_with_simplification.
18325         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18326         and gimple_simplified_result_is_gimple_val.
18327         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18328         gimple_simplify.
18329         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18330         with a gimple_match_op *.
18331         (vn_nary_build_or_lookup): Likewise.  Update call to
18332         vn_nary_build_or_lookup_1.
18333         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18334         gimple_match_op *.  Update calls to the gimple_resimplify routines
18335         and to gimple_simplified_result_is_gimple_val.
18336         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18337         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18338         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18339         (visit_nary_op): Likewise.
18340         (visit_reference_op_load): Likewise.
18342 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18344         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18345         modifier for printing the step amount.
18347 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18349         PR target/78849
18350         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18351         types.
18353 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18355         * doc/sourcebuild.texi (Endianness): New subsubsection.
18357 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18359         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18360         <prefetch_dynamic_strides>: New const bool field.
18361         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18362         prefetch_dynamic_strides.
18363         (exynosm1_prefetch_tune): Likewise.
18364         (thunderxt88_prefetch_tune): Likewise.
18365         (thunderx_prefetch_tune): Likewise.
18366         (thunderx2t99_prefetch_tune): Likewise.
18367         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18368         false.
18369         (aarch64_override_options_internal): Update to set
18370         PARAM_PREFETCH_DYNAMIC_STRIDES.
18371         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18372         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18373         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18374         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18375         prefetch-dynamic-strides setting.
18377 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18379         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18380         <minimum_stride>: New const int field.
18381         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18382         minimum_stride field defaulting to -1.
18383         (exynosm1_prefetch_tune): Likewise.
18384         (thunderxt88_prefetch_tune): Likewise.
18385         (thunderx_prefetch_tune): Likewise.
18386         (thunderx2t99_prefetch_tune): Likewise.
18387         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18388         <default_opt_level>: Set to 3.
18389         (aarch64_override_options_internal): Update to set
18390         PARAM_PREFETCH_MINIMUM_STRIDE.
18391         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18392         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18393         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18394         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18395         stride is constant and is below the minimum stride threshold.
18397 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18399         * config/arm/arm-cpus.in (mode26): Delete.
18400         (armv4): Delete mode26 reference.
18401         * config/arm/arm.c (arm_configure_build_target): Delete use of
18402         isa_bit_mode26.
18404 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18406         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18407         New insn pattern.
18408         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18409         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18410         for non-SSE modes.
18411         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18412         (floatunsdidf2): Ditto.
18414 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18416         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18417         (fixuns_trunc<mode>si2_avx512f): Ditto.
18418         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18419         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18420         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18422 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18424         PR rtl-optimization/79985
18425         * df-scan.c (df_insn_refs_collect): Remove special case for
18426         global registers and asm statements.
18428 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18430         * extend.texi (Global Register Variables): Rewrite the bullet list.
18431         Note that the register is available for allocation. Note that access
18432         via inline asm must use constraints. Add note about async-signal
18433         handlers. Remove paragraph about automagic register selection.
18435 2018-05-23  Richard Biener  <rguenther@suse.de>
18437         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18438         of fixed offset from memset VN.
18440 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18442         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18443         first_interp field.
18444         (alloc_cand_and_find_basis): Initialize first_interp field.
18445         (slsr_process_mul): Modify first_interp field.
18446         (slsr_process_add): Likewise.
18447         (slsr_process_cast): Modify first_interp field for each new
18448         interpretation.
18449         (slsr_process_copy): Likewise.
18450         (dump_candidate): Dump first_interp field.
18451         (replace_mult_candidate): Process all interpretations, not just
18452         subsequent ones.
18453         (replace_rhs_if_not_dup): Likewise.
18454         (replace_one_candidate): Likewise.
18456 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18458         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18459         Add new boolean.
18460         (aarch64_needs_frame_chain): New function.
18461         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18463 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18465         PR target/84882
18466         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18467         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18468         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18469         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18470         as true for strict-align.
18471         (aarch64_can_inline_p): Perform checks even when callee has no
18472         attributes to check for strict alignment.
18473         * doc/extend.texi (AArch64 Function Attributes): Document
18474         no-strict-align.
18475         * doc/invoke.texi: (AArch64 Options): Likewise.
18477 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18479         PR tree-optimization/85853
18480         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18481         the handling of the root of the node to...
18482         (vect_slp_analyze_node_operations_1): ...this new function,
18483         and run the whole thing with the child nodes' def types
18484         set according to their SLP node's def type.
18486 2018-05-23  Richard Biener  <rguenther@suse.de>
18488         PR middle-end/85874
18489         * tree-data-ref.c (create_runtime_alias_checks): Defer
18490         and ignore overflow warnings.
18492 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18494         PR tree-optimization/85822
18495         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18496         constants.
18498 2018-05-23  Richard Biener  <rguenther@suse.de>
18500         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18501         memset constants via native_interpret_expr.
18503 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18505         PR target/85345
18506         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18507         attribute.
18508         (cgraph_node::create_alias): Likewise.
18509         (cgraph_node::get_availability): Check ifunc_resolver instead
18510         of looking up ifunc attribute.
18511         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18512         * varasm.c (do_assemble_alias): Likewise.
18513         (assemble_alias): Likewise.
18514         (default_binds_local_p_3): Likewise.
18515         * cgraph.h (cgraph_node): Add ifunc_resolver.
18516         (cgraph_node::only_called_directly_or_aliased_p): Return false
18517         for IFUNC resolver.
18518         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18519         attribute.
18520         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18521         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18522         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18523         instead of looking up ifunc attribute.
18525 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18527         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18529 2018-05-22  Martin Sebor  <msebor@redhat.com>
18531         PR middle-end/85359
18532         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18533         only when expasion succeeds.
18534         (expand_builtin_strcmp): Same.
18535         (expand_builtin_strncmp): Same.
18537 2018-05-22  Martin Sebor  <msebor@redhat.com>
18539         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18541 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18542             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18544         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18545         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18546         code.
18547         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18548         Add check for legitimate_address.
18549         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18550         (aarch64_swap_ldrstr_operands): New.
18551         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18552         Define prototype.
18554 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18555             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18557         * config/aarch64/aarch64.md: New patterns to generate stp
18558         and ldp.
18559         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18560         single words and double words.
18561         (load_pair_sw, load_pair_dw): Likewise.
18562         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18563         Delete.
18564         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18565         Delete.
18566         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18567         for different mode ldpstp and add peephole for merged zero stores.
18568         Likewise for loads.
18569         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18570         Add size check.
18571         (aarch64_gen_store_pair): Rename calls to match new patterns.
18572         (aarch64_gen_load_pair): Rename calls to match new patterns.
18573         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18574         (load_pair<DREG:mode><DREG2:mode>): ... This.
18575         (store_pair<mode>): Rename to...
18576         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18577         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18578         New mode iterators.
18579         (V_INT_EQUIV): Handle SImode.
18580         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18581         New predicate.
18583 2018-05-22  Martin Sebor  <msebor@redhat.com>
18585         PR c/85623
18586         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18587         or ajust the presumed bound on an operation to avoid unnecessary
18588         warnings.
18590 2018-05-22  Martin Sebor  <msebor@redhat.com>
18592         PR tree-optimization/85826
18593         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18594         assuming that a DECL necesarily has a constant size.
18596 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18598         PR middle-end/85862
18599         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18601 2018-05-22  Richard Biener  <rguenther@suse.de>
18603         PR tree-optimization/85834
18604         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18605         non-constant and non-zero memset arguments.
18607 2018-05-22  Martin Liska  <mliska@suse.cz>
18609         PR ipa/85607
18610         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18612 2018-05-22  Richard Biener  <rguenther@suse.de>
18614         PR tree-optimization/85863
18615         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18616         comparisons when vectype is specified.
18617         (vectorizable_condition): Do not specify vectype for
18618         vect_is_simple_cond when SLP vectorizing.
18620 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18622         PR target/85657
18623         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18624         define __ibm128 as long double.
18625         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18626         as a distinct type when IEEE 128-bit support is enabled.
18627         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18628         128-bit types to use the correct functions.
18629         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18630         convert between 128-bit floating point types that have different
18631         modes but the same representation, instead of using gen_lowpart to
18632         makean alias.
18633         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18634         KFmode.
18635         (IFKF_reg): New attributes to give the register constraints for
18636         IFmode and KFmode.
18637         (extend<mode>tf2_internal): New insns to mark an explicit
18638         conversion between 128-bit floating point types that have a
18639         different mode but share the same representation.
18641 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18643         PR tree-optimization/85814
18644         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18645         a null return from get_strinfo when unsharing the next
18646         strinfo in the chain.
18648 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18650         PR gcc/84923
18651         * varasm.c (weak_finish): Clean up weak_decls.
18653 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18655         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18656         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18657         UNSPEC_UADALP values.
18658         * config/aarch64/iterators.md (ABAL): New int iterator.
18659         (ABDL2): Likewise.
18660         (ADALP): Likewise.
18661         (sur): Add mappings for the above.
18662         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18663         New define_insn.
18664         (aarch64_<sur>abal<mode>_4): Likewise.
18665         (aarch64_<sur>adalp<mode>_3): Likewise.
18666         (<sur>sadv16qi): New define_expand.
18668 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18670         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18671         (*movdf_internal): Ditto.
18672         (*rcpsf2_sse): Ditto.
18673         (*rsqrtsf2_sse): Ditto.
18674         (*sqrt<mode>2_sse): Ditto.
18676 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18678         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18679         eor3q<mode>4.
18680         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18681         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18682         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18683         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18684         vbcaxq_s64): New.
18685         * config/aarch64/arm_neon.h: Likewise.
18686         * config/aarch64/iterators.md (VQ_I): New.
18688 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18690         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18691         arc*-*-linux*.
18692         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18693         MULTILIB_DIRNAMES
18695 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18697         * config/nds32/constraints.md (S): New constraint.
18698         * config/nds32/nds32.md (call_internal): Use constraint S.
18699         (call_value_internal): Likewise.
18700         (sibcall_internal): Likewise.
18701         (sibcall_value_internal): Likewise.
18703 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18704             Chung-Ju Wu  <jasonwucj@gmail.com>
18706         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18707         into consideration.
18709 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18710             Chung-Ju Wu  <jasonwucj@gmail.com>
18712         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18713         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18714         (nds32_rtx_costs_impl): Simplify.
18715         (nds32_address_cost_impl): Simplify.
18716         (nds32_init_rtx_costs): New function.
18717         (nds32_rtx_costs_speed_prefer): Likewise.
18718         (nds32_rtx_costs_size_prefer): Likewise.
18719         (nds32_address_cost_speed_prefer): Likewise.
18720         (nds32_address_cost_speed_fwprop): Likewise.
18721         (nds32_address_cost_size_prefer): Likewise.
18722         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
18723         * config/nds32/nds32.c (nds32_option_override): Use
18724         nds32_init_rtx_costs function.
18726 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18728         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
18729         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
18730         (TARGET_PIPELINE_N8): Likewise.
18731         (TARGET_PIPELINE_N10): Likewise.
18732         (TARGET_PIPELINE_N13): Likewise.
18733         (TARGET_PIPELINE_GRAYWOLF): Likewise.
18735 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
18737         * config/nds32/nds32-fpu.md: Update copyright year.
18739 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18741         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18743 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18745         * config/nds32/nds32.c
18746         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
18747         * config/nds32/nds32.opt (minline-asm-r15): New option.
18749 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18751         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
18752         MASK_HW_ABS.
18753         * config/nds32/nds32.md (abssi2): New pattern.
18755 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
18757         * config/i386/i386.md (rex64namesuffix): New mode attribute.
18758         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
18759         Merge insn pattern from sse_cvtsi2ss<round_name> and
18760         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
18761         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
18762         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
18763         using SWI48 mode iterator.
18764         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
18765         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
18766         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
18767         pattern from sse_cvttss2si<round_saeonly_name>
18768         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
18769         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
18770         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
18771         using SWI48 mode iterator.
18772         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
18773         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
18774         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
18775         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
18776         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
18777         using SWI48 mode iterator.
18778         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
18779         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
18780         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
18781         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
18782         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
18783         SWI48 mode iterator.
18784         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
18785         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
18786         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
18787         pattern from sse_cvttsd2si<round_saeonly_name>
18788         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
18790 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18792         * config/nds32/nds32-md-auxiliary.c
18793         (nds32_valid_smw_lwm_base_p): Refine.
18794         (nds32_output_smw_single_word): Refine.
18795         (nds32_output_smw_double_word): New.
18796         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
18798 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18800         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
18801         (nds32_output_stack_pop): Refine.
18802         (nds32_expand_unaligned_load): Refine.
18803         (nds32_expand_unaligned_store): Refine.
18805 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18806             Chung-Ju Wu  <jasonwucj@gmail.com>
18808         * config/nds32/constants.md: Add TP_REGNUM constant.
18809         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
18810         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
18811         UNSPEC_ADD32.
18812         * config/nds32/nds32-doubleword.md: Consider flag_pic.
18813         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
18814         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
18815         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
18816         and PIC code generation.
18817         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
18818         code generation.
18819         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
18820         optimization.
18821         * config/nds32/nds32.md: Support TLS and PIC.
18822         * config/nds32/nds32.c: Support TLS and PIC.
18823         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
18824         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
18825         predicate.
18827 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18829         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
18830         mode with E_ prefix.
18832 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18833             Chung-Ju Wu  <jasonwucj@gmail.com>
18835         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
18836         * config/nds32/nds32-md-auxiliary.c
18837         (symbolic_reference_mentioned_p): New.
18838         (nds32_legitimize_ict_address): New.
18839         (nds32_expand_ict_move): New.
18840         (nds32_indirect_call_referenced_p): New.
18841         (nds32_symbol_binds_local_p): Delete.
18842         (nds32_long_call_p): Modify.
18843         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18844         * config/nds32/nds32-protos.h
18845         (symbolic_reference_mentioned_p): Declare.
18846         (nds32_legitimize_ict_address): Declare.
18847         (nds32_expand_ict_move): Declare.
18848         (nds32_indirect_call_referenced_p): Declare.
18849         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
18850         (nds32_relax_group): Use nds32_ict_const_p as condition.
18851         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
18852         (nds32_asm_file_start): Output ict_model directive in asm code.
18853         (nds32_legitimate_address_p): Consider indirect call.
18854         (nds32_print_operand): Consider indirect call.
18855         (nds32_print_operand_address): Consider indirect call.
18856         (nds32_insert_attributes): Handle "indirect_call" attribute.
18857         (TARGET_LEGITIMATE_ADDRESS_P): Define.
18858         (TARGET_LEGITIMATE_CONSTANT_P): Define.
18859         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
18860         (TARGET_DELEGITIMIZE_ADDRESS): Define.
18861         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
18862         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
18863         (TARGET_ICT_MODEL_SMALL): Define.
18864         (TARGET_ICT_MODEL_LARGE): Define.
18865         * config/nds32/nds32.md (movsi): Consider ict model.
18866         (call, call_value): Consider ict model.
18867         (sibcall, sibcall_value): Consider ict model.
18868         * config/nds32/nds32.opt (mict-model): New option.
18869         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
18870         model.
18872 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
18873             Monk Chiang  <sh.chiang04@gmail.com>
18874             Jim Wilson <jimw@sifive.com>
18876         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
18877         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
18878         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
18879         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
18880         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
18881         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
18882         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
18883         compute save_libcall_adjustment properly.
18884         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
18885         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
18886         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
18887         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
18888         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
18889         (ABI_SPEC): Handle mabi=ilp32e.
18890         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
18891         (RVE): Add RVE mask.
18892         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
18893         <-march>: Add rv32e as an example.
18895 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
18897         PR c++/82899
18898         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
18899         (intra_create_variable_infos): Handle C++ constructors.
18901 2018-05-18  Martin Liska  <mliska@suse.cz>
18903         * passes.def: Remove a redundant pass.
18905 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
18907         PR bootstrap/85838
18908         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
18910 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18912         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
18913         (ARMv4): Update.
18914         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
18915         (ARMv6m): Update.
18916         (armv2, armv2a, armv3, armv3m): Delete architectures.
18917         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
18918         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
18919         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
18920         Delete cpus.
18921         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
18922         (*mulsidi3adddi): Likewise.
18923         (mulsidi3): Likewise.
18924         (*mulsidi3_nov6): Likewise.
18925         (umulsidi3): Likewise.
18926         (umulsidi3_nov6): Likewise.
18927         (umaddsidi4): Likewise.
18928         (*umulsidi3adddi): Likewise.
18929         (smulsi3_highpart): Likewise.
18930         (*smulsi3_highpart_nov6): Likewise.
18931         (umulsi3_highpart): Likewise.
18932         (*umulsi3_highpart_nov6): Likewise.
18933         * config/arm/arm.h (arm_arch3m): Delete.
18934         * config/arm/arm.c (arm_arch3m): Delete.
18935         (arm_option_override_internal): Update armv3-related comment.
18936         (arm_configure_build_target): Delete use of isa_bit_mode32.
18937         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
18938         (arm_rtx_costs_internal): Delete check of arm_arch3m.
18939         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
18940         (mulsa3): Likewise.
18941         (mulusa3): Likewise.
18942         * config/arm/arm-protos.h (arm_arch3m): Delete.
18943         * config/arm/arm-tables.opt: Regenerate.
18944         * config/arm/arm-tune.md: Likewise.
18945         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
18946         deleted architectures.
18948 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18950         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
18951         (armv5t, armv5te): New features.
18952         (ARMv5, ARMv5e): Delete fgroups.
18953         (ARMv5t, ARMv5te): Adjust for above changes.
18954         (ARMv6m): Likewise.
18955         (armv5, armv5e): Delete arches.
18956         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
18957         arm_arch5.
18958         (*call_reg_arm): Likewise.
18959         (*call_value_reg_armv5): Likewise.
18960         (*call_value_reg_arm): Likewise.
18961         (*call_symbol): Likewise.
18962         (*call_value_symbol): Likewise.
18963         (*sibcall_insn): Likewise.
18964         (*sibcall_value_insn): Likewise.
18965         (clzsi2): Likewise.
18966         (prefetch): Likewise.
18967         (define_split and define_peephole2 dependent on arm_arch5):
18968         Likewise.
18969         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
18970         arm_arch5e.
18971         (TARGET_ARM_QBIT): Likewise.
18972         (TARGET_DSP_MULTIPLY): Likewise.
18973         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
18974         (arm_arch5, arm_arch5e): Delete.
18975         (arm_arch5t, arm_arch5te): Declare.
18976         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
18977         (arm_arch5t): Declare.
18978         (arm_option_reconfigure_globals): Update for the above.
18979         (arm_options_perform_arch_sanity_checks): Update comment, replace
18980         use of arm_arch5 with arm_arch5t.
18981         (use_return_insn): Likewise.
18982         (arm_emit_call_insn): Likewise.
18983         (output_return_instruction): Likewise.
18984         (arm_final_prescan_insn): Likewise.
18985         (arm_coproc_builtin_available): Likewise.
18986         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
18987         arm_arch5e with arm_arch5t and arm_arch5te.
18988         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
18989         (arm_arch5t, arm_arch5te): Declare.
18990         * config/arm/arm-tables.opt: Regenerate.
18991         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
18992         * config/arm/t-multilib: Likewise.
18993         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
18994         instead of arm_arch5.
18995         (*call_reg_thumb1): Likewise.
18996         (*call_value_reg_thumb1_v5): Likewise.
18997         (*call_value_reg_thumb1): Likewise.
18998         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
18999         unreachable path.
19000         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19002 2018-05-18  Martin Liska  <mliska@suse.cz>
19004         PR gcov-profile/84846
19005         * doc/gcov.texi: Document -t option of gcov tool.
19007 2018-05-18  Martin Liska  <mliska@suse.cz>
19009         PR gcov-profile/84846
19010         * gcov.c (print_usage): Add new -t option.
19011         (process_args): Handle the option.
19012         (generate_results): Use stdout as output when requested by
19013         the option.
19015 2018-05-18  Martin Liska  <mliska@suse.cz>
19017         PR gcov-profile/84846
19018         * coverage.c (coverage_init): Write PWD to .gcno file.
19019         * doc/gcov.texi: Document how working directory is printed.
19020         * gcov-dump.c (dump_gcov_file): Print PWD.
19021         * gcov.c (output_intermediate_file): Likewise.
19022         (read_graph_file): Read PWD string.
19023         (output_lines): Print PWD.
19025 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19027         PR middle-end/85817
19028         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19029         for retval and return false if all args to phi are zero.
19031 2018-05-18  Richard Biener  <rguenther@suse.de>
19033         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19034         method.
19035         (evrp_dom_walker::before_dom_children): Call it.
19037 2018-05-18  Richard Biener  <rguenther@suse.de>
19039         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19040         results when processing array refs with variable index.
19042 2018-05-18  Toon Moene  <toon@moene.org>
19044         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19045         directly after that of -floop-interchange. Indicate that both
19046         options are enabled by default when specifying -O3.
19048 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19050         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19051         iterator.  Delete separate integer-mode vec_set<mode> expander.
19052         (aarch64_simd_vec_setv2di): Delete.
19053         (vec_setv2di): Delete.
19054         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19055         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
19056         the "w, r" alternative.
19058 2018-05-18  Martin Liska  <mliska@suse.cz>
19060         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19061         * tree-pass.h (make_pass_lower_switch_O0): New function.
19062         * tree-switch-conversion.c (node_has_low_bound): Remove.
19063         (node_has_high_bound): Likewise.
19064         (node_is_bounded): Likewise.
19065         (class pass_lower_switch): Make it a template type and create
19066         two instances.
19067         (pass_lower_switch::execute): Add template argument.
19068         (make_pass_lower_switch): New function.
19069         (make_pass_lower_switch_O0): New function.
19070         (do_jump_if_equal): Remove.
19071         (emit_case_nodes): Simplify to just handle all 3 cases and leave
19072         all the hard work to tree optimization passes.
19074 2018-05-18  Martin Liska  <mliska@suse.cz>
19076         * dbgcnt.c (limit_low): Renamed from limit.
19077         (limit_high): New variable.
19078         (dbg_cnt_is_enabled): Check for upper limit.
19079         (dbg_cnt): Adjust dumping.
19080         (dbg_cnt_set_limit_by_index): Add new argument for high
19081         value.
19082         (dbg_cnt_set_limit_by_name): Likewise.
19083         (dbg_cnt_process_single_pair): Parse new format.
19084         (dbg_cnt_process_opt): Use strtok.
19085         (dbg_cnt_list_all_counters): Remove 'value' and add
19086         'limit_high'.
19087         * doc/invoke.texi: Document changes.
19089 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
19091         * doc/sourcebuild.texi (scalar_all_fma): Document.
19092         * tree.def (FMA_EXPR): Delete.
19093         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
19094         * internal-fn.c (ternary_direct): New macro.
19095         (expand_ternary_optab_fn): Likewise.
19096         (direct_ternary_optab_supported_p): Likewise.
19097         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
19098         * builtins.c (fold_builtin_fma): Delete.
19099         (fold_builtin_3): Don't call it.
19100         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
19101         * expr.c (expand_expr_real_2): Likewise.
19102         * fold-const.c (operand_equal_p): Likewise.
19103         (fold_ternary_loc): Likewise.
19104         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
19105         * gimple.c (DEFTREECODE): Likewise.
19106         * gimplify.c (gimplify_expr): Likewise.
19107         * optabs-tree.c (optab_for_tree_code): Likewise.
19108         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19109         * tree-eh.c (operation_could_trap_p): Likewise.
19110         (stmt_could_throw_1_p): Likewise.
19111         * tree-inline.c (estimate_operator_cost): Likewise.
19112         * tree-pretty-print.c (dump_generic_node): Likewise.
19113         (op_code_prio): Likewise.
19114         * tree-ssa-loop-im.c (stmt_cost): Likewise.
19115         * tree-ssa-operands.c (get_expr_operands): Likewise.
19116         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
19117         * fold-const-call.h (fold_fma): Delete.
19118         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
19119         CFN_FNMA and CFN_FNMS.
19120         (fold_fma): Delete.
19121         * genmatch.c (combined_fn): New enum.
19122         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
19123         (commutative_op): New function.
19124         (commutate): Use it.  Handle more than 2 operands.
19125         (dt_operand::gen_gimple_expr): Use commutative_op.
19126         (parser::parse_expr): Allow :c to be used with non-binary
19127         operators if the commutative operand is known.
19128         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
19129         CFN_FMS, CFN_FNMA and CFN_FNMS.
19130         (backprop::process_assign_use): Remove FMA_EXPR handling.
19131         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19132         (gen_hsa_fma): New function.
19133         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19134         IFN_FNMA and IFN_FNMS.
19135         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19136         * gimple-fold.h (follow_all_ssa_edges): Declare.
19137         * gimple-fold.c (follow_all_ssa_edges): New function.
19138         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19139         gimple_build interface and use follow_all_ssa_edges to fold the result.
19140         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19141         instead of checking for optabs directly.
19142         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19143         rather than FMA_EXPRs.
19144         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19145         call to IFN_FMA instead of an FMA_EXPR.
19147 2018-05-17  Jim Wilson  <jimw@sifive.com>
19149         * expr.c (do_tablejump): When converting index to Pmode, if we have a
19150         sign extended promoted subreg, and the range does not have the sign bit
19151         set, then do a sign extend.
19153         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19154         test, check for sign extended subreg and/or constant operands, and
19155         do a sign extend in that case.
19157 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
19159         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19160         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19161         Add untyped.
19162         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19163         Change logics_shift_reg to logics_shift_imm.
19164         (thunderx2t99_fp_loadpair_basic): Delete.
19165         (thunderx2t99_fp_storepair_basic): Delete.
19166         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19167         (thunderx2t99_asimd_polynomial): Delete.
19168         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19169         and neon_fp_mul_d_scalar_q.
19170         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19171         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19172         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19173         (thunderx2t99_asimd_lut): Add missing tbl types.
19174         (thunderx2t99_asimd_ext): Delete.
19175         (thunderx2t99_asimd_load1_1_mult): Delete.
19176         (thunderx2t99_asimd_load1_2_mult): Delete.
19177         (thunderx2t99_asimd_load1_ldp): New.
19178         (thunderx2t99_asimd_load1): New.
19179         (thunderx2t99_asimd_load2): Add missing *load2* types.
19180         (thunderx2t99_asimd_load3): New.
19181         (thunderx2t99_asimd_load4): New.
19182         (thunderx2t99_asimd_store1_1_mult): Delete.
19183         (thunderx2t99_asimd_store1_2_mult): Delete.
19184         (thunderx2t99_asimd_store2_mult): Delete.
19185         (thunderx2t99_asimd_store2_onelane): Delete.
19186         (thunderx2t99_asimd_store_stp): New.
19187         (thunderx2t99_asimd_store1): New.
19188         (thunderx2t99_asimd_store2): New.
19189         (thunderx2t99_asimd_store3): New.
19190         (thunderx2t99_asimd_store4): New.
19192 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
19194         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19195         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
19197 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
19198             Segher Boessenkool  <segher@kernel.crashing.org>
19200         PR target/85698
19201         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19202         operand.
19204 2018-05-17  Richard Biener  <rguenther@suse.de>
19206         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19207         for pruning loop and prune defs feeding only already visited PHIs.
19209 2018-05-17  Richard Biener  <rguenther@suse.de>
19211         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19213 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
19214             Richard Biener  <rguenther@suse.de>
19216         PR tree-optimization/85793
19217         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19218         for VMAT_ELEMENTWISE.
19220 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19222         * internal-fn.h (lookup_internal_fn): Declare
19223         * internal-fn.c (lookup_internal_fn): New function.
19224         * gimple.c (gimple_build_call_from_tree): Handle calls to
19225         internal functions.
19226         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19227         internal function names.
19228         * tree-pretty-print.c (dump_generic_node): Likewise.
19229         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19231 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19233         * gimple-fold.h (gimple_build): Make the function forms take
19234         combined_fn rather than built_in_function.
19235         (gimple_simplify): Likewise.
19236         * gimple-match-head.c (gimple_simplify): Likewise.
19237         * gimple-fold.c (gimple_build): Likewise.
19238         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19239         rather than gimple_build_call_internal.
19240         (get_initial_defs_for_reduction): Likewise.
19241         (vect_create_epilog_for_reduction): Likewise.
19242         (vectorizable_live_operation): Likewise.
19244 2018-05-17  Martin Liska  <mliska@suse.cz>
19246         * gimple-ssa-sprintf.c (format_directive): Do not use
19247         space in between 'G_' and '('.
19249 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19251         PR target/85323
19252         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19253         even if the mask is not all ones.
19255         PR target/85323
19256         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19257         vector.
19258         (ix86_gimple_fold_builtin): Likewise.
19260         PR target/85323
19261         * config/i386/i386.c: Include tree-vector-builder.h.
19262         (ix86_vector_shift_count): New function.
19263         (ix86_fold_builtin): Fold shift builtins by scalar count.
19264         (ix86_gimple_fold_builtin): Likewise.
19266         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19267         _mm512_setzero): New intrinsics.
19269 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19270             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19272         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19273         code generation for cases where splatting a value is not useful.
19274         * simplify-rtx.c (simplify_ternary_operation): Simplify
19275         vec_merge across a vec_duplicate and a paradoxical subreg forming
19276         a vector mode to a vec_concat.
19278 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19280         * config.gcc: Support "goldmont-plus".
19281         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19282         "goldmont-plus".
19283         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19284         PROCESSOR_GOLDMONT_PLUS.
19285         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19286         (processor_target_table): Add "goldmont-plus".
19287         (PTA_GOLDMONT_PLUS): Define.
19288         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19289         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19290         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19291         (fold_builtin_cpu): Add "goldmont-plus".
19292         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19293         (ix86_option_override_internal): Add "goldmont-plus".
19294         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19295         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19296         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19297         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19299 2018-05-17  Richard Biener  <rguenther@suse.de>
19301         PR tree-optimization/85757
19302         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19303         remove defs that only feed that PHI from further processing.
19305 2018-05-16  Jim Wilson  <jimw@sifive.com>
19307         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19308         asterisk to name.
19309         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19311 2018-05-16  Mark Wielaard  <mark@klomp.org>
19313         * dwarf2out.c (count_index_strings): New function.
19314         (output_indirect_strings): Call count_index_strings and generate
19315         header for dwarf_version >= 5.
19317 2018-05-16  Mark Wielaard  <mark@klomp.org>
19319         * dwarf2out.c (dwarf_FORM): New function.
19320         (set_indirect_string): Use dwarf_FORM.
19321         (reset_indirect_string): Likewise.
19322         (size_of_die): Likewise.
19323         (value_format): Likewise.
19324         (output_die): Likewise.
19325         (add_skeleton_AT_string): Likewise.
19326         (output_macinfo_op): Likewise.
19327         (index_string): Likewise.
19328         (output_index_string_offset): Likewise.
19329         (output_index_string): Likewise.
19330         (count_index_strings): Likewise.
19332 2018-05-16  Carl Love  <cel@us.ibm.com>
19334         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19335         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19337 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19339         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19340         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19342 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19344         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19345         (fnma<mode>4): Likewise.
19346         (fms<mode>4): Likewise.
19347         (fnms<mode>4): Likewise.
19348         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19349         (aarch64_fnma<mode>4): Likewise.
19350         (aarch64_fms<mode>4): Likewise.
19351         (aarch64_fnms<mode>4): Likewise.
19352         (aarch64_fnmadd<mode>4): Likewise.
19354 2018-05-16  Jason Merrill  <jason@redhat.com>
19356         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19358 2018-05-16  Richard Biener  <rguenther@suse.de>
19360         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19361         (dump_stmt_cost): Declare.
19362         (add_stmt_cost): Dump cost we add.
19363         (add_stmt_costs): New function.
19364         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19365         No longer exported.
19366         (vect_analyze_stmt): Adjust prototype.
19367         (vectorizable_condition): Likewise.
19368         (vectorizable_live_operation): Likewise.
19369         (vectorizable_reduction): Likewise.
19370         (vectorizable_induction): Likewise.
19371         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19372         cost vector to pass to vectorizable_ and record afterwards.
19373         (vect_model_reduction_cost): Take cost vector argument and adjust.
19374         (vect_model_induction_cost): Likewise.
19375         (vectorizable_reduction): Likewise.
19376         (vectorizable_induction): Likewise.
19377         (vectorizable_live_operation): Likewise.
19378         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19379         SLP_TREE_NUMBER_OF_VEC_STMTS.
19380         (vect_analyze_slp_cost_1): Remove.
19381         (vect_analyze_slp_cost): Likewise.
19382         (vect_slp_analyze_node_operations): Take visited args and
19383         a target cost vector.  Avoid processing already visited stmt sets.
19384         (vect_slp_analyze_operations): Use a local cost vector to gather
19385         costs and register those of non-discarded instances.
19386         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19387         (vect_schedule_slp_instance): Remove copying of
19388         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19389         zero.
19390         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19391         adding cost.  Record cost entry location.
19392         (vect_prologue_cost_for_slp_op): Function to compute cost of
19393         a constant or invariant generated for SLP vect in the prologue,
19394         split out from vect_analyze_slp_cost_1.
19395         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19396         (vect_model_promotion_demotion_cost): Likewise.
19397         (vect_model_store_cost): Likewise, make static.
19398         (vect_model_load_cost): Likewise.
19399         (vectorizable_bswap): Add cost vector arg and adjust.
19400         (vectorizable_call): Likewise.
19401         (vectorizable_simd_clone_call): Likewise.
19402         (vectorizable_conversion): Likewise.
19403         (vectorizable_assignment): Likewise.
19404         (vectorizable_shift): Likewise.
19405         (vectorizable_operation): Likewise.
19406         (vectorizable_store): Likewise.
19407         (vectorizable_load): Likewise.
19408         (vectorizable_condition): Likewise.
19409         (vectorizable_comparison): Likewise.
19410         (can_vectorize_live_stmts): Likewise.
19411         (vect_analyze_stmt): Likewise.
19412         (vect_transform_stmt): Adjust calls to vectorizable_*.
19413         * tree-vectorizer.c: Include gimple-pretty-print.h.
19414         (dump_stmt_cost): New function.
19416 2018-05-16  Richard Biener  <rguenther@suse.de>
19418         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19419         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19420         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19421         (check_name): New callback.
19422         (dse_classify_store): Track cycles via a visited bitmap of PHI
19423         defs and simplify handling of in-loop and across loop dead stores
19424         and properly fail for loop-variant refs.  Handle byte-tracking with
19425         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19426         limiting the walk.
19428 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19430         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19431         (vect_get_mask_type_for_stmt): Likewise.
19432         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19433         split out from...
19434         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19435         to determine the statement's vector type and the vector type that
19436         should be used for calculating nunits.  Deal with cases in which
19437         the type has to be deferred.
19438         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19439         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19440         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19441         (vect_determine_vf_for_stmt): New functions, split out from...
19442         (vect_determine_vectorization_factor): ...here.
19443         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19444         (vect_get_mask_type_for_stmt): New functions, split out from
19445         vect_determine_vectorization_factor.
19447 2018-05-16  Richard Biener  <rguenther@suse.de>
19449         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19450         verify the [VEC_]COND_EXPR embedded comparison.
19452 2018-05-15  Martin Sebor  <msebor@redhat.com>
19454         PR tree-optimization/85753
19455         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19456         RECORD_TYPE in addition to ARRAY_TYPE.
19458 2018-05-15  Martin Sebor  <msebor@redhat.com>
19460         PR middle-end/85643
19461         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19463 2018-05-15  Richard Biener  <rguenther@suse.de>
19465         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19466         add by_clobber_p one.  Change algorithm to collect all defs
19467         representing uses we need to walk and try reducing them to
19468         a single one before failing.
19469         (dse_dom_walker::dse_optimize_stmt): Adjust.
19471 2018-05-13  Mark Wielaard  <mark@klomp.org>
19473         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19474         (size_of_loc_descr): Likewise.
19475         (output_loc_operands): Likewise.
19476         (output_loc_operands_raw): Likewise.
19477         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19478         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19479         (hash_loc_operands): Likewise.
19480         (compare_loc_operands): Likewise.
19482 2018-05-14  Mark Wielaard  <mark@klomp.org>
19484         * dwarf2out.c (count_index_addrs): New function.
19485         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19487 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19489         PR tree-optimization/83648
19490         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19491         return value as malloc candidate.
19493 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19495         PR ipa/85734
19496         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19497         param as true in call to suggest_attribute.
19499 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19501         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19502         -mreadonly-in-sdata.
19504 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19506         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19507         New pattern.
19508         (aarch64_crypto_aesd_fused): Likewise.
19510 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19512         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19513         (movsi_aarch64): Likewise.
19514         (load_pairsi): Likewise.
19515         (load_pairdi): Likewise.
19516         (store_pairsi): Likewise.
19517         (store_pairdi): Likewise.
19518         (load_pairsf): Likewise.
19519         (load_pairdf): Likewise.
19520         (store_pairsf): Likewise.
19521         (store_pairdf): Likewise.
19522         (zero_extend): Likewise.
19523         (trunc): Swap alternatives.
19524         (fcvt_target): Add '?' to prefer w over r.
19526 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19528         PR target/85756
19529         * config/i386/i386.md: Disallow non-commutative arithmetics in
19530         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19531         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19532         in the peephole2 before it.
19534 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19536         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19537         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19538         (ix86_handle_option): Handle -mcldemote.
19539         * config.gcc: New header.
19540         * config/i386/cldemoteintrin.h: New file.
19541         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19542         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19543         -mcldemote.
19544         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19545         OPTION_MASK_ISA_CLDEMOTE.
19546         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19547         (ix86_valid_target_attribute_inner_p): Ditto.
19548         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19549         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19550         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19551         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19552         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19553         (cldemote): New.
19554         * config/i386/i386.opt: Add -mcldemote.
19555         * config/i386/x86intrin.h: New header.
19556         * doc/invoke.texi: Add -mcldemote.
19558 2018-05-14  Richard Biener  <rguenther@suse.de>
19560         * doc/match-and-simplify.texi: Adjust :s documentation.
19562 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19564         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19565         intended memcpy size.
19566         (REORDER_45): Likewise.
19568 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19570         * sort.cc: New file.
19571         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19572         * vec.c (qsort_chk): Use gcc_qsort.
19573         * Makefile.in (OBJS-libcommon): Add sort.o.
19574         (build/sort.o): New target.  Use it...
19575         (BUILD_RTL): ... here, and...
19576         (build/gencfn-macros): ... here, and...
19577         (build/genmatch): ... here.
19579 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19580             Chung-Ju Wu  <jasonwucj@gmail.com>
19582         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19583         * config/nds32/nds32-graywolf.md: New file.
19584         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19585         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19586         pipeline.
19587         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19588         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19589         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19590         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19591         * config/nds32/pipelines.md: Include n15 settings.
19593 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19594             Chung-Ju Wu  <jasonwucj@gmail.com>
19596         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19597         * config/nds32/nds32-n13.md: New file.
19598         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19599         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19600         pipeline.
19601         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19602         * config/nds32/nds32.md (pipeline_model): Add n13.
19603         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19604         * config/nds32/pipelines.md: Include n13 settings.
19606 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19607             Chung-Ju Wu  <jasonwucj@gmail.com>
19609         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19610         * config/nds32/nds32-n10.md: New file.
19611         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19612         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19613         pipeline.
19614         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19615         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19616         * config/nds32/nds32.md (pipeline_model): Add n10.
19617         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19618         * config/nds32/pipelines.md: Include n10 settings.
19620 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19621             Kito Cheng  <kito.cheng@gmail.com>
19622             Chung-Ju Wu  <jasonwucj@gmail.com>
19624         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19625         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19626         Add enum values for DSP extension instructions.
19627         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19628         New constraints.
19629         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19630         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19631         New code iterators.
19632         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19633         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19634         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19635         * config/nds32/nds32-intrinsic.md: Likewise.
19636         * config/nds32/nds32_intrinsic.h: Likewise.
19637         * config/nds32/nds32-md-auxiliary.c: Likewise.
19638         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19639         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19640         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19641         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19642         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19643         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19644         TYPE_DMAC in switch statement.
19645         * config/nds32/nds32.c: New checking and implementation for DSP
19646         extension instructions.
19647         * config/nds32/nds32.h: Likewise.
19648         * config/nds32/nds32.md: Likewise.
19649         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19650         * config/nds32/predicates.md: Implement new predicates for DSP
19651         extension.
19653 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19655         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19656         Reformat alternatives and attributes so it is easier to identify
19657         which constraints/attributes go with which instruction.
19658         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19659         (mov<mode>_softfloat32, FMOVE64): Likewise.
19660         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19661         (mov<mode>_softfloat64, FMOVE64): Likewise.
19663 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19665         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19666         subsection.
19667         (Basic PowerPC Built-in Functions): The new name of the
19668         subsection previously known as "PowerPC Built-in Functions".
19669         (Basic PowerPC Built-in Functions Available on all Configurations):
19670         New subsubsection.
19671         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19672         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19673         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19674         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19676 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19678         PR ipa/85655
19679         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19680         single const.
19682 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19684         PR target/85733
19685         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19687 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19689         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19690         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19691         (ix86_handle_option): Handle -mwaitpkg.
19692         * config.gcc: New header.
19693         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19694         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19695         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19696         function type.
19697         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19698         OPTION_MASK_ISA_WAITPKG.
19699         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19700         (ix86_option_override_internal): Add PTA_WAITPKG.
19701         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19702         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19703         IX86_BUILTIN_TPAUSE.
19704         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19705         __builtin_ia32_umwait and __builtin_ia32_tpause.
19706         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19707         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19708         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19709         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19710         UNSPECV_TPAUSE): New.
19711         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19712         * config/i386/i386.opt: Add -mwaitpkg.
19713         * config/i386/waitpkgintrin.h: New file.
19714         * config/i386/x86intrin.h: New header.
19715         * doc/invoke.texi: Add -mwaitpkg.
19717 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19719         PR target/85606
19720         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
19721         equivalent.
19722         (cortex-m0): Use armv6s-m isa.
19723         (cortex-m0plus): Likewise.
19724         (cortex-m1): Likewise.
19725         (cortex-m0.small-multiply): Likewise.
19726         (cortex-m0plus.small-multiply): Likewise.
19727         (cortex-m1.small-multiply): Likewise.
19729 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
19730             Jakub Jelinek  <jakub@redhat.com>
19732         PR tree-optimization/85692
19733         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
19734         source permute as well.
19736 2018-05-11  Martin Liska  <mliska@suse.cz>
19738         PR sanitizer/85556
19739         * doc/extend.texi: Document LLVM style format for no_sanitize
19740         attribute.
19742 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
19744         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
19745         mode_supports_vsx_dform_quad to mode_supports_dq_form.
19746         (mode_supports_vsx_dform_quad): Likewise.
19747         (mode_supports_vmx_dform): Move these functions to be next to the
19748         other mode_supports functions.
19749         (mode_supports_dq_form): Likewise.
19750         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
19751         mode_supports_dq_form.
19752         (reg_offset_addressing_ok_p): Likewise.
19753         (offsettable_ok_by_alignment): Likewise.
19754         (rs6000_legitimate_offset_address_p): Likewise.
19755         (legitimate_lo_sum_address_p): Likewise.
19756         (rs6000_legitimize_address): Likewise.
19757         (rs6000_legitimize_reload_address): Likewise.
19758         (rs6000_secondary_reload_inner): Likewise.
19759         (rs6000_preferred_reload_class): Likewise.
19760         (rs6000_output_move_128bit): Likewise.
19762 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19764         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
19765         Generate SImode target register for null target.
19766         <case IX86_BUILTIN_XGETBV>: Ditto.
19767         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
19768         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
19770 2018-05-10  Carl Love  <cel@us.ibm.com>
19772         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19773         dcbtt and dcbtstt if operands[2] is 0.
19775 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19777         PR target/85693
19778         * config/i386/sse.md (usadv64qi): New expander.
19780 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
19782         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
19783         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
19784         -maltivec=be support.
19785         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
19786         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
19787         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
19788         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
19789         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
19790         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
19791         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
19792         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
19793         altivec_vsumsws): Adjust.
19794         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
19795         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
19796         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
19797         support.
19798         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
19799         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
19800         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
19801         (altivec_lve<VI_char>x): Delete expand.
19802         (*altivec_lve<VI_char>x_internal): Rename to...
19803         (altivec_lve<VI_char>x): ... this.
19804         (altivec_lvxl_<mode>): Delete expand.
19805         (*altivec_lvxl_<mode>_internal): Rename to ...
19806         (altivec_lvxl_<mode>): ... this.
19807         (altivec_stvxl_<mode>): Delete expand.
19808         (*altivec_stvxl_<mode>_internal): Rename to ...
19809         (altivec_stvxl_<mode>): ... this.
19810         (altivec_stve<VI_char>x): Delete expand.
19811         (*altivec_stve<VI_char>x_internal): Rename to ...
19812         (altivec_stve<VI_char>x): ... this.
19813         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
19814         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
19815         reduc_plus_scal_<mode>): Adjust.
19816         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
19817         comment.
19818         (rs6000_cpu_cpp_builtins): Adjust.
19819         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
19820         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
19821         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
19822         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
19823         -maltivec=be support.
19824         (rs6000_split_vec_extract_var): Adjust.
19825         (rs6000_split_v4si_init): Adjust.
19826         (swap_selector_for_mode): Delete.
19827         (altivec_expand_lvx_be, altivec_expand_stvx_be,
19828         altivec_expand_stvex_be): Delete.
19829         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
19830         -maltivec=be support.
19831         (rs6000_gimple_fold_builtin): Ditto.
19832         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
19833         Adjust.
19834         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
19835         (TARGET_DIRECT_MOVE_64BIT): Adjust.
19836         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
19837         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
19838         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
19839         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
19840         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
19841         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
19842         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
19843         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
19844         anonymous split): Adjust.
19845         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
19846         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
19848 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
19850         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
19851         when --with-gxx-include-dir is also specified.
19852         * configure: Regenerate.
19854 2018-05-09  Jim Wilson  <jimw@sifive.com>
19856         PR target/84797
19857         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
19858         * config/riscv/t-withmultilib: New.
19859         * config/riscv/withmultilib.h: New.
19860         * doc/install.texi: Document RISC-V --with-multilib-list support.
19862 2018-05-09  Richard Biener  <rguenther@suse.de>
19864         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19865         vector.
19866         (vect_bb_vectorization_profitable_p): Adjust.  Compute
19867         actual scalar cost using the cost vector and the add_stmt_cost
19868         machinery.
19870 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19872         PR rtl-optimization/85645
19873         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
19874         in the REG_CFA_REGISTER note for LR, don't leave it empty.
19876 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19878         PR rtl-optimization/85645
19879         * shrink-wrap.c (spread_components): Return a boolean saying if
19880         anything was changed.
19881         (try_shrink_wrapping_separate): Iterate spread_components until
19882         nothing changes anymore.
19884 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19886         PR rtl-optimization/85645
19887         * regrename.c (build_def_use): Also kill the chains that include the
19888         destination of a REG_CFA_REGISTER note.
19890 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19892         PR rtl-optimization/85645
19893         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
19894         insn that has a REG_CFA_REGISTER note.
19896 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
19898         * cfgexpand.c (expand_clobber): New function.
19899         (expand_gimple_stmt_1): Use it.
19900         * tree-vect-stmts.c (vect_clobber_variable): New function,
19901         split out from...
19902         (vectorizable_simd_clone_call): ...here.
19903         (vectorizable_store): Emit a clobber either side of an
19904         IFN_STORE_LANES sequence.
19905         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
19907 2018-05-09  Tom de Vries  <tom@codesourcery.com>
19909         PR target/85626
19910         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
19911         (define_insn "trap_if_false"): Add exit after trap.
19913 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
19915         PR rtl-optimization/85638
19916         * bb-reorder.c: Include common/common-target.h.
19917         (create_forwarder_block): New function extracted from...
19918         (fix_up_crossing_landing_pad): ...here.  Rename into...
19919         (dw2_fix_up_crossing_landing_pad): ...this.
19920         (sjlj_fix_up_crossing_landing_pad): New function.
19921         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
19922         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
19923         from both partitions and exit the loop after one iteration.
19925 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19927         Revert:
19928         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19929         subsection.
19930         (Basic PowerPC Built-in Functions): The new name of the
19931         subsection previously known as "PowerPC Built-in Functions".
19932         (Basic PowerPC Built-in Functions Available on all Configurations):
19933         New subsubsection.
19934         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19935         subsubsection.
19936         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19937         subsubsection.
19938         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19939         subsubsection.
19940         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19941         subsubsection.
19943 2018-05-08  Jim Wilson  <jimw@sifive.com>
19945         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
19946         (LD_EMUL_SUFFIX): New.
19947         (LINK_SPEC): Use it.
19949 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19951         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19952         subsection.
19953         (Basic PowerPC Built-in Functions): The new name of the
19954         subsection previously known as "PowerPC Built-in Functions".
19955         (Basic PowerPC Built-in Functions Available on all Configurations):
19956         New subsubsection.
19957         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19958         subsubsection.
19959         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19960         subsubsection.
19961         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19962         subsubsection.
19963         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19964         subsubsection.
19966 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19968         PR target/85683
19969         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
19970         after cmpelim optimization.
19972 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
19974         * config.gcc: Support "goldmont".
19975         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
19976         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19977         PROCESSOR_GOLDMONT.
19978         * config/i386/i386.c (m_GOLDMONT): Define.
19979         (processor_target_table): Add "goldmont".
19980         (PTA_GOLDMONT): Define.
19981         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
19982         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
19983         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
19984         (fold_builtin_cpu): Add "goldmont".
19985         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
19986         (ix86_option_override_internal): Add "goldmont".
19987         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
19988         (processor_type): Add PROCESSOR_GOLDMONT.
19989         * config/i386/i386.md: Add CPU "glm".
19990         * config/i386/glm.md: New file.
19991         * config/i386/x86-tune.def: Add m_GOLDMONT.
19992         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
19994 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19996         PR target/85572
19997         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
19998         E_V4DImode.
19999         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20000         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
20001         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
20003         PR target/85317
20004         * config/i386/i386.c (ix86_fold_builtin): Handle
20005         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20007         PR target/85480
20008         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20009         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20011 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
20013         PR target/85658
20014         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20015         (check_arch): Likewise.
20016         (check_fpu): Return the result rather than printing it.
20017         (end arch): Fix operator precedence.
20018         (end cpu): Likewise.
20019         (END): Print the result from check_fpu.
20021 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20022             Alan Hayward  <alan.hayward@arm.com>
20023             David Sherwood  <david.sherwood@arm.com>
20025         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20026         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20027         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20028         (*fcmuo<mode>_and): New patterns.
20030 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20032         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20033         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20034         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20035         (cmp_op, sve_imm_con): New code attributes.
20036         (SVE_COND_INT_CMP, imm_con): Delete.
20037         (cmp_op): Remove above unspecs from int attribute.
20038         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20039         to...
20040         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
20041         comparison-specific unspecs.
20042         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20043         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20044         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20045         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20046         (*vec_fcm<cmp_op><mode>): Rename to...
20047         (*fcm<cmp_op><mode>): ...this and adjust likewise.
20048         (*vec_fcmuo<mode>): Rename to...
20049         (*fcmuo<mode>): ...this and adjust likewise.
20050         (*pred_fcm<cmp_op><mode>): New pattern.
20051         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20052         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20053         functions.
20054         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20055         and UNORDERED.
20056         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20057         (aarch64_emit_sve_predicated_cond): New function.
20058         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20059         (aarch64_emit_unspec_cond_or): Replace with...
20060         (aarch64_emit_sve_or_conds): ...this new function.  Use
20061         aarch64_emit_sve_ptrue_op for the individual comparisons and
20062         aarch64_emit_binop to OR them together.
20063         (aarch64_emit_inverted_unspec_cond): Replace with...
20064         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
20065         aarch64_emit_sve_ptrue_op for the comparison and
20066         aarch64_emit_unop to invert the result.
20067         (aarch64_expand_sve_vec_cmp_float): Update after the above
20068         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
20070 2018-05-07  Nathan Sidwell  <nathan@acm.org>
20072         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
20073         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
20074         (Backwards Compatibility): Likewise.
20076 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20078         PR bootstrap/85681
20079         Revert:
20080         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20082         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20083         <prefetch_dynamic_strides>: New const bool field.
20084         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20085         prefetch_dynamic_strides.
20086         (exynosm1_prefetch_tune): Likewise.
20087         (thunderxt88_prefetch_tune): Likewise.
20088         (thunderx_prefetch_tune): Likewise.
20089         (thunderx2t99_prefetch_tune): Likewise.
20090         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20091         to false.
20092         (aarch64_override_options_internal): Update to set
20093         PARAM_PREFETCH_DYNAMIC_STRIDES.
20094         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20095         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20096         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20097         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20098         prefetch-dynamic-strides setting.
20100         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20102         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20103         <minimum_stride>: New const int field.
20104         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20105         minimum_stride field.
20106         (exynosm1_prefetch_tune): Likewise.
20107         (thunderxt88_prefetch_tune): Likewise.
20108         (thunderx_prefetch_tune): Likewise.
20109         (thunderx2t99_prefetch_tune): Likewise.
20110         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20111         (aarch64_override_options_internal): Update to set
20112         PARAM_PREFETCH_MINIMUM_STRIDE.
20113         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20114         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20115         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20116         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20117         stride is constant and is below the minimum stride threshold.
20119 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20121         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
20122         to 512.
20124 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20126         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20127         <prefetch_dynamic_strides>: New const bool field.
20128         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20129         prefetch_dynamic_strides.
20130         (exynosm1_prefetch_tune): Likewise.
20131         (thunderxt88_prefetch_tune): Likewise.
20132         (thunderx_prefetch_tune): Likewise.
20133         (thunderx2t99_prefetch_tune): Likewise.
20134         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20135         to false.
20136         (aarch64_override_options_internal): Update to set
20137         PARAM_PREFETCH_DYNAMIC_STRIDES.
20138         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20139         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20140         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20141         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20142         prefetch-dynamic-strides setting.
20144 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20146         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20147         <minimum_stride>: New const int field.
20148         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20149         minimum_stride field.
20150         (exynosm1_prefetch_tune): Likewise.
20151         (thunderxt88_prefetch_tune): Likewise.
20152         (thunderx_prefetch_tune): Likewise.
20153         (thunderx2t99_prefetch_tune): Likewise.
20154         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20155         (aarch64_override_options_internal): Update to set
20156         PARAM_PREFETCH_MINIMUM_STRIDE.
20157         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20158         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20159         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20160         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20161         stride is constant and is below the minimum stride threshold.
20163 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
20165         PR c++/85659
20166         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20167         the type is addressable.  Don't force op into register if it has
20168         BLKmode.
20170 2018-05-05  Roland McGrath  <mcgrathr@google.com>
20172         PR other/77609
20173         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20174         any section for which we don't know a specific type it should have,
20175         regardless of name.  Previously this was done only for the exact
20176         names ".init_array", ".fini_array", and ".preinit_array".
20177         (default_elf_asm_named_section): Add comment about
20178         relationship with default_section_type_flags and SECTION_NOTYPE.
20179         (get_section): Don't consider it a type conflict if one side has
20180         SECTION_NOTYPE and the other doesn't, as long as neither has the
20181         SECTION_BSS et al used in the default_section_type_flags logic.
20183 2018-05-05  Tom de Vries  <tom@codesourcery.com>
20185         PR target/85653
20186         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20187         (workaround_barsyncs): New function.
20188         (nvptx_reorg): Use workaround_barsyncs.
20189         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20190         (define_expand "nvptx_membar_cta"): New define_expand.
20191         (define_insn "*nvptx_membar_cta"): New insn.
20193 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
20195         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20196         To improve optimization opportunities.
20197         * builtin-types.def: The new needed builtin types for the above.
20199 2018-05-04  Richard Biener  <rguenther@suse.de>
20201         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20202         * gimple-ssa-store-merging.c
20203         (imm_store_chain_info::output_merged_store): Remove redundant create,
20204         release split_store vector contents on failure.
20205         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20206         scalar stmt vector on cache hit.
20208 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
20210         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20211         Xilinx FP support.
20212         * config.gcc (powerpc-xilinx-eabi*): Remove.
20213         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20214         support.
20215         (fusion_addis_mem_combo_load): Ditto.
20216         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20217         FP support.
20218         (rs6000_cpu_cpp_builtins): Ditto.
20219         * config/rs6000/rs6000-linux.c
20220         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20221         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20222         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20223         support.
20224         (rs6000_setup_reg_addr_masks): Ditto.
20225         (rs6000_init_hard_regno_mode_ok): Ditto.
20226         (rs6000_option_override_internal): Ditto.
20227         (legitimate_lo_sum_address_p): Ditto.
20228         (rs6000_legitimize_address): Ditto.
20229         (rs6000_legitimize_reload_address): Ditto.
20230         (rs6000_legitimate_address_p): Ditto.
20231         (abi_v4_pass_in_fpr): Ditto.
20232         (setup_incoming_varargs): Ditto.
20233         (rs6000_gimplify_va_arg): Ditto.
20234         (rs6000_split_multireg_move): Ditto.
20235         (rs6000_savres_strategy): Ditto.
20236         (rs6000_emit_prologue_components): Ditto.
20237         (rs6000_emit_epilogue_components): Ditto.
20238         (rs6000_emit_prologue): Ditto.
20239         (rs6000_emit_epilogue): Ditto.
20240         (rs6000_elf_file_end): Ditto.
20241         (rs6000_function_value): Ditto.
20242         (rs6000_libcall_value): Ditto.
20243         * config/rs6000/rs6000.h: Ditto.
20244         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20245         (TARGET_MINMAX): ... this.  New.
20246         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20247         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20248         (*movsi_internal1_single): Delete.
20249         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20250         mfpu=, mxilinx-fpu): Delete.
20251         * config/rs6000/singlefp.h: Delete.
20252         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20253         * config/rs6000/t-rs6000: Ditto.
20254         * config/rs6000/t-xilinx: Delete.
20255         * config/rs6000/titan.md: Adjust for fp_type removal.
20256         * config/rs6000/vsx.md: Remove Xilinx FP support.
20257         (VStype_simple): Delete.
20258         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20259         * config/rs6000/xfpu.h: Delete.
20260         * config/rs6000/xfpu.md: Delete.
20261         * config/rs6000/xilinx.h: Delete.
20262         * config/rs6000/xilinx.opt: Delete.
20263         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20264         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20266 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20268         PR libgomp/85639
20269         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20270         if ignore == 0.
20272 2018-05-04  Richard Biener  <rguenther@suse.de>
20274         PR middle-end/85627
20275         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20276         (expand_complex_div_wide): Likewise.
20277         (expand_complex_operations_1): Likewise.
20278         (expand_complex_libcall): Preserve EH info of the original stmt.
20279         (tree_lower_complex): Handle removed blocks.
20280         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20281         on complex multiplication and division libcall builtins.
20283 2018-05-04  Richard Biener  <rguenther@suse.de>
20285         PR middle-end/85574
20286         * fold-const.c (negate_expr_p): Restrict negation of operand
20287         zero of a division to when we know that can happen without
20288         overflow.
20289         (fold_negate_expr_1): Likewise.
20291 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20293         PR libstdc++/85466
20294         * real.h (real_nextafter): Declare.
20295         * real.c (real_nextafter): New function.
20296         * fold-const-call.c (fold_const_nextafter): New function.
20297         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20298         CASE_CFN_NEXTTOWARD.
20299         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20300         even when arg1_mode is different from arg0_mode.
20302 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20304         * doc/extend.texi (Deprecated Features): Remove
20305         -ffriend-injection.
20306         (Backwards Compatibility): Likewise.
20307         * doc/invoke.texi (C++ Language Options): Likewise.
20308         (C++ Dialect Options): Likewise.
20310 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20312         PR target/85530
20313         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20314         _mm512_mask_mullox_epi64): New intrinsics.
20316 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20318         PR testsuite/85106
20319         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20320         dump files): Add offload-tree.
20322 2018-05-03  Richard Biener  <rguenther@suse.de>
20324         PR tree-optimization/85615
20325         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20326         to loops not nested in BBs loop father to avoid creating multi-entry
20327         loops.
20329 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20331         PR tree-optimization/70291
20332         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20333         arguments.  Change return type to tree.  Emit libcall as a new
20334         statement rather than replacing existing one when inplace_p is true.
20335         (expand_complex_multiplication_components): New function.
20336         (expand_complex_multiplication): Expand floating-point complex
20337         multiplication using the above.
20338         (expand_complex_division): Rename inner_type parameter to type.
20339         Update expand_complex_libcall call-site.
20340         (expand_complex_operations_1): Update expand_complex_multiplication
20341         and expand_complex_division call-sites.
20343 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20345         PR target/85582
20346         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20347         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20348         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20349         the highest significant bit of the shift count mask is clear.  In
20350         check whether and[sq]i3 is needed verify that all significant bits
20351         of the shift count other than the highest are set.
20353 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20355         PR libgomp/82428
20356         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20357         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20358         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20359         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20360         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20361         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20362         __builtin_goacc_parlevel_size.
20364 2018-05-02  Richard Biener  <rguenther@suse.de>
20366         PR tree-optimization/85597
20367         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20368         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20369         directly.
20371 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20373         PR testsuite/85106
20374         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20375         dump files): Add ltrans-tree.
20377 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20379         PR testsuite/85106
20380         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20381         dump files): Add wpa-ipa.
20383 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20385         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20386         powerpc*-*-linux*paired* target.
20387         * config/rs6000/750cl.h: Delete.
20388         * config/rs6000/paired.h: Delete.
20389         * config/rs6000/paired.md: Delete.
20390         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20391         float support.
20392         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20393         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20394         comment.  Remove paired float support.
20395         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20396         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20397         VECTOR_PAIRED.
20398         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20399         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20400         declarations.
20401         * config/rs6000/rs6000.c: Remove paired float support.
20402         (paired_expand_vector_init, paired_expand_vector_move,
20403         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20404         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20405         paired_expand_builtin, paired_expand_predicate_builtin,
20406         paired_init_builtins): Delete.
20407         * config/rs6000/rs6000.h: Remove paired float support.
20408         * config/rs6000/rs6000.md: Remove paired float support.
20409         (move_from_CR_ov_bit): Delete.
20410         * config/rs6000/rs6000.opt (mpaired): Delete.
20411         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20412         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20414 2018-05-02  Richard Biener  <rguenther@suse.de>
20416         PR middle-end/85567
20417         * gimplify.c (gimplify_save_expr): When in SSA form allow
20418         SAVE_EXPRs to compute to SSA vars.
20420 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20422         PR target/85582
20423         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20424         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20425         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20426         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20428 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20430         PR tree-optimization/85586
20431         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20432         exit early for statements in the same group if the accesses are
20433         not strided.
20435 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20437         PR lto/85451
20438         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20439         error message.
20441 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20443         PR tree-optimization/85143
20444         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20446 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20448         PR lto/85451
20449         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20450         not found" error message.
20452 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20454         PR other/83786
20455         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20456         * vec.c (test_ordered_remove_if): New function.
20457         (vec_c_tests): Call test_ordered_remove_if.
20458         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20459         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20460         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20461         VEC_ORDERED_REMOVE_IF.
20463 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20465         PR tree-optimization/82665
20466         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20467         pointer subtraction where arguments come from a memchr call.
20469 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20471         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20472         --push-state --as-needed and --pop-state instead of --as-needed and
20473         --no-as-needed if ld supports it.
20474         * configure: Regenerated.
20476         PR web/85578
20477         * doc/install.texi2html: Replace _002d with - and _002a with * in
20478         generated html files using sed.
20480 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20482         PR c++/85523
20483         * gcc-rich-location.c (blank_line_before_p): New function.
20484         (use_new_line): New function.
20485         (gcc_rich_location::add_fixit_insert_formatted): New function.
20486         * gcc-rich-location.h
20487         (gcc_rich_location::add_fixit_insert_formatted): New function.
20489 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20491         * selftest.c (assert_streq): Rename "expected" and "actual" to
20492         "val1" and "val2".  Extend NULL-handling to cover both inputs
20493         symmetrically, while still requiring both to be non-NULL for a pass.
20494         * selftest.h (assert_streq): Rename "expected" and "actual" to
20495         "val1" and "val2".
20496         (ASSERT_EQ): Likewise.
20497         (ASSERT_EQ_AT): Likewise.
20498         (ASSERT_KNOWN_EQ): Likewise.
20499         (ASSERT_KNOWN_EQ_AT): Likewise.
20500         (ASSERT_NE): Likewise.
20501         (ASSERT_MAYBE_NE): Likewise.
20502         (ASSERT_MAYBE_NE_AT): Likewise.
20503         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20504         the assertion to pass.
20505         (ASSERT_STREQ_AT): Likewise.
20507 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20509         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20510         interaction with -pie.
20512 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20514         * selftest.h: Fix alphabetization of per-source-file selftest
20515         declarations.
20517 2018-04-30  Jason Merrill  <jason@redhat.com>
20519         PR c++/61982 - dead stores to destroyed objects.
20520         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20521         of clobber.
20523 2018-04-30  Jason Merrill  <jason@redhat.com>
20525         * tree.c (build_clobber): New.
20526         * tree.h: Declare it.
20527         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20529 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20531         * diagnostic-show-locus.c (layout::layout): Update for
20532         location_get_source_line returning a char_span.
20533         (struct char_span): Move to input.h.
20534         (struct correction): Update for fields in char_span becoming
20535         private.
20536         (struct source_line): Update for location_get_source_line
20537         returning a char_span.
20538         (layout::print_line): Likewise.
20539         * edit-context.c (edited_file::print_content): Likewise.
20540         (edited_file::print_diff_hunk): Likewise.
20541         (edited_file::print_run_of_changed_lines): Likewise.
20542         (edited_file::get_num_lines): Likewise.
20543         (edited_line::edited_line): Likewise.
20544         * final.c (asm_show_source): Likewise.
20545         * input.c (location_get_source_line): Convert return type
20546         from const char * to char_span, losing the final "line_len"
20547         param.
20548         (dump_location_info): Update for the above.
20549         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20550         when handling the literal within the line.
20551         (test_reading_source_line): Update for location_get_source_line
20552         returning a char_span.
20553         * input.h (class char_span): Move here from
20554         diagnostic-show-locus.c, converting from a struct to a class.
20555         Make data members private.
20556         (char_span::operator bool): New.
20557         (char_span::length): New.
20558         (char_span::get_buffer): New.
20559         (char_span::operator[]): New.
20560         (char_span::subspan): Make const.
20561         (char_span::xstrdup): New.
20562         (location_get_source_line): Convert return type from const char *
20563         to char_span, losing the final "line_size" param.
20565 2018-04-30  Jan Hubicka  <jh@suse.cz>
20567         * lto-wrapper.c (ltrans_priorities): New static var.
20568         (cmp_priority): New.
20569         (run_gcc): Read priorities and if doing parallel build order
20570         the Makefile by them.
20572 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20574         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20576 2018-04-30  Richard Biener  <rguenther@suse.de>
20578         * tree-cfg.c (verify_address): Remove base argument, add
20579         flag whether to check TREE_ADDRESSABLE and do that.
20580         (verify_expr): Remove.
20581         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20582         (verify_gimple_assign_single): Likewise.
20583         (verify_gimple_switch): Likewise.
20584         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20585         piece from verify_expr.
20586         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20588 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20590         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20592 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20594         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20595         (small_data_pattern): Likewise.
20596         (arc_rewrite_small_data): Likewise.
20597         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20598         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20599         (get_symbol_alignment): New function.
20600         (legitimate_small_data_address_p): Likewise.
20601         (legitimate_scaled_address): Update, call
20602         legitimate_small_data_address_p.
20603         (output_sdata): New static variable.
20604         (arc_print_operand): Update how we handle small data operands.
20605         (arc_print_operand_address): Likewise.
20606         (arc_legitimate_address_p): Update, use
20607         legitimate_small_data_address_p.
20608         (arc_rewrite_small_data_p): Remove.
20609         (arc_rewrite_small_data_1): Likewise.
20610         (arc_rewrite_small_data): Likewise.
20611         (small_data_pattern): Likewise.
20612         (compact_sda_memory_operand): Update to use
20613         legitimate_small_data_address_p and get_symbol_alignment.
20614         (prepare_move_operands): Don't rewite sdata pattern.
20615         (prepare_extend_operands): Remove.
20616         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20617         pattern.
20618         (zero_extendqisi2): Likewise.
20619         (zero_extendhisi2): Likewise.
20620         (extendqihi2): Likewise.
20621         (extendqisi2): Likewise.
20622         (extendhisi2): Likewise.
20623         (addsi3): Likewise.
20624         (subsi3): Likewise.
20625         (andsi3): Likewise.
20626         * config/arc/constraints.md (Usd): Change it to memory constraint.
20628 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20630         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20631         as source of std instructions.
20632         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20633         allow 6-bit constants as source for store instructions.
20634         (movdi_insn): Update instruction pattern to allow 6-bit constants
20635         as source for store instructions.
20637 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20639         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20641 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20642             Sandra Loosemore <sandra@codesourcery.com>
20644         * dumpfile.c (dump_open): Allow '-' for stdout.
20645         * doc/invoke.texi (Developer Options): Document dump filename
20646         determination early.  Document stdin/stdout selection.
20648 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20650         Microblaze Target: PIC data text relative
20652         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20653         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20654         Add declaration.
20655         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20656         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20657         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20658         New addressing mode for data-text relative position indepenedent code.
20659         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20660         'ADDRESS_SYMBOLIC_TXT_REL'.
20661         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20662         (microblaze_legitimate_pic_operand): Exclude function calls from
20663         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20664         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20665         addresses cases.
20666         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20667         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20668         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20669         for 'address + offset'.
20670         (microblaze_expand_prologue): Add new function prologue call for
20671         'r20' assignation.
20672         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20673         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20674         table in case of TARGET_PIC_DATA_TEXT_REL.
20675         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20676         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20677         Add new macros 'UNSPEC_TEXT',
20678         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20679         + exclude function calls from 'UNSPEC_PLT' in case of data text
20680         relative mode.
20681         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20682         new target hook for generating address diff vector tables in case of
20683         flag_pic.
20684         * doc/tm.texi : Regenerate.
20685         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20686         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20687         of addr diff vector generation.
20688         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20689         target hook definition.
20690         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20691         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20692         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20693         option for data text relative.
20695 2018-04-30  Richard Biener  <rguenther@suse.de>
20697         * tree-chrec.h (evolution_function_is_constant_p): Remove
20698         redundant check.
20699         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20701 2018-04-30  Richard Biener  <rguenther@suse.de>
20703         PR bootstrap/85571
20704         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20706 2018-04-30  Richard Biener  <rguenther@suse.de>
20708         PR tree-optimization/28364
20709         PR tree-optimization/85275
20710         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20711         copying first exit test.
20713 2018-04-28  Mark Wielaard  <mark@klomp.org>
20715         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20716         dwarf_version >= 5.
20717         (dwarf_AT): Handle DW_AT_addr_base.
20718         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20720 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
20722         PR target/84431
20723         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
20724         (*ashl<dwi>3_doubleword_mask_1): Ditto.
20725         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
20726         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
20728 2018-04-28  Richard Biener  <rguenther@suse.de>
20730         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
20731         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
20732         to reflect use.  Only add interesting stmts.
20734 2018-04-27  Martin Jambor  <mjambor@suse.cz>
20736         PR ipa/85549
20737         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
20738         the jump function allows for passing through aggregate values.
20740 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
20742         * input.h (in_system_header_at): Convert from macro to inline
20743         function.
20744         (from_macro_expansion_at): Likewise.
20745         (from_macro_definition_at): Likewise.
20747 2018-04-27  Jeff Law  <law@redhat.com>
20749         * config.gcc: Mark tile* targets as deprecated/obsolete.
20751 2018-04-27  Richard Biener  <rguenther@suse.de>
20753         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
20754         fix for ILP32.
20756 2018-04-27  Richard Biener  <rguenther@suse.de>
20758         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
20760 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
20762         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
20763         with Yd constraint. Set "preferred_for_speed" attribute from
20764         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
20765         with Yd constraint.
20766         (*movdi_internal): Ditto.
20767         (movti_interunit splitters): Remove
20768         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
20769         (movdi_interunit splitters): Ditto.
20770         * config/i386/constraints.md (Ye): Remove.
20771         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
20773 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20775         PR target/85512
20776         * config/aarch64/constraints.md (Usg): Limit to 31.
20777         (Usj): Limit to 63.
20779 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
20781         PR tree-optimization/85529
20782         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
20783         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
20784         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
20785         zero extension or masking of the MSB bit.
20786         (optimize_range_tests): Add FIRST_BB argument, pass it through
20787         to optimize_range_tests_var_bound.
20788         (maybe_optimize_range_tests, reassociate_bb): Adjust
20789         optimize_range_tests callers.
20791 2018-04-26  Richard Biener  <rguenther@suse.de>
20792             Jakub Jelinek  <jakub@redhat.com>
20794         * cgraph.h (symbol_table): Just declare debug method here.
20795         * symtab.c (symbol_table::debug): Define.
20797 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
20799         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
20801 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
20803         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
20804         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
20805         (*movdi_internal): Substitute Yi and Yj constraint with x
20806         and Ym and Yn constraint with y constraint.  Update "isa"
20807         attribute and set "preferred_for_speed" attribute from
20808         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
20809         (*movsi_internal): Ditto.
20810         (*movdf_internal): Ditto.
20811         (*movsf_internal): Ditto.
20812         (*zero_extendsidi2): Ditto.
20813         * config/i386/sse.md (vec_set<mode>_0): Ditto.
20814         (sse2_loadld): Ditto.
20815         (*vec_extract<ssevecmodelower>_0): Ditto.
20816         (*vec_extractv4si_0_zext_sse4): Ditto.
20817         (vec_concatv2di): Ditto.
20818         (*vec_dup<mode>): Ditto.
20819         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
20820         * config/i386/constraints.md (Yi): Remove.
20821         (Yj): Remove.
20822         (Ym): Remove.
20823         (Yn): Remove.
20825 2018-04-26  Nathan Sidwell  <nathan@acm.org>
20827         * dumpfile.c (dump_open): New.
20828         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
20829         (dump_finish): Detect stdio/stderr by value not name.
20831 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
20833         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
20835 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20837         PR target/84952
20838         * config/nvptx/nvptx.c (verify_neutering_jumps)
20839         (verify_neutering_labels): New function
20840         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
20842 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20844         PR target/84025
20845         * config/nvptx/nvptx.c (needs_neutering_p): New function.
20846         (nvptx_single): Use needs_neutering_p to skip over insns that do not
20847         need neutering.
20849 2018-04-26  Richard Biener <rguenther@suse.de>
20850             Tom de Vries  <tom@codesourcery.com>
20852         PR lto/85422
20853         * lto-streamer-out.c (output_function): Fixup loops if required to match
20854         discovery done in the reader.
20856 2018-04-26  Richard Biener  <rguenther@suse.de>
20858         PR tree-optimization/85116
20859         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
20860         have a loop exit from the single latch predecessor.  Remove
20861         case of header with just condition.
20862         (ch_base::copy_headers): Exclude infinite loops from any
20863         processing.
20864         (pass_ch::execute): Record exits.
20866 2018-04-26  Richard Biener  <rguenther@suse.de>
20868         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
20869         prologue cost vector and pass it to vect_get_load_cost.
20870         (vect_get_peeling_costs_all_drs): Likewise.
20871         (vect_peeling_hash_get_lowest_cost): Likewise.
20872         (vect_enhance_data_refs_alignment): Likewise.
20874 2018-04-26  Richard Biener  <rguenther@suse.de>
20876         PR middle-end/85450
20877         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
20878         checking of integer<->pointer conversions.
20879         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
20880         sign-/zero-extending pointer types.
20881         (expand_omp_for_static_chunk): Likewise.
20883 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
20884             Jean Lee  <xiaoyur347@gmail.com>
20886         * config/mips/mips.c (mips_asan_shadow_offset): New function.
20887         (TARGET_ASAN_SHADOW_OFFSET): Define.
20888         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
20889         true for -fsanitize=address.
20891 2018-04-25  Mark Wielaard  <mark@klomp.org>
20893         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
20894         shorter ones.
20896 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20898         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
20899         than "alu", remove explicit "memory" and "imm_disp" attributes.
20900         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
20902         PR middle-end/85414
20903         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
20904         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
20905         gen_lowpart_no_emit.
20907 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
20909         PR target/85473
20910         * config/i386/i386.c (ix86_expand_builtin): Change memory
20911         operand to XI, extend p0 to Pmode.
20912         * config/i386/i386.md: Change unspec volatile and operand
20913         1 mode to XI, change operand 0 mode to P.
20915 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20917         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
20918         GET_MODE_MASK before any checking.
20919         (nds32_can_use_bset_p): Likewise.
20920         (nds32_can_use_btgl_p): Likewise.
20922 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20924         * config/nds32/nds32-doubleword.md: New define_split pattern for
20925         illegal register number.
20927 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20929         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
20931 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20933         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
20935 2018-04-25  Richard Biener  <rguenther@suse.de>
20937         * lto-streamer.h (LTO_major_version): Bump to 8.
20939 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20941         * BASE-VER: Set to 9.0.0.
20943 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
20945         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
20946         in __abskf2 and __powikf2.
20948 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20950         PR target/85512
20951         * config/aarch64/constraints.md (Usg, Usj): New constraints.
20952         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
20953         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
20954         Use the above on operand 2.  Reindent.
20955         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
20957 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20959         PR target/85485
20960         * common/config/i386/i386-common.c (ix86_handle_option): Don't
20961         handle OPT_mcet.
20962         * config/i386/i386.opt (mcet): Removed.
20963         * doc/install.texi: Remove -mcet documentation.
20964         * doc/invoke.texi: Likewise.
20966 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20968         PR target/85485
20969         * doc/install.texi: Remove -mcet from bootstrap-cet.
20971 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20973         PR target/85511
20974         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
20975         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
20976         if TARGET_64BIT.
20978         PR target/85503
20979         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
20980         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
20981         containing a CONST_VECTOR.
20983 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
20985         * doc/install.texi: Update newlib dependency for nvptx.
20987 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20989         PR target/85508
20990         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
20991         instead of INTVAL when shifting x left.
20993 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
20995         PR tree-optimization/85478
20996         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
20997         vect_grouped_store_supported for single element vectors.
20999 2018-04-24  Richard Biener  <rguenther@suse.de>
21001         PR target/85491
21002         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21003         load cost increase to the case of non-constant step.
21005 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21007         PR target/84828
21008         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21009         destination if any_malformed_asm.
21011 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
21013         PR middle-end/85496
21014         * expr.c (store_field): In the bitfield case, if the value comes from
21015         a function call and is returned in registers by means of a PARALLEL,
21016         do not change the mode of the temporary unless BLKmode and VOIDmode.
21018 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
21020         PR rtl-optimization/85423
21021         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21022         dependencies to debug insns when the previous insn is non-debug.
21024 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21026         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21027         enums into a single definition.
21028         (fls): Fix predicates and printing.
21029         (seti): Likewise.
21031 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21033         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21034         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21035         and short u6 immediate.
21036         (check_if_valid_sleep_operand): Remove.
21037         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21039 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21041         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21042         flag_always_save_lp condition.
21043         * config/nds32/nds32.opt (malways-save-lp): New option.
21045 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21047         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21048         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21049         * config/nds32/nds32.h
21050         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21051         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21053 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21055         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21056         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21058 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21059             Chung-Ju Wu  <jasonwucj@gmail.com>
21061         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21062         Declare.
21063         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21064         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21066 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21068         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
21070 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21072         * config/nds32/nds32-protos.h (nds32_data_alignment,
21073         nds32_local_alignment): Declare.
21074         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
21075         nds32_local_alignment): New functions.
21076         (TARGET_CONSTANT_ALIGNMENT): Define.
21077         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
21079 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21081         * config/nds32/nds32.c
21082         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
21083         (TARGET_MODES_TIEABLE_P): Likewise.
21085 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21087         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
21088         level Ofast and Og.
21090 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
21091             Chung-Ju Wu  <jasonwucj@gmail.com>
21093         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21094         for unaligned access.
21095         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21096         unaligned access.
21097         * config/nds32/nds32-intrinsic.md: Likewise.
21098         * config/nds32/nds32_intrinsic.h: Likewise.
21099         * config/nds32/nds32.h (nds32_builtins): Likewise.
21100         * config/nds32/nds32.opt (munaligned-access): New option.
21101         * config/nds32/nds32.c (nds32_asm_file_start): Display
21102         flag_unaligned_access status.
21104 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21106         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
21107         -mno-relax is present.
21108         * config/riscv/linux.h (LINK_SPEC): Ditto.
21110 2018-04-20  Martin Sebor  <msebor@redhat.com>
21112         PR c/85365
21113         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
21114         for null pointers.
21115         (gimple_fold_builtin_stxcpy_chk): Same.
21116         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
21118 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
21120         PR target/85456
21121         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
21122         __powikf2 when long double is IEEE 128-bit.
21124 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21126         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
21127         step to make sure stack always aligned.
21129 2018-04-20  Carl Love  <cel@us.ibm.com>
21131         PR target/83402
21132         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21133         size check for arg0.
21135 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
21136             Tom de Vries  <tom@codesourcery.com>
21138         PR target/85445
21139         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21140         Emit insns for calls too.
21141         (nvptx_find_par): Always look for worker-level predecessor insn.
21142         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
21143         calls.
21144         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21145         (nvptx_process_pars): Propagate frames for calls.
21147 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
21149         PR target/85469
21150         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21151         Removed.
21152         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21153         (ix86_handle_option): Don't handle OPT_mibt.
21154         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21155         __SHSTK__.
21156         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21157         has_ibt and ibt.
21158         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21159         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21160         (ix86_target_macros): Define __CET__ with flag_cf_protection
21161         for -fcf-protection.
21162         * config/i386/i386.c (isa2_opts): Remove -mibt.
21163         * config/i386/i386.h (TARGET_IBT): Removed.
21164         (TARGET_IBT_P): Likewise.
21165         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21166         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21167         * config/i386/i386.opt (mcet): Update help message.
21168         (mshstk): Likewise.
21169         (mibt): Removed.
21170         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
21171         -mcet as an alias for -mshstk.
21173 2018-04-20  Richard Biener <rguenther@suse.de>
21175         PR middle-end/85475
21176         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21177         complexity by forcing a single use of the multiply operand.
21179 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21181         ipa/85449
21182         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21183         recursion dependency to only apply to non-clones.
21185 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21187         ipa/85447
21188         * ipa-cp.c (create_specialized_node): Check that clones of
21189         self-recursive edges exist during IPA-CP.
21191 2018-04-19  Toon Moene  <toon@moene.org>
21193         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21194         by -O3.
21196 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21198         PR tree-optimization/85467
21199         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21200         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
21201         VECTOR_CST element to type.
21203 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21205         PR target/85397
21206         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21207         * config/i386/i386.md (builtin_setjmp_setup): Removed.
21208         (builtin_longjmp): Likewise.
21209         (save_stack_nonlocal): New pattern.
21210         (restore_stack_nonlocal): Likewise.
21212 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21214         PR target/85404
21215         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21216         Replace ASM_OUTPUT_LABEL with fprintf.
21218 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21220         PR target/85417
21221         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21222         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21223         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21224         define __IBT__ and __SHSTK__ for -fcf-protection.
21225         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21226         TARGET_IBT.
21227         (ix86_trampoline_init): Likewise.
21228         (x86_output_mi_thunk): Likewise.
21229         (ix86_notrack_prefixed_insn_p): Likewise.
21230         (ix86_option_override_internal): Don't disallow -fcf-protection.
21231         * config/i386/i386.md (rdssp<mode>): Also enable for
21232         -fcf-protection.
21233         (incssp<mode>): Likewise.
21234         (nop_endbr): Likewise.
21235         * config/i386/i386.opt (mcet): Change help message to built-in
21236         functions only.
21237         (mibt): Likewise.
21238         (mshstk): Likewise.
21239         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21240         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21241         enable CET built-in functions.
21243 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21245         * common/config/i386/i386-common.c
21246         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21247         OPTION_MASK_ISA_MOVDIRI_UNSET,
21248         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21249         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21250         * config.gcc (movdirintrin.h): New header.
21251         * config/i386/cpuid.h (bit_MOVDIRI,
21252         bit_MOVDIR64B): New bits.
21253         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21254         and -mmvodir64b.
21255         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21256         (VOID, PVOID, PCVOID)): New function types.
21257         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21258         __builtin_ia32_directstoreu_u64,
21259         __builtin_ia32_movdir64b): New builtins.
21260         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21261         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21262         and -mmovdiri.
21263         (ix86_valid_target_attribute_inner_p): Ditto.
21264         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21265         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21266         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21267         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21268         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21269         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21270         (movdiri<mode>, movdir64b_<mode>): New.
21271         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21272         * config/i386/immintrin.h: Include movdirintrin.h.
21273         * config/i386/movdirintrin.h: New file.
21274         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21276 2018-04-19  Richard Biener  <rguenther@suse.de>
21278         PR middle-end/85455
21279         * cfg.c (clear_bb_flags): When loop state says we have
21280         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21282 2018-04-19  Richard Biener  <rguenther@suse.de>
21284         PR tree-optimization/84737
21285         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21286         copying restrict info.
21287         (vect_setup_realignment): Use it.
21288         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21289         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21290         the first DR to all generated stores.
21291         (vectorizable_load): Likewise for loads.
21293 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21295         PR tree-optimization/85446
21296         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21297         the integral and pointer types to have the same precision.
21299         * doc/install.texi: Document --disable-cet being the default and
21300         --enable-cet=auto.
21302 2018-04-18  Martin Liska  <mliska@suse.cz>
21304         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21305         style.
21307 2018-04-18  Martin Liska  <mliska@suse.cz>
21309         Revert
21310         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21312         PR ipa/83983
21313         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21314         arguments if they are comparable.
21316 2018-04-18  Martin Liska  <mliska@suse.cz>
21318         Revert
21319         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21321         PR lto/84805
21322         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21323         incomplete types.
21325 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21327         PR target/85388
21328         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21329         ENDBR after calling __morestack.
21331 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21333         PR jit/85384
21334         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21335         by using gcc_base_ver to generate a gcc_driver_version, and use
21336         it when generating GCC_DRIVER_NAME.
21337         * configure: Regenerate.
21339 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21341         PR target/81084
21342         * config.gcc: Obsolete powerpc*-*-*spe*.
21344 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21346         PR debug/84637
21347         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21348         (stabstr_D): Change type of unum from unsigned int to
21349         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21350         type.
21352 2018-04-17  Jim Wilson  <jimw@sifive.com>
21354         PR 84856
21355         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21356         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21357         Set arg_pointer_offset after using pretend_args_size.
21359 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21361         PR rtl-optimization/85431
21362         * dse.c (record_store): Ignore zero width stores.
21364         PR sanitizer/85230
21365         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21366         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21367         __builtin_stack_restore rather than after it.
21368         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21369         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21370         argument instead of virtual_dynamic_stack_rtx.
21372 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21374         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21375         New prototype.
21376         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21377         Add note to error message to explain internal mapping of overloaded
21378         built-in function name to non-overloaded built-in function name.
21379         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21380         function.
21382 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21384         PR target/85424
21385         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21386         where the inputs overlap with the output.
21388 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21390         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21391         (=v, v) alternative and explicit "memory" attribute.
21392         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21393         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21394         attributes.
21395         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21396         "sselog1" type instead of "sselog".
21397         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21398         "sselog".  Remove explicit "memory" attribute.
21399         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21400         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21401         attributes.
21402         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21403         "sselog1" type instead of "sselog".  Remove explicit "memory"
21404         attribute.
21405         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21406         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21407         attribute.
21408         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21409         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21410         attributes.
21411         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21412         "sselog1" type instead of "sselog".  Remove explicit "memory"
21413         attribute.
21414         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21415         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21416         attribute.
21418         PR target/85430
21419         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21421         PR middle-end/85414
21422         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21423         on a SUBREG.
21425 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21427         PR ipa/85421
21428         * ipa-cp.c (create_specialized_node): Call
21429         expand_all_artificial_thunks if necessary.
21431 2018-04-17  Martin Liska  <mliska@suse.cz>
21433         PR lto/85405
21434         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21435         in message, remote space in between '_G' and '('.
21437 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21439         PR target/85281
21440         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21441         avx512f_vmcmp<mode>3<round_saeonly_name>,
21442         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21443         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21444         avx512f_rndscale<mode><round_saeonly_name>,
21445         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21446         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21447         Use %<iptr>2 instead of %2 for -masm=intel.
21448         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21449         avx512f_vcvttss2usi<round_saeonly_name>,
21450         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21451         -masm=intel.
21452         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21453         avx512f_vcvttsd2usi<round_saeonly_name>,
21454         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21455         Use %q1 instead of %1 for -masm=intel.
21456         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21457         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21458         of %3 for -masm=intel.
21459         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21460         -masm=intel.
21461         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21462         -masm=intel.
21463         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21464         -masm=intel.
21465         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21466         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21467         %g1.
21468         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21469         -masm=intel.
21470         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21471         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21472         %g1 and one with %0 and %1.
21473         (avx512er_vmrcp28<mode><round_saeonly_name>,
21474         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21475         %1 for -masm=intel.
21476         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21477         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21478         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21479         of %0 and %{%4%} for -masm=intel.
21480         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21481         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21482         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21483         order of %0 and %{%5%}%{z%} for -masm=intel.
21485 2018-04-17  Jan Hubicka  <jh@suse.cz>
21487         PR lto/85405
21488         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21490 2018-04-17  Martin Liska  <mliska@suse.cz>
21492         PR ipa/85329
21493         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21494         for target_clone error message.  Make default implementation
21495         clone to be a local declaration.
21496         (separate_attrs): Add new argument and check for an empty
21497         string.
21498         (expand_target_clones): Handle it.
21499         (ipa_target_clone): Make redirection just for target_clones
21500         functions.
21502 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21503             Tom de Vries  <tom@codesourcery.com>
21505         PR middle-end/84955
21506         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21507         tiled basic blocks without omp continue statements.
21509 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21511         PR target/83660
21512         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21513         vec_extract expression as having side effects to make sure it gets
21514         a cleanup point.
21516 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21518         PR target/85403
21519         * config/i386/i386.c (get_builtin_code_for_version): Check
21520         error_mark_node.
21522 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21524         PR target/84331
21525         * config.gcc: Support "skylake".
21526         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21527         PROCESSOR_SKYLAKE.
21528         * config/i386/i386.c (m_SKYLAKE): Define.
21529         (processor_target_table): Add "skylake".
21530         (ix86_option_override_internal): Add "skylake".
21531         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21532         PROCESSOR_CANNONLAKE.
21533         (get_builtin_code_for_version): Fix priority for
21534         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21535         PROCESSOR_SKYLAKE-AVX512.
21536         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21537         (processor_type): Add PROCESSOR_SKYLAKE.
21539 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21540             Jason Merrill  <jason@redhat.com>
21542         PR c++/85112
21543         * convert.c (convert_to_integer_1): Use direct recursion for
21544         enumeral types and types with a precision less than the number
21545         of bits in their mode.
21547 2018-04-16  Julia Koval  <julia.koval@intel.com>
21549         PR target/84413
21550         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21551         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21553 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21555         PR target/85293
21556         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21557         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21558         and -mno-direct-move.
21560 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21562         PR target/83402
21563         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21564         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21565         Ensure negative shifts result in {0}.
21567 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21569         PR rtl-optimization/79916
21570         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21571         regs (if any) to define how to gnerate SD moves when LRA is in
21572         progress.
21574 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21576         PR rtl-optimization/85393
21577         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21578         * except.c (expand_dw2_landing_pad_for_region): Make static.
21579         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21580         a label and unconditional jump to old_bb, rather than
21581         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21582         basic block.
21584         PR rtl-optimization/85376
21585         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21586         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21587         instead of a specific value.
21589 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21590             Bin Cheng  <bin.cheng@arm.com>
21592         PR tree-optimization/82965
21593         PR tree-optimization/83991
21594         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21595         by_profile_only parameter.
21596         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21597         information if the loop was predicted to iterate too many times.
21598         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21600 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21602         PR lto/71991
21603         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21604         always inline.
21606 2018-04-13  Martin Liska  <mliska@suse.cz>
21607             Jakub Jelinek  <jakub@redhat.com>
21609         PR middle-end/81657
21610         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21611         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21612         * builtins.c (expand_builtin_memory_copy_args): Use
21613         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21614         handle dest_addr == pc_rtx.
21616 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21618         PR target/85291
21619         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21620         asked to not generate direct moves.
21621         (fix_trunc<mode>si2_stfiwx): Similar.
21622         (fix_trunc<mode>si2_internal): Similar.
21624 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21626         PR debug/83157
21627         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21628         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21629         lookup if dest in some wider mode is known to be const0_rtx and
21630         if so, record permanent equivalence for it to be ZERO_EXTEND of
21631         the narrower mode destination.
21633 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21635         * lto-streamer-out.c (output_function): Revert 259346.
21636         * omp-expand.c (expand_oacc_for): Likewise.
21638 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21640         PR rtl-optimization/85354
21641         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21642         * sel-sched.c (sel_global_init): ... here.
21644 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21646         PR target/85238
21647         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21648         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21649         mode for PE-COFF targets.
21650         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21651         (i386_pe_asm_lto_end): Likewise.
21652         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21653         (TARGET_ASM_LTO_END): Likewise.
21654         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21655         (i386_pe_asm_lto_start): New function.
21656         (i386_pe_asm_lto_end): Likewise.
21658 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21659             Richard Biener  <rguenther@suse.de>
21661         PR middle-end/84955
21662         * lto-streamer-out.c (output_function): Fix CFG loop state before
21663         streaming out.
21664         * omp-expand.c (expand_oacc_for): Handle calls to internal
21665         functions like regular functions.
21667 2018-04-12  Richard Biener  <rguenther@suse.de>
21669         PR lto/85371
21670         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21671         for the early LTO debug to properly generate references to it
21672         during DIE emission.  Do not re-use that for the skeleton for
21673         split-dwarf.
21674         (dwarf2out_early_finish): Likewise.
21676 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21678         PR target/85328
21679         * config/i386/sse.md
21680         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21681         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21682         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21683         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21684         and output is a reg, avoid creating invalid lowpart subreg, but
21685         instead split into a 512-bit move.  Don't split if not AVX512VL,
21686         input is xmm16+ reg and output is a mem.
21687         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21688         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21689         xmm16+ reg and output is a mem.
21691 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21693         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21694         also for flag_dwarf2_cfi_asm.
21696 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21698         PR rtl-optimization/85342
21699         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21700         a bool scalar var inside of the loop instead.  Don't try to update
21701         recog_data.operand after failed apply_change_group.
21703 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21705         PR target/85296
21706         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21707         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21708         array with flexible array member as array without given dimension.
21709         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21710         argument for undefined param to true.
21712 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21714         PR target/85321
21715         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21716         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21717         from PowerPC section.
21718         * config/rs6000/sysv4.opt (mcall-): Improve help text.
21719         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21720         help text that is too long.
21721         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
21722         help text that is too long.
21723         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
21724         help text that is too long.
21726 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
21728         * config/alpha/alpha.md (stack_probe_internal): Rename
21729         from "probe_stack".  Update all callers.
21731 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21733         PR rtl-optimization/84566
21734         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
21735         sched_macro_fuse_insns.
21737 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21739         PR target/84301
21740         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
21741         (compute_block_dependences): ... from here.
21743 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21745         PR tree-optimization/85331
21746         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
21747         from int to HOST_WIDE_INT.
21749 2018-04-11  Martin Jambor  <mjambor@suse.cz>
21751         PR ipa/84149
21752         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
21753         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
21754         not the same as the source val.
21755         (cgraph_edge_brings_value_p): New parameter.
21756         (gather_edges_for_value): Pass destination value to
21757         cgraph_edge_brings_value_p.
21758         (perhaps_add_new_callers): Likewise.
21759         (get_info_about_necessary_edges): Likewise and exclude values brought
21760         only by self-recursive edges.
21761         (create_specialized_node): Redirect only clones of self-calling edges.
21762         (+self_recursive_pass_through_p): New function.
21763         (find_more_scalar_values_for_callers_subset): Use it.
21764         (find_aggregate_values_for_callers_subset): Likewise.
21765         (known_aggs_to_agg_replacement_list): Removed.
21766         (decide_whether_version_node): Re-calculate known constants for all
21767         remaining context clones.
21769 2018-04-11  Richard Biener  <rguenther@suse.de>
21771         PR lto/85339
21772         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
21773         from early DWARF output.
21774         (dwarf2out_early_finish): Output line info unconditionally into
21775         early DWARF and add reference to it.
21777 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21779         PR target/85281
21780         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
21781         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
21782         other than V2DFmode using iptr mode attribute.
21783         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
21785 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21787         PR rtl-optimization/84659
21788         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
21790 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21792         PR debug/85302
21793         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
21794         SIZEP is NULL.
21795         (output_loc_list): Pass address of a dummy size variable even in the
21796         locview handling loop.
21797         (index_location_lists): Add comment on why skip_loc_list_entry can't
21798         call size_of_locs.
21800 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21802         PR target/85261
21803         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
21804         into register.
21806 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
21808         PR target/85321
21809         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21810         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
21811         and -mstring-compare-inline-limit.
21813 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21815         PR target/85287
21816         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
21817         for stack clash protection in a register whenever we need it to be in
21818         a register.
21820 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21822         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
21823         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
21825 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21827         PR target/85321
21828         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
21829         the help text.
21830         (mlong-double-): Ditto.
21831         * config/rs6000/sysv4.opt (msdata=): Ditto.
21832         (mtls-size=): Ditto.
21834 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21836         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
21837         erroneous entries for
21838         "vector int vec_ldl (int, long int *)", and
21839         "vector unsigned int vec_ldl (int, unsigned long int *)".
21840         Add comments and entries for
21841         "vector bool char vec_ldl (int, bool char *)",
21842         "vector bool short vec_ldl (int, bool short *)",
21843         "vector bool int vec_ldl (int, bool int *)",
21844         "vector bool long long vec_ldl (int, bool long long *)",
21845         "vector pixel vec_ldl (int, pixel *)",
21846         "vector long long vec_ldl (int, long long *)",
21847         "vector unsigned long long vec_ldl (int, unsigned long long *)".
21848         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
21849         type tree bool_long_long_type_node and correct definition of
21850         bool_V2DI_type_node to make reference to this new type tree.
21851         (rs6000_mangle_type): Replace erroneous reference to
21852         bool_long_type_node with bool_long_long_type_node.
21853         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
21854         comments to emphasize sign distinctions for char and int types and
21855         replace RS6000_BTI_bool_long constant with
21856         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
21857         use of RS6000_BTI_pixel.
21858         (bool_long_type_node): Remove this macro definition.
21859         (bool_long_long_type_node): New macro definition
21861 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21863         PR rtl-optimization/85300
21864         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
21865         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
21866         simplify_unary_operation fails.
21868 2018-04-10  Martin Liska  <mliska@suse.cz>
21870         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
21871         cgraph_edge and ipa_ref.
21873 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21875         PR target/85177
21876         PR target/85255
21877         * config/i386/sse.md
21878         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
21879         computation of the VEC_MERGE selector from mask.
21880         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
21881         Fix decoding of the VEC_MERGE selector into mask.
21883 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21885         PR tree-optimization/85286
21886         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
21888 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21890         * final.c (final_1): Set insn_last_address as well as
21891         insn_current_address.
21893 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21895         PR target/85173
21896         * explow.c (emit_stack_probe): Call validize_mem on memory location
21897         before passing it to gen_probe_stack.  Create address operand and
21898         legitimize it for the probe_stack_address case.
21900 2018-04-09  Jan Hubicka  <jh@suse.cz>
21902         PR lto/85078
21903         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
21904         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
21905         * tree.c (free_lang_data_in_type): Fix handling of binfos;
21906         walk basetypes.
21907         (free_lang_data): Rebuild type inheritance graph.
21909 2018-04-09  Martin Sebor  <msebor@redhat.com>
21911         * invoke.texi (-finline-small-functions): Mention other optimization
21912         options.
21913         (-findirect-inlining, -fpartial-inlining): Same.
21914         (-finline-functions-called-once): Same.
21915         (-freorder-blocks-and-partition): Same.
21917 2018-04-09  Jan Hubicka  <jh@suse.cz>
21919         PR rtl/84058
21920         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
21921         jumps; choose last target that matches the criteria (i.e.
21922         no partition changes for non-crossing jumps).
21923         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
21924         support for redirecting crossing jumps to non-crossing.
21926 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
21928         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
21929         also for naked functions.
21931 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
21933         * config/arc/arc.md (add_shift): New pattern.
21934         (add_shift2): Likewise.
21935         (sub_shift): Likewise.
21936         (sub_shift_cmp0_noout): Likewise.
21937         (compare_si_ashiftsi): Likewise.
21938         (xbfu_cmp0_noout): New combine pattern.
21939         (xbfu_cmp0"): Likewise.
21940         (movsi_set_cc_insn): Place the predicable variant first.
21941         (commutative_binary_cmp0_noout): Remove clobber.
21942         (commutative_binary_cmp0): New pattern.
21943         (noncommutative_binary_cmp0): Likewise.
21944         (noncommutative_binary_cmp0_noout): Likewise.
21945         (noncommutative_binary_comparison_result_used): Removed.
21946         (rsub_cmp0): New pattern.
21947         (rsub_cmp0_noout): Likewise.
21948         (extzvsi): Changed, keep only meaningful variants.
21949         (SQH, SEZ): New iterators.
21950         (SQH_postfix): New mode attribute.
21951         (SEZ_prefix): New code attribute.
21952         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
21953         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
21954         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
21955         of numerical value.
21956         (noncommutative_operator): Check the availability of barrel
21957         shifter option.
21959 2018-04-09  Richard Biener  <rguenther@suse.de>
21961         PR tree-optimization/85284
21962         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
21963         Only use the niter constraining form of simple_iv when the exit
21964         is always executed.
21966 2018-04-09  Tom de Vries  <tom@codesourcery.com>
21968         PR target/84041
21969         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21970         (define_expand "*memory_barrier"): New define_expand.
21971         (define_insn "memory_barrier"): New insn.
21973 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21975         PR rtl-optimization/80463
21976         PR rtl-optimization/83972
21977         PR rtl-optimization/83480
21979         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
21980         correct producer for the insn.
21981         (tidy_control_flow): Fixup seqnos in case of debug insns.
21983 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21985         PR rtl-optimization/83913
21987         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
21988         different sched-times when merging exprs.
21990 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21992         PR rtl-optimization/83962
21994         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
21995         tidy_fallthru_edge and tidy_control_flow.
21997 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21999         PR rtl-optimization/83530
22001         * sel-sched.c (force_next_insn): New global variable.
22002         (remove_insn_for_debug): When force_next_insn is true, also leave only
22003         next insn in the ready list.
22004         (sel_sched_region): When the region wasn't scheduled, make another pass
22005         over it with force_next_insn set to 1.
22007 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22009         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22010         into tm_file.
22011         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22012         for interrupt control.
22013         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22014         functions for interrupt control.
22015         * config/nds32/nds32-intrinsic.md: Likewise.
22016         * config/nds32/nds32_intrinsic.h: Likewise.
22017         * config/nds32/nds32.h (nds32_builtins): Likewise.
22019 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22021         * config/nds32/nds32.c (nds32_init_machine_status,
22022         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22023         strict_aligned_p field.
22024         (nds32_expand_to_rtl_hook): New function.
22025         (TARGET_EXPAND_TO_RTL_HOOK): Define.
22026         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22028 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22029             Chung-Ju Wu  <jasonwucj@gmail.com>
22031         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22032         * config/nds32/nds32-n7.md: New file.
22033         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22034         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22035         pipeline.
22036         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22037         * config/nds32/nds32.md (pipeline_model): Add n7.
22038         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22039         * config/nds32/pipelines.md: Include n7 settings.
22041 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22042             Chung-Ju Wu  <jasonwucj@gmail.com>
22044         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22045         * config/nds32/nds32-e8.md: New file.
22046         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22047         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22048         pipeline.
22049         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22050         * config/nds32/nds32.md (pipeline_model): Add e8.
22051         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22052         * config/nds32/pipelines.md: Include e8 settings.
22054 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22055             Chung-Ju Wu  <jasonwucj@gmail.com>
22057         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22058         * config/nds32/nds32-n8.md: New file.
22059         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22060         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22061         pipeline.
22062         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22063         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22064         * config/nds32/nds32.md (pipeline_model): Add n8.
22065         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22066         * config/nds32/pipelines.md: Include n8 settings.
22068 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22069             Chung-Ju Wu  <jasonwucj@gmail.com>
22071         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
22072         * config/nds32/nds32-n9-2r1w.md: New file.
22073         * config/nds32/nds32-n9-3r2w.md: New file.
22074         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
22075         nds32_register_ports): New or modify for cpu n9.
22076         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
22077         pipeline.
22078         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
22079         * config/nds32/nds32-utils.c: New file.
22080         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
22081         TARGET_MUL_SLOW): Define.
22082         * config/nds32/nds32.md (pipeline_model): New attribute.
22083         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
22084         New options that support cpu n9.
22085         * config/nds32/pipelines.md: Include n9 settings.
22086         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
22088 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22090         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
22091         information if necessary.
22092         (output_cond_branch_compare_zero): Likewise.
22093         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
22094         (nds32_target_alignment): Refine for alignment.
22095         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
22096         (FUNCTION_BOUNDARY): Modify.
22097         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
22098         align case.
22099         * config/nds32/nds32.opt (malways-align, malign-functions): New.
22101 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22103         * config/nds32/constants.md (unspec_volatile_element): Add values for
22104         TLB operation and data prefetch.
22105         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22106         functions for TLB operation and data prefetch.
22107         * config/nds32/nds32-intrinsic.md: Likewise.
22108         * config/nds32/nds32_intrinsic.h: Likewise.
22109         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
22110         (nds32_print_operand): Likewise.
22111         * config/nds32/nds32.h (nds32_builtins): Likewise.
22113 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22114         Andrew Pinski <pinsika@gcc.gnu.org>
22116         PR middle-end/82976
22117         * match.pd: Use constant_boolean_node of correct type instead of
22118         boolean_true_node or boolean_false_node for simplifying
22119         pointer comparisons to zero.
22121 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22123         PR tree-optimization/80021
22124         * tree.c (verify_type_variant): Make error call in verify_variant_match
22125         translatable and remove final full stop.
22127 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22129         * config/nds32/constants.md (unspec_volatile_element): Add
22130         UNSPEC_VOLATILE_EH_RETURN.
22131         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22132         nds32_output_stack_pop): Support dwarf exception handling process.
22133         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22134         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22135         exception handling process.
22136         (nds32_compute_stack_frame): Likewise.
22137         (nds32_return_addr_rtx): Likewise.
22138         (nds32_initial_elimination_offset): Likewise.
22139         (nds32_expand_prologue): Likewise.
22140         (nds32_expand_epilogue): Likewise.
22141         (nds32_dynamic_chain_address): New function.
22142         * config/nds32/nds32.h (machine_function): Add fields for dwarf
22143         exception handling.
22144         (DYNAMIC_CHAIN_ADDRESS): Define.
22145         (EH_RETURN_DATA_REGNO): Define.
22146         (EH_RETURN_STACKADJ_RTX): Define.
22147         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22148         patterns for dwarf exception handling.
22150 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22152         * config/nds32/nds32.h: Clean up obsolete macros.
22154 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22156         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22157         Add enum values for particular instructions.
22158         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22159         particular intrinsic functions.
22160         * config/nds32/nds32-intrinsic.md: Likewise.
22161         * config/nds32/nds32_intrinsic.h: Likewise.
22162         * config/nds32/nds32.h (nds32_builtins): Likewise.
22163         * config/nds32/nds32.md (type): Add pbsad and pbsada.
22164         (btst, ave): New patterns for particular instructions.
22166 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22168         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22169         Add enum values for atomic load/store and memory sync.
22170         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22171         and memory sync.
22172         * config/nds32/nds32-intrinsic.md: Likewise.
22173         * config/nds32/nds32_intrinsic.h: Likewise.
22174         * config/nds32/nds32.h (nds32_builtins): Likewise.
22176 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22178         PR tree-optimization/85257
22179         * fold-const.c (native_encode_vector): If not all elts could fit
22180         and off is -1, return 0 rather than offset.
22181         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22182         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22183         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
22184         adjust buffer in native_interpret_expr call.
22186 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22188         * config/nds32/constants.md (unspec_volatile_element): Add cache
22189         control enum values.
22190         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22191         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22192         * config/nds32/nds32.c (nds32_cctl_names): New.
22193         (nds32_print_operand): Handle cache control register names.
22194         * config/nds32/nds32.h (nds32_builtins): New enum values.
22195         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22196         macros.
22197         * config/nds32/nds32.md (type): Add mmu.
22198         * config/nds32/pipelines.md (simple_insn): Add mmu.
22200 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22202         * config/nds32/nds32.md (type): Remove call.
22203         * config/nds32/pipelines.md (simple_insn): Likewise.
22205 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22207         * config/nds32/constants.md (unspec_volatile_element): Add
22208         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22209         UNSPEC_VOLATILE_FMFCFG.
22210         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22211         description for fmfcfg and fmfcsr.
22212         (bdesc_1arg): Add fmtcsr.
22213         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22214         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22215         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22216         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22217         unspec_fmfcfg): New patterns.
22218         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22219         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22220         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22221         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22222         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22223         __nds32__fmfcfg): Define.
22225 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22227         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22228         intrinsic register names.
22229         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22230         intrinsic register enum values and macros.
22232 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22234         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22235         for load/store addressing form.
22236         (nds32_print_operand_address): Likewise.
22238 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22240         PR target/85196
22241         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22242         based on LABEL_REF.  Remove useless assertion.
22243         (pic_address_needs_scratch): Fix formatting.
22244         (sparc_legitimize_pic_address): Minor tweaks.
22245         (sparc_delegitimize_address): Adjust assertion accordingly.
22246         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22247         into symbolic_operand.
22248         (movsi_high_pic_label_ref): Likewise.
22249         (movsi_lo_sum_pic_label_ref): Likewise.
22250         (movdi_pic_label_ref): Likewise.
22251         (movdi_high_pic_label_ref): Likewise.
22252         (movdi_lo_sum_pic_label_ref): Likewise.
22254 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22256         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22257         custom LIB_SPEC setup.
22259 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22260             Kito Cheng  <kito.cheng@gmail.com>
22262         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22263         * config/riscv/freebsd.h: New.
22265 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22267         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22268         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22269         file.
22271 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22272             Kito Cheng  <kito.cheng@gmail.com>
22274         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22275         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22276         * config/nds32/nds32-protos.h (nds32_output_call,
22277         nds32_output_return): Declare.
22278         * config/nds32/nds32.md: Refine all the call and return patterns.
22280 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22282         PR debug/85252
22283         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22284         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22286         PR rtl-optimization/84872
22287         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22288         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22289         EDGE_CROSSING edge.
22291 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22293         * expr.c (copy_blkmode_to_reg): Revert 254862.
22294         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22296 2018-04-06  Richard Biener  <rguenther@suse.de>
22298         PR middle-end/85244
22299         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22300         after seeing a component reference with an adjacent field.  Treat
22301         refs to arrays at struct end of external decls similar to
22302         refs to unconstrained commons.
22304 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22306         PR sanitizer/85213
22307         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22308         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22309         recursive calls.
22310         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22311         save_p here.
22313 2018-04-06  Richard Biener  <rguenther@suse.de>
22315         PR middle-end/85180
22316         * alias.c (find_base_term): New wrapper around find_base_term
22317         unwinding CSELIB_VAL_PTR changes.
22318         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22319         recursion.
22321 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22323         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22324         instructions.
22325         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22326         constant definitions.
22327         ("nop"): lr 0,0 -> nopr r0
22328         ("nop_lr0", "nop_lr1"): New insn definitions.
22330 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22332         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22333         NDS32_V3PUSH_AVAILABLE_P macro.
22335 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22336             Chung-Ju Wu  <jasonwucj@gmail.com>
22338         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22339         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22340         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22341         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22342         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22343         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22344         * config/nds32/constraints.md: New constraints and checking for hard
22345         float configuration.
22346         * config/nds32/iterators.md: New mode iterator and attribute for hard
22347         float configuration.
22348         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22349         patterns.
22350         * config/nds32/nds32-fpu.md: New file.
22351         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22352         deal with hard float code generation.
22353         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22354         ARCH_V3S.
22355         (abi_type, float_reg_number): New enum type.
22356         * config/nds32/nds32-predicates.c: New predicates for hard float.
22357         * config/nds32/nds32-protos.h: Declare functions for hard float.
22358         * config/nds32/nds32.c: Implementation for hard float configuration.
22359         * config/nds32/nds32.h: Definitions for hard float configuration.
22360         * config/nds32/nds32.md: Include hard float machine description and
22361         modify patterns for hard float configuration.
22362         * config/nds32/nds32.opt: New options for hard float configuration.
22363         * config/nds32/predicates.md: New predicates for hard float
22364         configuration.
22366 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22368         * common/config/nds32/nds32-common.c
22369         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22371 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22373         PR middle-end/85195
22374         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22375         CONSTRUCTOR_ELT (ctor, ...)->value.
22377 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22379         PR target/85193
22380         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22382 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22384         PR target/85204
22385         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22386         cond jump.
22388 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22389             Kito Cheng  <kito.cheng@gmail.com>
22391         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22392         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22393         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22394         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22396 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22397             Kito Cheng  <kito.cheng@gmail.com>
22399         * config/nds32/constraints.md (Ufe): New memory constraint.
22400         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22401         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22402         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22403         operands.
22404         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22405         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22407 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22409         * config/nds32/nds32.md: Use optimize_size in the condition for
22410         alu-shift instructions.
22412 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22414         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22416 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22418         * config/nds32/nds32.md (negsi2): Refine pattern.
22420 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22421             Chung-Ju Wu  <jasonwucj@gmail.com>
22423         * config/nds32/iterators.md (shift_rotate): New code iterator.
22424         (shift): New code attribute.
22425         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22426         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22427         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22428         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22429         bit-wise operations.
22430         (andsi3, *andsi3): Ditto.
22431         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22432         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22433         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22434         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22435         nds32_ior_operand, nds32_xor_operand): New predicates.
22437 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22439         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22440         (addsi3, subsi3): ... this.
22442 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22444         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22446 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22448         * config/nds32/nds32.md: Adjust indention.
22450 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22452         * config/nds32/nds32.md (feature): New attribute.
22454 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22456         * config/nds32/nds32.md (subtype): New attribute.
22458 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22460         PR target/85203
22461         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22462         expansion to perform a bitwise AND of the argument followed by a
22463         boolean negation of the result.
22465 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22467         PR rtl-optimization/84878
22468         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22469         the basic block.  Assert the use reference is not artificial and that
22470         it has an associated insn.
22472 2018-04-04  Michael Matz  <matz@suse.de>
22474         * builtins.c (compute_objsize): Pass correct operand
22475         to array_at_struct_end_p.
22477 2018-04-04  Richard Biener  <rguenther@suse.de>
22479         PR lto/85176
22480         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22481         from contexts for DINFO_LEVEL_TERSE and below.
22483 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22485         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22486         resiter_operand condition.
22487         * config/nds32/nds32.md (*move<mode>): Ditto.
22489 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22490             Monk Chiang  <sh.chiang04@gmail.com>
22492         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22494 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22496         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22498 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22499             Kito Cheng  <kito.cheng@gmail.com>
22501         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22502         nds32_cond_code_str, output_cond_branch,
22503         output_cond_branch_compare_zero, nds32_expand_cbranch,
22504         nds32_expand_cstore, nds32_expand_movcc,
22505         nds32_output_cbranchsi4_equality_zero,
22506         nds32_output_cbranchsi4_equality_reg,
22507         nds32_output_cbranchsi4_equality_reg_or_const_int,
22508         nds32_output_cbranchsi4_greater_less_zero: New functions.
22509         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22510         nds32_expand_cstore, nds32_expand_movcc,
22511         nds32_output_cbranchsi4_equality_zero,
22512         nds32_output_cbranchsi4_equality_reg,
22513         nds32_output_cbranchsi4_equality_reg_or_const_int,
22514         nds32_output_cbranchsi4_greater_less_zero): Declare.
22515         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22516         nds32_rimm11s_operand): New predicates.
22517         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22518         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22519         patterns.
22521 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22523         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22524         * config/nds32/nds32.md: Ditto.
22525         * config/nds32/pipelines.md: Ditto.
22527 2018-04-04  Richard Biener  <rguenther@suse.de>
22529         PR tree-optimization/85168
22530         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22531         propagating abnormals.
22533 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22535         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22537 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22538             Kito Cheng  <kito.cheng@gmail.com>
22540         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22541         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22542         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22543         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22544         * config/nds32/nds32.md (sibcall_internal): New.
22545         (sibcall_register): Remove.
22546         (sibcall_immediate): Remove.
22547         (sibcall_value_internal): New.
22548         (sibcall_value_register): Remove.
22549         (sibcall_value_immediate): Remove.
22550         * config/nds32/predicates.md (nds32_general_register_operand): New.
22551         (nds32_call_address_operand): New.
22553 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22555         PR rtl-optimization/85167
22556         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22557         bb_defs if *split_p, instead preinitialize it to NULL.
22559         PR tree-optimization/85156
22560         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22561         evaluating the argument multiple times.
22563 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22565         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22566         than vector.
22567         (_mm_cvtpd_ps): Likewise.
22568         (_mm_cvttpd_epi32): Likewise.
22569         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22570         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22571         vector, pixel, and bool following altivec.h include.
22573 2018-04-03  Martin Sebor  <msebor@redhat.com>
22575         * doc/extend.texi (Common Function Attributes): Clarify.
22576         (const attribute): Likewise.
22577         (pure attribute): Likewise.
22579 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22581         PR target/85169
22582         * config/i386/i386.c (ix86_expand_vector_set): Use
22583         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22585 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22587         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22588         instructions when changing rounding bits to preserve precision bits
22589         in the x87 control word.
22591 2018-04-03  Martin Liska  <mliska@suse.cz>
22593         PR tree-optimization/82491
22594         * rtl.h (strip_offset_and_add): Replace += suboffset with
22595         poly_uint64 () + suboffset.
22597 2018-03-29  Martin Liska  <mliska@suse.cz>
22598             Martin Jambor  <mjambor@suse.cz>
22600         PR ipa/84947
22601         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22602         param_type is not an integral or pointer type.
22604 2018-04-03  Richard Biener  <rguenther@suse.de>
22606         * sese.h (recompute_all_dominators): Remove.
22608 2018-04-02  Martin Sebor  <msebor@redhat.com>
22610         * doc/invoke.texi (-Wrestrict): Fix typos.
22612 2018-04-02  Jim Wilson  <jimw@sifive.com>
22614         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22615         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22616         (<optab>di3, <optab>si3_extend): Likewise.
22617         (<optab>si3_mask, <optab>si3_mask_1): New.
22618         (<optab>di3_mask, <optab>di3_mask_1): New.
22619         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22620         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22621         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22623 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22625         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22626         example.
22628 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22630         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22631         (nds32_canonicalize_comparison): New function.
22633 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22634             Kito Cheng  <kito.cheng@gmail.com>
22635             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22637         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22638         * config/nds32/constants.md (unspec_volatile_element): Add
22639         UNSPEC_VOLATILE_RELAX_GROUP.
22640         * config/nds32/nds32-relax-opt.c: New file.
22641         * config/nds32/nds32-predicates.c
22642         (nds32_symbol_load_store_p): New function.
22643         * config/nds32/nds32-protos.h
22644         (nds32_symbol_load_store_p): Declare function.
22645         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22646         * config/nds32/nds32.c
22647         (nds32_register_pass): New function to register pass.
22648         (nds32_register_passes): New function to register passes.
22649         * config/nds32/nds32.md (relax_group): New pattern.
22650         * config/nds32/nds32.opt (mrelax-hint): New option.
22651         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22653 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22655         * config/nds32/t-nds32: Modify files dependency.
22657 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22659         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22660         (PROFILE_HOOK): Define its implementation.
22662 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22664         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22665         type and 32-bit size.
22667 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22669         PR middle-end/85090
22670         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22671         (V_128_256): New mode iterator.
22672         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22673         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22674         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22675         of V.
22676         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22677         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22679 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22681         PR target/83315
22682         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22683         NaN inputs correctly.
22685 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22687         PR target/80546
22688         * config/rs6000/vsx.md (??r): New mode attribute.
22689         (*vsx_mov<mode>_64bit): Use it.
22690         (*vsx_mov<mode>_32bit): Likewise.
22692 2018-03-30  Martin Sebor  <msebor@redhat.com>
22694         PR tree-optimization/84818
22695         * builtins.c (check_access): Use warning_n.
22697 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22699         PR target/83822
22700         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22701         condition.
22702         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22703         condition.
22705 2018-03-30  Julia Koval  <julia.koval@intel.com>
22707         PR target/84413
22708         * x86-tune.def (movx, partial_reg_dependency): Enable for
22709         m_SKYLAKE_AVX512.
22711 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22713         PR inline-asm/84985
22714         * lra-constraints.c (process_alt_operands): Move setting
22715         this_alternative_matches below.
22717 2018-03-29  Martin Liska  <mliska@suse.cz>
22719         PR lto/84995.
22720         * doc/invoke.texi: Document how LTO works with debug info.
22721         Describe auto-load support of binutils.  Mention 'x86-64'
22722         as valid option value of -march option.
22724 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
22726         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
22728         PR c/85094
22729         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
22730         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
22731         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
22732         checking.
22734 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22736         PR target/84912
22737         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
22738         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
22739         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
22740         for RS6000_BTM_POWERPC64.
22741         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
22742         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
22743         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
22744         definition.
22745         (DIVDE): Use it.
22746         (DIVDEU): Likewise.
22748 2018-03-28  Carl Love  <cel@us.ibm.com>
22750         Revert
22751         2017-09-27  Carl Love  <cel@us.ibm.com>
22753         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
22754         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
22755         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
22756         fctiw instruction.
22758 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22760         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
22761         instead of __vector bool.
22762         (_mm_max_pu8): Likewise.
22763         (_mm_min_pi16): Likewise.
22765 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22767         PR target/84912
22768         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
22769         (DIVWEUO): Likewise.
22770         (DIVDEO): Likewise.
22771         (DIVDEUO): Likewise.
22772         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
22773         DIVWEUO and DIVDEUO.
22774         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
22775         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
22776         (div_extend): Likewise.
22777         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
22778         builtin function.
22779         (__builtin_divweuo): Likewise.
22780         (__builtin_divdeo): Likewise.
22781         (__builtin_divdeuo): Likewise.
22783 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22785         PR target/85095
22786         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
22787         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
22789         PR tree-optimization/82004
22790         * gimple-match-head.c (optimize_pow_to_exp): New function.
22791         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
22792         Don't fold to exp if optimize_pow_to_exp is false.
22794 2018-03-28  Martin Liska  <mliska@suse.cz>
22796         PR other/84819
22797         * calls.c (initialize_argument_information): Fix trailing space.
22798         * common.opt: Fix typo and provide better explanation for
22799         -fsanitize-coverage option.
22800         * config/i386/i386.opt: Fix typo.
22802 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22803             Martin Liska  <mliska@suse.cz>
22805         PR sanitizer/85081
22806         * gimplify.c (asan_poison_variable): Don't do the check for
22807         gimplify_omp_ctxp here.
22808         (gimplify_decl_expr): Do it here.
22809         (gimplify_target_expr): Likewise.
22811 2018-03-28  Martin Liska  <mliska@suse.cz>
22813         PR target/84988
22814         * config/i386/i386.c (ix86_function_arg_advance): Do not call
22815         chkp_type_bounds_count if MPX is not enabled.
22817 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
22819         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
22821 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
22823         PR target/84914
22824         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
22825         function to create the function decl for complex long double
22826         multiply and divide for -mabi=ieeelongdouble.
22827         (init_float128_ieee): Call it.
22829 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
22831         PR target/85044
22832         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
22833         -fcf-protection=branch -mibt.
22834         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
22836 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22838         PR target/81863
22839         * config/arm/arm.c (arm_valid_symbolic_address): Handle
22840         arm_word_relocations.
22842 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
22844         PR target/85056
22845         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
22846         extern array declarations.
22848 2018-03-27  Richard Biener  <rguenther@suse.de>
22850         PR middle-end/84067
22851         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
22852         explicit single_use checks.
22854 2018-03-27  Richard Biener  <rguenther@suse.de>
22856         PR tree-optimization/85082
22857         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
22858         Valueize the VUSE.
22860 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22862         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
22863         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
22864         Turn on fasynchronous-unwind-tables and funwind-tables.
22866 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
22868         PR target/85073
22869         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
22870         (*bmi_blsr_<mode>_ccz): Ditto.
22872 2018-03-26  Tom de Vries  <tom@codesourcery.com>
22874         PR tree-optimization/85063
22875         * omp-general.c (offloading_function_p): New function.  Factor out
22876         of ...
22877         * omp-offload.c (pass_omp_target_link::gate): ... here.
22878         * omp-general.h (offloading_function_p): Declare.
22879         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
22880         with attribute omp declare target for offloading functions.
22882 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
22884         PR tree-optimization/84005
22885         * tree-data-ref.h (get_base_for_alignment): Declare.
22886         * tree-data-ref.c (get_base_for_alignment_1): New function.
22887         (get_base_for_alignment): Likewise.
22888         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
22889         get_base_for_alignment to find a suitable base object, instead
22890         of always using drb->base_address.
22892 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22894         PR inline-asm/85022
22895         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
22896         known size by default.
22898 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
22900         PR inline-asm/85030
22901         * lra-constraints.c (process_alt_operands): Don't match BLKmode
22902         and non BLKmode operands.
22904 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22906         PR target/85026
22907         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
22908         Clean up attributes.
22910 2018-03-23  Richard Biener  <rguenther@suse.de>
22912         PR debug/85020
22913         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
22914         we are going to emit early debug for LTO.
22916 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22918         PR inline-asm/85034
22919         * function.c (match_asm_constraints_1): Don't optimize if input
22920         doesn't satisfy general_operand predicate for output's mode.
22922         PR inline-asm/85022
22923         * alias.c (write_dependence_p): Don't require for x_canonicalized
22924         non-VOIDmode if x has VOIDmode.
22926         PR sanitizer/85029
22927         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
22928         just don't try to optimize it rather than assert it never happens.
22930 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22932         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
22933         macro expansions for definition of ST_INTERNAL_<mode> and
22934         LD_INTERNAL_<mode> builtins.
22935         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
22936         Remove prototype.
22937         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
22938         function.
22939         (altivec_expand_st_builtin): Likewise.
22940         (altivec_expand_builtin): Remove calls to deleted functions.
22941         (rs6000_address_for_altivec): Delete this function.
22942         * config/rs6000/vector.md: Remove expands for
22943         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
22945 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
22947         PR target/84826
22948         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
22949         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
22950         re-computing once computed.
22951         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
22952         (arm_init_machine_status): Initialize
22953         machine->static_chain_stack_bytes.
22955 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22957         PR target/84760
22958         * doc/extend.texi: Add four new prototypes for vec_ld.
22959         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
22960         definitions for more logical presentation.
22961         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
22962         entries for V1TI variants of __builtin_altivec_ld builtin.
22963         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
22964         handling of V1TI variant of LVX icode pattern.
22965         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
22966         (rs6000_gimple_fold_builtin): Likewise.
22967         (altivec_init_builtins): Add code to define
22968         __builtin_altivec_lvx_v1ti function.
22970 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22972         PR inline-asm/84941
22973         * function.c (match_asm_constraints_1): Don't do the optimization
22974         if input isn't a REG, SUBREG, MEM or constant.
22976 2018-03-22  Tom de Vries  <tom@codesourcery.com>
22978         PR tree-optimization/84956
22979         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
22980         bb_has_abnormal_pred.
22982 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22984         PR sanitizer/85018
22985         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
22986         DECL_INITIAL (decl) to decl at the end.
22987         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
22988         adjust the comment.
22990 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
22992         * doc/extend.texi (__builtin_tgmath): Document when complex
22993         integer types are treated as _Complex _Float64.
22995 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22997         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
22999 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23001         PR tree-optimization/84960
23002         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23003         if it is ENTRY block, move them into single succ of ENTRY in that case.
23005 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
23007         PR tree-optimization/84811
23008         * poly-int.h (poly_span_traits): Remove the T3 parameter and
23009         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23010         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23011         (known_subrange_p): Update accordingly.  Cast each value involved
23012         in the size comparison, rather than casting the result of the
23013         subtraction.
23015 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23017         PR tree-optimization/84982
23018         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23019         by flipping the least significant bit rather than all bits from
23020         bitpos to bitpos + bitsize - 1.
23022 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23024         * doc/extend.texi (Deprecated Features): Remove mention of
23025         long-deleted deprecations.
23027 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23029         PR jit/84288
23030         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23031         * configure: Regenerate.
23033 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23035         PR tree-optimization/83126
23036         * tree-parloops.c (num_phis): New function.
23037         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23039 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23041         * doc/extend.texi (Deprecated Features): Update deprecated flags,
23042         mention anon-struct/union members and trailing attributes.
23044 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
23046         PR tree-optimization/84969
23047         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23048         builtin memset partitions if they set different rhs values.
23050 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23052         PR rtl-optimization/84989
23053         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23054         VEC_DUPLICATE with scalar result mode.
23056 2018-03-21  Martin Liska  <mliska@suse.cz>
23058         PR ipa/84963
23059         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23060         not intended return statement.
23062 2018-03-21  Martin Liska  <mliska@suse.cz>
23064         PR target/84988
23065         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23066         (chkp_find_bound_slots_1): Limit number of iterations.
23068 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
23070         PR target/84838
23071         * Minor grammar fixes for x86 options.
23073 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23075         PR debug/84875
23076         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
23077         holding REG_CFA_RESTORE notes, instead turn them into a USE.
23079 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
23081         PR target/83789
23082         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
23083         (altivec_lvx_<mode>_1op): Likewise.
23084         (altivec_stvx_<mode>_2op): Likewise.
23085         (altivec_stvx_<mode>_1op): Likewise.
23086         (altivec_lvx_<VM2:mode>): New define_expand.
23087         (altivec_stvx_<VM2:mode>): Likewise.
23088         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
23089         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23090         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
23091         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23092         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
23093         (rs6000_gen_lvx): Likewise.
23094         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
23095         (altivec_expand_stv_builtin): Likewise.
23096         (altivec_expand_builtin): Likewise.
23097         * config/rs6000/vector.md: Likewise.
23099 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23101         PR target/82518
23102         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
23103         BYTES_BIG_ENDIAN.
23105 2018-03-20  Richard Biener  <rguenther@suse.de>
23107         PR target/84986
23108         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
23109         sign-conversions as zero, fall back to standard scalar_stmt
23110         cost for the rest.
23112 2018-03-20  Martin Liska  <mliska@suse.cz>
23114         PR ipa/84825
23115         * predict.c (rebuild_frequencies): Handle case when we have
23116         PROFILE_ABSENT, but flag_guess_branch_prob is false.
23118 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23120         PR target/84990
23121         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
23122         flag_section_anchors.
23123         * varasm.c (use_blocks_for_decl_p): Remove hack for
23124         dw2_force_const_mem.
23126         PR target/84845
23127         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
23128         to ...
23129         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
23130         be created, use lowpart_subreg of operands[0] rather than operands[0]
23131         itself.
23132         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23133         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23134         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23135         and n constraint instead of aarch64_shift_imm_di and Usd.
23136         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23137         (*aarch64_<optab>_reg_minus<mode>3): ... this.
23139 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
23141         PR target/82989
23142         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23143         to favor GPR over NEON registers.
23144         (<shift>di3_neon): Likewise.
23146 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23148         PR target/84952
23149         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23150         (nvptx_process_pars): Emit bar.sync asap and alap.
23152 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23154         PR target/84954
23155         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23156         seen_label if seen_label is already set.
23158 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23160         PR target/84945
23161         * config/i386/i386.c (fold_builtin_cpu): For features above 31
23162         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23163         Use 1U instead of 1.  Formatting fixes.
23165         PR c/84953
23166         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23167         instead of TREE_TYPE (s1) for the return value.
23169 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
23171         PR tree-optimization/84946
23172         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23173         bitsize + bitsize in poly_uint64 rather than poly_int64.
23175         PR sanitizer/78651
23176         * dwarf2asm.c: Include fold-const.c.
23177         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23178         of decl rather than decl itself.
23180         PR rtl-optimization/84643
23181         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23183 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
23185         PR sanitizer/78651
23186         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23187         calling assemble_variable.
23189 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
23191         PR target/81647
23192         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23193         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23195 2018-03-19  Jim Wilson  <jimw@sifive.com>
23197         PR bootstrap/84856
23198         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23199         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23200         (riscv_first_stack_step): Likewise.
23201         (riscv_option_override): Use STACK_BOUNDARY instead of
23202         MIN_STACK_BOUNDARY.
23203         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23204         MIN_STACK_BOUNDARY.
23205         (BIGGEST_ALIGNMENT): Set to 128.
23206         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23207         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23208         STACK_BOUNDARY.
23210 2018-03-19  Richard Biener  <rguenther@suse.de>
23212         PR tree-optimization/84933
23213         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23214         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23216 2018-03-19  Richard Biener  <rguenther@suse.de>
23218         PR tree-optimization/84859
23219         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23220         (cond_if_else_store_replacement): Perform sinking operation on
23221         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23222         Generalize what a BB with a single eligible store is.
23224 2018-03-19  Richard Biener  <rguenther@suse.de>
23226         PR tree-optimization/84929
23227         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23228         chrec_is_positive against non-chrec arg.
23230 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23232         PR target/84711
23233         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23235 2018-03-18  Martin Liska  <mliska@suse.cz>
23237         PR rtl-optimization/84635
23238         * regrename.c (build_def_use): Use matches_mode only when
23239         matches >= 0.
23241 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23243         PR tree-optimization/84913
23244         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23245         vectorize chains of COND_EXPRs.
23247 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23249         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23251 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23253         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23255 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23257         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23259 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23260             Kito Cheng  <kito.cheng@gmail.com>
23262         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23263         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23264         (nds32_adjust_reg_alloc_order): New function.
23265         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23267 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23269         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23270         nds32_print_operand, nds32_print_operand_address): Use
23271         HOST_WIDE_INT_PRINT_DEC instead.
23273 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23275         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23277 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23279         PR target/84902
23280         * config/i386/i386.c (initial_ix86_tune_features,
23281         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23282         unsigned long long.
23283         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23284         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23285         rather than 1u << ix86_tune.  Formatting fix.
23286         (ix86_option_override_internal): Change ix86_arch_mask from
23287         unsigned int to unsigned HOST_WIDE_INT, initialize to
23288         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23289         (ix86_function_specific_restore): Likewise.
23291 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23293         PR target/84899
23294         * postreload.c (reload_combine_recognize_pattern): Perform
23295         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23296         truncate_int_for_mode the result for the destination's mode.
23298         PR c/84909
23299         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23300         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23301         Likewise.
23302         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23303         Likewise.
23305 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23307         PR target/84876
23308         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23309         regno_allocno_class_array and sorted_pseudos.
23310         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23311         insns where regno is used.
23313 2018-03-16  Martin Liska  <mliska@suse.cz>
23315         PR ipa/84833
23316         * multiple_target.c (create_dispatcher_calls): Redirect
23317         reference in the symbol table.
23319 2018-03-16  Martin Liska  <mliska@suse.cz>
23321         PR ipa/84722
23322         * multiple_target.c (create_dispatcher_calls): Redirect also
23323         an alias.
23325 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23327         PR c++/79937
23328         PR c++/82410
23329         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23330         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23331         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23333 2018-03-16  Julia Koval  <julia.koval@intel.com>
23335         * doc/invoke.texi (Skylake Server): Add CLWB.
23336         Cannonlake): Remove CLWB.
23338 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23340         PR tree-optimization/84841
23341         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23342         1 << 3.
23343         (FLOAT_ONE_CONST_TYPE): Define.
23344         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23345         (sort_by_operand_rank): Put entries with higher constant_type last
23346         rather than first to match comments.
23348 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23350         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23351         split predicate.
23353 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23355         PR c++/79085
23356         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23357         check and use address of target always.
23359 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23361         PR target/84574
23362         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23363         (indirect_thunk_bnd_needed): Likewise.
23364         (indirect_thunks_used): Likewise.
23365         (indirect_thunks_bnd_used): Likewise.
23366         (indirect_return_needed): New.
23367         (indirect_return_bnd_needed): Likewise.
23368         (output_indirect_thunk_function): Add a bool argument for
23369         function return.
23370         (output_indirect_thunk_function): Don't generate alias for
23371         function return thunk.
23372         (ix86_code_end): Call output_indirect_thunk_function to generate
23373         function return thunks.
23374         (ix86_output_function_return): Set indirect_return_bnd_needed
23375         and indirect_return_needed instead of indirect_thunk_bnd_needed
23376         and indirect_thunk_needed.
23378 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23380         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23381         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23382         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23384 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23385             Paul Hua <paul.hua.gm@gmail.com>
23387         PR c/84852
23388         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23390 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23392         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23393         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23394         resp. SFmode cases.
23396 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23398         PR target/84711
23399         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23400         instead of GET_MODE_SIZE when comparing Units.
23402 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23404         PR target/68256
23405         * varasm.c (hash_section): Return an unchangeble hash value
23406         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23407         Return !aarch64_can_use_per_function_literal_pools_p ().
23409 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23411         PR target/84860
23412         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23413         rather than address of cmode as last argument to prepare_cmp_insn.
23415 2018-03-15  Julia Koval  <julia.koval@intel.com>
23417         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23418         F_AVX512VNNI, F_AVX512BITALG): New.
23420 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23422         PR target/83451
23423         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23424         insn for floating-point loads and stores.
23426 2018-03-14  Carl Love  <cel@us.ibm.com>
23428         * config/rs6000/rs6000-c.c: Add macro definitions for
23429         ALTIVEC_BUILTIN_VEC_PERMXOR.
23430         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23431         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23432         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23433         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23434         UNSPEC_VPERMXOR.
23435         * config/doc/extend.texi: Add prototypes for vec_permxor.
23437 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23439         PR c/84852
23440         * diagnostic-show-locus.c (class layout_point): Convert m_line
23441         from int to linenum_type.
23442         (line_span::comparator): Use linenum "compare" function when
23443         comparing line numbers.
23444         (test_line_span): New function.
23445         (layout_range::contains_point): Convert param "row" from int to
23446         linenum_type.
23447         (layout_range::intersects_line_p): Likewise.
23448         (layout::will_show_line_p): Likewise.
23449         (layout::print_source_line): Likewise.
23450         (layout::should_print_annotation_line_p): Likewise.
23451         (layout::print_annotation_line): Likewise.
23452         (layout::print_leading_fixits): Likewise.
23453         (layout::annotation_line_showed_range_p): Likewise.
23454         (struct line_corrections): Likewise for field m_row.
23455         (line_corrections::line_corrections): Likewise for param "row".
23456         (layout::print_trailing_fixits): Likewise.
23457         (layout::get_state_at_point): Likewise.
23458         (layout::get_x_bound_for_row): Likewise.
23459         (layout::print_line): Likewise.
23460         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23461         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23462         * input.c (selftest::test_linenum_comparisons): New function.
23463         (selftest::input_c_tests): Call it.
23464         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23465         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23466         * selftest.h (ASSERT_GT): New macro.
23467         (ASSERT_GT_AT): New macro.
23468         (ASSERT_LT): New macro.
23469         (ASSERT_LT_AT): New macro.
23471 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23473         PR rtl-optimization/84780
23474         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23476 2018-03-14  Martin Liska  <mliska@suse.cz>
23478         * tree.c (record_node_allocation_statistics): Use
23479         get_stats_node_kind.
23480         (get_stats_node_kind): New function extracted from
23481         record_node_allocation_statistics.
23482         (free_node): Use get_stats_node_kind.
23484 2018-03-14  Richard Biener  <rguenther@suse.de>
23486         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23487         that the value-set of ANTIC_IN doesn't grow.
23489         Revert
23490         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23491         member.
23492         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23493         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23495 2018-03-14  Julia Koval  <julia.koval@intel.com>
23497         * config.gcc (icelake-client, icelake-server): New.
23498         (icelake): Remove.
23499         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23500         (initial_ix86_arch_features): Ditto.
23501         (PTA_SKYLAKE): Add SGX.
23502         (PTA_ICELAKE): Remove.
23503         (PTA_ICELAKE_CLIENT): New.
23504         (PTA_ICELAKE_SERVER): New.
23505         (ix86_option_override_internal): Split up icelake on icelake client and
23506         icelake server.
23507         (get_builtin_code_for_version): Ditto.
23508         (fold_builtin_cpu): Ditto.
23509         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23510         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23511         * config/i386/i386.h (processor_type): Ditto.
23512         * doc/invoke.texi: Ditto.
23514 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23516         PR sanitizer/83392
23517         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23518         INTEGER_CST offset, add it together with bitpos / 8 and
23519         sign extend based on POINTER_SIZE.
23521         PR target/84844
23522         Revert
23523         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23525         PR target/78090
23526         * config/i386/constraints.md (Yc): New register constraint.
23527         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23528         Use Yc constraint for alternative 2 of operand 0.  Remove
23529         preferred_for_speed attribute.
23531 2018-03-14  Richard Biener  <rguenther@suse.de>
23533         PR tree-optimization/84830
23534         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23535         with the old one to avoid oscillations.
23537 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23539         PR target/83712
23540         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23541         pseudos.
23542         (assign_by_spills): Return a flag of reload assignment failure.
23543         Do not process the reload assignment failures.  Do not spill other
23544         reload pseudos if they has the same reg class.  Update n if
23545         necessary.
23546         (lra_assign): Add a return arg.  Set up from the result of
23547         assign_by_spills call.
23548         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23549         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23550         usage_insns if it is not NULL.
23551         (spill_hard_reg_in_range): New function.
23552         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23553         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23554         function prototypes.
23555         (lra_assign): Change prototype.
23556         * lra.c (lra): Add code to deal with fails by splitting hard reg
23557         live ranges.
23559 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23561         * config/riscv/riscv.opt (mrelax): New option.
23562         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23563         "norelax" when riscv_mrelax is disabled.
23564         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23566 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23568         PR target/84743
23569         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23570         reassociation for int modes.
23572 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23574         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23575         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23576         for big-endian.
23577         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23578         * config/aarch64/aarch64-sve.md
23579         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23580         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23581         (*extend<mode><Vwide>2): Rename to...
23582         (aarch64_sve_extend<mode><Vwide>2): ...this.
23583         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23584         renaming the old pattern to...
23585         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23586         unsigned packs.
23587         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23588         define_expand, renaming the old pattern to...
23589         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23590         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23591         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23592         account when deciding which SVE instruction the optab should use.
23593         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23595 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23597         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23598         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23599         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23600         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23601         (tlsdesc_small_<mode>): Turn a define_expand and use
23602         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23603         (tlsdesc_small_advsimd_<mode>): ...this.
23604         (tlsdesc_small_sve_<mode>): New pattern.
23606 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23608         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23609         (UNSPEC_UMUL_HIGHPART): New constants.
23610         (MUL_HIGHPART): New int iteraor.
23611         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23612         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23613         define_expand.
23614         (*<su>mul<mode>3_highpart): New define_insn.
23616 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23618         PR lto/84805
23619         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23620         incomplete types.
23622 2018-03-13  Martin Liska  <mliska@suse.cz>
23624         PR ipa/84658.
23625         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23626         vector.
23627         (sem_item_optimizer::~sem_item_optimizer): Release it.
23628         (sem_item_optimizer::merge_classes): Register variable aliases.
23629         (sem_item_optimizer::fixup_pt_set): New function.
23630         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23631         * ipa-icf.h: Declare new variables and functions.
23633 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23635         PR middle-end/84834
23636         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23637         integer_pow2p@2 and test integer_pow2p in condition.
23638         (A < 0 ? C : 0): Similarly for @1.
23640         PR middle-end/84831
23641         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23642         characters starting at p contain '\0' character, don't look beyond
23643         that.
23645         PR target/84827
23646         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23647         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23649         PR target/84828
23650         * reg-stack.c (change_stack): Change update_end var from int to
23651         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23652         also call set_block_for_insn on the newly added insns and rescan.
23654         PR target/84786
23655         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23656         on the last operand.
23658         PR c++/84704
23659         * tree.c (stabilize_reference_1): Return save_expr (e) for
23660         STATEMENT_LIST even if it doesn't have side-effects.
23662 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23664         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23666 2018-03-12  Renlin Li  <renlin.li@arm.com>
23668         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23669         aarch64_output_scalar_simd_mov_immediate.
23671 2018-03-12  Martin Sebor  <msebor@redhat.com>
23673         PR tree-optimization/83456
23674         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23675         for perfectly overlapping calls to memcpy.
23676         (gimple_fold_builtin_memory_chk): Same.
23677         (gimple_fold_builtin_strcpy): Handle no-warning.
23678         (gimple_fold_builtin_stxcpy_chk): Same.
23679         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23681 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23683         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23684         parameter.  Use it for SFmode.
23685         (rs6000_function_arg_advance_1): Adjust.
23686         (rs6000_function_arg): Adjust.
23687         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23689 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23691         PR rtl-optimization/84169
23692         PR rtl-optimization/84780
23693         * combine.c (can_combine_p): Check for a 2-insn combination whether
23694         the destination register is used between the two insns, too.
23696 2018-03-12  Richard Biener  <rguenther@suse.de>
23698         PR tree-optimization/84803
23699         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23700         for refs DR analysis didn't process.
23702 2018-03-12  Richard Biener  <rguenther@suse.de>
23704         PR tree-optimization/84777
23705         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23706         force-vectorize loops ignore whether we are optimizing for size.
23708 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23710         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23711         (TARGET_MD_ASM_ADJUST): Define.
23713 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23714             Kito Cheng  <kito.cheng@gmail.com>
23715             Chung-Ju Wu  <jasonwucj@gmail.com>
23717         * config/nds32/nds32.c (nds32_compute_stack_frame,
23718         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23719         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23720         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
23721         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
23722         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
23723         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
23724         * config/nds32/nds32.md (prologue, epilogue): Use macro
23725         NDS32_V3PUSH_AVAILABLE_P to do checking.
23727 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
23729         PR debug/58150
23730         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
23731         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
23732         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
23733         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
23734         addition of most attributes on !orig_type_die or the attribute not
23735         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
23737 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23738             Chung-Ju Wu  <jasonwucj@gmail.com>
23740         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
23741         __NDS32_VH__ macro.
23742         * config/nds32/nds32.opt (mvh): New option.
23744 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23745             Chung-Ju Wu  <jasonwucj@gmail.com>
23747         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
23748         function.
23749         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
23750         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
23751         definition.
23753 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23754             Chung-Ju Wu  <jasonwucj@gmail.com>
23756         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
23757         function.
23758         * config/nds32/nds32-multiple.md (strlensi): New pattern.
23759         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
23761 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
23762             Kito Cheng  <kito.cheng@gmail.com>
23763             Chung-Ju Wu  <jasonwucj@gmail.com>
23765         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
23766         UNSPEC_FFMISM and UNSPEC_FLMISM.
23767         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
23768         for ffb, ffmism and flmism.
23769         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
23770         (unspec_ffmism): Ditto.
23771         (unspec_flmism): Ditto.
23772         (nds32_expand_builtin_impl): Check if string extension is available.
23773         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
23774         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
23776 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
23778         Reverting patch:
23779         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23781         PR target/83712
23782         * lra-assigns.c (assign_by_spills): Return a flag of reload
23783         assignment failure.  Do not process the reload assignment
23784         failures.  Do not spill other reload pseudos if they has the same
23785         reg class.
23786         (lra_assign): Add a return arg.  Set up from the result of
23787         assign_by_spills call.
23788         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23789         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23790         usage_insns if it is not NULL.
23791         (spill_hard_reg_in_range): New function.
23792         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23793         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23794         function prototypes.
23795         (lra_assign): Change prototype.
23796         * lra.c (lra): Add code to deal with fails by splitting hard reg
23797         live ranges.
23799 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
23801         PR target/84807
23802         * config/i386/i386.opt: Replace Enforcment with Enforcement.
23804 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
23806         PR debug/84620
23807         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
23808         (dw_val_node): Add val_symbolic_view.
23809         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
23810         (symview_upper_bound): New.
23811         (new_line_info_table): Initialize symviews_since_reset.
23812         (dwarf2out_source_line): Count symviews_since_reset and set
23813         symview_upper_bound.
23814         (dw_val_equal_p): Handle symview.
23815         (add_AT_symview): New.
23816         (print_dw_val): Handle symview.
23817         (attr_checksum, attr_checksum_ordered): Likewise.
23818         (same_dw_val_p, size_of_die): Likewise.
23819         (value_format, output_die): Likewise.
23820         (add_high_low_attributes): Use add_AT_symview for entry_view.
23821         (dwarf2out_finish): Reset symview_upper_bound, clear
23822         zero_view_p.
23824 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
23826         PR target/83969
23827         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
23828         Add strict argument and use it.
23829         (rs6000_split_multireg_move): Update for new strict argument.
23830         (mem_operand_gpr): Disallow all non-offsettable addresses.
23831         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
23833 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
23835         PR target/84772
23836         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
23837         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
23838         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
23840         PR c++/84767
23841         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
23842         decl, use remap_type if we want to use the type.
23844 2018-03-09  Martin Sebor  <msebor@redhat.com>
23846         PR tree-optimization/84526
23847         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23848         Remove dead code.
23849         (builtin_access::generic_overlap): Be prepared to handle non-array
23850         base objects.
23852 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
23854         PR rtl-optimization/84682
23855         * lra-constraints.c (process_address_1): Check is_address flag
23856         for address constraints.
23857         (process_alt_operands): Likewise.
23858         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
23859         preprocess_constraints.
23860         * recog.h (preprocess_constraints): Add oploc parameter.
23861         Adjust callers.
23862         * recog.c (preprocess_constraints): Test address_operand for
23863         CT_ADDRESS constraints.
23865 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23867         PR target/83712
23868         * lra-assigns.c (assign_by_spills): Return a flag of reload
23869         assignment failure.  Do not process the reload assignment
23870         failures.  Do not spill other reload pseudos if they has the same
23871         reg class.
23872         (lra_assign): Add a return arg.  Set up from the result of
23873         assign_by_spills call.
23874         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23875         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23876         usage_insns if it is not NULL.
23877         (spill_hard_reg_in_range): New function.
23878         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23879         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23880         function prototypes.
23881         (lra_assign): Change prototype.
23882         * lra.c (lra): Add code to deal with fails by splitting hard reg
23883         live ranges.
23885 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23887         PR target/83193
23888         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
23889         Accept complain bool parameter.  Only emit errors if it is true.
23890         (arm_parse_cpu_option_name): Likewise.
23891         (arm_target_thumb_only): Adjust callers of the above.
23892         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
23893         prototype to take a default true bool parameter.
23894         (arm_parse_arch_option_name): Likewise.
23896 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
23897             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
23899         PR jit/64089
23900         PR jit/84288
23901         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
23902         * configure: Regenerate.
23903         * configure.ac ("linker --version-script option"): New.
23904         ("linker soname option"): New.
23906 2018-03-09  Richard Biener  <rguenther@suse.de>
23908         PR tree-optimization/84775
23909         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
23910         immediate uses of predicate stmts and mark them modified.
23912         Revert
23913         PR tree-optimization/84178
23914         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23915         to caller.
23916         (version_loop_for_if_conversion): Delay update_ssa call.
23917         (tree_if_conversion): Delay update_ssa until after predicate
23918         insertion.
23920 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
23922         PR target/84763
23923         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
23924         when the function accesses prior frames.
23926 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23928         PR debug/84456
23929         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
23930         gen_llsym, otherwise call maybe_gen_llsym.
23932         PR inline-asm/84742
23933         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
23934         has ',' character inside of it.
23936 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23938         PR target/84748
23939         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
23940         as clobbering CC_REGNUM.
23942 2018-03-08  Richard Biener  <rguenther@suse.de>
23944         PR middle-end/84552
23945         * tree-scalar-evolution.c: Include tree-into-ssa.h.
23946         (follow_copies_to_constant): Do not follow SSA names registered
23947         for update.
23949 2018-03-08  Richard Biener  <rguenther@suse.de>
23951         PR tree-optimization/84178
23952         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23953         to caller.
23954         (version_loop_for_if_conversion): Delay update_ssa call.
23955         (tree_if_conversion): Delay update_ssa until after predicate
23956         insertion.
23958 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
23960         PR tree-optimization/84178
23961         * tree-if-conv.c (release_bb_predicate): Remove the
23962         the assertion that the stmts have NULL use_ops.
23963         Discard the statements, asserting that they haven't
23964         yet been added to a BB.
23966 2018-03-08  Richard Biener  <rguenther@suse.de>
23968         PR tree-optimization/84746
23969         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
23970         (phi_translate): Pass in destination ANTIC_OUT set.
23971         (phi_translate_1): Likewise.  For a simplified result lookup
23972         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
23973         (phi_translate_set): Adjust.
23974         (do_pre_regular_insertion): Likewise.
23975         (do_pre_partial_partial_insertion): Likewise.
23977 2018-03-08  Martin Liska  <mliska@suse.cz>
23979         PR gcov-profile/84735
23980         * doc/gcov.texi: Document usage of profile files.
23981         * gcov-io.h: Document changes in the format.
23983 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
23985         PR debug/84404
23986         PR debug/84408
23987         * dwarf2out.c (struct dw_line_info_table): Update comments for
23988         view == -1.
23989         (FORCE_RESET_NEXT_VIEW): New.
23990         (FORCE_RESETTING_VIEW_P): New.
23991         (RESETTING_VIEW_P): Check for -1 too.
23992         (ZERO_VIEW_P): Likewise.
23993         (new_line_info_table): Force-reset next view.
23994         (dwarf2out_begin_function): Likewise.
23995         (dwarf2out_source_line): Simplify zero_view_p initialization.
23996         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
23997         view directly.  Omit view when omitting .loc at line 0.
23999 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24001         PR tree-optimization/84740
24002         * tree-switch-conversion.c (process_switch): Call build_constructors
24003         only if info.phi_count is non-zero.
24005         PR tree-optimization/84739
24006         * tree-tailcall.c (find_tail_calls): Check call arguments against
24007         DECL_ARGUMENTS (current_function_decl) rather than
24008         DECL_ARGUMENTS (func) when checking for tail recursion.
24010 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24012         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24013         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24014         Volker Reichelt's entry and add entries for people that perform
24015         GCC fuzzy testing and report numerous bugs.
24017 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
24019         PR target/82411
24020         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24021         readonly data in sdata, if that is disabled.
24022         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24023         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24024         -mreadonly-in-sdata option.
24026 2018-03-07  Martin Sebor  <msebor@redhat.com>
24028         PR tree-optimization/84468
24029         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24030         basic block when looking for nul assignment.
24032 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
24034         PR target/84277
24035         * except.h (output_function_exception_table): Adjust prototype.
24036         * except.c (output_function_exception_table): Remove FNNAME parameter
24037         and add SECTION parameter.  Ouput one part of the table at a time.
24038         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24039         the first part of the exception table and emit unwind directives.
24040         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24041         (i386_pe_seh_cold_init): Likewise.
24042         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24043         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24044         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24045         (ix86_output_call_insn): Emit a nop in one more case for SEH.
24046         * config/i386/winnt.c: Include except.h.
24047         (struct seh_frame_state): Add reg_offset, after_prologue and
24048         in_cold_section fields.
24049         (i386_pe_seh_end_prologue): Set seh->after_prologue.
24050         (i386_pe_seh_cold_init): New function.
24051         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24052         to seh->in_cold_section.
24053         (seh_emit_push): Record the offset of the push.
24054         (seh_emit_save): Record the offet of the save.
24055         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24056         Test seh->after_prologue to disregard the epilogue.
24057         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24058         (i386_pe_end_cold_function): New function.
24060 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24062         PR fortran/84565
24063         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24064         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24066         PR c++/84704
24067         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24068         on tmp_var.
24069         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
24070         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
24072         PR middle-end/84723
24073         * multiple_target.c: Include tree-inline.h and intl.h.
24074         (expand_target_clones): Diagnose and fail if node->definition and
24075         !tree_versionable_function_p (node->decl).
24077 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
24079         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
24080         sprint_ul.
24081         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
24082         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
24083         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
24085 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24087         PR target/84710
24088         * combine.c (try_combine): Use reg_or_subregno instead of handling
24089         just paradoxical SUBREGs and REGs.
24091 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
24093         * config/arc/arc.c (arc_finalize_pic): Remove function.
24094         (arc_must_save_register): We use single base PIC register, remove
24095         checks to save/restore the PIC register.
24096         (arc_expand_prologue): Likewise.
24097         * config/arc/arc-protos.h (arc_set_default_type_attributes):
24098         Remove.
24099         (arc_verify_short): Likewise.
24100         (arc_attr_type): Likewise.
24101         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
24102         (walk_stores): Likewise.
24103         (arc_address_cost): Make it static.
24104         (arc_verify_short): Likewise.
24105         (branch_dest): Likewise.
24106         (arc_attr_type): Likewise.
24107         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24108         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
24109         (arc_final_prescan_insn): Remove inserting the nops due to
24110         hardware hazards.  It is done in reorg step.
24111         (insn_length_variant_t): Remove.
24112         (insn_length_parameters_t): Likewise.
24113         (arc_insn_length_parameters): Likewise.
24114         (arc_get_insn_variants): Likewise.
24115         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
24117 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24119         PR inline-asm/84683
24120         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
24121         assertion failure.
24123         PR tree-optimization/84687
24124         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
24125         on new_node->decl.
24126         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
24128 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24130         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
24131         Rename to ppc_speculation_barrier.
24132         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24133         __builtin_ppc_speculation_barrier.
24135 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24137         PR target/84700
24138         * combine.c (combine_simplify_rtx): Don't try to simplify if
24139         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24140         are equal to x.
24142 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
24144         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24145         to 32 bytes when compiling for POWER9.
24147 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24149         PR target/84564
24150         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24151         regparm >= 3 with no arg reg available also for calls with
24152         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
24154         PR target/84524
24155         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24156         orig,vex.
24157         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
24159 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
24161         PR target/84264
24162         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24164 2018-03-05  Richard Biener  <rguenther@suse.de>
24166         PR tree-optimization/84486
24167         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24168         When inserting a __builtin_assume_aligned call set the LHS
24169         SSA name alignment info accordingly.
24171 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
24173         PR tree-optimization/84114
24174         * config/aarch64/aarch64.c (aarch64_reassociation_width)
24175         Avoid reassociation of FLOAT_MODE addition.
24177 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
24179         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24180         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24181         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24182         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24183         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24184         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24185         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24186         and -mwbnoinvd.
24187         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24188         __builtin_ia32_wbinvd): New builtins.
24189         (SPECIAL_ARGS2): New.
24190         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24191         (SPECIAL_ARGS2): New.
24192         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24193         (ix86_valid_target_attribute_inner_p): Ditto.
24194         (ix86_init_mmx_sse_builtins): Add special_args2.
24195         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24196         TARGET_WBNOINVD_P): New.
24197         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24198         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24199         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24200         * config/i386/immintrin.h (_wbinvd): New intrinsic.
24201         * config/i386/pconfigintrin.h: New file.
24202         * config/i386/wbnoinvdintrin.h: Ditto.
24203         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24204         wbnoinvdintrin.h.
24205         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24207 2018-03-05  Richard Biener  <rguenther@suse.de>
24209         PR tree-optimization/84670
24210         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24211         member.
24212         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24213         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24214         (compute_antic_aux): Only assert the number of values in ANTIC_IN
24215         doesn't grow if all successors (recursively) were visited at least
24216         once.
24218 2018-03-05  Richard Biener  <rguenther@suse.de>
24220         PR tree-optimization/84650
24221         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24222         if executed in the loop pipeline.
24224 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24226         * doc/configfiles.texi (Configuration Files): Move info about
24227         conditionalizing $target-protos.h to...
24228         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24229         differs from $target-protos.h.
24231 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24232             Chung-Ju Wu  <jasonwucj@gmail.com>
24234         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24235         * config/nds32/nds32-multiple.md (setmemsi): Define.
24236         * config/nds32/nds32-memory-manipulation.c
24237         (nds32_gen_dup_4_byte_to_word_value): New.
24238         (emit_setmem_word_loop): New.
24239         (emit_setmem_byte_loop): New.
24240         (nds32_expand_setmem_loop): New.
24241         (nds32_expand_setmem_loop_v3m): New.
24242         (nds32_expand_setmem_unroll): New.
24243         (nds32_expand_setmem): New.
24245 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24246             Chung-Ju Wu  <jasonwucj@gmail.com>
24248         * config/nds32/nds32-memory-manipulation.c
24249         (nds32_emit_load_store): New.
24250         (nds32_emit_post_inc_load_store): New.
24251         (nds32_emit_mem_move): New.
24252         (nds32_emit_mem_move_block): New.
24253         (nds32_expand_movmemsi_loop_unknown_size): New.
24254         (nds32_expand_movmemsi_loop_known_size): New.
24255         (nds32_expand_movmemsi_loop): New.
24256         (nds32_expand_movmemsi_unroll): New.
24257         (nds32_expand_movmemqi): Rename ...
24258         (nds32_expand_movmemsi): ... to this.
24259         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24260         (movmemsi): ... to this.
24261         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24262         (nds32_expand_movmemsi): ... to this.
24264 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24265             Monk Chiang  <sh.chiang04@gmail.com>
24266             Chung-Ju Wu  <jasonwucj@gmail.com>
24268         * config/nds32/nds32-protos.h
24269         (nds32_expand_load_multiple): New arguments.
24270         (nds32_expand_store_multiple): Ditto.
24271         (nds32_valid_multiple_load_store): Rename ...
24272         (nds32_valid_multiple_load_store_p): ... to this.
24273         * config/nds32/nds32-memory-manipulation.c
24274         (nds32_expand_load_multiple): Refine implementation.
24275         (nds32_expand_store_multiple): Ditto.
24276         * config/nds32/nds32-multiple.md
24277         (load_multiple): Update nds32_expand_load_multiple interface.
24278         (store_multiple): Update nds32_expand_store_multiple interface.
24279         * config/nds32/nds32-predicates.c
24280         (nds32_valid_multiple_load_store): Rename ...
24281         (nds32_valid_multiple_load_store_p): ... to this and refine
24282         implementation.
24283         * config/nds32/predicates.md
24284         (nds32_load_multiple_and_update_address_operation): New predicate.
24285         (nds32_store_multiple_and_update_address_operation): New predicate.
24287 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24288             Chung-Ju Wu  <jasonwucj@gmail.com>
24290         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24291         (combo): New attribute.
24292         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24294 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24296         * config/nds32/nds32.opt: Change -mcmodel= default value.
24298 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24299             Monk Chiang  <sh.chiang04@gmail.com>
24300             Chung-Ju Wu  <jasonwucj@gmail.com>
24302         * config/nds32/constants.md (unspec_element): New enum.
24303         * config/nds32/constraints.md (Umw): New constraint.
24304         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24305         * config/nds32/nds32-intrinsic.md: Likewise.
24306         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24307         (nds32_valid_smw_lwm_base_p): New.
24308         (nds32_output_smw_single_word): New.
24309         (nds32_output_lmw_single_word): New.
24310         (nds32_expand_unaligned_load): New.
24311         (nds32_expand_unaligned_store): New.
24312         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24313         (nds32_output_smw_single_word): Declare.
24314         (nds32_output_lmw_single_word): Declare.
24315         (nds32_expand_unaligned_load): Declare.
24316         (nds32_expand_unaligned_store): Declare.
24317         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24318         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24319         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24320         NDS32_BUILTIN_UASTORE_DW.
24321         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24322         predicate.
24324 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24325             Kito Cheng  <kito.cheng@gmail.com>
24326             Chung-Ju Wu  <jasonwucj@gmail.com>
24328         * config/nds32/nds32-intrinsic.c
24329         (nds32_expand_builtin_null_ftype_reg): Delete.
24330         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24331         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24332         (nds32_read_argument): New.
24333         (nds32_legitimize_target): Ditto.
24334         (nds32_legitimize_argument): Ditto.
24335         (nds32_check_constant_argument): Ditto.
24336         (nds32_expand_unop_builtin): Ditto.
24337         (nds32_expand_unopimm_builtin): Ditto.
24338         (nds32_expand_binop_builtin): Ditto.
24339         (nds32_builtin_decl_impl): Ditto.
24340         (builtin_description): Ditto.
24341         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24342         (nds32_init_builtins_impl): Ditto.
24343         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24344         (nds32_builtin_decl): New.
24345         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24346         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24348 2018-03-02  Jeff Law  <law@redhat.com>
24350         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24351         (redundant_insn, fill_simple_delay_slots): Likewise.
24352         (fill_slots_from_thread): Likewise.
24353         * resource.c (mark_referenced_resources): Likewise.
24354         (mark_set_resources, find_dead_or_set_registers): Likewise.
24356 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24358         * substring-locations.h (format_warning_va): Formatting fix for
24359         ATTRIBUTE_GCC_DIAG.
24360         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24361         argument.
24362         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24363         * substring-locations.c: Include intl.h.
24364         (format_warning_va): Turned into small wrapper around
24365         format_warning_n_va, renamed to ...
24366         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24367         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24368         use ngettext.
24369         (format_warning_at_substring_n): New function.
24370         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24371         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24372         format_warning_at_substring with just a shorter name instead of
24373         const function pointer.
24374         (fmtwarn_n): New function.
24375         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24376         appropriate, get rid of all the fmtstr temporaries, move conditionals
24377         with G_() wrapped string literals directly into fmtwarn arguments,
24378         cast dir.len to (int), formatting fixes.
24380 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24382         * doc/invoke.texi: Remove "Cilk Plus" references.
24384 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24385             Richard Biener  <rguenther@suse.de>
24387         PR ipa/84628
24388         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24389         for error or warning attributes if CALL_FROM_THUNK_P is set.
24390         Formatting fixes.
24392 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24394         PR target/56540
24395         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24396         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24398         PR target/56540
24399         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24400         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24402         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24403         instead of -1U in last predictors element's probability member.
24405 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24407         PR ipa/83983
24408         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24409         arguments if they are comparable.
24411 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24413         PR tree-optimization/84634
24414         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24415         masks and masked_loop_p with a single loop_masks, making sure it's
24416         null for bb vectorization.
24418 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24420         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24421         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24422         loop->force_vectorize to check whether there is no alias.
24424 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24426         PR target/84614
24427         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24428         prototypes.
24429         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24430         comments.
24431         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24432         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24433         instead of a loop around prev_real_insn.
24434         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24435         prev_real_insn.
24437         PR inline-asm/84625
24438         * config/i386/i386.c (ix86_print_operand): Use conditional
24439         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24440         zero vector.
24442 2018-03-02  Richard Biener  <rguenther@suse.de>
24444         PR tree-optimization/84427
24445         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24446         (bitmap_set_subtract_values): Rewrite to handle multiple
24447         exprs per value.
24448         (clean): Likewise.
24449         (prune_clobbered_mems): Likewise.
24450         (phi_translate): Take edge instead of pred/phiblock.
24451         (phi_translate_1): Likewise.
24452         (phi_translate_set): Likewise.  Insert all translated
24453         exprs for a value into the set, keeping possibly multiple
24454         expressions per value.
24455         (compute_antic_aux): Adjust for phi_translate changes.
24456         When intersecting union the expressions and prune those
24457         not in the final value set, keeping possibly multiple
24458         expressions per value.  Do not use value-insertion
24459         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24460         all expressions.  Add verification that the value-sets
24461         only shrink during iteration.
24462         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24463         (do_pre_regular_insertion): Likewise.
24464         (do_pre_partial_partial_insertion): Likewise.
24466 2018-03-02  Richard Biener  <rguenther@suse.de>
24468         PR target/82005
24469         * config/darwin.c (saved_debug_info_level): New static global.
24470         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24471         (darwin_asm_lto_end): Restore debug info generation settings.
24473 2018-03-01  Martin Liska  <mliska@suse.cz>
24475         PR sanitizer/82484
24476         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24477         volatile arguments.
24479 2018-03-01  Richard Biener  <rguenther@suse.de>
24481         PR debug/84645
24482         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24483         type annotation with LTO when debug was disabled at compile-time.
24485 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24487         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24488         XINT with INTVAL.
24489         (mips_final_postscan_insn): Likewise.
24491 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24493         PR rtl-optimization/84528
24494         * alias.c (init_alias_target): Add commentary.
24495         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24496         a unique base value if the frame pointer is not eliminated
24497         to the stack pointer.
24499 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24501         PR rtl-optimization/83327
24502         * lra-int.h (hard_regs_spilled_into): Declare.
24503         * lra.c (hard_regs_spilled_into): Define.
24504         (init_reg_info): Init hard_regs_spilled_into.
24505         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24506         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24507         (process_bb_lives): Handle hard_regs_spilled_into.
24508         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24509         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24511 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24513         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24514         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24515         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24516         * config/rs6000/aix72.h: New file.
24518 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24520         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24521         instead of warning_at with conditional singular and plural messages
24522         where possible.
24524         PR target/52991
24525         * stor-layout.c (update_alignment_for_field): For
24526         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24527         && !DECL_PACKED (field), do the alignment update, just use
24528         only desired_align instead of MAX (type_align, desired_align)
24529         as the alignment.
24530         (place_field): Don't do known_align < desired_align handling
24531         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24532         is non-NULL, instead do it after rli->prev_field handling and
24533         only if not within a bitfield word.  For DECL_PACKED (field)
24534         use type_align of BITS_PER_UNIT.
24536 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24538         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24539         superfluous parentheses and trailing spaces.
24541 2018-02-28  Richard Biener  <rguenther@suse.de>
24543         PR tree-optimization/84584
24544         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24545         SCoPs with fake exit edge.
24547 2018-02-28  Martin Liska  <mliska@suse.cz>
24549         PR testsuite/84597
24550         * timevar.c (timer::print): Fix format to properly print 100%
24551         values.
24553 2018-02-28  Richard Biener  <rguenther@suse.de>
24555         PR middle-end/84607
24556         * genmatch.c (capture_info::walk_match): Do not mark
24557         captured expressions without operands as expr_p given
24558         they act more like predicates and should be subject to
24559         "lost tail" side-effect preserving.
24561 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24563         PR rtl-optimization/81611
24564         * auto-inc-dec.c (attempt_change): Move dead note from
24565         mem_insn if it's the next use of regno
24566         (find_address): Take address use of reg holding
24567         non-incremented value.  Add parm to limit search to the named
24568         reg only.
24569         (merge_in_block): Attempt to use a mem insn that is the next
24570         use of the original regno.
24572 2018-02-27  Martin Sebor  <msebor@redhat.com>
24574         PR c++/83871
24575         * doc/invoke.texi (-Wmissing-attributes): New option.
24576         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24578 2018-02-27  Martin Sebor  <msebor@redhat.com>
24580         PR translation/84207
24581         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24582         n argument to unsigned HOST_WIDE_INT.
24583         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24584         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24585         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24586         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24588 2018-02-27  Richard Biener  <rguenther@suse.de>
24590         PR tree-optimization/84512
24591         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24592         Do not use the estimate returned from record_stmt_cost for
24593         the scalar iteration cost but sum properly using add_stmt_cost.
24595 2018-02-27  Richard Biener  <rguenther@suse.de>
24597         PR tree-optimization/84466
24598         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24599         Adjust last change to less strictly validate use operands.
24601 2018-02-27  Martin Liska  <mliska@suse.cz>
24603         PR gcov-profile/84548
24604         * gcov.c (process_file): Allow partial overlap and consider it
24605         also as group functions.
24606         (output_lines): Properly calculate range of lines for a group.
24608 2018-02-27  Martin Liska  <mliska@suse.cz>
24610         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24611         'ggc' suffixes.  Change first column width.
24612         (timer::print): Fix formatting of the column.
24614 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24616         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24617         preserve inline entry blocks for the sake of debug inline
24618         entry point markers alone.
24619         (remove_unused_locals): Suggest in comments a better place to
24620         force the preservation of inline entry blocks that are
24621         otherwise unused, but do not preserve them.
24623 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24625         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24627 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24629         PR target/84039
24630         * config/i386/constraints.md (Bs): Replace
24631         ix86_indirect_branch_register with
24632         TARGET_INDIRECT_BRANCH_REGISTER.
24633         (Bw): Likewise.
24634         * config/i386/i386.md (indirect_jump): Likewise.
24635         (tablejump): Likewise.
24636         (*sibcall_memory): Likewise.
24637         (*sibcall_value_memory): Likewise.
24638         Peepholes of indirect call and jump via memory: Likewise.
24639         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24640         (*sibcall_value_GOT_32): Likewise.
24641         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24642         (GOT_memory_operand): Likewise.
24643         (call_insn_operand): Likewise.
24644         (sibcall_insn_operand): Likewise.
24645         (GOT32_symbol_operand): Likewise.
24646         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24648 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24650         PR rtl-optimization/83496
24651         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24652         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24653         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24654         redundant insn, if any.
24655         (relax_delay_slots): Likewise.
24656         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24658 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24660         PR tree-optimization/83965
24661         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24662         that grouped statements are part of a reduction chain.  Return
24663         true if the statement is not marked as a reduction itself but
24664         is part of a group.
24665         (vect_recog_dot_prod_pattern): Don't check whether the statement
24666         is part of a group here.
24667         (vect_recog_sad_pattern): Likewise.
24668         (vect_recog_widen_sum_pattern): Likewise.
24670 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24672         PR debug/84545
24673         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24675 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24677         PR target/84530
24678         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24679         the bool argument.
24680         (ix86_output_indirect_function_return): New prototype.
24681         (ix86_split_simple_return_pop_internal): Likewise.
24682         * config/i386/i386.c (indirect_return_via_cx): New.
24683         (indirect_return_via_cx_bnd): Likewise.
24684         (indirect_thunk_name): Handle return va CX_REG.
24685         (output_indirect_thunk_function): Create alias for
24686         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24687         (ix86_output_indirect_jmp): Remove the bool argument.
24688         (ix86_output_indirect_function_return): New function.
24689         (ix86_split_simple_return_pop_internal): Likewise.
24690         * config/i386/i386.md (*indirect_jump): Don't pass false
24691         to ix86_output_indirect_jmp.
24692         (*tablejump_1): Likewise.
24693         (simple_return_pop_internal): Change it to define_insn_and_split.
24694         Call ix86_split_simple_return_pop_internal to split it for
24695         -mfunction-return=.
24696         (simple_return_indirect_internal): Call
24697         ix86_output_indirect_function_return instead of
24698         ix86_output_indirect_jmp.
24700 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24702         PR bootstrap/84405
24703         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24704         memset and value initialization afterwards.
24706 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24708         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24710 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24712         PR target/84521
24713         * common/config/aarch64/aarch64-common.c
24714         (aarch_option_optimization_table[]): Switch
24715         off fomit-frame-pointer
24717 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24718             Chung-Ju Wu  <jasonwucj@gmail.com>
24720         * config/nds32/nds32-multiple.md (load_multiple): Disallow
24721         volatile memory.
24722         (store_multiple): Ditto.
24724 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24726         * config.gcc: Add --with-cpu support for nds32 target.
24727         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
24728         * config/nds32/nds32.opt: Add -mcpu= option.
24730 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
24732         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
24733         isel=yes): Warn for these deprecated options.
24735 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
24737         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
24738         ISA_2_5_MASKS_EMBEDDED.
24740 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
24742         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
24743         p->max as pointers rather than using iterative_hash_expr.
24745 2018-02-23  Carl Love  <cel@us.ibm.com>
24747         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
24748         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
24749         BU_P8V_OVERLOAD_2.
24750         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
24751         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
24752         P8V_BUILTIN_VEC_VUNSIGNED2.
24754 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
24756         PR target/81572
24757         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
24758         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
24759         LRA_UNKNOWN_ALT.
24760         * lra-constraints.c (curr_insn_transform): Set up
24761         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
24762         LRA_UNKNOWN_ALT.
24763         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
24764         * lra-eliminations.c (spill_pseudos): Ditto.
24765         (process_insn_for_elimination): Ditto.
24766         * lra-lives.c (reg_early_clobber_p): Use the new macros.
24767         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
24768         LRA_NON_CLOBBERED_ALT.
24770 2018-02-22  Martin Sebor  <msebor@redhat.com>
24772         PR tree-optimization/84480
24773         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
24774         to maybe_diag_stxncpy_trunc.  Call it.
24775         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
24776         from gimple_fold_builtin_strcpy.  Print inlining stack.
24777         (handle_builtin_stxncpy): Print inlining stack.
24778         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
24780 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
24782         PR target/84176
24783         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
24784         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
24785         and -fcheck-pointer-bounds are used together.
24786         (indirect_thunk_prefix): New enum.
24787         (indirect_thunk_need_prefix): New function.
24788         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
24789         "_nt" instead of "_bnd" for NOTRACK prefix.
24790         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
24791         (output_indirect_thunk_function): Likewise.
24792         (): Likewise.
24793         (ix86_code_end): Update output_indirect_thunk_function calls.
24794         (ix86_output_indirect_branch_via_reg): Replace
24795         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
24796         (ix86_output_indirect_branch_via_push): Likewise.
24797         (ix86_output_function_return): Likewise.
24798         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
24799         incompatible with -fcf-protection=branch and
24800         -fcheck-pointer-bounds.
24802 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24804         PR target/83335
24805         * config/aarch64/aarch64.c (aarch64_print_address_internal):
24806         Change gcc_assert call to output_operand_lossage.
24808 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24810         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
24812 2018-02-22  DJ Delorie  <dj@redhat.com>
24813             Sebastian Perta  <sebastian.perta@renesas.com>
24814             Oleg Endo  <olegendo@gcc.gnu.org>
24816         * config/rx/rx.c (rx_rtx_costs): New function.
24817         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
24819 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24821         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
24823 2018-02-22  Martin Liska  <mliska@suse.cz>
24825         PR driver/83193
24826         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
24827         Add "native" as a possible value.
24829 2018-02-22  Martin Liska  <mliska@suse.cz>
24831         PR driver/83193
24832         * config/i386/i386.c (ix86_option_override_internal):
24833         Add "native" as a possible value for -march and -mtune.
24835 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
24837         PR target/84502
24838         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
24839         to all type variants.
24841         PR tree-optimization/84503
24842         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
24843         width as info->bitpos + info->bitsize - start.
24844         (merged_store_group::merge_overlapping): Simplify width computation.
24845         (check_no_overlap): New function.
24846         (imm_store_chain_info::try_coalesce_bswap): Compute expected
24847         start + width and last_order of the group, fail if check_no_overlap
24848         fails.
24849         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
24850         to group if check_no_overlap fails.
24852 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24854         * config/rs6000/altivec.md: Delete contraint arguments to
24855         define_expand, define_split, and define_peephole2, and in
24856         define_insn_and_split if always unused.
24857         * config/rs6000/darwin.md: Ditto.
24858         * config/rs6000/dfp.md: Ditto.
24859         * config/rs6000/rs6000.md: Ditto.
24860         * config/rs6000/sync.md: Ditto.
24861         * config/rs6000/vector.md: Ditto.
24862         * config/rs6000/vsx.md: Ditto.
24864 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24866         * config/rs6000/altivec.md: Write output control strings as braced
24867         blocks instead of double-quoted strings.
24868         * config/rs6000/darwin.md: Ditto.
24869         * config/rs6000/rs6000.md: Ditto.
24870         * config/rs6000/vector.md: Ditto.
24871         * config/rs6000/vsx.md: Ditto.
24873 2018-02-21  Jason Merrill  <jason@redhat.com>
24875         PR c++/84314 - ICE with templates and fastcall attribute.
24876         * attribs.c (build_type_attribute_qual_variant): Remove assert.
24878 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24880         * ipa-cp.c (determine_versionability): Fix comment typos.
24882 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24884         PR c/84229
24885         * ipa-cp.c (determine_versionability): Do not version functions caling
24886         va_arg_pack.
24888 2018-02-21  Martin Liska  <mliska@suse.cz>
24890         PR driver/83193
24891         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
24892         Add "native" as a possible value.
24893         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
24894         the macro when native cpu detection is available.
24896 2018-02-21  Martin Liska  <mliska@suse.cz>
24898         PR driver/83193
24899         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
24900         Add "native" as a possible value.
24901         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
24902         when native cpu detection is available.
24904 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
24905             Martin Sebor  <msebor@redhat.com>
24907         PR tree-optimization/84478
24908         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
24909         false.
24910         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
24911         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
24912         support which is conservatively correct, for 2 only stay conservative
24913         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
24914         argument to the 2 argument get_range_strlen, adjust 6 arg
24915         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
24916         false.
24917         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
24918         (gimple_fold_builtin_strlen): Pass true as last argument to
24919         get_range_strlen.
24921 2018-02-20  Martin Sebor  <msebor@redhat.com>
24923         PR middle-end/84095
24924         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
24925         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
24926         (builtin_memref::builtin_memref): Factor out parts into
24927         set_base_and_offset and call it.
24929 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24931         PR middle-end/84406
24932         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
24933         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
24934         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
24935         search at the associated MODE_INT.
24937 2018-02-20  Jeff Law  <law@redhat.com>
24939         PR middle-end/82123
24940         PR tree-optimization/81592
24941         PR middle-end/79257
24942         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
24943         for range data rather than using global data.
24944         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
24945         range data rather than using global data.
24946         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
24947         pass it to children as needed.
24948         (struct directive::fmtresult): Similarly.
24949         (struct directive::set_width): Similarly.
24950         (struct directive::set_precision): Similarly.
24951         (format_integer, format_directive, parse_directive): Similarly.
24952         (format_none): Accept unnamed vr_values parameter.
24953         (format_percent, format_floating, format_character): Similarly.
24954         (format_string, format_plain): Similarly.
24955         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
24956         the EVRP range analyzer for range data rather than using global data.
24957         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
24958         gimple-ssa-evrp-analyze.h
24959         (class sprintf_dom_walker): Add after_dom_children member function.
24960         Add evrp_range_analyzer member.
24961         (sprintf_dom_walker::before_dom_children): Call into the EVRP
24962         range analyzer as needed.
24963         (sprintf_dom_walker::after_dom_children): New member function.
24964         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
24965         if not optimizing.
24966         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
24967         (evrp_range_analyzer::pop_to_marker): Likewise.
24969 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24971         PR tree-optimization/84419
24972         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
24973         with the required type if its current type is compatible but
24974         different.
24976 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
24978         PR middle-end/82004
24979         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
24980         after vectorization.
24982 2018-02-20  Martin Liska  <mliska@suse.cz>
24984         PR driver/83193
24985         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
24986         possible values if we don't have a hint.
24988 2018-02-20  Martin Liska  <mliska@suse.cz>
24990         PR c/84310
24991         PR target/79747
24992         * final.c (shorten_branches): Build align_tab array with one
24993         more element.
24994         * opts.c (finish_options): Add alignment option limit check.
24995         (MAX_CODE_ALIGN): Likewise.
24996         (MAX_CODE_ALIGN_VALUE): Likewise.
24997         * doc/invoke.texi: Document maximum allowed option value for
24998         all -falign-* options.
25000 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
25002         PR target/84146
25003         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25004         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25005         * var-tracking.c (emit_note_insn_var_location): Remove all references
25006         to NOTE_INSN_CALL_ARG_LOCATION.
25007         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25008         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25009         Use copy_rtx_if_shared.
25010         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25011         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25012         (dwarf2out_var_location): Remove handling of
25013         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25014         on call_insn.
25015         * final.c (final_scan_insn): Remove all references to
25016         NOTE_INSN_CALL_ARG_LOCATION.
25017         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
25018         before dumping final insns.
25019         * except.c (emit_note_eh_region_end): Remove all references to
25020         NOTE_INSN_CALL_ARG_LOCATION.
25021         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25022         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25023         * config/arc/arc.c (hwloop_optimize): Likewise.
25024         * config/arm/arm.c (create_fix_barrier): Likewise.
25025         * config/s390/s390.c (s390_chunkify_start): Likewise.
25026         * config/sh/sh.c (find_barrier): Likewise.
25027         * config/i386/i386.c (rest_of_insert_endbranch,
25028         ix86_seh_fixup_eh_fallthru): Likewise.
25029         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25030         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25031         * config/frv/frv.c (frv_function_prologue): Likewise.
25032         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
25033         reg note.
25034         (note_outside_basic_block_p): Remove all references to
25035         NOTE_INSN_CALL_ARG_LOCATION.
25036         * gengtype.c (adjust_field_rtx_def): Likewise.
25037         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25038         Likewise.
25039         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25040         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25042         PR c++/84444
25043         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25044         is ADDR_EXPR.
25046         PR tree-optimization/84452
25047         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25048         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25049         is NULL.
25051 2018-02-19  Martin Liska  <mliska@suse.cz>
25053         PR sanitizer/82183
25054         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25056 2018-02-19  Martin Liska  <mliska@suse.cz>
25057             Richard Sandiford  <richard.sandiford@linaro.org>
25059         PR tree-optimization/82491
25060         * gimple-fold.c (get_base_constructor): Make earlier bail out
25061         to prevent ubsan.
25063 2018-02-19  Carl Love  <cel@us.ibm.com>
25065         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25066         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25067         BU_P8V_OVERLOAD_1.
25068         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25069         P8V_BUILTIN_VEC_NEG.
25071 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
25073         * config/rl78/rl78.md (movdf): New define expand.
25075 2018-02-19  Martin Liska  <mliska@suse.cz>
25077         PR other/80589
25078         * doc/invoke.texi: Fix typo.
25079         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
25081 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
25083         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
25084         handle rs6000_single_float and rs6000_double_float specially for
25085         e500 family CPUs.
25087 2018-02-16  Jeff Law  <law@redhat.com>
25089         * config/rx/rx.c (add_pop_cfi_notes): New function.;
25090         (pop_regs): Use it.
25092 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25094         PR ipa/84425
25095         * ipa-inline.c (inline_small_functions): Fix a typo.
25097 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25099         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
25101 2018-02-16  Carl Love  <cel@us.ibm.com>
25103         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
25104         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
25105         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
25106         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
25107         expansion to P8V_BUILTIN_VEC_FLOAT2.
25109 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
25111         PR rtl-optimization/70023
25112         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
25113         src_regno into account.
25115 2018-02-16  Carl Love  <cel@us.ibm.com>
25117         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
25118         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
25119         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
25120         * config/rs6000/rs6000.c: Remove case statements for
25121         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
25122         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
25123         and P9V_BUILTIN_VEC_VINSERT4B.
25124         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
25125         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
25126         * config/rs6000/vsx.md:
25127         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
25128         vec_insert4b.
25130 2018-02-16  Carl Love  <cel@us.ibm.com>
25132         * config/rs6000/altivec.h: Add builtin names vec_extract4b
25133         vec_insert4b.
25134         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25135         definitions.
25136         * config/rs6000/rs6000-c.c: Add the definitions for
25137         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25138         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25139         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25140         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
25141         definition for insert4b and define insn *insert3b_internal.
25142         * doc/extend.texi: Add documentation for vec_extract4b.
25144 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25146         * doc/extend.texi (Backwards Compatibility): Mention friend
25147         injection.  Note for-scope is deprecated.
25148         * doc/invoke.texi (-ffriend-injection): Deprecate.
25150 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
25152         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25153         that moved to I2, also allow destinations that are a paradoxical
25154         subreg (instead of a normal reg).
25156 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
25158         PR target/83831
25159         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25160         to QImode.
25162 2018-02-16  Richard Biener  <rguenther@suse.de>
25164         PR tree-optimization/84037
25165         PR tree-optimization/84016
25166         PR target/82862
25167         * config/i386/i386.c (ix86_builtin_vectorization_cost):
25168         Adjust vec_construct for the fact we need additional higher latency
25169         128bit inserts for AVX256 and AVX512 vector builds.
25170         (ix86_add_stmt_cost): Scale vector construction cost for
25171         elementwise loads.
25173 2018-02-16  Richard Biener  <rguenther@suse.de>
25175         PR tree-optimization/84417
25176         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25177         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25178         (non_rewritable_lvalue_p): Likewise, use poly-ints.
25180 2018-02-16  Martin Liska  <mliska@suse.cz>
25182         PR sanitizer/84307
25183         * internal-fn.def (ASAN_CHECK): Set proper flags.
25184         (ASAN_MARK): Likewise.
25186 2018-02-16  Julia Koval  <julia.koval@intel.com>
25188         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25189         from PTA_CANNONLAKE.
25191 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25193         PR target/84272
25194         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25195         Use ++iter rather than iter++ for std::list iterators.
25196         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
25197         defer deleting them until all nodes in the forest are processed.  Do
25198         free even leaf nodes.  Change to_process into auto_vec.
25200         PR bootstrap/84405
25201         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25202         * vec.h (vec_default_construct): Use memset instead of placement new
25203         if BROKEN_VALUE_INITIALIZATION is defined.
25204         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25205         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25206         is defined.
25208         PR rtl-optimization/83723
25209         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25210         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
25211         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
25212         recursive calls.
25213         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25214         callers.
25215         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25217 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
25219         PR rtl-optimization/81443
25220         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25221         from inner REGs to paradoxical SUBREGs.
25223 2018-02-16  Richard Biener  <rguenther@suse.de>
25225         PR tree-optimization/84399
25226         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25227         For operands we can analyze at their definition make sure we can
25228         analyze them at each use as well.
25230 2018-02-16  Richard Biener  <rguenther@suse.de>
25232         PR tree-optimization/84190
25233         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25234         volatile accesses if the decl isn't volatile.
25236 2018-02-15  Jason Merrill  <jason@redhat.com>
25238         PR c++/84314 - ICE with templates and fastcall attribute.
25239         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25240         TYPE_CANONICAL on an existing type.
25242 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25244         PR tree-optimization/84383
25245         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25246         dstoff nor call operand_equal_p if dstbase is NULL.
25248         PR tree-optimization/84334
25249         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25250         also a CONSTANT_CLASS_P, punt.
25252 2018-02-14  Jim Wilson  <jimw@sifive.com>
25254         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25255         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25256         to where locals are set.  Add TARGET_RVC support.
25257         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25259 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25261         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25263 2018-02-14  Martin Sebor  <msebor@redhat.com>
25265         PR tree-optimization/83698
25266         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25267         arrays constrain the offset range to their bounds.
25268         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25269         (builtin_access::overlap): Avoid setting the size of overlap if it's
25270         already been set.
25271         (maybe_diag_overlap): Also consider arrays when deciding what values
25272         of offsets to include in diagnostics.
25274 2018-02-14  Martin Sebor  <msebor@redhat.com>
25276         PR c/84108
25277         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25278         that correspond to the kind of a declaration.
25280 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25282         PR target/83984
25283         * config/pa/pa.md: Load address of PIC label using the linkage table
25284         if the label is nonlocal.
25286 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25288         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25289         warning message if user requests -maltivec=be.
25290         * doc/invoke.texi: Document deprecation of -maltivec=be.
25292 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25294         PR target/84220
25295         * config/rs6000/rs6000-c.c: Update definitions for
25296         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25297         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25299 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25301         PR target/84239
25302         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25303         add _get_ssp intrinsics. Remove argument from
25304         __builtin_ia32_rdssp[d|q].
25305         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25306         * config/i386/i386-builtin.def: Remove argument from
25307         __builtin_ia32_rdssp[d|q].
25308         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25309         ix86_expand_special_args_builtin for _rdssp[d|q].
25310         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25311         Clear register before usage.
25312         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25313         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25315 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25317         PR tree-optimization/84357
25318         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25319         operand 1 of an ARRAY_REF too.
25321 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25323         PR target/83831
25324         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25325         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25326         declarations.
25327         (set_of_reg): New struct.
25328         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25329         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25330         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25331         functions.
25332         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25333         Split into bitclr, bitset, bitinvert patterns if appropriate.
25334         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25335         use rx_fuse_in_memory_bitop.
25336         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25337         to named insn, correct maximum insn length.
25339 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25341         PR target/79242
25342         * machmode.def: Define a complex mode for PARTIAL_INT.
25343         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25344         MODE_PARTIAL_INT.
25345         * doc/rtl.texi: Document CSPImode.
25346         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25347         handling.
25348         (msp430_hard_regno_nregs_with_padding): Likewise.
25350 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25352         PR target/84279
25353         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25355 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25357         PR rtl-optimization/84169
25358         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25359         we generated a parallel as new i3 and we split that to new i2 and i3
25360         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25361         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25362         those to i2, not i1.  Partially rewrite this scan code.
25364 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25366         PR c/82210
25367         * stor-layout.c (place_field): For variable length fields, adjust
25368         offset_align afterwards not just based on the field's alignment,
25369         but also on the size.
25371         PR middle-end/84309
25372         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25373         of exps and logs in the use_exp2 case.
25375 2018-02-13  Jeff Law  <law@redhat.com>
25377         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25378         entry for "vector".
25380         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25381         ARGS as unused.
25383 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25385         PR debug/84342
25386         PR debug/84319
25387         * common.opt (gas-loc-support, gas-locview-support): New.
25388         (ginline-points, ginternal-reset-location-views): New.
25389         * doc/invoke.texi: Document them.  Use @itemx where intended.
25390         (gvariable-location-views): Adjust.
25391         * target.def (reset_location_view): New.
25392         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25393         (TARGET_RESET_LOCATION_VIEW): New.
25394         * doc/tm.texi: Rebuilt.
25395         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25396         (dwarf2out_default_as_locview_support): New.
25397         (output_asm_line_debug_info): Use option variables.
25398         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25399         (output_loc_list): Likewise.
25400         (add_high_low_attributes): Check option variables.
25401         Don't output entry view attribute in strict mode.
25402         (gen_inlined_subroutine_die): Check option variables.
25403         (dwarf2out_inline_entry): Likewise.
25404         (init_sections_and_labels): Likewise.
25405         (dwarf2out_early_finish): Likewise.
25406         (maybe_reset_location_view): New, from...
25407         (dwarf2out_var_location): ... here.  Call it.
25408         * debug.h (dwarf2out_default_as_loc_support): Declare.
25409         (dwarf2out_default_as_locview_support): Declare.
25410         * hooks.c (hook_int_rtx_insn_0): New.
25411         * hooks.h (hook_int_rtx_insn_0): Declare.
25412         * toplev.c (process_options): Take -gas-loc-support and
25413         -gas-locview-support from dwarf2out.  Enable
25414         -gvariable-location-views by default only with locview
25415         assembler support.  Enable -ginternal-reset-location-views by
25416         default only if the target defines the corresponding hook.
25417         Enable -ginline-points by default if location views are
25418         enabled; force it disabled if statement frontiers are
25419         disabled.
25420         * tree-inline.c (expand_call_inline): Check option variables.
25421         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25423 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25425         PR tree-optimization/84321
25426         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25427         handling.  Also check whether the anti-range contains any values
25428         that satisfy the mask; switch to a VR_RANGE if not.
25430 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25432         PR sanitizer/84340
25433         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25435 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25437         PR c++/83990
25438         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25439         of call statements, also set location of a load to a temporary.
25441 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25443         * config/rl78/rl78.c (add_vector_labels): New function.
25444         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25445         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25446         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25447         which checks that no arguments are passed.
25448         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25449         * doc/extend.texi: Documentation for the new attribute.
25451 2018-02-13  Andreas Schwab  <schwab@suse.de>
25453         * config/riscv/linux.h (CPP_SPEC): Define.
25455 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25457         PR target/84335
25458         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25459         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25460         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25461         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25462         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25463         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25464         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25465         temporarily for AES and PCLMUL builtins.
25467         PR tree-optimization/84339
25468         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25469         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25470         Formatting fixes.
25472         PR middle-end/84309
25473         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25474         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25475         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25476         inline function.
25477         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25478         inline function.
25479         * omp-simd-clone.h: New file.
25480         * omp-simd-clone.c: Include omp-simd-clone.h.
25481         (expand_simd_clones): No longer static.
25482         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25483         cgraph.h and omp-simd-clone.h.
25484         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25485         (vect_recog_widen_shift_pattern): Formatting fix.
25486         (vect_pattern_recog_1): Don't check optab for calls.
25488         PR target/84336
25489         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25490         operands[2] into a REG before using gen_lowpart on it.
25492 2018-02-12  Jeff Law  <law@redhat.com>
25494         PR target/83760
25495         * config/sh/sh.c (find_barrier): Consider a sibling call
25496         a barrier as well.
25498         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25499         successfully back substituting a reg.
25501 2018-02-12  Richard Biener  <rguenther@suse.de>
25503         PR tree-optimization/84037
25504         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25505         parameter, move visited init to caller.
25506         (vect_slp_analyze_operations): Separate cost from validity
25507         check, initialize visited once for all instances.
25508         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25509         for all instances.
25510         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25511         out an assert.
25512         (vect_model_promotion_demotion_cost): Likewise.
25513         (vectorizable_bswap): Guard cost modeling with !slp_node
25514         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25515         SLP stmts.
25516         (vectorizable_call): Likewise.
25517         (vectorizable_conversion): Likewise.
25518         (vectorizable_assignment): Likewise.
25519         (vectorizable_shift): Likewise.
25520         (vectorizable_operation): Likewise.
25521         (vectorizable_store): Likewise.
25522         (vectorizable_load): Likewise.
25523         (vectorizable_condition): Likewise.
25524         (vectorizable_comparison): Likewise.
25526 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25528         PR sanitizer/84307
25529         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25530         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25531         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25533 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25535         PR middle-end/83665
25536         * params.def (inline-min-speedup): Increase from 8 to 15.
25537         (max-inline-insns-auto): Decrease from 40 to 30.
25538         * ipa-split.c (consider_split): Add some buffer for function to
25539         be considered inlining candidate.
25540         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25541         default values.
25543 2018-02-12  Richard Biener  <rguenther@suse.de>
25545         PR tree-optimization/84037
25546         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25547         matched stmts if we cannot swap the non-matched ones.
25549 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25551         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25552         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25553         _mm_maskz_scalef_round_ss): New intrinsics.
25554         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25555         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25556         __builtin_ia32_scalefss_round): Remove.
25557         (__builtin_ia32_scalefsd_mask_round,
25558         __builtin_ia32_scalefss_mask_round): New intrinsics.
25559         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25560         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25561         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25562         "<round_constraint>")): Changed to ...
25563         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25564         "<round_scalar_constraint>")): ... this.
25565         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25566         %0, %1, %2<round_op3>}"): Changed to ...
25567         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25568         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25569         %2<round_scalar_mask_op3>}"): ... this.
25570         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25572 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25574         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25575         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25576         (_mm_maskz_sqrt_round_ss): New intrinsics.
25577         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25578         (__builtin_ia32_sqrtsd_mask_round)
25579         (__builtin_ia32_sqrtss_mask_round): New builtins.
25580         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25581         (__builtin_ia32_sqrtss_round): Remove.
25582         (__builtin_ia32_sqrtsd_mask_round)
25583         (__builtin_ia32_sqrtss_mask_round): New builtins.
25584         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25585         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25586         ((match_operand:VF_128 1 "vector_operand"
25587         "xBm,<round_constraint>")): Changed to ...
25588         ((match_operand:VF_128 1 "vector_operand"
25589         "xBm,<round_scalar_constraint>")): ... this.
25590         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25591         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25592         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25593         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25594         %<iptr>1<round_scalar_mask_op3>}): ... this.
25595         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25596         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25598 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25600         PR target/84266
25601         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25602         Cast vec_cmpeq result to correct type.
25603         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25604         Cast vec_cmpgt result to correct type.
25606 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25608         * final.c (final_scan_insn_1): Renamed from...
25609         (final_scan_insn): ... this.  New wrapper, to recover
25610         seen from the outermost call in recursive ones.
25611         * config/sparc/sparc.c (output_return): Drop seen from call.
25612         (output_sibcall): Likewise.
25613         * config/visium/visium.c (output_branch): Likewise.
25615 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25617         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25618         function label.
25620 2018-02-10  Alan Modra  <amodra@gmail.com>
25622         PR target/84300
25623         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25624         Specify LR as an input.
25626 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25628         PR sanitizer/83987
25629         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25630         remove_member_access_dummy_vars): New functions.
25631         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25632         lower_omp_1, execute_lower_omp): Use them.
25634         PR rtl-optimization/84308
25635         * shrink-wrap.c (spread_components): Release todo vector.
25637 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25639         PR rtl-optimization/57193
25640         * ira-color.c (struct allocno_color_data): Add member
25641         conflict_allocno_hard_prefs.
25642         (update_conflict_allocno_hard_prefs): New.
25643         (bucket_allocno_compare_func): Add a preference based on
25644         conflict_allocno_hard_prefs.
25645         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25646         (color_allocnos): Remove a dead code.  Initiate
25647         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25649 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25651         PR target/84226
25652         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25653         constraint from =wa to wa.  Avoid a subreg on the output operand,
25654         instead use a pseudo and subreg it in a move.
25655         (p9_xxbrd_<mode>): Changed to ...
25656         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25657         (p9_xxbrd_v2df): New expander.
25658         (p9_xxbrw_<mode>): Changed to ...
25659         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25660         (p9_xxbrw_v4sf): New expander.
25662 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25664         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25666 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25668         PR target/83926
25669         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25670         multiply in 32-bit mode.
25671         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25672         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25673         mode.
25675 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25677         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25678         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25679         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25680         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25682 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25684         PR lto/84213
25685         * dwarf2out.c (is_trivial_indirect_ref): New function.
25686         (dwarf2out_late_global_decl): Do not generate a location
25687         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25688         and that are not defined in the current unit.
25690 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25692         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25693         instead of a libcall for UNORDERED.
25695 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25697         PR target/82641
25698         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25699         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25701 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25703         PR target/PR84295
25704         * config/s390/s390.c (s390_set_current_function): Invoke
25705         s390_indirect_branch_settings also if fndecl didn't change.
25707 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25709         * config/rs6000/rs6000.md (blockage): Set length to zero.
25711 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25713         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25715 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25717         PR sanitizer/84285
25718         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25719         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25720         -static-lib*san.
25722         PR debug/84252
25723         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
25724         PARALLEL incoming that failed vt_get_decl_and_offset check.
25726         PR middle-end/84237
25727         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
25728         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
25729         TREE_READONLY bit.
25730         (get_variable_section): For decls in named .bss* sections pass true as
25731         second argument to bss_initializer_p.
25733 2018-02-09  Marek Polacek  <polacek@redhat.com>
25734             Jakub Jelinek  <jakub@redhat.com>
25736         PR c++/83659
25737         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
25738         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
25739         Sync some changes from cxx_fold_indirect_ref.
25741 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25743         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
25744         markers.
25745         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
25746         (BLOCK_INLINE_ENTRY_LABEL): New.
25747         (dwarf2out_var_location): Disregard inline entry markers.
25748         (inline_entry_data): New struct.
25749         (inline_entry_data_hasher): New hashtable type.
25750         (inline_entry_data_hasher::hash): New.
25751         (inline_entry_data_hasher::equal): New.
25752         (inline_entry_data_table): New variable.
25753         (add_high_low_attributes): Add DW_AT_entry_pc and
25754         DW_AT_GNU_entry_view attributes if a pending entry is found
25755         in inline_entry_data_table.  Add old entry_pc attribute only
25756         if debug nonbinding markers are disabled.
25757         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
25758         markers are enabled.
25759         (block_within_block_p, dwarf2out_inline_entry): New.
25760         (dwarf2out_finish): Check that no entries remained in
25761         inline_entry_data_table.
25762         * final.c (reemit_insn_block_notes): Handle inline entry notes.
25763         (final_scan_insn, notice_source_line): Likewise.
25764         (rest_of_clean_state): Skip inline entry markers.
25765         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
25766         markers.
25767         * gimple.c (gimple_build_debug_inline_entry): New.
25768         * gimple.h (enum gimple_debug_subcode): Add
25769         GIMPLE_DEBUG_INLINE_ENTRY.
25770         (gimple_build_debug_inline_entry): Declare.
25771         (gimple_debug_inline_entry_p): New.
25772         (gimple_debug_nonbind_marker_p): Adjust.
25773         * insn-notes.def (INLINE_ENTRY): New.
25774         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
25775         inline entry marker notes.
25776         (print_insn): Likewise.
25777         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
25778         (INSN_DEBUG_MARKER_KIND): Likewise.
25779         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
25780         * tree-inline.c (expand_call_inline): Build and insert
25781         debug_inline_entry stmt.
25782         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
25783         inline entry blocks early, if nonbind markers are enabled.
25784         (dump_scope_block): Dump fragment info.
25785         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
25786         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
25787         (gimple_build_debug_inline_entry): New.
25788         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
25789         Enable/disable inline entry points too.
25790         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
25791         (DEBUG_INSN): Describe inline entry markers.
25793         * common.opt (gvariable-location-views): New.
25794         (gvariable-location-views=incompat5): New.
25795         * config.in: Rebuilt.
25796         * configure: Rebuilt.
25797         * configure.ac: Test assembler for view support.
25798         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
25799         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
25800         * dwarf2out.c (var_loc_view): New typedef.
25801         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
25802         (dwarf2out_locviews_in_attribute): New.
25803         (dwarf2out_locviews_in_loclist): New.
25804         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
25805         (enum dw_line_info_opcode): Add LI_adv_address.
25806         (struct dw_line_info_table): Add view.
25807         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
25808         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
25809         (zero_view_p): New variable.
25810         (ZERO_VIEW_P): New macro.
25811         (output_asm_line_debug_info): New.
25812         (struct var_loc_node): Add view.
25813         (add_AT_view_list, AT_loc_list): New.
25814         (add_var_loc_to_decl): Add view param.  Test it against last.
25815         (new_loc_list): Add view params.  Record them.
25816         (AT_loc_list_ptr): Handle loc and view lists.
25817         (view_list_to_loc_list_val_node): New.
25818         (print_dw_val): Handle dw_val_class_view_list.
25819         (size_of_die): Likewise.
25820         (value_format): Likewise.
25821         (loc_list_has_views): New.
25822         (gen_llsym): Set vl_symbol too.
25823         (maybe_gen_llsym, skip_loc_list_entry): New.
25824         (dwarf2out_maybe_output_loclist_view_pair): New.
25825         (output_loc_list): Output view list or entries too.
25826         (output_view_list_offset): New.
25827         (output_die): Handle dw_val_class_view_list.
25828         (output_dwarf_version): New.
25829         (output_compilation_unit_header): Use it.
25830         (output_skeleton_debug_sections): Likewise.
25831         (output_rnglists, output_line_info): Likewise.
25832         (output_pubnames, output_aranges): Update version comments.
25833         (output_one_line_info_table): Output view numbers in asm comments.
25834         (dw_loc_list): Determine current endview, pass it to new_loc_list.
25835         Call maybe_gen_llsym.
25836         (loc_list_from_tree_1): Adjust.
25837         (add_AT_location_description): Create view list attribute if
25838         needed, check it's absent otherwise.
25839         (convert_cfa_to_fb_loc_list): Adjust.
25840         (maybe_emit_file): Call output_asm_line_debug_info for test.
25841         (dwarf2out_var_location): Reset views as needed.  Precompute
25842         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
25843         attribute.  Set view.
25844         (new_line_info_table): Reset next view.
25845         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
25846         (dwarf2out_source_line): Likewise.  Output view resets and labels to
25847         the assembler, or select appropriate line info opcodes.
25848         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
25849         (optimize_string_length): Catch it.  Adjust.
25850         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
25851         dw_val_class_view_list, and remove it if no longer needed.
25852         (hash_loc_list): Hash view numbers.
25853         (loc_list_hasher::equal): Compare them.
25854         (optimize_location_lists): Check whether a view list symbol is
25855         needed, and whether the locview attribute is present, and
25856         whether they match.  Remove the locview attribute if no longer
25857         needed.
25858         (index_location_lists): Call skip_loc_list_entry for test.
25859         (dwarf2out_finish): Call output_asm_line_debug_info for test.
25860         Use output_dwarf_version.
25861         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
25862         (struct dw_val_node): Add val_view_list.
25863         * final.c (SEEN_NEXT_VIEW): New.
25864         (set_next_view_needed): New.
25865         (clear_next_view_needed): New.
25866         (maybe_output_next_view): New.
25867         (final_start_function): Rename to...
25868         (final_start_function_1): ... this.  Take pointer to FIRST,
25869         add SEEN parameter.  Emit param bindings in the initial view.
25870         (final_start_function): Reintroduce SEEN-less interface.
25871         (final): Rename to...
25872         (final_1): ... this.  Take SEEN parameter.  Output final pending
25873         next view at the end.
25874         (final): Reintroduce seen-less interface.
25875         (final_scan_insn): Output pending next view before switching
25876         sections or ending a block.  Mark the next view as needed when
25877         outputting variable locations.  Notify debug backend of section
25878         changes, and of location view changes.
25879         (rest_of_handle_final): Adjust.
25880         * toplev.c (process_options): Autodetect value for debug variable
25881         location views option.  Warn on incompat5 without -gdwarf-5.
25882         * doc/invoke.texi (gvariable-location-views): New.
25883         (gvariable-location-views=incompat5): New.
25884         (gno-variable-location-views): New.
25886 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
25888         PR tree-optimization/84136
25889         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
25890         that the result of find_edge is non-NULL.
25892 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25894         PR target/83008
25895         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
25896         storing integer register in SImode.  Fix cost of 256 and 512
25897         byte aligned SSE register store.
25899 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25901         * config/i386/i386.c (ix86_multiplication_cost): Fix
25902         multiplication cost for TARGET_AVX512DQ.
25904 2018-02-08  Marek Polacek  <polacek@redhat.com>
25906         PR tree-optimization/84238
25907         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
25908         get_range_strlen.
25910 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25912         PR tree-optimization/84265
25913         * tree-vect-stmts.c (vectorizable_store): Don't treat
25914         VMAT_CONTIGUOUS accesses as grouped.
25915         (vectorizable_load): Likewise.
25917 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25919         PR tree-optimization/81635
25920         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
25921         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
25922         (test_round_for_mask): New functions.
25923         (wide_int_cc_tests): Call test_round_for_mask.
25924         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
25925         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
25926         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
25927         range returned by get_range_info.
25929 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25931         PR ipa/81360
25932         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
25933         * symtab.c: Include builtins.h
25934         (symtab_node::output_to_lto_symbol_table_p): Move here
25935         from lto-streamer-out.c:output_symbol_p.
25936         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
25937         (output_symbol_p): Move all logic to symtab.c
25938         (produce_symtab): Update.
25940 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25942         * config/s390/s390-opts.h (enum indirect_branch): Define.
25943         * config/s390/s390-protos.h (s390_return_addr_from_memory)
25944         (s390_indirect_branch_via_thunk)
25945         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
25946         (enum s390_indirect_branch_type): Define.
25947         * config/s390/s390.c (struct s390_frame_layout, struct
25948         machine_function): Remove.
25949         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
25950         (indirect_branch_table_label_no, indirect_branch_table_name):
25951         Define variables.
25952         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
25953         (enum s390_indirect_branch_option): Define.
25954         (s390_return_addr_from_memory): New function.
25955         (s390_handle_string_attribute): New function.
25956         (s390_attribute_table): Add new attribute handler.
25957         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
25958         (s390_indirect_branch_via_thunk): New function.
25959         (s390_indirect_branch_via_inline_thunk): New function.
25960         (s390_function_ok_for_sibcall): When jumping via thunk disallow
25961         sibling call optimization for non z10 compiles.
25962         (s390_emit_call): Force indirect branch target to be a single
25963         register.  Add r1 clobber for non-z10 compiles.
25964         (s390_emit_epilogue): Emit return jump via return_use expander.
25965         (s390_reorg): Handle JUMP_INSNs as execute targets.
25966         (s390_option_override_internal): Perform validity checks for the
25967         new command line options.
25968         (s390_indirect_branch_attrvalue): New function.
25969         (s390_indirect_branch_settings): New function.
25970         (s390_set_current_function): Invoke s390_indirect_branch_settings.
25971         (s390_output_indirect_thunk_function):  New function.
25972         (s390_code_end): Implement target hook.
25973         (s390_case_values_threshold): Implement target hook.
25974         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
25975         macros.
25976         * config/s390/s390.h (struct s390_frame_layout)
25977         (struct machine_function): Move here from s390.c.
25978         (TARGET_INDIRECT_BRANCH_NOBP_RET)
25979         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
25980         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
25981         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
25982         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
25983         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
25984         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
25985         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
25986         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
25987         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
25988         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
25989         (mnemonic attribute): Add values which aren't recognized
25990         automatically.
25991         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
25992         pattern for branch conversion.  Fix mnemonic attribute.
25993         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
25994         indirect branch via thunk if requested.
25995         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
25996         ("*indirect_jump"): Disable for branch conversion using out of
25997         line thunks.
25998         ("indirect_jump_via_thunk<mode>_z10")
25999         ("indirect_jump_via_thunk<mode>")
26000         ("indirect_jump_via_inlinethunk<mode>_z10")
26001         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26002         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26003         ("casesi_jump_via_inlinethunk<mode>_z10")
26004         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26005         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26006         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26007         ("*indirect2_jump"): Disable for branch conversion.
26008         ("casesi_jump"): Turn into expander and expand patterns for branch
26009         conversion.
26010         ("return_use"): New expander.
26011         ("*return"): Emit return via thunk and rename it to ...
26012         ("*return<mode>"): ... this one.
26013         * config/s390/s390.opt: Add new options and and enum for the
26014         option values.
26016 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26018         * lra-constraints.c (match_reload): Unconditionally use
26019         gen_lowpart_SUBREG, rather than selecting between that
26020         and equivalent gen_rtx_SUBREG code.
26022 2018-02-08  Richard Biener  <rguenther@suse.de>
26024         PR tree-optimization/84233
26025         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26026         changed flag instead of boguously re-using phi_inserted.
26028 2018-02-08  Martin Jambor  <mjambor@suse.cz>
26030         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26031         static local variables.
26033 2018-02-08  Richard Biener  <rguenther@suse.de>
26035         PR tree-optimization/84278
26036         * tree-vect-stmts.c (vectorizable_store): When looking for
26037         smaller vector types to perform grouped strided loads/stores
26038         make sure the mode is supported by the target.
26039         (vectorizable_load): Likewise.
26041 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26043         * config/aarch64/aarch64.c (aarch64_components_for_bb):
26044         Increase LDP/STP opportunities by adding adjacent callee-saves.
26046 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26048         PR rtl-optimization/84068
26049         PR rtl-optimization/83459
26050         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26052 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
26054         PR tree-optimization/84224
26055         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26056         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26057         non-zero arguments.
26059 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
26061         PR target/84113
26062         * config/rs6000/altivec.md (*restore_world): Remove LR use.
26063         * config/rs6000/predicates.md (restore_world_operation): Adjust op
26064         count, remove one USE.
26066 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
26068         * doc/install.texi (Configuration): Document the
26069         --with-long-double-format={ibm,ieee} PowerPC configuration
26070         options.
26072         PR target/84154
26073         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
26074         Convert from define_expand to be define_insn_and_split.  Rework
26075         float/double/_Float128 conversions to QI/HI/SImode to work with
26076         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
26077         conversions to QI/HImode types did a store and then a load to
26078         truncate the value.  For conversions to VSX registers, don't split
26079         the insn, instead emit the code directly.  Use the code iterator
26080         any_fix to combine signed and unsigned conversions.
26081         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
26082         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
26083         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26084         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
26085         (fix_<mode>di2_hw): Likewise.
26086         (fixuns_<mode>di2_hw): Likewise.
26087         (fix_<mode>si2_hw): Likewise.
26088         (fixuns_<mode>si2_hw): Likewise.
26089         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
26090         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26091         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
26092         fix<uns>_trunc<SFDF:mode>si2_p8.
26093         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
26094         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
26095         (fix<uns>_<mode>_mem): Likewise.
26096         (fctiw<u>z_<mode>_mem): Likewise.
26097         (fix<uns>_<mode>_mem): Likewise.
26098         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
26099         the register allocator from doing a direct move to the GPRs to do
26100         a store, and instead use the ISA 3.0 store byte/half-word from
26101         vector register instruction.  For IEEE 128-bit floating point,
26102         also optimize stores of 32-bit ints.
26103         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
26105 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
26107         * genextract.c (push_pathstr_operand): New function to support
26108         [a-zA-Z].
26109         (walk_rtx): Call push_pathstr_operand.
26110         (print_path): Support [a-zA-Z].
26112 2018-02-07  Richard Biener  <rguenther@suse.de>
26114         PR tree-optimization/84037
26115         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
26116         (cse_and_gimplify_to_preheader): Declare.
26117         (vect_get_place_in_interleaving_chain): Likewise.
26118         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26119         ivexpr_map.
26120         (_loop_vec_info::~_loop_vec_info): Delete it.
26121         (cse_and_gimplify_to_preheader): New function.
26122         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
26123         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
26124         (vectorizable_load): Likewise.  For grouped stores always base
26125         the IV on the first element.
26126         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
26127         condition before gimplifying.
26129 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26131         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26132         *DIV_EXPR and *MOD_EXPR.
26134 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
26136         PR target/84248
26137         * config/i386/i386.c (ix86_option_override_internal): Mask out
26138         the CF_SET bit when checking -fcf-protection.
26140 2018-02-07  Tom de Vries  <tom@codesourcery.com>
26142         PR libgomp/84217
26143         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26144         enough.
26146 2018-02-07  Richard Biener  <rguenther@suse.de>
26148         PR tree-optimization/84204
26149         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26150         this place.
26152         PR tree-optimization/84205
26153         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26154         special-case isl_ast_op_zdiv_r.
26156         PR tree-optimization/84223
26157         * graphite-scop-detection.c (gather_bbs::before_dom_children):
26158         Only add conditions from within the region.
26159         (gather_bbs::after_dom_children): Adjust.
26161 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
26163         PR target/84209
26164         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26165         * config/avr/avr.md: Only post-reload split REG-REG moves if
26166         either register is GENERAL_REG_P.
26168 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26170         PR tree-optimization/84235
26171         * tree-ssa-scopedtables.c
26172         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26173         if the subtraction is performed in floating point type where NaNs are
26174         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26175         build 1.  Formatting fix.
26177 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
26179         PR target/84146
26180         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26181         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26182         and skip it regardless of bb boundaries.  Use CALL_P macro,
26183         don't test INSN_P (insn) together with CALL_P or JUMP_P check
26184         unnecessarily, formatting fix.
26186 2018-02-06  Michael Collison  <michael.collison@arm.com>
26188         * config/arm/thumb2.md:
26189         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26190         (*thumb_mov_notscc): Ditto.
26192 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
26194         PR target/84154
26195         * config/rs6000/rs6000.md (su code attribute): Use "u" for
26196         unsigned_fix, not "s".
26198 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26200         * configure.ac (gcc_fn_eh_frame_ro): New function.
26201         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26202         correct .eh_frame permissions.
26203         * configure: Regenerate.
26205 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
26207         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26208         irrelevant options.
26210 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26212         * config/rs6000/rs6000.c (rs6000_option_override_internal):
26213         Display warning message for -mno-speculate-indirect-jumps.
26215 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
26217         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26218         Undocumented.
26219         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26221 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26223         PR tree-optimization/84225
26224         * tree-eh.c (find_trapping_overflow): Only call
26225         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26227 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26229         PR target/84145
26230         * config/i386/i386.c: Reimplement the check of possible options
26231         -mibt/-mshstk conbination. Change error messages.
26232         * doc/invoke.texi: Fix a typo: remove extra '='.
26234 2018-02-06  Marek Polacek  <polacek@redhat.com>
26236         PR tree-optimization/84228
26237         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26239 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26241         PR target/82641
26242         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26243         emitted arch directives.
26244         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26245         __ARM_FEATURE_COPROC before changing architectures.
26247 2018-02-06  Richard Biener  <rguenther@suse.de>
26249         * config/i386/i386.c (print_reg): Fix typo.
26250         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26252 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26254         * configure: Regenerate.
26256 2018-02-05  Martin Sebor  <msebor@redhat.com>
26258         PR tree-optimization/83369
26259         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26260         inlining context.
26262 2018-02-05  Martin Liska  <mliska@suse.cz>
26264         * doc/invoke.texi: Cherry-pick upstream r323995.
26266 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26268         * ira.c (ira_init_register_move_cost): Adjust comment.
26270 2018-02-05  Martin Liska  <mliska@suse.cz>
26272         PR gcov-profile/84137
26273         * doc/gcov.texi: Fix typo in documentation.
26275 2018-02-05  Martin Liska  <mliska@suse.cz>
26277         PR gcov-profile/83879
26278         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26279         using dlopen functionality.
26281 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26283         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26284         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26285         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26286         _mm_maskz_range_round_ss): New intrinsics.
26287         (__builtin_ia32_rangesd128_round)
26288         (__builtin_ia32_rangess128_round): Remove.
26289         (__builtin_ia32_rangesd128_mask_round,
26290         __builtin_ia32_rangess128_mask_round): New builtins.
26291         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26292         __builtin_ia32_rangess128_round): Remove.
26293         (__builtin_ia32_rangesd128_mask_round,
26294         __builtin_ia32_rangess128_mask_round): New builtins.
26295         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26296         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26297         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26298         "<round_saeonly_constraint>")): Changed to ...
26299         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26300         "<round_saeonly_scalar_constraint>")): ... this.
26301         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26302         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26303         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26304         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26305         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26307 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26309         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26310         options.
26311         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26312         Remove all values except native, 8540 and 8548.
26314 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26316         * config/i386/i386.c (ix86_output_function_return): Pass
26317         INVALID_REGNUM, instead of -1, as invalid register number to
26318         indirect_thunk_name and output_indirect_thunk.
26320 2018-02-02  Julia Koval  <julia.koval@intel.com>
26322         * config.gcc: Add -march=icelake.
26323         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26324         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26325         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26326         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26327         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26328         (processor_target_table): Add icelake.
26329         (ix86_option_override_internal): Handle new PTAs.
26330         (get_builtin_code_for_version): Handle icelake.
26331         (M_INTEL_COREI7_ICELAKE): New.
26332         (fold_builtin_cpu): Handle icelake.
26333         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26334         * doc/invoke.texi: Add -march=icelake.
26336 2018-02-02  Julia Koval  <julia.koval@intel.com>
26338         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26339         to wide_int_bitmask.
26340         * wide-int-bitmask.h: New.
26342 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26344         PR target/84066
26345         * config/i386/i386.md: Replace Pmode with word_mode in
26346         builtin_setjmp_setup and builtin_longjmp to support x32.
26348 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26350         PR target/56010
26351         PR target/83743
26352         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26353         #include "opts.h".
26354         (rs6000_supported_cpu_names): New static variable.
26355         (linux_cpu_translation_table): Likewise.
26356         (elf_platform) <cpu>: Define new static variable and use it.
26357         Translate kernel AT_PLATFORM name to canonical name if needed.
26358         Error if platform name is unknown.
26360 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26362         PR target/84089
26363         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26365 2018-02-01  Jeff Law  <law@redhat.com>
26367         PR target/84128
26368         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26369         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26370         the scratch if RELEASE_VIA_POP is false.
26371         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26372         If we have to save a temporary register, decrement SIZE appropriately.
26373         Pass new arguments to release_scratch_register_on_entry.
26374         (ix86_adjust_stack_and_probe): Likewise.
26375         (ix86_emit_probe_stack_range): Pass new arguments to
26376         release_scratch_register_on_entry.
26378 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26380         PR rtl-optimization/84157
26381         * combine.c (change_zero_ext): Use REG_P predicate in
26382         front of HARD_REGISTER_P predicate.
26384 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26386         * config/avr/avr.c (avr_option_override): Move disabling of
26387         -fdelete-null-pointer-checks to...
26388         * common/config/avr/avr-common.c (avr_option_optimization_table):
26389         ...here.
26391 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26393         PR tree-optimization/81635
26394         * tree-data-ref.c (split_constant_offset_1): For types that
26395         wrap on overflow, try to use range info to prove that wrapping
26396         cannot occur.
26398 2018-02-01  Renlin Li  <renlin.li@arm.com>
26400         PR target/83370
26401         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26402         TAILCALL_ADDR_REGS.
26403         (aarch64_register_move_cost): Likewise.
26404         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26405         TAILCALL_ADDR_REGS.
26406         (REG_CLASS_NAMES): Likewise.
26407         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26408         TAILCALL_ADDR_REGS. Remove IP registers.
26409         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26411 2018-02-01  Richard Biener  <rguenther@suse.de>
26413         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26414         for specifying RPO order and allow NULL for that.
26415         * domwalk.c (dom_walker::dom_walker): Likewise.
26416         (dom_walker::walk): Handle NULL RPO order.
26417         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26418         in RPO order.
26419         (rewrite_update_dom_walker): Likewise.
26420         (mark_def_dom_walker): Likewise.
26422 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26424         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26425         (aarch64_maybe_expand_sve_subreg_move): Declare.
26426         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26427         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26428         predicate.
26429         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26430         that are semantically a reverse operation.
26431         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26432         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26433         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26434         functions.
26435         (aarch64_can_change_mode_class): For big-endian, forbid changes
26436         between two SVE modes if they have different element sizes.
26438 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26440         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26441         the TImode handling for big-endian targets.
26443 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26445         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26446         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26447         not just bytes.
26448         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26449         Remove BSWAP handing for big-endian targets and use the form of
26450         LD1RQ appropariate for the mode.
26452 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26454         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26455         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26456         duplicated element.
26458 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26460         PR tearget/83845
26461         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26462         check for operands that need to go through aarch64_sve_reload_be.
26464 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26466         PR tree-optimization/81661
26467         PR tree-optimization/84117
26468         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26469         * tree-eh.c: Include gimplify.h.
26470         (find_trapping_overflow, replace_trapping_overflow,
26471         rewrite_to_non_trapping_overflow): New functions.
26472         * tree-vect-loop.c: Include tree-eh.h.
26473         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26474         * tree-data-ref.c: Include tree-eh.h.
26475         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26477 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26479         PR rtl-optimization/84123
26480         * combine.c (change_zero_ext): Check if hard register satisfies
26481         can_change_dest_mode before calling gen_lowpart_SUBREG.
26483 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26485         PR target/82444
26486         * ira.c (ira_init_register_move_cost): Remove assert.
26488 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26490         PR rtl-optimization/84071
26491         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26492         * doc/tm.texi: Regenerate.
26494 2018-01-31  Richard Biener  <rguenther@suse.de>
26496         PR tree-optimization/84132
26497         * tree-data-ref.c (analyze_miv_subscript): Properly
26498         check whether evolution_function_is_affine_multivariate_p
26499         before calling gcd_of_steps_may_divide_p.
26501 2018-01-31  Julia Koval  <julia.koval@intel.com>
26503         PR target/83618
26504         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26505         * config/i386/i386.md (rdpid_rex64) New.
26506         (rdpid): Make 32bit only.
26508 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26510         PR lto/84105
26511         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26512         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26514 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26516         Revert
26517         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26519         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26521 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26523         PR rtl-optimization/84071
26524         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26525         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26527 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26529         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26530         (arc_attribute_table): Add 'aux' attribute.
26531         (arc_in_small_data_p): Consider aux like variables.
26532         (arc_is_aux_reg_p): New function.
26533         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26534         (arc_get_aux_arg): New function.
26535         (prepare_move_operands): Handle aux-register access.
26536         (arc_handle_aux_attribute): New function.
26537         * doc/extend.texi (ARC Variable attributes): Add subsection.
26539 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26541         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26542         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26543         (arc_attribute_table): Add 'uncached' attribute.
26544         (arc_print_operand): Print '.di' flag for uncached memory
26545         accesses.
26546         (arc_in_small_data_p): Do not consider for small data the uncached
26547         types.
26548         (arc_is_uncached_mem_p): New function.
26549         * config/arc/predicates.md (compact_store_memory_operand): Check
26550         for uncached memory accesses.
26551         (nonvol_nonimm_operand): Likewise.
26552         * doc/extend.texi (ARC Type Attribute): New subsection.
26554 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26556         PR c/84100
26557         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26558         falign-loops=): Add Optimization flag.
26560 2018-01-30  Jeff Law  <law@redhat.com>
26562         PR target/84064
26563         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26564         INT_REGISTERS_SAVED.  Check it prior to calling
26565         get_scratch_register_on_entry.
26566         (ix86_adjust_stack_and_probe): Similarly.
26567         (ix86_emit_probe_stack_range): Similarly.
26568         (ix86_expand_prologue): Corresponding changes.
26570 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26572         PR target/40411
26573         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26574         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26576 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26578         PR target/84112
26579         * lra-constraints.c (curr_insn_transform): Process AND in the
26580         address.
26582 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26584         PR rtl-optimization/83986
26585         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26586         dependence against last_pending_memory_flush in addition to
26587         pending_jump_insns.
26589 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26591         PR tree-optimization/81611
26592         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26593         copies.
26595 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26597         PR target/83758
26598         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26599         a reg rtx.
26601 2018-01-30  Richard Biener  <rguenther@suse.de>
26602             Jakub Jelinek  <jakub@redhat.com>
26604         PR tree-optimization/84111
26605         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26606         inner loops added during recursion, as they don't have up-to-date
26607         SSA form.
26609 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26611         PR ipa/81360
26612         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26613         (can_inline_edge_by_limits_p): ... here.
26614         (can_early_inline_edge_p, check_callers,
26615         update_caller_keys, update_callee_keys, recursive_inlining,
26616         add_new_edges_to_heap, speculation_useful_p,
26617         inline_small_functions,
26618         inline_small_functions, flatten_function,
26619         inline_to_all_callers_1): Update.
26621 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26623         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26624         zeros correctly.
26626 2018-01-30  Richard Biener  <rguenther@suse.de>
26628         PR tree-optimization/83008
26629         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26630         invariant and constant vector uses in stmts when they need
26631         more than one stmt.
26633 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26635         PR bootstrap/84017
26636         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26637         * configure: Regenerate.
26639 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26641         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26642         pattern.
26643         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26644         Use gen_rtx_REG rather than gen_lowpart.
26646 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26648         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26649         rather than 0 when creating partial subregs.
26651 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26653         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26654         of usage.
26656 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26658         PR target/81550
26659         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26660         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26661         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26662         flags.  This restores the settings used before the 2017-07-24.
26663         Turning off pre increment/decrement/modify allows IVOPTS to
26664         optimize DF/SF loops where the index is an int.
26666 2018-01-29  Richard Biener  <rguenther@suse.de>
26667             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26669         PR bootstrap/80867
26670         * tree-vect-stmts.c (vectorizable_call): Don't call
26671         targetm.vectorize_builtin_md_vectorized_function if callee is
26672         NULL.
26674 2018-01-22  Carl Love  <cel@us.ibm.com>
26676         * doc/extend.tex: Fix typo in second arg in
26677         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26679 2018-01-29  Richard Biener  <rguenther@suse.de>
26681         PR tree-optimization/84086
26682         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26683         (flush_ssaname_freelist): When SSA names were released reset
26684         the SCEV hash table.
26686 2018-01-29  Richard Biener  <rguenther@suse.de>
26688         PR tree-optimization/84057
26689         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26690         removed paths when removing edges.
26692 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26694         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26695         -mfunction-return=@var{choice}.
26697 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26699         PR diagnostic/84034
26700         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26701         Handle CR like TAB.
26702         (layout::print_source_line): Likewise.
26703         (test_get_line_width_without_trailing_whitespace): Add test cases.
26705 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26707         PR middle-end/84040
26708         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26709         debug insns.
26711 2018-01-26  Jim Wilson  <jimw@sifive.com>
26713         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26715         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26716         specified.
26718 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26720         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
26721         and CMP + SUB-immediate -> SUBS.
26723 2018-01-26  Martin Sebor  <msebor@redhat.com>
26725         PR tree-optimization/83896
26726         * tree-ssa-strlen.c (get_string_len): Rename...
26727         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
26728         Avoid assuming length is constant.
26729         (handle_char_store): Use HOST_WIDE_INT for string length.
26731 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
26733         PR target/81763
26734         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
26735         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
26737 2018-01-26  Richard Biener  <rguenther@suse.de>
26739         PR rtl-optimization/84003
26740         * dse.c (record_store): Only record redundant stores when
26741         the earlier store aliases at least all accesses the later one does.
26743 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
26745         PR rtl-optimization/83985
26746         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
26747         REG_CFA_RESTORE insns.
26748         (delete_unmarked_insns): Don't ignore separate shrink wrapping
26749         REG_CFA_RESTORE insns here.
26751         PR c/83989
26752         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
26753         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
26755 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26757         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
26758         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
26759         (arc_init): Likewise.
26760         (arc_override_options): Likewise.
26761         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
26762         value.
26763         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
26764         support.
26765         * config/arc/arc.h (TARGET_DBNZ): Define.
26766         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
26767         properly set the tune attribute.
26768         (dbnz): Use TARGET_DBNZ guard.
26769         * config/arc/arc.opt (mtune): Add core3 option.
26771 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26773         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
26774         recognize new pic like addresses.
26775         (arc_delegitimize_address): Clean up.
26777 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26779         * config/arc/arc-arches.def: Option mrf16 valid for all
26780         architectures.
26781         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
26782         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
26783         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
26784         * config/arc/arc-tables.opt: Regenerate.
26785         * config/arc/arc.c (arc_conditional_register_usage): Handle
26786         reduced register file case.
26787         (arc_file_start): Set must have build attributes.
26788         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
26789         mrf16 option value.
26790         * config/arc/arc.opt (mrf16): Add new option.
26791         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
26792         * config/arc/genmultilib.awk: Handle new mrf16 option.
26793         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
26794         * config/arc/t-multilib: Regenerate.
26795         * doc/invoke.texi (ARC Options): Document mrf16 option.
26797 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26799         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
26800         * config/arc/arc.c (arc_handle_secure_attribute): New function.
26801         (arc_attribute_table): Add 'secure_call' attribute.
26802         (arc_print_operand): Print secure call operand.
26803         (arc_function_ok_for_sibcall): Don't optimize tail calls when
26804         secure.
26805         (arc_is_secure_call_p): New function.  * config/arc/arc.md
26806         (call_i): Add support for sjli instruction.
26807         (call_value_i): Likewise.
26808         * config/arc/constraints.md (Csc): New constraint.
26810 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26811             John Eric Martin  <John.Martin@emmicro-us.com>
26813         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
26814         * config/arc/arc.c (_arc_jli_section): New struct.
26815         (arc_jli_section): New type.
26816         (rc_jli_sections): New static variable.
26817         (arc_handle_jli_attribute): New function.
26818         (arc_attribute_table): Add jli_always and jli_fixed attribute.
26819         (arc_file_end): New function.
26820         (TARGET_ASM_FILE_END): Define.
26821         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
26822         (arc_add_jli_section): New function.
26823         (jli_call_scan): Likewise.
26824         (arc_reorg): Call jli_call_scan.
26825         (arc_output_addsi): Remove 'S' from printing asm operand.
26826         (arc_is_jli_call_p): New function.
26827         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
26828         operand.
26829         (movhi_insn): Likewise.
26830         (movsi_insn): Likewise.
26831         (movsi_set_cc_insn): Likewise.
26832         (loadqi_update): Likewise.
26833         (load_zeroextendqisi_update): Likewise.
26834         (load_signextendqisi_update): Likewise.
26835         (loadhi_update): Likewise.
26836         (load_zeroextendhisi_update): Likewise.
26837         (load_signextendhisi_update): Likewise.
26838         (loadsi_update): Likewise.
26839         (loadsf_update): Likewise.
26840         (movsicc_insn): Likewise.
26841         (bset_insn): Likewise.
26842         (bxor_insn): Likewise.
26843         (bclr_insn): Likewise.
26844         (bmsk_insn): Likewise.
26845         (bicsi3_insn): Likewise.
26846         (cmpsi_cc_c_insn): Likewise.
26847         (movsi_ne): Likewise.
26848         (movsi_cond_exec): Likewise.
26849         (clrsbsi2): Likewise.
26850         (norm_f): Likewise.
26851         (normw): Likewise.
26852         (swap): Likewise.
26853         (divaw): Likewise.
26854         (flag): Likewise.
26855         (sr): Likewise.
26856         (kflag): Likewise.
26857         (ffs): Likewise.
26858         (ffs_f): Likewise.
26859         (fls): Likewise.
26860         (call_i): Remove 'S' asm letter, add jli instruction.
26861         (call_value_i): Likewise.
26862         * config/arc/arc.op (mjli-always): New option.
26863         * config/arc/constraints.md (Cji): New constraint.
26864         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
26865         operand.
26866         (subsf3_fpx): Likewise.
26867         (mulsf3_fpx): Likewise.
26868         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
26869         asm operand.
26870         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
26871         function attrbutes.
26872         * doc/invoke.texi (ARC): Document mjli-always option.
26874 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
26876         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
26877         avoid addition with 0 and use incw and decw where possible.
26879 2018-01-26  Richard Biener  <rguenther@suse.de>
26881         PR tree-optimization/81082
26882         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
26883         association if it requires casting to unsigned.
26884         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
26885         from fold_plusminus_mult_expr to catch important cases late when
26886         range info is available.
26888 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26890         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
26891         * configure.ac (hidden_linkonce): New test.
26892         * configure: Regenerate.
26893         * config.in: Regenerate.
26895 2018-01-26  Julia Koval  <julia.koval@intel.com>
26897         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
26898         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
26899         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
26900         _mm_mask_bitshuffle_epi64_mask): Fix type.
26901         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
26902         USI_FTYPE_V4DI_V4DI_USI): Remove.
26903         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
26904         __builtin_ia32_vpshufbitqmb256_mask,
26905         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
26906         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
26907         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
26909 2018-01-26  Alan Modra  <amodra@gmail.com>
26911         PR target/84033
26912         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
26913         UNSPEC_VBPERMQ.  Sort other unspecs.
26915 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
26917         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
26919 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
26921         PR middle-end/83055
26922         * predict.c (drop_profile): Do not push/pop cfun; update also
26923         node->count.
26924         (handle_missing_profiles): Fix logic looking for zero profiles.
26926 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
26928         PR middle-end/83977
26929         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
26930         on functions with #pragma omp declare simd or functions with simd
26931         attribute.
26932         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
26933         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
26934         Remove trailing \n from warning_at calls.
26936 2018-01-25  Tom de Vries  <tom@codesourcery.com>
26938         PR target/84028
26939         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26940         for neutered workers.
26942 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
26944         PR target/68467
26945         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
26946         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
26948 2018-01-24  Jeff Law  <law@redhat.com>
26950         PR target/83994
26951         * i386.c (get_probe_interval): Move to earlier point.
26952         (ix86_compute_frame_layout): If -fstack-clash-protection and
26953         the frame is larger than the probe interval, then use pushes
26954         to save registers rather than reg->mem moves.
26955         (ix86_expand_prologue): Remove conditional for int_registers_saved
26956         assertion.
26958 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
26960         PR target/84014
26961         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
26962         min/max for never referenced object.
26964 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
26966         PR middle-end/83977
26967         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
26968         here.
26969         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
26970         attributes from DECL_ATTRIBUTES (decl) without affecting
26971         DECL_ATTRIBUTES (current_function_decl).
26972         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
26973         functions with non-NULL DECL_ABSTRACT_ORIGIN.
26975 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
26977         PR tree-optimization/83979
26978         * fold-const.c (fold_comparison): Use constant_boolean_node
26979         instead of boolean_{true,false}_node.
26981 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
26983         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
26984         with zero counts.
26986 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26988         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26989         Simplify the clause that sets the length attribute.
26990         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26991         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
26992         clause that sets the length attribute.
26993         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26995 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26997         PR target/83589
26998         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
26999         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27000         Add strict parameter.
27001         (prevent_branch_around_nothing): Insert dummy insn between branch to
27002         label and label with no ptx insn inbetween.
27003         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27005 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27007         PR target/81352
27008         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27009         for neutered threads in warp.
27010         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27012 2018-01-24  Richard Biener  <rguenther@suse.de>
27014         PR tree-optimization/83176
27015         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27016         operands.
27018 2018-01-24  Richard Biener  <rguenther@suse.de>
27020         PR tree-optimization/82819
27021         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27022         code generating pluses that are no-ops in the target precision.
27024 2018-01-24  Richard Biener  <rguenther@suse.de>
27026         PR middle-end/84000
27027         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27029 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27031         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27032         to merge probabilities.
27033         * predict.c (probably_never_executed): Also mark as cold functions
27034         with global 0 profile and guessed local profile.
27035         * profile-count.c (profile_probability::combine_with_count): New
27036         member function.
27037         * profile-count.h (profile_probability::operator*,
27038         profile_probability::operator*=, profile_probability::operator/,
27039         profile_probability::operator/=): Reduce precision to adjusted
27040         and set value to guessed on contradictory divisions.
27041         (profile_probability::combine_with_freq): Remove.
27042         (profile_probability::combine_wiht_count): Declare.
27043         (profile_count::force_nonzero):: Set to adjusted.
27044         (profile_count::probability_in):: Set quality to adjusted.
27045         * tree-ssa-tail-merge.c (replace_block_by): Use
27046         combine_with_count.
27048 2018-01-23  Andrew Waterman  <andrew@sifive.com>
27049             Jim Wilson  <jimw@sifive.com>
27051         * config/riscv/riscv.c (riscv_stack_boundary): New.
27052         (riscv_option_override): Set riscv_stack_boundary.  Handle
27053         riscv_preferred_stack_boundary_arg.
27054         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27055         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27056         (STACK_BOUNDARY): Set to riscv_stack_boundary.
27057         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27058         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27059         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27061 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
27063         PR target/83905
27064         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27065         of struct ix86_frame.
27066         (ix86_expand_epilogue): Likewise.  Add a local variable for
27067         the reg_save_offset field in struct ix86_frame.
27069 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
27071         PR tree-optimization/82604
27072         * tree-loop-distribution.c (enum partition_kind): New enum item
27073         PKIND_PARTIAL_MEMSET.
27074         (partition_builtin_p): Support above new enum item.
27075         (generate_code_for_partition): Ditto.
27076         (compute_access_range): Differentiate cases that equality can be
27077         proven at all loops, the innermost loops or no loops.
27078         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
27079         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
27080         (finalize_partitions, distribute_loop): Don't fuse partition of
27081         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
27082         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
27083         parloop is enabled.
27085 2018-01-23  Martin Liska  <mliska@suse.cz>
27087         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
27088         order to ignore the predictor.
27089         (PRED_POLYMORPHIC_CALL): Likewise.
27090         (PRED_RECURSIVE_CALL): Likewise.
27092 2018-01-23  Martin Liska  <mliska@suse.cz>
27094         * tree-profile.c (tree_profiling): Print function header to
27095         aware reader which function we are working on.
27096         * value-prof.c (gimple_find_values_to_profile): Do not print
27097         not interesting value histograms.
27099 2018-01-23  Martin Liska  <mliska@suse.cz>
27101         * profile-count.h (enum profile_quality): Add
27102         profile_uninitialized as the first value. Do not number values
27103         as they are zero based.
27104         (profile_count::verify): Update sanity check.
27105         (profile_probability::verify): Likewise.
27107 2018-01-23  Nathan Sidwell  <nathan@acm.org>
27109         * doc/invoke.texi (ffor-scope): Deprecate.
27111 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27113         PR tree-optimization/83510
27114         * domwalk.c (set_all_edges_as_executable): New function.
27115         (dom_walker::dom_walker): Convert bool param
27116         "skip_unreachable_blocks" to enum reachability.  Move setup of
27117         edge flags to set_all_edges_as_executable and only do it when
27118         reachability is REACHABLE_BLOCKS.
27119         * domwalk.h (enum dom_walker::reachability): New enum.
27120         (dom_walker::dom_walker): Convert bool param
27121         "skip_unreachable_blocks" to enum reachability.
27122         (set_all_edges_as_executable): New decl.
27123         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
27124         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
27125         "reachability".
27126         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
27127         but converting true to REACHABLE_BLOCKS.
27128         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
27129         * tree-vrp.c
27130         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
27131         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27132         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27133         REACHABLE_BLOCKS.
27134         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27135         if check_all_array_refs will be called.
27137 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27139         * tree.c (selftest::test_location_wrappers): Add more test
27140         coverage.
27142 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27144         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27145         (selftest::test_bit_in_range): Likewise.
27147 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27149         PR testsuite/83888
27150         * doc/sourcebuild.texi (vect_float): Say that the selector
27151         only describes the situation when -funsafe-math-optimizations is on.
27152         (vect_float_strict): Document.
27154 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27156         PR tree-optimization/83965
27157         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27158         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27159         instead of checking only for a reduction.
27160         (vect_recog_widen_sum_pattern): Likewise.
27162 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27164         * predict.c (probably_never_executed): Only use precise profile info.
27165         (compute_function_frequency): Skip after inlining hack since we now
27166         have quality checking.
27168 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27170         * profile-count.h (profile_probability::very_unlikely,
27171         profile_probability::unlikely, profile_probability::even): Set
27172         precision to guessed.
27174 2018-01-23  Richard Biener  <rguenther@suse.de>
27176         PR tree-optimization/83963
27177         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27178         Properly terminate dominator walk when crossing the exit edge not
27179         when visiting its source block.
27181 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
27183         PR c++/83918
27184         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27185         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27187 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
27189         PR tree-optimization/83957
27190         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
27191         semicolon after for body surrounded by braces.
27193         PR tree-optimization/83081
27194         * profile-count.h (profile_probability::split): New method.
27195         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27196         Use profile_probability::split.
27197         (do_compare_rtx_and_jump): Fix adjustment of probabilities
27198         when splitting a single conditional jump into 2.
27200 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
27202         PR tree-optimization/69452
27203         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27204         decl.
27206 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27208         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27209         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27210         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27212 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27214         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27215         declaration.
27216         * config/rl78/rl78.md (movdi): New define_expand.
27217         * config/rl78/rl78.c (rl78_split_movdi): New function.
27219 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27221         PR target/83862
27222         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27223         no longer used.
27224         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27225         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27226         128-bit to produce an UNSPEC move to get the double word with the
27227         signbit and then a shift directly to do signbit.
27228         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27229         implementation with a new version that just does either a direct
27230         move or a regular move.  Move memory interface to separate insns.
27231         Move insns so they are next to the expander.
27232         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27233         with signbit move.  Split big and little endian case.
27234         (signbit<mode>2_dm_mem_le): Likewise.
27235         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27236         (signbit<mode>2_dm2): Likewise.
27238 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27240         * config/rl78/rl78.md (anddi3): New define_expand.
27242 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27244         * config/rl78/rl78.md (umindi3): New define_expand.
27246 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27248         * config/rl78/rl78.md (smindi3): New define_expand.
27250 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27252         * config/rl78/rl78.md (smaxdi3): New define_expand.
27254 2018-01-22  Carl Love  <cel@us.ibm.com>
27256         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27257         LVX_V1TI): Add macro expansion.
27258         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27259         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27260         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27261         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27262         Change check to determine if the instruction is a byte reversing
27263         entry.  Fix typo in comment.
27264         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27265         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27266         Add def_builtin calls for new builtins.
27267         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27268         Add define_insn expansion.
27270 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27272         * config/rl78/rl78.md (umaxdi3): New define_expand.
27274 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27276         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27277         for non-QImode registers.
27279 2018-01-22  Richard Biener  <rguenther@suse.de>
27281         PR tree-optimization/83963
27282         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27283         including the loop exit block.
27284         (scop_detection::merge_sese): Likewise.
27285         (scop_detection::add_scop): Do it here instead.
27287 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27289         * doc/sourcebuild.texi (arm_softfloat): Document.
27291 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27293         PR gcc/77734
27294         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27295         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27296         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27298 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27299             David Edelsohn  <dje.gcc@gmail.com>
27301         PR target/83946
27302         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27303         Change "crset eq" to "crset 2".
27304         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27305         (*call_indirect_aix<mode>_nospec): Likewise.
27306         (*call_value_indirect_aix<mode>_nospec): Likewise.
27307         (*call_indirect_elfv2<mode>_nospec): Likewise.
27308         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27309         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27310         change assembly output from . to $.
27311         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27312         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27313         (*tablejump<mode>_internal1_nospec): Likewise.
27315 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27317         PR target/80870
27318         * config/sh/sh_optimize_sett_clrt.cc:
27319         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27321 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27323         PR tree-optimization/83940
27324         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27325         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27326         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27327         use it to pass back the definition type.
27328         (vect_check_store_rhs): Likewise rhs_dt_out.
27329         (vect_build_gather_load_calls): Add a mask_dt argument and use
27330         it instead of a call to vect_is_simple_use.
27331         (vectorizable_store): Update calls to vect_check_load_store_mask
27332         and vect_check_store_rhs.  Use the dt returned by the latter instead
27333         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27334         instead of calls to vect_is_simple_use.  Pass the scalar rather
27335         than the vector operand to vect_is_simple_use when handling
27336         second and subsequent copies of an rhs value.
27337         (vectorizable_load): Update calls to vect_check_load_store_mask
27338         and vect_build_gather_load_calls.  Use the cached mask_dt and
27339         gs_info.offset_dt instead of calls to vect_is_simple_use.
27341 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27343         PR middle-end/83945
27344         * tree-emutls.c: Include gimplify.h.
27345         (lower_emutls_2): New function.
27346         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27347         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27348         it before further processing.
27350         PR target/83930
27351         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27352         UINTVAL (trueop1) instead of INTVAL (op1).
27354 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27356         PR debug/81570
27357         PR debug/83728
27358         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27359         INCOMING_FRAME_SP_OFFSET if not defined.
27360         (scan_trace): Add ENTRY argument.  If true and
27361         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27362         emit a note to adjust the CFA offset.
27363         (create_cfi_notes): Adjust scan_trace callers.
27364         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27365         INCOMING_FRAME_SP_OFFSET in the CIE.
27366         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27367         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27368         Likewise.
27369         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27370         * doc/tm.texi: Regenerated.
27372 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27374         PR rtl-optimization/83147
27375         * lra-constraints.c (remove_inheritance_pseudos): Use
27376         lra_substitute_pseudo_within_insn.
27378 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27379             Cesar Philippidis  <cesar@codesourcery.com>
27381         PR target/83920
27382         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27384 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27386         PR target/83790
27387         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27388         spaces for function labels.
27390 2018-01-19  Martin Liska  <mliska@suse.cz>
27392         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27393         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27394         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27395         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27396         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27397         (PRED_CONST_RETURN): Change from 69 to 65.
27398         (PRED_NULL_RETURN): Change from 91 to 71.
27399         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27400         (PRED_LOOP_GUARD): Change from 66 to 73.
27402 2018-01-19  Martin Liska  <mliska@suse.cz>
27404         * predict.c (predict_insn_def): Add new assert.
27405         (struct branch_predictor): Change type to signed integer.
27406         (test_prediction_value_range): Amend test to cover
27407         PROB_UNINITIALIZED.
27408         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27409         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27410         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27411         (PRED_LOOP_IV_COMPARE): Likewise.
27412         * predict.h (PROB_UNINITIALIZED): Define new constant.
27414 2018-01-19  Martin Liska  <mliska@suse.cz>
27416         * predict.c (dump_prediction): Add new format for
27417         analyze_brprob.py script which is enabled with -details
27418         suboption.
27419         * profile-count.h (precise_p): New function.
27421 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27423         PR tree-optimization/83922
27424         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27425         there are no statements that need masking.
27426         (vect_active_double_reduction_p): New function.
27427         (vect_analyze_loop_operations): Use it when handling phis that
27428         are not in the loop header.
27430 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27432         PR tree-optimization/83914
27433         * tree-vect-loop.c (vectorizable_induction): Don't convert
27434         init_expr or apply the peeling adjustment for inductions
27435         that are nested within the vectorized loop.
27437 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27439         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27440         instead of NEG.
27442 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27444         PR sanitizer/81715
27445         PR testsuite/83882
27446         * function.h (gimplify_parameters): Add gimple_seq * argument.
27447         * function.c: Include gimple.h and options.h.
27448         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27449         for the added local temporaries if needed.
27450         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27451         if there are any parameter cleanups, wrap whole body into a
27452         try/finally with the cleanups.
27454 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27456         PR target/82964
27457         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27458         Use GET_MODE_CLASS for scalar floating point.
27460 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27462         PR ipa/82256
27463         patch by PaX Team
27464         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27465         Fix call of call_cgraph_insertion_hooks.
27467 2018-01-18  Martin Sebor  <msebor@redhat.com>
27469         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27471 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27473         PR ipa/83619
27474         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27475         frequencies.
27477 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27479         PR other/70268
27480         * common.opt: (-ffile-prefix-map): New option.
27481         * opts.c (common_handle_option): Defer it.
27482         * opts-global.c (handle_common_deferred_options): Handle it.
27483         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27484         * file-prefix-map.h: New file.
27485         (remap_debug_filename, add_debug_prefix_map): ...here.
27486         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27487         * final.c (debug_prefix_map, add_debug_prefix_map
27488         remap_debug_filename): Move to...
27489         * file-prefix-map.c: New file.
27490         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27491         generalize, get rid of alloca(), use strrchr() instead of strchr().
27492         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27493         Implement in terms of add_prefix_map().
27494         (remap_macro_filename, remap_debug_filename): Implement in term of
27495         remap_filename().
27496         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27497         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27498         * dbxout.c: Include file-prefix-map.h.
27499         * varasm.c: Likewise.
27500         * vmsdbgout.c: Likewise.
27501         * xcoffout.c: Likewise.
27502         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27503         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27504         * doc/invoke.texi (-ffile-prefix-map): Document.
27505         (-fdebug-prefix-map): Update description.
27507 2018-01-18  Martin Liska  <mliska@suse.cz>
27509         * config/i386/i386.c (indirect_thunk_name): Document that also
27510         lfence is emitted.
27511         (output_indirect_thunk): Document why both instructions
27512         (pause and lfence) are generated.
27514 2018-01-18  Richard Biener  <rguenther@suse.de>
27516         PR tree-optimization/83887
27517         * graphite-scop-detection.c
27518         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27519         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27520         (scop_detection::merge_sese): Re-implement with a flood-fill
27521         algorithm that properly finds a SESE region if it exists.
27523 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27525         PR c/61240
27526         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27527         pointer_diff optimizations use view_convert instead of convert.
27529 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27531         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27532         Generate different code for -mno-speculate-indirect-jumps.
27533         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27534         (*call_indirect_aix<mode>): Disable for
27535         -mno-speculate-indirect-jumps.
27536         (*call_indirect_aix<mode>_nospec): New define_insn.
27537         (*call_value_indirect_aix<mode>): Disable for
27538         -mno-speculate-indirect-jumps.
27539         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27540         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27541         -mno-speculate-indirect-jumps.
27542         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27544 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27546         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27547         long double type, set the flags for noting the default long double
27548         type, even if we don't pass or return a long double type.
27550 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27552         PR ipa/83051
27553         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27554         failure.
27556 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27558         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27559         support for merge[hl].
27560         (fold_mergehl_helper): New helper function.
27561         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27562         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27563         (altivec_vmrglw_direct): Add xxmrglw insn.
27565 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27567         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27568         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27570 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27572         PR lto/83121
27573         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27574         call the lto_location_cache before reading the
27575         DECL_SOURCE_LOCATION of the types.
27577 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27578             Richard Sandiford  <richard.sandiford@linaro.org>
27580         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27581         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27582         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27583         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27584         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27585         Add declaration.
27586         * config/aarch64/constraints.md (aarch64_movti_operand):
27587         Limit immediates.
27588         * config/aarch64/predicates.md (Uti): Add new constraint.
27590 2018-01-17  Carl Love  <cel@us.ibm.com>
27592         * config/rs6000/vsx.md (define_expand xl_len_r,
27593         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27594         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27595         lxvll.
27596         (define_expand, define_insn): Move the shift left from  the
27597         define_insn to the define_expand for lxvl and stxvl instructions.
27598         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27599         and XL_LEN_R definitions to PURE.
27601 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27603         * config/i386/i386.c (indirect_thunk_name): Declare regno
27604         as unsigned int.  Compare regno with INVALID_REGNUM.
27605         (output_indirect_thunk): Ditto.
27606         (output_indirect_thunk_function): Ditto.
27607         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27608         in the call to output_indirect_thunk_function.
27610 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27612         PR middle-end/83884
27613         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27614         rather than the size of inner_type to determine the stack slot size
27615         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27617 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27619         PR target/83546
27620         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27621         to PTA_SILVERMONT.
27623 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27625         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27626         endian Linux systems to optionally enable multilibs for selecting
27627         the long double type if the user configured an explicit type.
27628         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27629         have no long double multilibs if not defined.
27630         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27631         warn if the user used -mabi={ieee,ibm}longdouble and we built
27632         multilibs for long double.
27633         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27634         appropriate multilib option.
27635         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27636         multilib options.
27637         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27638         for building long double multilibs.
27639         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27641 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27643         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27644         copies.
27646         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27647         64 bits.
27648         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27649         128 bits.
27651         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27652         variables.
27654         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27655         return value.
27657 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27659         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27660         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27662 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27664         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27665         different rtl trees depending on TARGET_64BIT.
27666         (rs6000_gen_lvx): Likewise.
27668 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27670         * config/visium/visium.md (nop): Tweak comment.
27671         (hazard_nop): Likewise.
27673 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27675         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27676         -mspeculate-indirect-jumps.
27677         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27678         for -mno-speculate-indirect-jumps.
27679         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27680         (*call_value_indirect_elfv2<mode>): Disable for
27681         -mno-speculate-indirect-jumps.
27682         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27683         (indirect_jump): Emit different RTL for
27684         -mno-speculate-indirect-jumps.
27685         (*indirect_jump<mode>): Disable for
27686         -mno-speculate-indirect-jumps.
27687         (*indirect_jump<mode>_nospec): New define_insn.
27688         (tablejump): Emit different RTL for
27689         -mno-speculate-indirect-jumps.
27690         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27691         (tablejumpsi_nospec): New define_expand.
27692         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27693         (tablejumpdi_nospec): New define_expand.
27694         (*tablejump<mode>_internal1): Disable for
27695         -mno-speculate-indirect-jumps.
27696         (*tablejump<mode>_internal1_nospec): New define_insn.
27697         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27698         option.
27700 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27702         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27703         callers updated.
27705 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27706             Richard Biener  <rguenth@suse.de>
27708         PR libgomp/83590
27709         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27710         return early, inline manually is_gimple_sizepos.  Make sure if we
27711         call gimplify_expr we don't end up with a gimple constant.
27712         * tree.c (variably_modified_type_p): Don't return true for
27713         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27714         * gimplify.h (is_gimple_sizepos): Remove.
27716 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27718         PR tree-optimization/83857
27719         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27720         vectorizable_live_operation for pure SLP statements.
27721         (vectorizable_live_operation): Handle PHIs.
27723 2018-01-16  Richard Biener  <rguenther@suse.de>
27725         PR tree-optimization/83867
27726         * tree-vect-stmts.c (vect_transform_stmt): Precompute
27727         nested_in_vect_loop_p since the scalar stmt may get invalidated.
27729 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27731         PR c/83844
27732         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
27733         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
27734         If off is not INTEGER_CST, issue a may not be aligned warning
27735         rather than isn't aligned.  Use isn%'t rather than isn't.
27736         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
27737         into MULT_EXPR.
27738         <case MULT_EXPR>: Improve the case when bottom and one of the
27739         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
27740         operand, in that case check if the other operand is multiple of
27741         bottom divided by the INTEGER_CST operand.
27743 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27745         PR target/83858
27746         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
27747         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
27748         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
27749         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
27750         * config/pa/pa.c (pa_function_arg_advance): Likewise.
27751         (pa_function_arg, pa_arg_partial_bytes): Likewise.
27752         (pa_function_arg_size): New function.
27754 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27756         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
27757         in a separate statement.
27759 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27761         PR tree-optimization/83847
27762         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
27763         group gathers and scatters.
27765 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27767         PR rtl-optimization/86620
27768         * params.def (max-sched-ready-insns): Bump minimum value to 1.
27770         PR rtl-optimization/83213
27771         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
27772         to last if both are JUMP_INSNs.
27774         PR tree-optimization/83843
27775         * gimple-ssa-store-merging.c
27776         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
27777         store_immediate_info for bswap/nop orig_stores.
27779 2018-01-15  Andrew Waterman  <andrew@sifive.com>
27781         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
27782         !TARGET_MUL.
27783         <UDIV>: Increase cost if !TARGET_DIV.
27785 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
27787         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
27788         (define_attr "cr_logical_3op"): New.
27789         (cceq_ior_compare): Adjust.
27790         (cceq_ior_compare_complement): Adjust.
27791         (*cceq_rev_compare): Adjust.
27792         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
27793         (is_cracked_insn): Adjust.
27794         (insn_must_be_first_in_group): Adjust.
27795         * config/rs6000/40x.md: Adjust.
27796         * config/rs6000/440.md: Adjust.
27797         * config/rs6000/476.md: Adjust.
27798         * config/rs6000/601.md: Adjust.
27799         * config/rs6000/603.md: Adjust.
27800         * config/rs6000/6xx.md: Adjust.
27801         * config/rs6000/7450.md: Adjust.
27802         * config/rs6000/7xx.md: Adjust.
27803         * config/rs6000/8540.md: Adjust.
27804         * config/rs6000/cell.md: Adjust.
27805         * config/rs6000/e300c2c3.md: Adjust.
27806         * config/rs6000/e500mc.md: Adjust.
27807         * config/rs6000/e500mc64.md: Adjust.
27808         * config/rs6000/e5500.md: Adjust.
27809         * config/rs6000/e6500.md: Adjust.
27810         * config/rs6000/mpc.md: Adjust.
27811         * config/rs6000/power4.md: Adjust.
27812         * config/rs6000/power5.md: Adjust.
27813         * config/rs6000/power6.md: Adjust.
27814         * config/rs6000/power7.md: Adjust.
27815         * config/rs6000/power8.md: Adjust.
27816         * config/rs6000/power9.md: Adjust.
27817         * config/rs6000/rs64.md: Adjust.
27818         * config/rs6000/titan.md: Adjust.
27820 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27822         * config/i386/predicates.md (indirect_branch_operand): Rewrite
27823         ix86_indirect_branch_register logic.
27825 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27827         * config/i386/constraints.md (Bs): Update
27828         ix86_indirect_branch_register check.  Don't check
27829         ix86_indirect_branch_register with GOT_memory_operand.
27830         (Bw): Likewise.
27831         * config/i386/predicates.md (GOT_memory_operand): Don't check
27832         ix86_indirect_branch_register here.
27833         (GOT32_symbol_operand): Likewise.
27835 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27837         * config/i386/predicates.md (constant_call_address_operand):
27838         Rewrite ix86_indirect_branch_register logic.
27839         (sibcall_insn_operand): Likewise.
27841 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27843         * config/i386/constraints.md (Bs): Replace
27844         ix86_indirect_branch_thunk_register with
27845         ix86_indirect_branch_register.
27846         (Bw): Likewise.
27847         * config/i386/i386.md (indirect_jump): Likewise.
27848         (tablejump): Likewise.
27849         (*sibcall_memory): Likewise.
27850         (*sibcall_value_memory): Likewise.
27851         Peepholes of indirect call and jump via memory: Likewise.
27852         * config/i386/i386.opt: Likewise.
27853         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27854         (GOT_memory_operand): Likewise.
27855         (call_insn_operand): Likewise.
27856         (sibcall_insn_operand): Likewise.
27857         (GOT32_symbol_operand): Likewise.
27859 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27861         PR middle-end/83837
27862         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
27863         type rather than type addr's type points to.
27864         (expand_omp_atomic_mutex): Likewise.
27865         (expand_omp_atomic): Likewise.
27867 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27869         PR target/83839
27870         * config/i386/i386.c (output_indirect_thunk_function): Use
27871         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
27872         for  __x86_return_thunk.
27874 2018-01-15  Richard Biener  <rguenther@suse.de>
27876         PR middle-end/83850
27877         * expmed.c (extract_bit_field_1): Fix typo.
27879 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27881         PR target/83687
27882         * config/arm/iterators.md (VF): New mode iterator.
27883         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
27884         Remove integer-related logic from pattern.
27885         (neon_vabd<mode>_3): Likewise.
27887 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27889         PR middle-end/82694
27890         * common.opt (fstrict-overflow): No longer an alias.
27891         (fwrapv-pointer): New option.
27892         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
27893         also for pointer types based on flag_wrapv_pointer.
27894         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
27895         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
27896         opts->x_flag_wrapv got set.
27897         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
27898         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
27899         POINTER_TYPE_OVERFLOW_UNDEFINED.
27900         * match.pd: Likewise in address comparison pattern.
27901         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
27903 2018-01-15  Richard Biener  <rguenther@suse.de>
27905         PR lto/83804
27906         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
27907         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
27908         Reset type names to their identifier if their TYPE_DECL doesn't
27909         have linkage (and thus is used for ODR and devirt).
27910         (save_debug_info_for_decl): Remove.
27911         (save_debug_info_for_type): Likewise.
27912         (add_tree_to_fld_list): Adjust.
27913         * tree-pretty-print.c (dump_generic_node): Make dumping of
27914         type names more robust.
27916 2018-01-15  Richard Biener  <rguenther@suse.de>
27918         * BASE-VER: Bump to 8.0.1.
27920 2018-01-14  Martin Sebor  <msebor@redhat.com>
27922         PR other/83508
27923         * builtins.c (check_access): Avoid warning when the no-warning bit
27924         is set.
27926 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
27928         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
27929         * ira-color (allocno_hard_regs_compare): Likewise.
27931 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
27933         PR target/83013
27934         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
27935         Use .pushsection/.popsection.
27937 2018-01-14  Martin Sebor  <msebor@redhat.com>
27939         PR c++/81327
27940         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
27942 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
27944         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
27945         entry from extra_headers.
27946         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
27947         extra_headers, make the list bitwise identical to the i?86-*-* one.
27949 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27951         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
27952         -mcmodel=large with -mindirect-branch=thunk,
27953         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
27954         -mfunction-return=thunk-extern.
27955         * doc/invoke.texi: Document -mcmodel=large is incompatible with
27956         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
27957         -mfunction-return=thunk and -mfunction-return=thunk-extern.
27959 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27961         * config/i386/i386.c (print_reg): Print the name of the full
27962         integer register without '%'.
27963         (ix86_print_operand): Handle 'V'.
27964         * doc/extend.texi: Document 'V' modifier.
27966 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27968         * config/i386/constraints.md (Bs): Disallow memory operand for
27969         -mindirect-branch-register.
27970         (Bw): Likewise.
27971         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27972         (GOT_memory_operand): Likewise.
27973         (call_insn_operand): Likewise.
27974         (sibcall_insn_operand): Likewise.
27975         (GOT32_symbol_operand): Likewise.
27976         * config/i386/i386.md (indirect_jump): Call convert_memory_address
27977         for -mindirect-branch-register.
27978         (tablejump): Likewise.
27979         (*sibcall_memory): Likewise.
27980         (*sibcall_value_memory): Likewise.
27981         Disallow peepholes of indirect call and jump via memory for
27982         -mindirect-branch-register.
27983         (*call_pop): Replace m with Bw.
27984         (*call_value_pop): Likewise.
27985         (*sibcall_pop_memory): Replace m with Bs.
27986         * config/i386/i386.opt (mindirect-branch-register): New option.
27987         * doc/invoke.texi: Document -mindirect-branch-register option.
27989 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27991         * config/i386/i386-protos.h (ix86_output_function_return): New.
27992         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
27993         set function_return_type.
27994         (indirect_thunk_name): Add ret_p to indicate thunk for function
27995         return.
27996         (output_indirect_thunk_function): Pass false to
27997         indirect_thunk_name.
27998         (ix86_output_indirect_branch_via_reg): Likewise.
27999         (ix86_output_indirect_branch_via_push): Likewise.
28000         (output_indirect_thunk_function): Create alias for function
28001         return thunk if regno < 0.
28002         (ix86_output_function_return): New function.
28003         (ix86_handle_fndecl_attribute): Handle function_return.
28004         (ix86_attribute_table): Add function_return.
28005         * config/i386/i386.h (machine_function): Add
28006         function_return_type.
28007         * config/i386/i386.md (simple_return_internal): Use
28008         ix86_output_function_return.
28009         (simple_return_internal_long): Likewise.
28010         * config/i386/i386.opt (mfunction-return=): New option.
28011         (indirect_branch): Mention -mfunction-return=.
28012         * doc/extend.texi: Document function_return function attribute.
28013         * doc/invoke.texi: Document -mfunction-return= option.
28015 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28017         * config/i386/i386-opts.h (indirect_branch): New.
28018         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28019         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28020         with local indirect jump when converting indirect call and jump.
28021         (ix86_set_indirect_branch_type): New.
28022         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28023         (indirectlabelno): New.
28024         (indirect_thunk_needed): Likewise.
28025         (indirect_thunk_bnd_needed): Likewise.
28026         (indirect_thunks_used): Likewise.
28027         (indirect_thunks_bnd_used): Likewise.
28028         (INDIRECT_LABEL): Likewise.
28029         (indirect_thunk_name): Likewise.
28030         (output_indirect_thunk): Likewise.
28031         (output_indirect_thunk_function): Likewise.
28032         (ix86_output_indirect_branch_via_reg): Likewise.
28033         (ix86_output_indirect_branch_via_push): Likewise.
28034         (ix86_output_indirect_branch): Likewise.
28035         (ix86_output_indirect_jmp): Likewise.
28036         (ix86_code_end): Call output_indirect_thunk_function if needed.
28037         (ix86_output_call_insn): Call ix86_output_indirect_branch if
28038         needed.
28039         (ix86_handle_fndecl_attribute): Handle indirect_branch.
28040         (ix86_attribute_table): Add indirect_branch.
28041         * config/i386/i386.h (machine_function): Add indirect_branch_type
28042         and has_local_indirect_jump.
28043         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28044         to true.
28045         (tablejump): Likewise.
28046         (*indirect_jump): Use ix86_output_indirect_jmp.
28047         (*tablejump_1): Likewise.
28048         (simple_return_indirect_internal): Likewise.
28049         * config/i386/i386.opt (mindirect-branch=): New option.
28050         (indirect_branch): New.
28051         (keep): Likewise.
28052         (thunk): Likewise.
28053         (thunk-inline): Likewise.
28054         (thunk-extern): Likewise.
28055         * doc/extend.texi: Document indirect_branch function attribute.
28056         * doc/invoke.texi: Document -mindirect-branch= option.
28058 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
28060         PR ipa/83051
28061         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28063 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
28065         * ipa-inline.c (want_inline_small_function_p): Return false if
28066         inlining has already failed with CIF_FINAL_ERROR.
28067         (update_caller_keys): Call want_inline_small_function_p before
28068         can_inline_edge_p.
28069         (update_callee_keys): Likewise.
28071 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28073         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
28074         New function.
28075         (rs6000_quadword_masked_address_p): Likewise.
28076         (quad_aligned_load_p): Likewise.
28077         (quad_aligned_store_p): Likewise.
28078         (const_load_sequence_p): Add comment to describe the outer-most loop.
28079         (mimic_memory_attributes_and_flags): New function.
28080         (rs6000_gen_stvx): Likewise.
28081         (replace_swapped_aligned_store): Likewise.
28082         (rs6000_gen_lvx): Likewise.
28083         (replace_swapped_aligned_load): Likewise.
28084         (replace_swapped_load_constant): Capitalize argument name in
28085         comment describing this function.
28086         (rs6000_analyze_swaps): Add a third pass to search for vector loads
28087         and stores that access quad-word aligned addresses and replace
28088         with stvx or lvx instructions when appropriate.
28089         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
28090         New function prototype.
28091         (rs6000_quadword_masked_address_p): Likewise.
28092         (rs6000_gen_lvx): Likewise.
28093         (rs6000_gen_stvx): Likewise.
28094         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
28095         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
28096         when memory address is aligned.
28097         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
28098         this split to select lvx instruction when memory address is aligned.
28099         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
28100         instruction when memory address is aligned.
28101         (*vsx_le_perm_load_v16qi): Likewise.
28102         (four unnamed splitters): Modify to select the stvx instruction
28103         when memory is aligned.
28105 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
28107         * predict.c (determine_unlikely_bbs): Handle correctly BBs
28108         which appears in the queue multiple times.
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         * tree-vectorizer.h (vec_lower_bound): New structure.
28115         (_loop_vec_info): Add check_nonzero and lower_bounds.
28116         (LOOP_VINFO_CHECK_NONZERO): New macro.
28117         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
28118         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
28119         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
28120         fields.  Make seg_len the distance travelled, not including the
28121         access size.
28122         (dr_direction_indicator): Declare.
28123         (dr_zero_step_indicator): Likewise.
28124         (dr_known_forward_stride_p): Likewise.
28125         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
28126         tree-ssanames.h.
28127         (runtime_alias_check_p): Allow runtime alias checks with
28128         variable strides.
28129         (operator ==): Compare access_size and align.
28130         (prune_runtime_alias_test_list): Rework for new distinction between
28131         the access_size and seg_len.
28132         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
28133         segment lengths.
28134         (get_segment_min_max): New function.
28135         (create_intersect_range_checks): Use it.
28136         (dr_step_indicator): New function.
28137         (dr_direction_indicator): Likewise.
28138         (dr_zero_step_indicator): Likewise.
28139         (dr_known_forward_stride_p): Likewise.
28140         * tree-loop-distribution.c (data_ref_segment_size): Return
28141         DR_STEP * (niters - 1).
28142         (compute_alias_check_pairs): Update call to the dr_with_seg_len
28143         constructor.
28144         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28145         (vect_preserves_scalar_order_p): New function, split out from...
28146         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
28147         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28148         (vect_vfa_access_size): New function.
28149         (vect_vfa_align): Likewise.
28150         (vect_compile_time_alias): Take access_size_a and access_b arguments.
28151         (dump_lower_bound): New function.
28152         (vect_check_lower_bound): Likewise.
28153         (vect_small_gap_p): Likewise.
28154         (vectorizable_with_step_bound_p): Likewise.
28155         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28156         depencies if the vectorization factor is 1.  Convert the checks
28157         for nonzero steps into checks on the bounds of DR_STEP.  Try using
28158         a bunds check for variable steps if the minimum required step is
28159         relatively small. Update calls to the dr_with_seg_len
28160         constructor and to vect_compile_time_alias.
28161         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28162         function.
28163         (vect_loop_versioning): Call it.
28164         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28165         when retrying.
28166         (vect_estimate_min_profitable_iters): Account for any bounds checks.
28168 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28169             Alan Hayward  <alan.hayward@arm.com>
28170             David Sherwood  <david.sherwood@arm.com>
28172         * doc/sourcebuild.texi (vect_scatter_store): Document.
28173         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28174         optabs.
28175         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28176         Document.
28177         * genopinit.c (main): Add supports_vec_scatter_store and
28178         supports_vec_scatter_store_cached to target_optabs.
28179         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28180         IFN_MASK_SCATTER_STORE.
28181         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28182         functions.
28183         * internal-fn.h (internal_store_fn_p): Declare.
28184         (internal_fn_stored_value_index): Likewise.
28185         * internal-fn.c (scatter_store_direct): New macro.
28186         (expand_scatter_store_optab_fn): New function.
28187         (direct_scatter_store_optab_supported_p): New macro.
28188         (internal_store_fn_p): New function.
28189         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28190         IFN_MASK_SCATTER_STORE.
28191         (internal_fn_mask_index): Likewise.
28192         (internal_fn_stored_value_index): New function.
28193         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28194         for scatter stores.
28195         * optabs-query.h (supports_vec_scatter_store_p): Declare.
28196         * optabs-query.c (supports_vec_scatter_store_p): New function.
28197         * tree-vectorizer.h (vect_get_store_rhs): Declare.
28198         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28199         true for scatter stores.
28200         (vect_gather_scatter_fn_p): Handle scatter stores too.
28201         (vect_check_gather_scatter): Consider using scatter stores if
28202         supports_vec_scatter_store_p.
28203         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28204         scatter stores too.
28205         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28206         internal_fn_stored_value_index.
28207         (check_load_store_masking): Handle scatter stores too.
28208         (vect_get_store_rhs): Make public.
28209         (vectorizable_call): Use internal_store_fn_p.
28210         (vectorizable_store): Handle scatter store internal functions.
28211         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28212         when deciding whether the end of the group has been reached.
28213         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28214         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28215         (mask_scatter_store<mode>): New insns.
28217 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28218             Alan Hayward  <alan.hayward@arm.com>
28219             David Sherwood  <david.sherwood@arm.com>
28221         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28222         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28223         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28224         function.
28225         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28226         Use vect_truncate_gather_scatter_offset if we can't treat the
28227         operation as a normal gather load or scatter store.
28228         (get_group_load_store_type): Take the gather_scatter_info
28229         as argument.  Try using a gather load or scatter store for
28230         single-element groups.
28231         (get_load_store_type): Update calls to get_group_load_store_type
28232         and vect_use_strided_gather_scatters_p.
28234 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28235             Alan Hayward  <alan.hayward@arm.com>
28236             David Sherwood  <david.sherwood@arm.com>
28238         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28239         optional tree argument.
28240         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28241         null target hooks.
28242         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28243         but continue to use the current value as a fallback.
28244         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28245         to compare the updates.
28246         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28247         (get_load_store_type): Use it when handling a strided access.
28248         (vect_get_strided_load_store_ops): New function.
28249         (vect_get_data_ptr_increment): Likewise.
28250         (vectorizable_load): Handle strided gather loads.  Always pass
28251         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28253 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28254             Alan Hayward  <alan.hayward@arm.com>
28255             David Sherwood  <david.sherwood@arm.com>
28257         * doc/md.texi (gather_load@var{m}): Document.
28258         (mask_gather_load@var{m}): Likewise.
28259         * genopinit.c (main): Add supports_vec_gather_load and
28260         supports_vec_gather_load_cached to target_optabs.
28261         * optabs-tree.c (init_tree_optimization_optabs): Use
28262         ggc_cleared_alloc to allocate target_optabs.
28263         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28264         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28265         functions.
28266         * internal-fn.h (internal_load_fn_p): Declare.
28267         (internal_gather_scatter_fn_p): Likewise.
28268         (internal_fn_mask_index): Likewise.
28269         (internal_gather_scatter_fn_supported_p): Likewise.
28270         * internal-fn.c (gather_load_direct): New macro.
28271         (expand_gather_load_optab_fn): New function.
28272         (direct_gather_load_optab_supported_p): New macro.
28273         (direct_internal_fn_optab): New function.
28274         (internal_load_fn_p): Likewise.
28275         (internal_gather_scatter_fn_p): Likewise.
28276         (internal_fn_mask_index): Likewise.
28277         (internal_gather_scatter_fn_supported_p): Likewise.
28278         * optabs-query.c (supports_at_least_one_mode_p): New function.
28279         (supports_vec_gather_load_p): Likewise.
28280         * optabs-query.h (supports_vec_gather_load_p): Declare.
28281         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28282         and memory_type field.
28283         (NUM_PATTERNS): Bump to 15.
28284         * tree-vect-data-refs.c: Include internal-fn.h.
28285         (vect_gather_scatter_fn_p): New function.
28286         (vect_describe_gather_scatter_call): Likewise.
28287         (vect_check_gather_scatter): Try using internal functions for
28288         gather loads.  Recognize existing calls to a gather load function.
28289         (vect_analyze_data_refs): Consider using gather loads if
28290         supports_vec_gather_load_p.
28291         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28292         (vect_get_gather_scatter_offset_type): Likewise.
28293         (vect_convert_mask_for_vectype): Likewise.
28294         (vect_add_conversion_to_patterm): Likewise.
28295         (vect_try_gather_scatter_pattern): Likewise.
28296         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28297         (vect_vect_recog_func_ptrs): Add it.
28298         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28299         internal_fn_mask_index and internal_gather_scatter_fn_p.
28300         (check_load_store_masking): Take the gather_scatter_info as an
28301         argument and handle gather loads.
28302         (vect_get_gather_scatter_ops): New function.
28303         (vectorizable_call): Check internal_load_fn_p.
28304         (vectorizable_load): Likewise.  Handle gather load internal
28305         functions.
28306         (vectorizable_store): Update call to check_load_store_masking.
28307         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28308         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28309         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28310         (aarch64_gather_scale_operand_d): New predicates.
28311         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28312         (mask_gather_load<mode>): New insns.
28314 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28315             Alan Hayward  <alan.hayward@arm.com>
28316             David Sherwood  <david.sherwood@arm.com>
28318         * optabs.def (fold_left_plus_optab): New optab.
28319         * doc/md.texi (fold_left_plus_@var{m}): Document.
28320         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28321         * internal-fn.c (fold_left_direct): Define.
28322         (expand_fold_left_optab_fn): Likewise.
28323         (direct_fold_left_optab_supported_p): Likewise.
28324         * fold-const-call.c (fold_const_fold_left): New function.
28325         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28326         * tree-parloops.c (valid_reduction_p): New function.
28327         (gather_scalar_reductions): Use it.
28328         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28329         (vect_finish_replace_stmt): Declare.
28330         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28331         (needs_fold_left_reduction_p): New function, split out from...
28332         (vect_is_simple_reduction): ...here.  Accept reductions that
28333         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28334         (vect_force_simple_reduction): Also store the reduction type in
28335         the assignment's STMT_VINFO_REDUC_TYPE.
28336         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28337         (merge_with_identity): New function.
28338         (vect_expand_fold_left): Likewise.
28339         (vectorize_fold_left_reduction): Likewise.
28340         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28341         scalar phi in place for it.  Check for target support and reject
28342         cases that would reassociate the operation.  Defer the transform
28343         phase to vectorize_fold_left_reduction.
28344         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28345         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28346         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28348 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28350         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28351         call to ifc_temp_var.
28353 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28354             Alan Hayward  <alan.hayward@arm.com>
28355             David Sherwood  <david.sherwood@arm.com>
28357         * target.def (legitimize_address_displacement): Take the original
28358         offset as a poly_int.
28359         * targhooks.h (default_legitimize_address_displacement): Update
28360         accordingly.
28361         * targhooks.c (default_legitimize_address_displacement): Likewise.
28362         * doc/tm.texi: Regenerate.
28363         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28364         as an argument, moving assert of ad->disp == ad->disp_term to...
28365         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28366         Try calling targetm.legitimize_address_displacement before expanding
28367         the address rather than afterwards, and adjust for the new interface.
28368         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28369         Match the new hook interface.  Handle SVE addresses.
28370         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28371         new hook interface.
28373 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28375         * Makefile.in (OBJS): Add early-remat.o.
28376         * target.def (select_early_remat_modes): New hook.
28377         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28378         * doc/tm.texi: Regenerate.
28379         * targhooks.h (default_select_early_remat_modes): Declare.
28380         * targhooks.c (default_select_early_remat_modes): New function.
28381         * timevar.def (TV_EARLY_REMAT): New timevar.
28382         * passes.def (pass_early_remat): New pass.
28383         * tree-pass.h (make_pass_early_remat): Declare.
28384         * early-remat.c: New file.
28385         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28386         function.
28387         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28389 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28390             Alan Hayward  <alan.hayward@arm.com>
28391             David Sherwood  <david.sherwood@arm.com>
28393         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28394         vfm1 with a bound_epilog parameter.
28395         (vect_do_peeling): Update calls accordingly, and move the prologue
28396         call earlier in the function.  Treat the base bound_epilog as 0 for
28397         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28398         this base when peeling for gaps.
28399         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28400         with fully-masked loops.
28401         (vect_estimate_min_profitable_iters): Handle the single peeled
28402         iteration in that case.
28404 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28405             Alan Hayward  <alan.hayward@arm.com>
28406             David Sherwood  <david.sherwood@arm.com>
28408         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28409         single-element interleaving even if the size is not a power of 2.
28410         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28411         accesses for single-element interleaving if the group size is
28412         not a power of 2.
28414 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28415             Alan Hayward  <alan.hayward@arm.com>
28416             David Sherwood  <david.sherwood@arm.com>
28418         * doc/md.texi (fold_extract_last_@var{m}): Document.
28419         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28420         * optabs.def (fold_extract_last_optab): New optab.
28421         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28422         * internal-fn.c (fold_extract_direct): New macro.
28423         (expand_fold_extract_optab_fn): Likewise.
28424         (direct_fold_extract_optab_supported_p): Likewise.
28425         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28426         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28427         EXTRACT_LAST_REDUCTION.
28428         (get_initial_def_for_reduction): Do not create an initial vector
28429         for EXTRACT_LAST_REDUCTION reductions.
28430         (vectorizable_reduction): Leave the scalar phi in place for
28431         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28432         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28433         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28434         transform phase to vectorizable_condition.
28435         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28436         split out from...
28437         (vect_finish_stmt_generation): ...here.
28438         (vect_finish_replace_stmt): New function.
28439         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28440         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28441         pattern.
28442         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28444 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28445             Alan Hayward  <alan.hayward@arm.com>
28446             David Sherwood  <david.sherwood@arm.com>
28448         * doc/md.texi (extract_last_@var{m}): Document.
28449         * optabs.def (extract_last_optab): New optab.
28450         * internal-fn.def (EXTRACT_LAST): New internal function.
28451         * internal-fn.c (cond_unary_direct): New macro.
28452         (expand_cond_unary_optab_fn): Likewise.
28453         (direct_cond_unary_optab_supported_p): Likewise.
28454         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28455         loops using EXTRACT_LAST.
28456         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28457         (extract_last_<mode>): ...this optab.
28458         (vec_extract<mode><Vel>): Update accordingly.
28460 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28461             Alan Hayward  <alan.hayward@arm.com>
28462             David Sherwood  <david.sherwood@arm.com>
28464         * target.def (empty_mask_is_expensive): New hook.
28465         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28466         * doc/tm.texi: Regenerate.
28467         * targhooks.h (default_empty_mask_is_expensive): Declare.
28468         * targhooks.c (default_empty_mask_is_expensive): New function.
28469         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28470         if the target says that empty masks are expensive.
28471         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28472         New function.
28473         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28475 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28476             Alan Hayward  <alan.hayward@arm.com>
28477             David Sherwood  <david.sherwood@arm.com>
28479         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28480         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28481         (vect_use_loop_mask_for_alignment_p): New function.
28482         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28483         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28484         niters_skip argument.  Make sure that the first niters_skip elements
28485         of the first iteration are inactive.
28486         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28487         Update call to vect_set_loop_masks_directly.
28488         (get_misalign_in_elems): New function, split out from...
28489         (vect_gen_prolog_loop_niters): ...here.
28490         (vect_update_init_of_dr): Take a code argument that specifies whether
28491         the adjustment should be added or subtracted.
28492         (vect_update_init_of_drs): Likewise.
28493         (vect_prepare_for_masked_peels): New function.
28494         (vect_do_peeling): Skip prologue peeling if we're using a mask
28495         instead.  Update call to vect_update_inits_of_drs.
28496         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28497         mask_skip_niters.
28498         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28499         alignment.  Do not include the number of peeled iterations in
28500         the minimum threshold in that case.
28501         (vectorizable_induction): Adjust the start value down by
28502         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28503         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28504         Take the number of skipped iterations into account when calculating
28505         the loop bounds.
28506         * tree-vect-stmts.c (vect_gen_while_not): New function.
28508 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28509             Alan Hayward  <alan.hayward@arm.com>
28510             David Sherwood  <david.sherwood@arm.com>
28512         * doc/sourcebuild.texi (vect_fully_masked): Document.
28513         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28514         default value to 0.
28515         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28516         split out from...
28517         (vect_analyze_loop_2): ...here. Don't check the vectorization
28518         factor against the number of loop iterations if the loop is
28519         fully-masked.
28521 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28522             Alan Hayward  <alan.hayward@arm.com>
28523             David Sherwood  <david.sherwood@arm.com>
28525         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28526         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28527         (dump_groups): Update accordingly.
28528         (iv_use::mem_type): New member variable.
28529         (address_p): New function.
28530         (record_use): Add a mem_type argument and initialize the new
28531         mem_type field.
28532         (record_group_use): Add a mem_type argument.  Use address_p.
28533         Remove obsolete null checks of base_object.  Update call to record_use.
28534         (find_interesting_uses_op): Update call to record_group_use.
28535         (find_interesting_uses_cond): Likewise.
28536         (find_interesting_uses_address): Likewise.
28537         (get_mem_type_for_internal_fn): New function.
28538         (find_address_like_use): Likewise.
28539         (find_interesting_uses_stmt): Try find_address_like_use before
28540         calling find_interesting_uses_op.
28541         (addr_offset_valid_p): Use the iv mem_type field as the type
28542         of the addressed memory.
28543         (add_autoinc_candidates): Likewise.
28544         (get_address_cost): Likewise.
28545         (split_small_address_groups_p): Use address_p.
28546         (split_address_groups): Likewise.
28547         (add_iv_candidate_for_use): Likewise.
28548         (autoinc_possible_for_pair): Likewise.
28549         (rewrite_groups): Likewise.
28550         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28551         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28552         (get_alias_ptr_type_for_ptr_address): New function.
28553         (rewrite_use_address): Rewrite address uses in calls that were
28554         identified by find_address_like_use.
28556 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28557             Alan Hayward  <alan.hayward@arm.com>
28558             David Sherwood  <david.sherwood@arm.com>
28560         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28561         TARGET_MEM_REFs.
28562         * gimple-expr.h (is_gimple_addressable: Likewise.
28563         * gimple-expr.c (is_gimple_address): Likewise.
28564         * internal-fn.c (expand_call_mem_ref): New function.
28565         (expand_mask_load_optab_fn): Use it.
28566         (expand_mask_store_optab_fn): Likewise.
28568 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28569             Alan Hayward  <alan.hayward@arm.com>
28570             David Sherwood  <david.sherwood@arm.com>
28572         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28573         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28574         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28575         (cond_umax@var{mode}): Document.
28576         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28577         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28578         (cond_umin_optab, cond_umax_optab): New optabs.
28579         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28580         (COND_IOR, COND_XOR): New internal functions.
28581         * internal-fn.h (get_conditional_internal_fn): Declare.
28582         * internal-fn.c (cond_binary_direct): New macro.
28583         (expand_cond_binary_optab_fn): Likewise.
28584         (direct_cond_binary_optab_supported_p): Likewise.
28585         (get_conditional_internal_fn): New function.
28586         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28587         Cope with reduction statements that are vectorized as calls rather
28588         than assignments.
28589         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28590         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28591         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28592         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28593         (UNSPEC_COND_EOR): New unspecs.
28594         (optab): Add mappings for them.
28595         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28596         (sve_int_op, sve_fp_op): New int attributes.
28598 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28599             Alan Hayward  <alan.hayward@arm.com>
28600             David Sherwood  <david.sherwood@arm.com>
28602         * optabs.def (while_ult_optab): New optab.
28603         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28604         * internal-fn.def (WHILE_ULT): New internal function.
28605         * internal-fn.h (direct_internal_fn_supported_p): New override
28606         that takes two types as argument.
28607         * internal-fn.c (while_direct): New macro.
28608         (expand_while_optab_fn): New function.
28609         (convert_optab_supported_p): Likewise.
28610         (direct_while_optab_supported_p): New macro.
28611         * wide-int.h (wi::udiv_ceil): New function.
28612         * tree-vectorizer.h (rgroup_masks): New structure.
28613         (vec_loop_masks): New typedef.
28614         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28615         and fully_masked_p.
28616         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28617         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28618         (vect_max_vf): New function.
28619         (slpeel_make_loop_iterate_ntimes): Delete.
28620         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28621         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28622         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28623         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28624         internal-fn.h, stor-layout.h and optabs-query.h.
28625         (vect_set_loop_mask): New function.
28626         (add_preheader_seq): Likewise.
28627         (add_header_seq): Likewise.
28628         (interleave_supported_p): Likewise.
28629         (vect_maybe_permute_loop_masks): Likewise.
28630         (vect_set_loop_masks_directly): Likewise.
28631         (vect_set_loop_condition_masked): Likewise.
28632         (vect_set_loop_condition_unmasked): New function, split out from
28633         slpeel_make_loop_iterate_ntimes.
28634         (slpeel_make_loop_iterate_ntimes): Rename to..
28635         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28636         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28637         (vect_do_peeling): Update call accordingly.
28638         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28639         loops.
28640         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28641         mask_compare_type, can_fully_mask_p and fully_masked_p.
28642         (release_vec_loop_masks): New function.
28643         (_loop_vec_info): Use it to free the loop masks.
28644         (can_produce_all_loop_masks_p): New function.
28645         (vect_get_max_nscalars_per_iter): Likewise.
28646         (vect_verify_full_masking): Likewise.
28647         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28648         retries, and free the mask rgroups before retrying.  Check loop-wide
28649         reasons for disallowing fully-masked loops.  Make the final decision
28650         about whether use a fully-masked loop or not.
28651         (vect_estimate_min_profitable_iters): Do not assume that peeling
28652         for the number of iterations will be needed for fully-masked loops.
28653         (vectorizable_reduction): Disable fully-masked loops.
28654         (vectorizable_live_operation): Likewise.
28655         (vect_halve_mask_nunits): New function.
28656         (vect_double_mask_nunits): Likewise.
28657         (vect_record_loop_mask): Likewise.
28658         (vect_get_loop_mask): Likewise.
28659         (vect_transform_loop): Handle the case in which the final loop
28660         iteration might handle a partial vector.  Call vect_set_loop_condition
28661         instead of slpeel_make_loop_iterate_ntimes.
28662         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28663         (check_load_store_masking): New function.
28664         (prepare_load_store_mask): Likewise.
28665         (vectorizable_store): Handle fully-masked loops.
28666         (vectorizable_load): Likewise.
28667         (supportable_widening_operation): Use vect_halve_mask_nunits for
28668         booleans.
28669         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28670         (vect_gen_while): New function.
28671         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28672         (aarch64_uqdec<mode>): New insn.
28674 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28675             Alan Hayward  <alan.hayward@arm.com>
28676             David Sherwood  <david.sherwood@arm.com>
28678         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28679         (reduc_xor_scal_optab): New optabs.
28680         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28681         (reduc_xor_scal_@var{m}): Document.
28682         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28683         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28684         internal functions.
28685         * fold-const-call.c (fold_const_call): Handle them.
28686         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28687         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28688         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28689         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28690         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28691         (UNSPEC_XORV): New unspecs.
28692         (optab): Add entries for them.
28693         (BITWISEV): New int iterator.
28694         (bit_reduc_op): New int attributes.
28696 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28697             Alan Hayward  <alan.hayward@arm.com>
28698             David Sherwood  <david.sherwood@arm.com>
28700         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28701         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28702         * optabs.def (vec_shl_insert_optab): New optab.
28703         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28704         (duplicate_and_interleave): Likewise.
28705         * tree-vect-loop.c: Include internal-fn.h.
28706         (neutral_op_for_slp_reduction): New function, split out from
28707         get_initial_defs_for_reduction.
28708         (get_initial_def_for_reduction): Handle option 2 for variable-length
28709         vectors by loading the neutral value into a vector and then shifting
28710         the initial value into element 0.
28711         (get_initial_defs_for_reduction): Replace the code argument with
28712         the neutral value calculated by neutral_op_for_slp_reduction.
28713         Use gimple_build_vector for constant-length vectors.
28714         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28715         but the first group_size elements have a neutral value.
28716         Use duplicate_and_interleave otherwise.
28717         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28718         Update call to get_initial_defs_for_reduction.  Handle SLP
28719         reductions for variable-length vectors by creating one vector
28720         result for each scalar result, with the elements associated
28721         with other scalar results stubbed out with the neutral value.
28722         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
28723         Require IFN_VEC_SHL_INSERT for double reductions on
28724         variable-length vectors, or SLP reductions that have
28725         a neutral value.  Require can_duplicate_and_interleave_p
28726         support for variable-length unchained SLP reductions if there
28727         is no neutral value, such as for MIN/MAX reductions.  Also require
28728         the number of vector elements to be a multiple of the number of
28729         SLP statements when doing variable-length unchained SLP reductions.
28730         Update call to vect_create_epilog_for_reduction.
28731         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
28732         and remove initial values.
28733         (duplicate_and_interleave): Make public.
28734         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
28735         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
28737 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28738             Alan Hayward  <alan.hayward@arm.com>
28739             David Sherwood  <david.sherwood@arm.com>
28741         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
28742         (can_duplicate_and_interleave_p): New function.
28743         (vect_get_and_check_slp_defs): Take the vector of statements
28744         rather than just the current one.  Remove excess parentheses.
28745         Restriction rejectinon of vect_constant_def and vect_external_def
28746         for variable-length vectors to boolean types, or types for which
28747         can_duplicate_and_interleave_p is false.
28748         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
28749         (duplicate_and_interleave): New function.
28750         (vect_get_constant_vectors): Use gimple_build_vector for
28751         constant-length vectors and suitable variable-length constant
28752         vectors.  Use duplicate_and_interleave for other variable-length
28753         vectors.  Don't defer the update when inserting new statements.
28755 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28756             Alan Hayward  <alan.hayward@arm.com>
28757             David Sherwood  <david.sherwood@arm.com>
28759         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
28760         min_profitable_iters doesn't go negative.
28762 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28763             Alan Hayward  <alan.hayward@arm.com>
28764             David Sherwood  <david.sherwood@arm.com>
28766         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
28767         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
28768         * optabs.def (vec_mask_load_lanes_optab): New optab.
28769         (vec_mask_store_lanes_optab): Likewise.
28770         * internal-fn.def (MASK_LOAD_LANES): New internal function.
28771         (MASK_STORE_LANES): Likewise.
28772         * internal-fn.c (mask_load_lanes_direct): New macro.
28773         (mask_store_lanes_direct): Likewise.
28774         (expand_mask_load_optab_fn): Handle masked operations.
28775         (expand_mask_load_lanes_optab_fn): New macro.
28776         (expand_mask_store_optab_fn): Handle masked operations.
28777         (expand_mask_store_lanes_optab_fn): New macro.
28778         (direct_mask_load_lanes_optab_supported_p): Likewise.
28779         (direct_mask_store_lanes_optab_supported_p): Likewise.
28780         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
28781         parameter.
28782         (vect_load_lanes_supported): Likewise.
28783         * tree-vect-data-refs.c (strip_conversion): New function.
28784         (can_group_stmts_p): Likewise.
28785         (vect_analyze_data_ref_accesses): Use it instead of checking
28786         for a pair of assignments.
28787         (vect_store_lanes_supported): Take a masked_p parameter.
28788         (vect_load_lanes_supported): Likewise.
28789         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
28790         vect_store_lanes_supported and vect_load_lanes_supported.
28791         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
28792         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
28793         parameter.  Don't allow gaps for masked accesses.
28794         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
28795         and vect_load_lanes_supported.
28796         (get_load_store_type): Take a masked_p parameter and update
28797         call to get_group_load_store_type.
28798         (vectorizable_store): Update call to get_load_store_type.
28799         Handle IFN_MASK_STORE_LANES.
28800         (vectorizable_load): Update call to get_load_store_type.
28801         Handle IFN_MASK_LOAD_LANES.
28803 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28804             Alan Hayward  <alan.hayward@arm.com>
28805             David Sherwood  <david.sherwood@arm.com>
28807         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
28808         modes for SVE.
28809         * config/aarch64/aarch64-protos.h
28810         (aarch64_sve_struct_memory_operand_p): Declare.
28811         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
28812         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
28813         (VPRED, vpred): Handle SVE structure modes.
28814         * config/aarch64/constraints.md (Utx): New constraint.
28815         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
28816         (aarch64_sve_struct_nonimmediate_operand): New predicates.
28817         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
28818         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
28819         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
28820         structure modes.  Split into pieces after RA.
28821         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
28822         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
28823         New patterns.
28824         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
28825         SVE structure modes.
28826         (aarch64_classify_address): Likewise.
28827         (sizetochar): Move earlier in file.
28828         (aarch64_print_operand): Handle SVE register lists.
28829         (aarch64_array_mode): New function.
28830         (aarch64_sve_struct_memory_operand_p): Likewise.
28831         (TARGET_ARRAY_MODE): Redefine.
28833 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28834             Alan Hayward  <alan.hayward@arm.com>
28835             David Sherwood  <david.sherwood@arm.com>
28837         * target.def (array_mode): New target hook.
28838         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
28839         * doc/tm.texi: Regenerate.
28840         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
28841         * hooks.c (hook_optmode_mode_uhwi_none): New function.
28842         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
28843         targetm.array_mode.
28844         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
28845         type sizes.
28847 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28848             Alan Hayward  <alan.hayward@arm.com>
28849             David Sherwood  <david.sherwood@arm.com>
28851         * fold-const.c (fold_binary_loc): Check the argument types
28852         rather than the result type when testing for a vector operation.
28854 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28856         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
28857         * doc/tm.texi: Regenerate.
28859 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28860             Alan Hayward  <alan.hayward@arm.com>
28861             David Sherwood  <david.sherwood@arm.com>
28863         * doc/invoke.texi (-msve-vector-bits=): Document new option.
28864         (sve): Document new AArch64 extension.
28865         * doc/md.texi (w): Extend the description of the AArch64
28866         constraint to include SVE vectors.
28867         (Upl, Upa): Document new AArch64 predicate constraints.
28868         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
28869         enum.
28870         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
28871         (msve-vector-bits=): New option.
28872         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
28873         SVE when these are disabled.
28874         (sve): New extension.
28875         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
28876         modes.  Adjust their number of units based on aarch64_sve_vg.
28877         (MAX_BITSIZE_MODE_ANY_MODE): Define.
28878         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
28879         aarch64_addr_query_type.
28880         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
28881         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
28882         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
28883         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
28884         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
28885         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
28886         (aarch64_simd_imm_zero_p): Delete.
28887         (aarch64_check_zero_based_sve_index_immediate): Declare.
28888         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28889         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28890         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28891         (aarch64_sve_float_mul_immediate_p): Likewise.
28892         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28893         rather than an rtx.
28894         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
28895         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
28896         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
28897         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
28898         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
28899         (aarch64_regmode_natural_size): Likewise.
28900         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
28901         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
28902         left one place.
28903         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
28904         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
28905         for VG and the SVE predicate registers.
28906         (V_ALIASES): Add a "z"-prefixed alias.
28907         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
28908         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
28909         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
28910         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
28911         (REG_CLASS_NAMES): Add entries for them.
28912         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
28913         and the predicate registers.
28914         (aarch64_sve_vg): Declare.
28915         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
28916         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
28917         (REGMODE_NATURAL_SIZE): Define.
28918         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
28919         SVE macros.
28920         * config/aarch64/aarch64.c: Include cfgrtl.h.
28921         (simd_immediate_info): Add a constructor for series vectors,
28922         and an associated step field.
28923         (aarch64_sve_vg): New variable.
28924         (aarch64_dbx_register_number): Handle VG and the predicate registers.
28925         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
28926         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
28927         (VEC_ANY_DATA, VEC_STRUCT): New constants.
28928         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
28929         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
28930         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
28931         (aarch64_get_mask_mode): New functions.
28932         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
28933         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
28934         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
28935         predicate modes and predicate registers.  Explicitly restrict
28936         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
28937         to store a vector mode if it is recognized by
28938         aarch64_classify_vector_mode.
28939         (aarch64_regmode_natural_size): New function.
28940         (aarch64_hard_regno_caller_save_mode): Return the original mode
28941         for predicates.
28942         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
28943         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
28944         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
28945         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
28946         functions.
28947         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
28948         does not overlap dest if the function is frame-related.  Handle
28949         SVE constants.
28950         (aarch64_split_add_offset): New function.
28951         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
28952         them aarch64_add_offset.
28953         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
28954         and update call to aarch64_sub_sp.
28955         (aarch64_add_cfa_expression): New function.
28956         (aarch64_expand_prologue): Pass extra temporary registers to the
28957         functions above.  Handle the case in which we need to emit new
28958         DW_CFA_expressions for registers that were originally saved
28959         relative to the stack pointer, but now have to be expressed
28960         relative to the frame pointer.
28961         (aarch64_output_mi_thunk): Pass extra temporary registers to the
28962         functions above.
28963         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
28964         IP0 and IP1 values for SVE frames.
28965         (aarch64_expand_vec_series): New function.
28966         (aarch64_expand_sve_widened_duplicate): Likewise.
28967         (aarch64_expand_sve_const_vector): Likewise.
28968         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
28969         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
28970         into the register, rather than emitting a SET directly.
28971         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
28972         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
28973         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
28974         (offset_9bit_signed_scaled_p): New functions.
28975         (aarch64_replicate_bitmask_imm): New function.
28976         (aarch64_bitmask_imm): Use it.
28977         (aarch64_cannot_force_const_mem): Reject expressions involving
28978         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
28979         (aarch64_classify_index): Handle SVE indices, by requiring
28980         a plain register index with a scale that matches the element size.
28981         (aarch64_classify_address): Handle SVE addresses.  Assert that
28982         the mode of the address is VOIDmode or an integer mode.
28983         Update call to aarch64_classify_symbol.
28984         (aarch64_classify_symbolic_expression): Update call to
28985         aarch64_classify_symbol.
28986         (aarch64_const_vec_all_in_range_p): New function.
28987         (aarch64_print_vector_float_operand): Likewise.
28988         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
28989         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
28990         and the FP immediates 1.0 and 0.5.
28991         (aarch64_print_address_internal): Handle SVE addresses.
28992         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
28993         (aarch64_regno_regclass): Handle predicate registers.
28994         (aarch64_secondary_reload): Handle big-endian reloads of SVE
28995         data modes.
28996         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
28997         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
28998         (aarch64_convert_sve_vector_bits): New function.
28999         (aarch64_override_options): Use it to handle -msve-vector-bits=.
29000         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29001         rather than an rtx.
29002         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29003         Handle SVE vector and predicate modes.  Accept VL-based constants
29004         that need only one temporary register, and VL offsets that require
29005         no temporary registers.
29006         (aarch64_conditional_register_usage): Mark the predicate registers
29007         as fixed if SVE isn't available.
29008         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29009         Return true for SVE vector and predicate modes.
29010         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29011         rather than an unsigned int.  Handle SVE modes.
29012         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
29013         SVE modes.
29014         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29015         if SVE is enabled.
29016         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29017         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29018         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29019         (aarch64_sve_float_mul_immediate_p): New functions.
29020         (aarch64_sve_valid_immediate): New function.
29021         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29022         Explicitly reject structure modes.  Check for INDEX constants.
29023         Handle PTRUE and PFALSE constants.
29024         (aarch64_check_zero_based_sve_index_immediate): New function.
29025         (aarch64_simd_imm_zero_p): Delete.
29026         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29027         vector modes.  Accept constants in the range of CNT[BHWD].
29028         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29029         ask for an Advanced SIMD mode.
29030         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29031         (aarch64_simd_vector_alignment): Handle SVE predicates.
29032         (aarch64_vectorize_preferred_vector_alignment): New function.
29033         (aarch64_simd_vector_alignment_reachable): Use it instead of
29034         the vector size.
29035         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29036         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29037         functions.
29038         (MAX_VECT_LEN): Delete.
29039         (expand_vec_perm_d): Add a vec_flags field.
29040         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29041         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29042         (aarch64_evpc_ext): Don't apply a big-endian lane correction
29043         for SVE modes.
29044         (aarch64_evpc_rev): Rename to...
29045         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
29046         (aarch64_evpc_rev_global): New function.
29047         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29048         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29049         MAX_VECT_LEN.
29050         (aarch64_evpc_sve_tbl): New function.
29051         (aarch64_expand_vec_perm_const_1): Update after rename of
29052         aarch64_evpc_rev.  Handle SVE permutes too, trying
29053         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29054         than aarch64_evpc_tbl.
29055         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29056         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29057         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29058         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29059         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29060         (aarch64_expand_sve_vcond): New functions.
29061         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29062         of aarch64_vector_mode_p.
29063         (aarch64_dwarf_poly_indeterminate_value): New function.
29064         (aarch64_compute_pressure_classes): Likewise.
29065         (aarch64_can_change_mode_class): Likewise.
29066         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29067         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29068         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29069         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
29070         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
29071         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
29072         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
29073         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
29074         constraints.
29075         (Dn, Dl, Dr): Accept const as well as const_vector.
29076         (Dz): Likewise.  Compare against CONST0_RTX.
29077         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
29078         of "vector" where appropriate.
29079         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
29080         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
29081         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
29082         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
29083         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
29084         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
29085         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
29086         (v_int_equiv): Extend to SVE modes.
29087         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
29088         mode attributes.
29089         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
29090         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
29091         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
29092         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
29093         (SVE_COND_FP_CMP): New int iterators.
29094         (perm_hilo): Handle the new unpack unspecs.
29095         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
29096         attributes.
29097         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
29098         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
29099         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
29100         (aarch64_equality_operator, aarch64_constant_vector_operand)
29101         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
29102         (aarch64_sve_nonimmediate_operand): Likewise.
29103         (aarch64_sve_general_operand): Likewise.
29104         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
29105         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
29106         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
29107         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
29108         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
29109         (aarch64_sve_float_arith_immediate): Likewise.
29110         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
29111         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
29112         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
29113         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
29114         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
29115         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
29116         (aarch64_sve_float_arith_operand): Likewise.
29117         (aarch64_sve_float_arith_with_sub_operand): Likewise.
29118         (aarch64_sve_float_mul_operand): Likewise.
29119         (aarch64_sve_vec_perm_operand): Likewise.
29120         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
29121         (aarch64_mov_operand): Accept const_poly_int and const_vector.
29122         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
29123         as well as const_vector.
29124         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
29125         in file.  Use CONST0_RTX and CONSTM1_RTX.
29126         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
29127         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
29128         Use aarch64_simd_imm_zero.
29129         * config/aarch64/aarch64-sve.md: New file.
29130         * config/aarch64/aarch64.md: Include it.
29131         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29132         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29133         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29134         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29135         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29136         (sve): New attribute.
29137         (enabled): Disable instructions with the sve attribute unless
29138         TARGET_SVE.
29139         (movqi, movhi): Pass CONST_POLY_INT operaneds through
29140         aarch64_expand_mov_immediate.
29141         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29142         CNT[BHSD] immediates.
29143         (movti): Split CONST_POLY_INT moves into two halves.
29144         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29145         Split additions that need a temporary here if the destination
29146         is the stack pointer.
29147         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29148         (*add<mode>3_poly_1): New instruction.
29149         (set_clobber_cc): New expander.
29151 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29153         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29154         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
29155         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29156         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29157         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
29158         Change innermode from fixed_mode_size to machine_mode.
29159         (simplify_subreg): Update call accordingly.  Handle a constant-sized
29160         subreg of a variable-length CONST_VECTOR.
29162 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29163             Alan Hayward  <alan.hayward@arm.com>
29164             David Sherwood  <david.sherwood@arm.com>
29166         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29167         (add_offset_to_base): New function, split out from...
29168         (create_mem_ref): ...here.  When handling a scale other than 1,
29169         check first whether the address is valid without the offset.
29170         Add it into the base if so, leaving the index and scale as-is.
29172 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29174         PR c++/83778
29175         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29176         fold_for_warn before checking if arg2 is INTEGER_CST.
29178 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
29180         * config/rs6000/predicates.md (load_multiple_operation): Delete.
29181         (store_multiple_operation): Delete.
29182         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29183         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29184         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29185         guarded by TARGET_STRING.
29186         (rs6000_output_load_multiple): Delete.
29187         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29188         OPTION_MASK_STRING / TARGET_STRING handling.
29189         (print_operand) <'N', 'O'>: Add comment that these are unused now.
29190         (const rs6000_opt_masks) <"string">: Change mask to 0.
29191         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29192         (MASK_STRING): Delete.
29193         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29194         parts.  Simplify.
29195         (load_multiple): Delete.
29196         (*ldmsi8): Delete.
29197         (*ldmsi7): Delete.
29198         (*ldmsi6): Delete.
29199         (*ldmsi5): Delete.
29200         (*ldmsi4): Delete.
29201         (*ldmsi3): Delete.
29202         (store_multiple): Delete.
29203         (*stmsi8): Delete.
29204         (*stmsi7): Delete.
29205         (*stmsi6): Delete.
29206         (*stmsi5): Delete.
29207         (*stmsi4): Delete.
29208         (*stmsi3): Delete.
29209         (movmemsi_8reg): Delete.
29210         (corresponding unnamed define_insn): Delete.
29211         (movmemsi_6reg): Delete.
29212         (corresponding unnamed define_insn): Delete.
29213         (movmemsi_4reg): Delete.
29214         (corresponding unnamed define_insn): Delete.
29215         (movmemsi_2reg): Delete.
29216         (corresponding unnamed define_insn): Delete.
29217         (movmemsi_1reg): Delete.
29218         (corresponding unnamed define_insn): Delete.
29219         * config/rs6000/rs6000.opt (mno-string): New.
29220         (mstring): Replace by deprecation warning stub.
29221         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29223 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29225         * regrename.c (regrename_do_replace): If replacing the same
29226         reg multiple times, try to reuse last created gen_raw_REG.
29228         PR debug/81155
29229         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29230         main to workaround a bug in GDB.
29232 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29234         PR target/83737
29235         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29237 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29239         PR rtl-optimization/80481
29240         * ira-color.c (get_cap_member): New function.
29241         (allocnos_conflict_by_live_ranges_p): Use it.
29242         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29243         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29245 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29247         PR target/83628
29248         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29249         (*saddl_se_1): Ditto.
29250         (*ssubsi_1): Ditto.
29251         (*ssubl_se_1): Ditto.
29253 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29255         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29256         rather than wi::to_widest for DR_INITs.
29257         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29258         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29259         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29260         INTEGER_CSTs.
29261         (vect_analyze_group_access_1): Note that here.
29263 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29265         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29266         polynomial type sizes.
29268 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29270         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29271         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29272         (gimple_add_tmp_var): Likewise.
29274 2018-01-12  Martin Liska  <mliska@suse.cz>
29276         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29277         (gimple_alloc_sizes): Likewise.
29278         (dump_gimple_statistics): Use PRIu64 in printf format.
29279         * gimple.h: Change uint64_t to int.
29281 2018-01-12  Martin Liska  <mliska@suse.cz>
29283         * tree-core.h: Use uint64_t instead of int.
29284         * tree.c (tree_node_counts): Likewise.
29285         (tree_node_sizes): Likewise.
29286         (dump_tree_statistics): Use PRIu64 in printf format.
29288 2018-01-12  Martin Liska  <mliska@suse.cz>
29290         * Makefile.in: As qsort_chk is implemented in vec.c, add
29291         vec.o to linkage of gencfn-macros.
29292         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29293         passing the info to record_node_allocation_statistics.
29294         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29295         and pass the info.
29296         * ggc-common.c (struct ggc_usage): Add operator== and use
29297         it in operator< and compare function.
29298         * mem-stats.h (struct mem_usage): Likewise.
29299         * vec.c (struct vec_usage): Remove operator< and compare
29300         function. Can be simply inherited.
29302 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29304         PR target/81616
29305         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29306         * tree-ssa-math-opts.c: Include domwalk.h.
29307         (convert_mult_to_fma_1): New function.
29308         (fma_transformation_info): New type.
29309         (fma_deferring_state): Likewise.
29310         (cancel_fma_deferring): New function.
29311         (result_of_phi): Likewise.
29312         (last_fma_candidate_feeds_initial_phi): Likewise.
29313         (convert_mult_to_fma): Added deferring logic, split actual
29314         transformation to convert_mult_to_fma_1.
29315         (math_opts_dom_walker): New type.
29316         (math_opts_dom_walker::after_dom_children): New method, body moved
29317         here from pass_optimize_widening_mul::execute, added deferring logic
29318         bits.
29319         (pass_optimize_widening_mul::execute): Moved most of code to
29320         math_opts_dom_walker::after_dom_children.
29321         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29322         * config/i386/i386.c (ix86_option_override_internal): Added
29323         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29325 2018-01-12  Richard Biener  <rguenther@suse.de>
29327         PR debug/83157
29328         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29329         inline instance vars.
29331 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29333         PR target/81819
29334         * config/rx/rx.c (rx_is_restricted_memory_address):
29335         Handle SUBREG case.
29337 2018-01-12  Richard Biener  <rguenther@suse.de>
29339         PR tree-optimization/80846
29340         * target.def (split_reduction): New target hook.
29341         * targhooks.c (default_split_reduction): New function.
29342         * targhooks.h (default_split_reduction): Declare.
29343         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29344         target requests first reduce vectors by combining low and high
29345         parts.
29346         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29347         (get_vectype_for_scalar_type_and_size): Export.
29348         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29349         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29350         * doc/tm.texi: Regenerate.
29351         * config/i386/i386.c (ix86_split_reduction): Implement
29352         TARGET_VECTORIZE_SPLIT_REDUCTION.
29354 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29356         PR target/83368
29357         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29358         in PIC mode except for TARGET_VXWORKS_RTP.
29359         * config/sparc/sparc.c: Include cfgrtl.h.
29360         (TARGET_INIT_PIC_REG): Define.
29361         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29362         (sparc_pic_register_p): New predicate.
29363         (sparc_legitimate_address_p): Use it.
29364         (sparc_legitimize_pic_address): Likewise.
29365         (sparc_delegitimize_address): Likewise.
29366         (sparc_mode_dependent_address_p): Likewise.
29367         (gen_load_pcrel_sym): Remove 4th parameter.
29368         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29369         (sparc_expand_prologue): Do not call load_got_register here.
29370         (sparc_flat_expand_prologue): Likewise.
29371         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29372         (sparc_use_pseudo_pic_reg): New function.
29373         (sparc_init_pic_reg): Likewise.
29374         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29375         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29377 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29379         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29380         Add item for branch_cost.
29382 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29384         PR rtl-optimization/83565
29385         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29386         not extend the result to a larger mode for rotate operations.
29387         (num_sign_bit_copies1): Likewise.
29389 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29391         PR target/40411
29392         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29393         -symbolic.
29394         Use values-Xc.o for -pedantic.
29395         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29397 2018-01-12  Martin Liska  <mliska@suse.cz>
29399         PR ipa/83054
29400         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29401         New function.
29402         (possible_polymorphic_call_targets): Use it.
29403         (ipa_devirt): Likewise.
29405 2018-01-12  Martin Liska  <mliska@suse.cz>
29407         * profile-count.h (enum profile_quality): Use 0 as invalid
29408         enum value of profile_quality.
29410 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29412         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29413         -mext-string options.
29415 2018-01-12  Richard Biener  <rguenther@suse.de>
29417         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29418         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29419         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29420         Likewise.
29421         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29423 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29425         * configure.ac (--with-long-double-format): Add support for the
29426         configuration option to change the default long double format on
29427         PowerPC systems.
29428         * config.gcc (powerpc*-linux*-*): Likewise.
29429         * configure: Regenerate.
29430         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29431         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29432         used without modification.
29434 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29436         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29437         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29438         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29439         MISC_BUILTIN_SPEC_BARRIER.
29440         (rs6000_init_builtins): Likewise.
29441         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29442         enum value.
29443         (speculation_barrier): New define_insn.
29444         * doc/extend.texi: Document __builtin_speculation_barrier.
29446 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29448         PR target/83203
29449         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29450         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29451         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29452         iterators.
29453         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29454         integral modes instead of "ss" and "sd".
29455         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29456         vectors with 32-bit and 64-bit elements.
29457         (vecdupssescalarmodesuffix): New mode attribute.
29458         (vec_dup<mode>): Use it.
29460 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29462         PR target/83330
29463         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29464         frame if argument is passed on stack.
29466 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29468         PR target/82682
29469         * ree.c (combine_reaching_defs): Optimize also
29470         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29471         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29473 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29475         PR middle-end/83189
29476         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29478 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29480         PR middle-end/83718
29481         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29482         after they are computed.
29484 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29486         PR tree-optimization/83695
29487         * gimple-loop-linterchange.cc
29488         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29489         reset cached scev information after interchange.
29490         (pass_linterchange::execute): Remove call to scev_reset_htab.
29492 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29494         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29495         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29496         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29497         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29498         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29499         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29500         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29501         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29502         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29503         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29504         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29505         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29506         (V_lane_reg): Likewise.
29507         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29508         New define_expand.
29509         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29510         (vfmal_lane_low<mode>_intrinsic,
29511         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29512         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29513         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29514         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29515         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29516         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29518 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29520         * config/arm/arm-cpus.in (fp16fml): New feature.
29521         (ALL_SIMD): Add fp16fml.
29522         (armv8.2-a): Add fp16fml as an option.
29523         (armv8.3-a): Likewise.
29524         (armv8.4-a): Add fp16fml as part of fp16.
29525         * config/arm/arm.h (TARGET_FP16FML): Define.
29526         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29527         when appropriate.
29528         * config/arm/arm-modes.def (V2HF): Define.
29529         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29530         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29531         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29532         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29533         vfmsl_low, vfmsl_high): New set of builtins.
29534         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29535         (vfml_op): New code attribute.
29536         (VFMLHALVES): New int iterator.
29537         (VFML, VFMLSEL): New mode attributes.
29538         (V_reg): Define mapping for V2HF.
29539         (V_hi, V_lo): New mode attributes.
29540         (VF_constraint): Likewise.
29541         (vfml_half, vfml_half_selector): New int attributes.
29542         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29543         define_expand.
29544         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29545         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29546         New define_insn.
29547         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29548         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29549         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29550         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29551         documentation.
29552         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29553         Document new effective target and option set.
29555 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29557         * config/arm/arm-cpus.in (armv8_4): New feature.
29558         (ARMv8_4a): New fgroup.
29559         (armv8.4-a): New arch.
29560         * config/arm/arm-tables.opt: Regenerate.
29561         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29562         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29563         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29564         Add matching rules for -march=armv8.4-a and extensions.
29565         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29567 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29569         PR target/81821
29570         * config/rx/rx.md (BW): New mode attribute.
29571         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29573 2018-01-11  Richard Biener  <rguenther@suse.de>
29575         PR tree-optimization/83435
29576         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29577         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29578         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29580 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29581             Alan Hayward  <alan.hayward@arm.com>
29582             David Sherwood  <david.sherwood@arm.com>
29584         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29585         field.
29586         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29587         (aarch64_print_address_internal): Use it to check for a zero offset.
29589 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29590             Alan Hayward  <alan.hayward@arm.com>
29591             David Sherwood  <david.sherwood@arm.com>
29593         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29594         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29595         Return a poly_int64 rather than a HOST_WIDE_INT.
29596         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29597         rather than a HOST_WIDE_INT.
29598         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29599         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29600         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29601         final_offset from HOST_WIDE_INT to poly_int64.
29602         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29603         to_constant when getting the number of units in an Advanced SIMD
29604         mode.
29605         (aarch64_builtin_vectorized_function): Check for a constant number
29606         of units.
29607         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29608         GET_MODE_SIZE.
29609         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29610         attribute instead of GET_MODE_NUNITS.
29611         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29612         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29613         GET_MODE_SIZE for fixed-size registers.
29614         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29615         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29616         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29617         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29618         (aarch64_print_operand, aarch64_print_address_internal)
29619         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29620         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29621         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29622         Handle polynomial GET_MODE_SIZE.
29623         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29624         wider than SImode without modification.
29625         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29626         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29627         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29628         passing and returning SVE modes.
29629         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29630         rather than GEN_INT.
29631         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29632         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29633         (aarch64_allocate_and_probe_stack_space): Likewise.
29634         (aarch64_layout_frame): Cope with polynomial offsets.
29635         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29636         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29637         polynomial offsets.
29638         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29639         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29640         poly_int64 rather than a HOST_WIDE_INT.
29641         (aarch64_get_separate_components, aarch64_process_components)
29642         (aarch64_expand_prologue, aarch64_expand_epilogue)
29643         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29644         (aarch64_anchor_offset): New function, split out from...
29645         (aarch64_legitimize_address): ...here.
29646         (aarch64_builtin_vectorization_cost): Handle polynomial
29647         TYPE_VECTOR_SUBPARTS.
29648         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29649         GET_MODE_NUNITS.
29650         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29651         number of elements from the PARALLEL rather than the mode.
29652         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29653         rather than GET_MODE_BITSIZE.
29654         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29655         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29656         (aarch64_expand_vec_perm_const_1): Handle polynomial
29657         d->perm.length () and d->perm elements.
29658         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29659         Apply to_constant to d->perm elements.
29660         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29661         polynomial CONST_VECTOR_NUNITS.
29662         (aarch64_move_pointer): Take amount as a poly_int64 rather
29663         than an int.
29664         (aarch64_progress_pointer): Avoid temporary variable.
29665         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29666         the mode attribute instead of GET_MODE.
29668 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29669             Alan Hayward  <alan.hayward@arm.com>
29670             David Sherwood  <david.sherwood@arm.com>
29672         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29673         x exists before using it.
29674         (aarch64_add_constant_internal): Rename to...
29675         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29676         src and dest rtxes.  Handle the case in which they're different,
29677         including when the offset is zero.  Replace scratchreg with an rtx.
29678         Use 2 additions if there is no spare register into which we can
29679         move a 16-bit constant.
29680         (aarch64_add_constant): Delete.
29681         (aarch64_add_offset): Replace reg with separate src and dest
29682         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29683         Use aarch64_add_offset_1.
29684         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29685         an rtx rather than an int.  Take the delta as a poly_int64
29686         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29687         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29688         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29689         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29690         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29691         and aarch64_add_sp.
29692         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29693         aarch64_add_constant.
29695 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29697         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29698         Use scalar_float_mode.
29700 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29702         * config/aarch64/aarch64-simd.md
29703         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29704         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29705         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29706         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29707         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29708         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29709         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29710         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29711         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29712         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29714 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29716         PR target/83514
29717         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29718         targ_options->x_arm_arch_string is non NULL.
29720 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
29722         * config/aarch64/aarch64.h
29723         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
29725 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
29727         PR target/82096
29728         * expmed.c (emit_store_flag_force): Swap if const op0
29729         and change VOIDmode to mode of op0.
29731 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29733         PR rtl-optimization/83761
29734         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
29735         than bytes to mode_for_size.
29737 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29739         PR middle-end/83189
29740         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
29741         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
29742         profile.
29744 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29746         PR middle-end/83575
29747         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
29748         when in layout mode.
29749         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
29750         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
29751         partition fixup.
29753 2018-01-10  Michael Collison  <michael.collison@arm.com>
29755         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
29756         * config/aarch64/aarch64-option-extension.def: Add
29757         AARCH64_OPT_EXTENSION of 'fp16fml'.
29758         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29759         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
29760         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
29761         * config/aarch64/constraints.md (Ui7): New constraint.
29762         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
29763         (VFMLA_SEL_W): Ditto.
29764         (f16quad): Ditto.
29765         (f16mac1): Ditto.
29766         (VFMLA16_LOW): New int iterator.
29767         (VFMLA16_HIGH): Ditto.
29768         (UNSPEC_FMLAL): New unspec.
29769         (UNSPEC_FMLSL): Ditto.
29770         (UNSPEC_FMLAL2): Ditto.
29771         (UNSPEC_FMLSL2): Ditto.
29772         (f16mac): New code attribute.
29773         * config/aarch64/aarch64-simd-builtins.def
29774         (aarch64_fmlal_lowv2sf): Ditto.
29775         (aarch64_fmlsl_lowv2sf): Ditto.
29776         (aarch64_fmlalq_lowv4sf): Ditto.
29777         (aarch64_fmlslq_lowv4sf): Ditto.
29778         (aarch64_fmlal_highv2sf): Ditto.
29779         (aarch64_fmlsl_highv2sf): Ditto.
29780         (aarch64_fmlalq_highv4sf): Ditto.
29781         (aarch64_fmlslq_highv4sf): Ditto.
29782         (aarch64_fmlal_lane_lowv2sf): Ditto.
29783         (aarch64_fmlsl_lane_lowv2sf): Ditto.
29784         (aarch64_fmlal_laneq_lowv2sf): Ditto.
29785         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
29786         (aarch64_fmlalq_lane_lowv4sf): Ditto.
29787         (aarch64_fmlsl_lane_lowv4sf): Ditto.
29788         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
29789         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
29790         (aarch64_fmlal_lane_highv2sf): Ditto.
29791         (aarch64_fmlsl_lane_highv2sf): Ditto.
29792         (aarch64_fmlal_laneq_highv2sf): Ditto.
29793         (aarch64_fmlsl_laneq_highv2sf): Ditto.
29794         (aarch64_fmlalq_lane_highv4sf): Ditto.
29795         (aarch64_fmlsl_lane_highv4sf): Ditto.
29796         (aarch64_fmlalq_laneq_highv4sf): Ditto.
29797         (aarch64_fmlsl_laneq_highv4sf): Ditto.
29798         * config/aarch64/aarch64-simd.md:
29799         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
29800         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29801         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
29802         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29803         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
29804         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
29805         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
29806         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
29807         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
29808         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
29809         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
29810         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
29811         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
29812         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
29813         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
29814         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
29815         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
29816         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
29817         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
29818         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
29819         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
29820         (vfmlsl_low_u32): Ditto.
29821         (vfmlalq_low_u32): Ditto.
29822         (vfmlslq_low_u32): Ditto.
29823         (vfmlal_high_u32): Ditto.
29824         (vfmlsl_high_u32): Ditto.
29825         (vfmlalq_high_u32): Ditto.
29826         (vfmlslq_high_u32): Ditto.
29827         (vfmlal_lane_low_u32): Ditto.
29828         (vfmlsl_lane_low_u32): Ditto.
29829         (vfmlal_laneq_low_u32): Ditto.
29830         (vfmlsl_laneq_low_u32): Ditto.
29831         (vfmlalq_lane_low_u32): Ditto.
29832         (vfmlslq_lane_low_u32): Ditto.
29833         (vfmlalq_laneq_low_u32): Ditto.
29834         (vfmlslq_laneq_low_u32): Ditto.
29835         (vfmlal_lane_high_u32): Ditto.
29836         (vfmlsl_lane_high_u32): Ditto.
29837         (vfmlal_laneq_high_u32): Ditto.
29838         (vfmlsl_laneq_high_u32): Ditto.
29839         (vfmlalq_lane_high_u32): Ditto.
29840         (vfmlslq_lane_high_u32): Ditto.
29841         (vfmlalq_laneq_high_u32): Ditto.
29842         (vfmlslq_laneq_high_u32): Ditto.
29843         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
29844         (AARCH64_FL_FOR_ARCH8_4): New.
29845         (AARCH64_ISA_F16FML): New ISA flag.
29846         (TARGET_F16FML): New feature flag for fp16fml.
29847         (doc/invoke.texi): Document new fp16fml option.
29849 2018-01-10  Michael Collison  <michael.collison@arm.com>
29851         * config/aarch64/aarch64-builtins.c:
29852         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
29853         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29854         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
29855         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
29856         (AARCH64_ISA_SHA3): New ISA flag.
29857         (TARGET_SHA3): New feature flag for sha3.
29858         * config/aarch64/iterators.md (sha512_op): New int attribute.
29859         (CRYPTO_SHA512): New int iterator.
29860         (UNSPEC_SHA512H): New unspec.
29861         (UNSPEC_SHA512H2): Ditto.
29862         (UNSPEC_SHA512SU0): Ditto.
29863         (UNSPEC_SHA512SU1): Ditto.
29864         * config/aarch64/aarch64-simd-builtins.def
29865         (aarch64_crypto_sha512hqv2di): New builtin.
29866         (aarch64_crypto_sha512h2qv2di): Ditto.
29867         (aarch64_crypto_sha512su0qv2di): Ditto.
29868         (aarch64_crypto_sha512su1qv2di): Ditto.
29869         (aarch64_eor3qv8hi): Ditto.
29870         (aarch64_rax1qv2di): Ditto.
29871         (aarch64_xarqv2di): Ditto.
29872         (aarch64_bcaxqv8hi): Ditto.
29873         * config/aarch64/aarch64-simd.md:
29874         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
29875         (aarch64_crypto_sha512su0qv2di): Ditto.
29876         (aarch64_crypto_sha512su1qv2di): Ditto.
29877         (aarch64_eor3qv8hi): Ditto.
29878         (aarch64_rax1qv2di): Ditto.
29879         (aarch64_xarqv2di): Ditto.
29880         (aarch64_bcaxqv8hi): Ditto.
29881         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
29882         (vsha512h2q_u64): Ditto.
29883         (vsha512su0q_u64): Ditto.
29884         (vsha512su1q_u64): Ditto.
29885         (veor3q_u16): Ditto.
29886         (vrax1q_u64): Ditto.
29887         (vxarq_u64): Ditto.
29888         (vbcaxq_u16): Ditto.
29889         * config/arm/types.md (crypto_sha512): New type attribute.
29890         (crypto_sha3): Ditto.
29891         (doc/invoke.texi): Document new sha3 option.
29893 2018-01-10  Michael Collison  <michael.collison@arm.com>
29895         * config/aarch64/aarch64-builtins.c:
29896         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
29897         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29898         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
29899         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
29900         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
29901         (AARCH64_ISA_SM4): New ISA flag.
29902         (TARGET_SM4): New feature flag for sm4.
29903         * config/aarch64/aarch64-simd-builtins.def
29904         (aarch64_sm3ss1qv4si): Ditto.
29905         (aarch64_sm3tt1aq4si): Ditto.
29906         (aarch64_sm3tt1bq4si): Ditto.
29907         (aarch64_sm3tt2aq4si): Ditto.
29908         (aarch64_sm3tt2bq4si): Ditto.
29909         (aarch64_sm3partw1qv4si): Ditto.
29910         (aarch64_sm3partw2qv4si): Ditto.
29911         (aarch64_sm4eqv4si): Ditto.
29912         (aarch64_sm4ekeyqv4si): Ditto.
29913         * config/aarch64/aarch64-simd.md:
29914         (aarch64_sm3ss1qv4si): Ditto.
29915         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
29916         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
29917         (aarch64_sm4eqv4si): Ditto.
29918         (aarch64_sm4ekeyqv4si): Ditto.
29919         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
29920         (sm3part_op): Ditto.
29921         (CRYPTO_SM3TT): Ditto.
29922         (CRYPTO_SM3PART): Ditto.
29923         (UNSPEC_SM3SS1): New unspec.
29924         (UNSPEC_SM3TT1A): Ditto.
29925         (UNSPEC_SM3TT1B): Ditto.
29926         (UNSPEC_SM3TT2A): Ditto.
29927         (UNSPEC_SM3TT2B): Ditto.
29928         (UNSPEC_SM3PARTW1): Ditto.
29929         (UNSPEC_SM3PARTW2): Ditto.
29930         (UNSPEC_SM4E): Ditto.
29931         (UNSPEC_SM4EKEY): Ditto.
29932         * config/aarch64/constraints.md (Ui2): New constraint.
29933         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
29934         * config/arm/types.md (crypto_sm3): New type attribute.
29935         (crypto_sm4): Ditto.
29936         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
29937         (vsm3tt1aq_u32): Ditto.
29938         (vsm3tt1bq_u32): Ditto.
29939         (vsm3tt2aq_u32): Ditto.
29940         (vsm3tt2bq_u32): Ditto.
29941         (vsm3partw1q_u32): Ditto.
29942         (vsm3partw2q_u32): Ditto.
29943         (vsm4eq_u32): Ditto.
29944         (vsm4ekeyq_u32): Ditto.
29945         (doc/invoke.texi): Document new sm4 option.
29947 2018-01-10  Michael Collison  <michael.collison@arm.com>
29949         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
29950         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
29951         (AARCH64_FL_FOR_ARCH8_4): New.
29952         (AARCH64_FL_V8_4): New flag.
29953         (doc/invoke.texi): Document new armv8.4-a option.
29955 2018-01-10  Michael Collison  <michael.collison@arm.com>
29957         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29958         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
29959         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
29960         * config/aarch64/aarch64-option-extension.def: Add
29961         AARCH64_OPT_EXTENSION of 'sha2'.
29962         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
29963         (crypto): Disable sha2 and aes if crypto disabled.
29964         (crypto): Enable aes and sha2 if enabled.
29965         (simd): Disable sha2 and aes if simd disabled.
29966         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
29967         New flags.
29968         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
29969         (TARGET_SHA2): New feature flag for sha2.
29970         (TARGET_AES): New feature flag for aes.
29971         * config/aarch64/aarch64-simd.md:
29972         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
29973         conditional on TARGET_AES.
29974         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
29975         (aarch64_crypto_sha1hsi): Make pattern conditional
29976         on TARGET_SHA2.
29977         (aarch64_crypto_sha1hv4si): Ditto.
29978         (aarch64_be_crypto_sha1hv4si): Ditto.
29979         (aarch64_crypto_sha1su1v4si): Ditto.
29980         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
29981         (aarch64_crypto_sha1su0v4si): Ditto.
29982         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
29983         (aarch64_crypto_sha256su0v4si): Ditto.
29984         (aarch64_crypto_sha256su1v4si): Ditto.
29985         (doc/invoke.texi): Document new aes and sha2 options.
29987 2018-01-10  Martin Sebor  <msebor@redhat.com>
29989         PR tree-optimization/83781
29990         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
29991         as string arrays.
29993 2018-01-11  Martin Sebor  <msebor@gmail.com>
29994             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29996         PR tree-optimization/83501
29997         PR tree-optimization/81703
29999         * tree-ssa-strlen.c (get_string_cst): Rename...
30000         (get_string_len): ...to this.  Handle global constants.
30001         (handle_char_store): Adjust.
30003 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
30004             Jim Wilson  <jimw@sifive.com>
30006         * config/riscv/riscv-protos.h (riscv_output_return): New.
30007         * config/riscv/riscv.c (struct machine_function): New naked_p field.
30008         (riscv_attribute_table, riscv_output_return),
30009         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30010         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30011         (riscv_compute_frame_info): Only compute frame->mask if not a naked
30012         function.
30013         (riscv_expand_prologue): Add early return for naked function.
30014         (riscv_expand_epilogue): Likewise.
30015         (riscv_function_ok_for_sibcall): Return false for naked function.
30016         (riscv_set_current_function): New.
30017         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30018         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30019         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30020         * doc/extend.texi (RISC-V Function Attributes): New.
30022 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
30024         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30025         check for 128-bit long double before checking TCmode.
30026         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30027         128-bit long doubles before checking TFmode or TCmode.
30028         (FLOAT128_IBM_P): Likewise.
30030 2018-01-10  Martin Sebor  <msebor@redhat.com>
30032         PR tree-optimization/83671
30033         * builtins.c (c_strlen): Unconditionally return zero for the empty
30034         string.
30035         Use -Warray-bounds for warnings.
30036         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30037         for non-constant array indices with COMPONENT_REF, arrays of
30038         arrays, and pointers to arrays.
30039         (gimple_fold_builtin_strlen): Determine and set length range for
30040         non-constant character arrays.
30042 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
30044         PR middle-end/81897
30045         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30046         empty blocks.
30048 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
30050         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30052 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30054         PR target/83399
30055         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30056         VECTOR_MEM_ALTIVEC_OR_VSX_P.
30057         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30058         indexed_or_indirect_operand predicate.
30059         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30060         (*vsx_le_perm_load_v8hi): Likewise.
30061         (*vsx_le_perm_load_v16qi): Likewise.
30062         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30063         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30064         (*vsx_le_perm_store_v8hi): Likewise.
30065         (*vsx_le_perm_store_v16qi): Likewise.
30066         (eight unnamed splitters): Likewise.
30068 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30070         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
30071         * config/rs6000/emmintrin.h: Likewise.
30072         * config/rs6000/mmintrin.h: Likewise.
30073         * config/rs6000/xmmintrin.h: Likewise.
30075 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
30077         PR c++/43486
30078         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
30079         "public_flag".
30080         * tree.c (tree_nop_conversion): Return true for location wrapper
30081         nodes.
30082         (maybe_wrap_with_location): New function.
30083         (selftest::check_strip_nops): New function.
30084         (selftest::test_location_wrappers): New function.
30085         (selftest::tree_c_tests): Call it.
30086         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
30087         (maybe_wrap_with_location): New decl.
30088         (EXPR_LOCATION_WRAPPER_P): New macro.
30089         (location_wrapper_p): New inline function.
30090         (tree_strip_any_location_wrapper): New inline function.
30092 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
30094         PR target/83735
30095         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
30096         stack_realign_offset for the largest alignment of stack slot
30097         actually used.
30098         (ix86_find_max_used_stack_alignment): New function.
30099         (ix86_finalize_stack_frame_flags): Use it.  Set
30100         max_used_stack_alignment if we don't realign stack.
30101         * config/i386/i386.h (machine_function): Add
30102         max_used_stack_alignment.
30104 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
30106         * config/arm/arm.opt (-mbranch-cost): New option.
30107         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
30108         account.
30110 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
30112         PR target/83629
30113         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
30114         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
30116 2018-01-10  Richard Biener  <rguenther@suse.de>
30118         PR debug/83765
30119         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
30120         early out so it also covers the case where we have a non-NULL
30121         origin.
30123 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30125         PR tree-optimization/83753
30126         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
30127         for non-strided grouped accesses if the number of elements is 1.
30129 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30131         PR target/81616
30132         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30133         * i386.h (TARGET_USE_GATHER): Define.
30134         * x86-tune.def (X86_TUNE_USE_GATHER): New.
30136 2018-01-10  Martin Liska  <mliska@suse.cz>
30138         PR bootstrap/82831
30139         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30140         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30141         partitioning.
30142         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30143         CLEANUP_NO_PARTITIONING is not set.
30145 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30147         * doc/rtl.texi: Remove documentation of (const ...) wrappers
30148         for vectors, as a partial revert of r254296.
30149         * rtl.h (const_vec_p): Delete.
30150         (const_vec_duplicate_p): Don't test for vector CONSTs.
30151         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30152         * expmed.c (make_tree): Likewise.
30154         Revert:
30155         * common.md (E, F): Use CONSTANT_P instead of checking for
30156         CONST_VECTOR.
30157         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30158         checking for CONST_VECTOR.
30160 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30162         PR middle-end/83575
30163         * predict.c (force_edge_cold): Handle in more sane way edges
30164         with no prediction.
30166 2018-01-09  Carl Love  <cel@us.ibm.com>
30168         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30169         V4SI, V4SF types.
30170         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30171         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30172         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30173         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
30174         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30175         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
30176         * config/rs6000/rs6000-protos.h: Add extern defition for
30177         rs6000_generate_float2_double_code.
30178         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30179         function.
30180         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30181         (float2_v2df): Add define_expand.
30183 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
30185         PR target/83628
30186         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30187         op_mode in the force_to_mode call.
30189 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30191         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30192         instead of checking each element individually.
30193         (aarch64_evpc_uzp): Likewise.
30194         (aarch64_evpc_zip): Likewise.
30195         (aarch64_evpc_ext): Likewise.
30196         (aarch64_evpc_rev): Likewise.
30197         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30198         instead of checking each element individually.  Return true without
30199         generating rtl if
30200         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30201         whether all selected elements come from the same input, instead of
30202         checking each element individually.  Remove calls to gen_rtx_REG,
30203         start_sequence and end_sequence and instead assert that no rtl is
30204         generated.
30206 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30208         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30209         order of HIGH and CONST checks.
30211 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30213         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30214         if the destination isn't an SSA_NAME.
30216 2018-01-09  Richard Biener  <rguenther@suse.de>
30218         PR tree-optimization/83668
30219         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30220         move prologue...
30221         (canonicalize_loop_form): ... here, renamed from ...
30222         (canonicalize_loop_closed_ssa_form): ... this and amended to
30223         swap successor edges for loop exit blocks to make us use
30224         the RPO order we need for initial schedule generation.
30226 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30228         PR tree-optimization/64811
30229         * match.pd: When optimizing comparisons with Inf, avoid
30230         introducing or losing exceptions from comparisons with NaN.
30232 2018-01-09  Martin Liska  <mliska@suse.cz>
30234         PR sanitizer/82517
30235         * asan.c (shadow_mem_size): Add gcc_assert.
30237 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30239         Don't save registers in main().
30241         PR target/83738
30242         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30243         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30244         * config/avr/avr.c (avr_set_current_function): Don't error if
30245         naked, OS_task or OS_main are specified at the same time.
30246         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30247         OS_main.
30248         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30249         attribute.
30250         * common/config/avr/avr-common.c (avr_option_optimization_table):
30251         Switch on -mmain-is-OS_task for optimizing compilations.
30253 2018-01-09  Richard Biener  <rguenther@suse.de>
30255         PR tree-optimization/83572
30256         * graphite.c: Include cfganal.h.
30257         (graphite_transform_loops): Connect infinite loops to exit
30258         and remove fake edges at the end.
30260 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30262         * ipa-inline.c (edge_badness): Revert accidental checkin.
30264 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30266         PR ipa/80763
30267         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30268         symbols; not inline clones.
30270 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30272         PR target/83507
30273         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30274         hard registers.  Formatting fixes.
30276         PR preprocessor/83722
30277         * gcc.c (try_generate_repro): Pass
30278         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30279         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30280         do_report_bug.
30282 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30283             Kito Cheng  <kito.cheng@gmail.com>
30285         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30286         (riscv_leaf_function_p): Delete.
30287         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30289 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30291         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30292         function.
30293         (do_ifelse): New function.
30294         (do_isel): New function.
30295         (do_sub3): New function.
30296         (do_add3): New function.
30297         (do_load_mask_compare): New function.
30298         (do_overlap_load_compare): New function.
30299         (expand_compare_loop): New function.
30300         (expand_block_compare): Call expand_compare_loop() when appropriate.
30301         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30302         option description.
30303         (-mblock-compare-inline-loop-limit): New option.
30305 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30307         PR target/83677
30308         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30309         Reverse order of second and third operands in first alternative.
30310         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30311         of first and second elements in UNSPEC_VPERMR vector.
30312         (altivec_expand_vec_perm_le): Likewise.
30314 2018-01-08  Jeff Law  <law@redhat.com>
30316         PR rtl-optimizatin/81308
30317         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30318         (process_switch): If group_case_labels makes a change, then set
30319         cfg_altered.
30320         (pass_convert_switch::execute): If a switch is converted, then
30321         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30323         PR rtl-optimization/81308
30324         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30325         splitting insns.
30327 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30329         PR target/83663 - Revert r255946
30330         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30331         generation for cases where splatting a value is not useful.
30332         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30333         across a vec_duplicate and a paradoxical subreg forming a vector
30334         mode to a vec_concat.
30336 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30338         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30339         -march=armv8.3-a variants.
30340         * config/arm/t-multilib: Likewise.
30341         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30343 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30345         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30346         to generate rtl.
30347         (cceq_ior_compare_complement): Give it a name so I can use it, and
30348         change boolean_or_operator predicate to boolean_operator so it can
30349         be used to generate a crand.
30350         (eqne): New code iterator.
30351         (bd/bd_neg): New code_attrs.
30352         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30353         a single define_insn.
30354         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30355         decrement (bdnzt/bdnzf/bdzt/bdzf).
30356         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30357         with the new names of the branch decrement patterns, and added the
30358         names of the branch decrement conditional patterns.
30360 2018-01-08  Richard Biener  <rguenther@suse.de>
30362         PR tree-optimization/83563
30363         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30364         cache.
30366 2018-01-08  Richard Biener  <rguenther@suse.de>
30368         PR middle-end/83713
30369         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30371 2018-01-08  Richard Biener  <rguenther@suse.de>
30373         PR tree-optimization/83685
30374         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30375         references to abnormals.
30377 2018-01-08  Richard Biener  <rguenther@suse.de>
30379         PR lto/83719
30380         * dwarf2out.c (output_indirect_strings): Handle empty
30381         skeleton_debug_str_hash.
30382         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30384 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30386         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30387         (emit_store_direct): Likewise.
30388         (arc_trampoline_adjust_address): Likewise.
30389         (arc_asm_trampoline_template): New function.
30390         (arc_initialize_trampoline): Use asm_trampoline_template.
30391         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30392         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30393         * config/arc/arc.md (flush_icache): Delete pattern.
30395 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30397         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30398         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30399         munaligned-access.
30401 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30403         PR target/83681
30404         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30405         by not USED_FOR_TARGET.
30406         (make_pass_resolve_sw_modes): Likewise.
30408 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30410         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30411         USED_FOR_TARGET.
30413 2018-01-08  Richard Biener  <rguenther@suse.de>
30415         PR middle-end/83580
30416         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30418 2018-01-08  Richard Biener  <rguenther@suse.de>
30420         PR middle-end/83517
30421         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30423 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30425         PR middle-end/81897
30426         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30427         basic blocks with a small number of successors.
30428         (convert_control_dep_chain_into_preds): Improve handling of
30429         forwarder blocks.
30430         (dump_predicates): Split apart into...
30431         (dump_pred_chain): ...here...
30432         (dump_pred_info): ...and here.
30433         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30434         (can_chain_union_be_invalidated_p): Improve check for invalidation
30435         of paths.
30436         (uninit_uses_cannot_happen): Avoid unnecessary if
30437         convert_control_dep_chain_into_preds yielded nothing.
30439 2018-01-06  Martin Sebor  <msebor@redhat.com>
30441         PR tree-optimization/83640
30442         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30443         subtracting negative offset from size.
30444         (builtin_access::overlap): Adjust offset bounds of the access to fall
30445         within the size of the object if possible.
30447 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30449         PR rtl-optimization/83699
30450         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30451         extract_bit_field_as_subreg to cases in which the extracted
30452         value is also a vector.
30454         * lra-constraints.c (process_alt_operands): Test for the equivalence
30455         substitutions when detecting a possible reload cycle.
30457 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30459         PR debug/83480
30460         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30461         by default if flag_selective_schedling{,2}.  Formatting fixes.
30463         PR rtl-optimization/83682
30464         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30465         if it has non-VECTOR_MODE element mode.
30466         (vec_duplicate_p): Likewise.
30468         PR middle-end/83694
30469         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30470         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30472 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30474         PR target/83604
30475         * config/i386/i386-builtin.def
30476         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30477         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30478         Require also OPTION_MASK_ISA_AVX512F in addition to
30479         OPTION_MASK_ISA_GFNI.
30480         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30481         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30482         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30483         to OPTION_MASK_ISA_GFNI.
30484         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30485         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30486         OPTION_MASK_ISA_AVX512BW.
30487         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30488         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30489         addition to OPTION_MASK_ISA_GFNI.
30490         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30491         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30492         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30493         to OPTION_MASK_ISA_GFNI.
30494         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30495         a requirement for all ISAs rather than any of them with a few
30496         exceptions.
30497         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30498         processing.
30499         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30500         bitmasks to be enabled with 3 exceptions, instead of requiring any
30501         enabled ISA with lots of exceptions.
30502         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30503         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30504         Change avx512bw in isa attribute to avx512f.
30505         * config/i386/sgxintrin.h: Add license boilerplate.
30506         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30507         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30508         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30509         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30510         defined.
30511         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30512         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30513         temporarily sse2 rather than sse if not enabled already.
30515         PR target/83604
30516         * config/i386/sse.md (VI248_VLBW): Rename to ...
30517         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30518         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30519         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30520         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30521         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30522         mode iterator instead of VI248_VLBW.
30524 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30526         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30527         (record_modified): Skip clobbers; add debug output.
30528         (param_change_prob): Use sreal frequencies.
30530 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30532         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30533         punt for user-aligned variables.
30535 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30537         * tree-chrec.c (chrec_contains_symbols): Return true for
30538         POLY_INT_CST.
30540 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30542         PR target/82439
30543         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30544         of (x|y) == x for BICS pattern.
30546 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30548         PR tree-optimization/83605
30549         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30550         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30551         can throw.
30553 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30555         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30556         * config/epiphany/rtems.h: New file.
30558 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30559             Uros Bizjak  <ubizjak@gmail.com>
30561         PR target/83554
30562         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30563         QIreg_operand instead of register_operand predicate.
30564         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30565         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30566         comments instead of -fmitigate[-_]rop.
30568 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30570         PR bootstrap/81926
30571         * cgraphunit.c (symbol_table::compile): Switch to text_section
30572         before calling assembly_start debug hook.
30573         * run-rtl-passes.c (run_rtl_passes): Likewise.
30574         Include output.h.
30576 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30578         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30579         range_int_cst_p rather than !symbolic_range_p before calling
30580         extract_range_from_multiplicative_op_1.
30582 2018-01-04  Jeff Law  <law@redhat.com>
30584         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30585         redundant test in assertion.
30587 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30589         * doc/rtl.texi: Document machine_mode wrapper classes.
30591 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30593         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30594         using tree_to_uhwi.
30596 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30598         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30599         the VEC_PERM_EXPR fold to fail.
30601 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30603         PR debug/83585
30604         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30605         to switched_sections.
30607 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30609         PR target/83680
30610         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30611         test for d.testing.
30613 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30615         PR target/83387
30616         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30617         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30619 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30621         PR debug/83666
30622         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30623         is BLKmode and bitpos not zero or mode change is needed.
30625 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30627         PR target/83675
30628         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30629         TARGET_VIS2.
30631 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30633         PR target/83628
30634         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30635         instead of MULT rtx.  Update all corresponding splitters.
30636         (*saddl_se): Ditto.
30637         (*ssub<modesuffix>): Ditto.
30638         (*ssubl_se): Ditto.
30639         (*cmp_sadd_di): Update split patterns.
30640         (*cmp_sadd_si): Ditto.
30641         (*cmp_sadd_sidi): Ditto.
30642         (*cmp_ssub_di): Ditto.
30643         (*cmp_ssub_si): Ditto.
30644         (*cmp_ssub_sidi): Ditto.
30645         * config/alpha/predicates.md (const23_operand): New predicate.
30646         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30647         Look for ASHIFT, not MULT inner operand.
30648         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30650 2018-01-04  Martin Liska  <mliska@suse.cz>
30652         PR gcov-profile/83669
30653         * gcov.c (output_intermediate_file): Add version to intermediate
30654         gcov file.
30655         * doc/gcov.texi: Document new field 'version' in intermediate
30656         file format. Fix location of '-k' option of gcov command.
30658 2018-01-04  Martin Liska  <mliska@suse.cz>
30660         PR ipa/82352
30661         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30663 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30665         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30667 2018-01-03  Martin Sebor  <msebor@redhat.com>
30669         PR tree-optimization/83655
30670         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30671         checking calls with invalid arguments.
30673 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30675         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30676         (vectorizable_mask_load_store): Delete.
30677         (vectorizable_call): Return false for masked loads and stores.
30678         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30679         instead of gimple_assign_rhs1.
30680         (vectorizable_load): Handle IFN_MASK_LOAD.
30681         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30683 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30685         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30686         split out from..,
30687         (vectorizable_mask_load_store): ...here.
30688         (vectorizable_load): ...and here.
30690 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30692         * tree-vect-stmts.c (vect_build_all_ones_mask)
30693         (vect_build_zero_merge_argument): New functions, split out from...
30694         (vectorizable_load): ...here.
30696 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30698         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30699         split out from...
30700         (vectorizable_mask_load_store): ...here.
30701         (vectorizable_store): ...and here.
30703 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30705         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30706         split out from...
30707         (vectorizable_mask_load_store): ...here.
30709 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30711         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30712         (vect_model_store_cost): Take a vec_load_store_type instead of a
30713         vect_def_type.
30714         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30715         (vect_model_store_cost): Take a vec_load_store_type instead of a
30716         vect_def_type.
30717         (vectorizable_mask_load_store): Update accordingly.
30718         (vectorizable_store): Likewise.
30719         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
30721 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30723         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
30724         IFN_MASK_LOAD calls here rather than...
30725         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
30727 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30728             Alan Hayward  <alan.hayward@arm.com>
30729             David Sherwood  <david.sherwood@arm.com>
30731         * expmed.c (extract_bit_field_1): For vector extracts,
30732         fall back to extract_bit_field_as_subreg if vec_extract
30733         isn't available.
30735 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30736             Alan Hayward  <alan.hayward@arm.com>
30737             David Sherwood  <david.sherwood@arm.com>
30739         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
30740         they are variable or constant sized.
30741         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
30742         slots for constant-sized data.
30744 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30745             Alan Hayward  <alan.hayward@arm.com>
30746             David Sherwood  <david.sherwood@arm.com>
30748         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
30749         handling COND_EXPRs with boolean comparisons, try to find a better
30750         basis for the mask type than the boolean itself.
30752 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30754         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
30755         is calculated and how it can be overridden.
30756         * genmodes.c (max_bitsize_mode_any_mode): New variable.
30757         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
30758         if defined.
30759         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
30760         if nonzero.
30762 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30763             Alan Hayward  <alan.hayward@arm.com>
30764             David Sherwood  <david.sherwood@arm.com>
30766         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
30767         Remove the mode argument.
30768         (aarch64_simd_valid_immediate): Remove the mode and inverse
30769         arguments.
30770         * config/aarch64/iterators.md (bitsize): New iterator.
30771         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
30772         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
30773         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
30774         aarch64_simd_valid_immediate.
30775         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
30776         (aarch64_reg_or_bic_imm): Likewise.
30777         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
30778         with an insn_type enum and msl with a modifier_type enum.
30779         Replace element_width with a scalar_mode.  Change the shift
30780         to unsigned int.  Add constructors for scalar_float_mode and
30781         scalar_int_mode elements.
30782         (aarch64_vect_float_const_representable_p): Delete.
30783         (aarch64_can_const_movi_rtx_p)
30784         (aarch64_simd_scalar_immediate_valid_for_move)
30785         (aarch64_simd_make_constant): Update call to
30786         aarch64_simd_valid_immediate.
30787         (aarch64_advsimd_valid_immediate_hs): New function.
30788         (aarch64_advsimd_valid_immediate): Likewise.
30789         (aarch64_simd_valid_immediate): Remove mode and inverse
30790         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
30791         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
30792         and aarch64_float_const_representable_p on the result.
30793         (aarch64_output_simd_mov_immediate): Remove mode argument.
30794         Update call to aarch64_simd_valid_immediate and use of
30795         simd_immediate_info.
30796         (aarch64_output_scalar_simd_mov_immediate): Update call
30797         accordingly.
30799 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30800             Alan Hayward  <alan.hayward@arm.com>
30801             David Sherwood  <david.sherwood@arm.com>
30803         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
30804         (mode_nunits): Likewise CONST_MODE_NUNITS.
30805         * machmode.def (ADJUST_NUNITS): Document.
30806         * genmodes.c (mode_data::need_nunits_adj): New field.
30807         (blank_mode): Update accordingly.
30808         (adj_nunits): New variable.
30809         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
30810         parameter.
30811         (emit_mode_size_inline): Set need_bytesize_adj for all modes
30812         listed in adj_nunits.
30813         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
30814         listed in adj_nunits.  Don't emit case statements for such modes.
30815         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
30816         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
30817         nothing if adj_nunits is nonnull.
30818         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
30819         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
30820         (emit_mode_fbit): Update use of print_maybe_const_decl.
30821         (emit_move_size): Likewise.  Treat the array as non-const
30822         if adj_nunits.
30823         (emit_mode_adjustments): Handle adj_nunits.
30825 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30827         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
30828         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
30829         (VECTOR_MODES): Use it.
30830         (make_vector_modes): Take the prefix as an argument.
30832 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30833             Alan Hayward  <alan.hayward@arm.com>
30834             David Sherwood  <david.sherwood@arm.com>
30836         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
30837         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
30838         for MODE_VECTOR_BOOL.
30839         * machmode.def (VECTOR_BOOL_MODE): Document.
30840         * genmodes.c (VECTOR_BOOL_MODE): New macro.
30841         (make_vector_bool_mode): New function.
30842         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
30843         MODE_VECTOR_BOOL.
30844         * lto-streamer-in.c (lto_input_mode_table): Likewise.
30845         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
30846         Likewise.
30847         * stor-layout.c (int_mode_for_mode): Likewise.
30848         * tree.c (build_vector_type_for_mode): Likewise.
30849         * varasm.c (output_constant_pool_2): Likewise.
30850         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
30851         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
30852         for MODE_VECTOR_BOOL.
30853         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
30854         of mode class checks.
30855         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
30856         instead of a list of mode class checks.
30857         (expand_vector_scalar_condition): Likewise.
30858         (type_for_widest_vector_mode): Handle BImode as an inner mode.
30860 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30861             Alan Hayward  <alan.hayward@arm.com>
30862             David Sherwood  <david.sherwood@arm.com>
30864         * machmode.h (mode_size): Change from unsigned short to
30865         poly_uint16_pod.
30866         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
30867         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30868         or if measurement_type is not polynomial.
30869         (fixed_size_mode::includes_p): Check for constant-sized modes.
30870         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
30871         return a poly_uint16 rather than an unsigned short.
30872         (emit_mode_size): Change the type of mode_size from unsigned short
30873         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
30874         (emit_mode_adjustments): Cope with polynomial vector sizes.
30875         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30876         for GET_MODE_SIZE.
30877         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30878         for GET_MODE_SIZE.
30879         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
30880         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
30881         * caller-save.c (setup_save_areas): Likewise.
30882         (replace_reg_with_saved_mem): Likewise.
30883         * calls.c (emit_library_call_value_1): Likewise.
30884         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
30885         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
30886         (gen_lowpart_for_combine): Likewise.
30887         * convert.c (convert_to_integer_1): Likewise.
30888         * cse.c (equiv_constant, cse_insn): Likewise.
30889         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
30890         (cselib_subst_to_values): Likewise.
30891         * dce.c (word_dce_process_block): Likewise.
30892         * df-problems.c (df_word_lr_mark_ref): Likewise.
30893         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
30894         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
30895         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
30896         (rtl_for_decl_location): Likewise.
30897         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
30898         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
30899         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
30900         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
30901         (expand_expr_real_1): Likewise.
30902         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
30903         (pad_below): Likewise.
30904         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30905         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
30906         * ira.c (get_subreg_tracking_sizes): Likewise.
30907         * ira-build.c (ira_create_allocno_objects): Likewise.
30908         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
30909         (ira_sort_regnos_for_alter_reg): Likewise.
30910         * ira-costs.c (record_operand_costs): Likewise.
30911         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
30912         (resolve_simple_move): Likewise.
30913         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
30914         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
30915         (lra_constraints): Likewise.
30916         (CONST_POOL_OK_P): Reject variable-sized modes.
30917         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
30918         (add_pseudo_to_slot, lra_spill): Likewise.
30919         * omp-low.c (omp_clause_aligned_alignment): Likewise.
30920         * optabs-query.c (get_best_extraction_insn): Likewise.
30921         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30922         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
30923         (expand_mult_highpart, valid_multiword_target_p): Likewise.
30924         * recog.c (offsettable_address_addr_space_p): Likewise.
30925         * regcprop.c (maybe_mode_change): Likewise.
30926         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
30927         * regrename.c (build_def_use): Likewise.
30928         * regstat.c (dump_reg_info): Likewise.
30929         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
30930         (find_reloads, find_reloads_subreg_address): Likewise.
30931         * reload1.c (eliminate_regs_1): Likewise.
30932         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
30933         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
30934         (simplify_binary_operation_1, simplify_subreg): Likewise.
30935         * targhooks.c (default_function_arg_padding): Likewise.
30936         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
30937         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
30938         (verify_gimple_assign_ternary): Likewise.
30939         * tree-inline.c (estimate_move_cost): Likewise.
30940         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30941         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
30942         (get_address_cost_ainc): Likewise.
30943         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
30944         (vect_supportable_dr_alignment): Likewise.
30945         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30946         (vectorizable_reduction): Likewise.
30947         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
30948         (vectorizable_operation, vectorizable_load): Likewise.
30949         * tree.c (build_same_sized_truth_vector_type): Likewise.
30950         * valtrack.c (cleanup_auto_inc_dec): Likewise.
30951         * var-tracking.c (emit_note_insn_var_location): Likewise.
30952         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
30953         (ADDR_VEC_ALIGN): Likewise.
30955 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30956             Alan Hayward  <alan.hayward@arm.com>
30957             David Sherwood  <david.sherwood@arm.com>
30959         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
30960         unsigned short.
30961         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30962         or if measurement_type is polynomial.
30963         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
30964         * combine.c (make_extraction): Likewise.
30965         * dse.c (find_shift_sequence): Likewise.
30966         * dwarf2out.c (mem_loc_descriptor): Likewise.
30967         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
30968         (extract_bit_field, extract_low_bits): Likewise.
30969         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
30970         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
30971         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
30972         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
30973         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30974         * reload.c (find_reloads): Likewise.
30975         * reload1.c (alter_reg): Likewise.
30976         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
30977         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
30978         * tree-if-conv.c (predicate_mem_writes): Likewise.
30979         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
30980         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
30981         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
30982         * valtrack.c (dead_debug_insert_temp): Likewise.
30983         * varasm.c (mergeable_constant_section): Likewise.
30984         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
30986 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30987             Alan Hayward  <alan.hayward@arm.com>
30988             David Sherwood  <david.sherwood@arm.com>
30990         * expr.c (expand_assignment): Cope with polynomial mode sizes
30991         when assigning to a CONCAT.
30993 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30994             Alan Hayward  <alan.hayward@arm.com>
30995             David Sherwood  <david.sherwood@arm.com>
30997         * machmode.h (mode_precision): Change from unsigned short to
30998         poly_uint16_pod.
30999         (mode_to_precision): Return a poly_uint16 rather than an unsigned
31000         short.
31001         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31002         or if measurement_type is not polynomial.
31003         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
31004         in which the mode is already known to be a scalar_int_mode.
31005         * genmodes.c (emit_mode_precision): Change the type of mode_precision
31006         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
31007         initializer.
31008         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31009         for GET_MODE_PRECISION.
31010         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31011         for GET_MODE_PRECISION.
31012         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31013         as polynomial.
31014         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31015         (expand_field_assignment, make_extraction): Likewise.
31016         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31017         (get_last_value): Likewise.
31018         * convert.c (convert_to_integer_1): Likewise.
31019         * cse.c (cse_insn): Likewise.
31020         * expr.c (expand_expr_real_1): Likewise.
31021         * lra-constraints.c (simplify_operand_subreg): Likewise.
31022         * optabs-query.c (can_atomic_load_p): Likewise.
31023         * optabs.c (expand_atomic_load): Likewise.
31024         (expand_atomic_store): Likewise.
31025         * ree.c (combine_reaching_defs): Likewise.
31026         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31027         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31028         * tree.h (type_has_mode_precision_p): Likewise.
31029         * ubsan.c (instrument_si_overflow): Likewise.
31031 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31032             Alan Hayward  <alan.hayward@arm.com>
31033             David Sherwood  <david.sherwood@arm.com>
31035         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31036         polynomial numbers of units.
31037         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31038         (valid_vector_subparts_p): New function.
31039         (build_vector_type): Remove temporary shim and take the number
31040         of units as a poly_uint64 rather than an int.
31041         (build_opaque_vector_type): Take the number of units as a
31042         poly_uint64 rather than an int.
31043         * tree.c (build_vector_from_ctor): Handle polynomial
31044         TYPE_VECTOR_SUBPARTS.
31045         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31046         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31047         (build_vector_from_val): If the number of units is variable,
31048         use build_vec_duplicate_cst for constant operands and
31049         VEC_DUPLICATE_EXPR otherwise.
31050         (make_vector_type): Remove temporary is_constant ().
31051         (build_vector_type, build_opaque_vector_type): Take the number of
31052         units as a poly_uint64 rather than an int.
31053         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31054         VECTOR_CST_NELTS.
31055         * cfgexpand.c (expand_debug_expr): Likewise.
31056         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31057         (store_constructor, expand_expr_real_1): Likewise.
31058         (const_scalar_mask_from_tree): Likewise.
31059         * fold-const-call.c (fold_const_reduction): Likewise.
31060         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31061         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31062         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31063         (fold_relational_const): Likewise.
31064         (native_interpret_vector): Likewise.  Change the size from an
31065         int to an unsigned int.
31066         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31067         TYPE_VECTOR_SUBPARTS.
31068         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31069         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
31070         duplicating a non-constant operand into a variable-length vector.
31071         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
31072         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
31073         * ipa-icf.c (sem_variable::equals): Likewise.
31074         * match.pd: Likewise.
31075         * omp-simd-clone.c (simd_clone_subparts): Likewise.
31076         * print-tree.c (print_node): Likewise.
31077         * stor-layout.c (layout_type): Likewise.
31078         * targhooks.c (default_builtin_vectorization_cost): Likewise.
31079         * tree-cfg.c (verify_gimple_comparison): Likewise.
31080         (verify_gimple_assign_binary): Likewise.
31081         (verify_gimple_assign_ternary): Likewise.
31082         (verify_gimple_assign_single): Likewise.
31083         * tree-pretty-print.c (dump_generic_node): Likewise.
31084         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31085         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
31086         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
31087         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
31088         (vect_shift_permute_load_chain): Likewise.
31089         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
31090         (expand_vector_condition, optimize_vector_constructor): Likewise.
31091         (lower_vec_perm, get_compute_type): Likewise.
31092         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31093         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
31094         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
31095         (vect_recog_mask_conversion_pattern): Likewise.
31096         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
31097         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
31098         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31099         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
31100         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
31101         (vectorizable_shift, vectorizable_operation, vectorizable_store)
31102         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
31103         (supportable_widening_operation): Likewise.
31104         (supportable_narrowing_operation): Likewise.
31105         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
31106         Likewise.
31107         * varasm.c (output_constant): Likewise.
31109 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31110             Alan Hayward  <alan.hayward@arm.com>
31111             David Sherwood  <david.sherwood@arm.com>
31113         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
31114         so that both the length == 3 and length != 3 cases set up their
31115         own permute vectors.  Add comments explaining why we know the
31116         number of elements is constant.
31117         (vect_permute_load_chain): Likewise.
31119 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31120             Alan Hayward  <alan.hayward@arm.com>
31121             David Sherwood  <david.sherwood@arm.com>
31123         * machmode.h (mode_nunits): Change from unsigned char to
31124         poly_uint16_pod.
31125         (ONLY_FIXED_SIZE_MODES): New macro.
31126         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
31127         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
31128         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
31129         New typedefs.
31130         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
31131         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31132         or if measurement_type is not polynomial.
31133         * genmodes.c (ZERO_COEFFS): New macro.
31134         (emit_mode_nunits_inline): Make mode_nunits_inline return a
31135         poly_uint16.
31136         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31137         Use ZERO_COEFFS when emitting initializers.
31138         * data-streamer.h (bp_pack_poly_value): New function.
31139         (bp_unpack_poly_value): Likewise.
31140         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31141         for GET_MODE_NUNITS.
31142         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31143         for GET_MODE_NUNITS.
31144         * tree.c (make_vector_type): Remove temporary shim and make
31145         the real function take the number of units as a poly_uint64
31146         rather than an int.
31147         (build_vector_type_for_mode): Handle polynomial nunits.
31148         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31149         * emit-rtl.c (const_vec_series_p_1): Likewise.
31150         (gen_rtx_CONST_VECTOR): Likewise.
31151         * fold-const.c (test_vec_duplicate_folding): Likewise.
31152         * genrecog.c (validate_pattern): Likewise.
31153         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31154         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31155         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31156         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31157         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31158         * rtlanal.c (subreg_get_info): Likewise.
31159         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31160         (vect_grouped_load_supported): Likewise.
31161         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31162         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31163         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31164         (simplify_const_unary_operation, simplify_binary_operation_1)
31165         (simplify_const_binary_operation, simplify_ternary_operation)
31166         (test_vector_ops_duplicate, test_vector_ops): Likewise.
31167         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31168         instead of CONST_VECTOR_NUNITS.
31169         * varasm.c (output_constant_pool_2): Likewise.
31170         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31171         explicit-encoded elements in the XVEC for variable-length vectors.
31173 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31175         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31177 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31178             Alan Hayward  <alan.hayward@arm.com>
31179             David Sherwood  <david.sherwood@arm.com>
31181         * coretypes.h (fixed_size_mode): Declare.
31182         (fixed_size_mode_pod): New typedef.
31183         * builtins.h (target_builtins::x_apply_args_mode)
31184         (target_builtins::x_apply_result_mode): Change type to
31185         fixed_size_mode_pod.
31186         * builtins.c (apply_args_size, apply_result_size, result_vector)
31187         (expand_builtin_apply_args_1, expand_builtin_apply)
31188         (expand_builtin_return): Update accordingly.
31190 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31192         * cse.c (hash_rtx_cb): Hash only the encoded elements.
31193         * cselib.c (cselib_hash_rtx): Likewise.
31194         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31195         CONST_VECTOR encoding.
31197 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31198             Jeff Law  <law@redhat.com>
31200         PR target/83641
31201         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31202         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31203         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31204         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31206         PR target/83641
31207         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31208         explicitly probe *sp in a noreturn function if there were any callee
31209         register saves or frame pointer is needed.
31211 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31213         PR debug/83621
31214         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31215         BLKmode for ternary, binary or unary expressions.
31217         PR debug/83645
31218         * var-tracking.c (delete_vta_debug_insn): New inline function.
31219         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31220         insns from get_insns () to NULL instead of each bb separately.
31221         Use delete_vta_debug_insn.  No longer static.
31222         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31223         delete_vta_debug_insns callers.
31224         * rtl.h (delete_vta_debug_insns): Declare.
31225         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31226         instead of variable_tracking_main.
31228 2018-01-03  Martin Sebor  <msebor@redhat.com>
31230         PR tree-optimization/83603
31231         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31232         arguments past the endof the argument list in functions declared
31233         without a prototype.
31234         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31235         Avoid checking when arguments are null.
31237 2018-01-03  Martin Sebor  <msebor@redhat.com>
31239         PR c/83559
31240         * doc/extend.texi (attribute const): Fix a typo.
31241         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31242         issuing -Wsuggest-attribute for void functions.
31244 2018-01-03  Martin Sebor  <msebor@redhat.com>
31246         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31247         offset_int::from instead of wide_int::to_shwi.
31248         (maybe_diag_overlap): Remove assertion.
31249         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31250         * gimple-ssa-sprintf.c (format_directive): Same.
31251         (parse_directive): Same.
31252         (sprintf_dom_walker::compute_format_length): Same.
31253         (try_substitute_return_value): Same.
31255 2018-01-03  Jeff Law  <law@redhat.com>
31257         PR middle-end/83654
31258         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31259         non-constant residual for zero at runtime and avoid probing in
31260         that case.  Reorganize code for trailing problem to mirror handling
31261         of the residual.
31263 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31265         PR tree-optimization/83501
31266         * tree-ssa-strlen.c (get_string_cst): New.
31267         (handle_char_store): Call get_string_cst.
31269 2018-01-03  Martin Liska  <mliska@suse.cz>
31271         PR tree-optimization/83593
31272         * tree-ssa-strlen.c: Include tree-cfg.h.
31273         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31274         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31275         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31276         to false.
31277         (strlen_dom_walker::before_dom_children): Call
31278         gimple_purge_dead_eh_edges. Dump tranformation with details
31279         dump flags.
31280         (strlen_dom_walker::before_dom_children): Update call by adding
31281         new argument cleanup_eh.
31282         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31284 2018-01-03  Martin Liska  <mliska@suse.cz>
31286         PR ipa/83549
31287         * cif-code.def (VARIADIC_THUNK): New enum value.
31288         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31289         thunks.
31291 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31293         * sse.md (mov<mode>_internal): Tighten condition for when to use
31294         vmovdqu<ssescalarsize> for TI and OI modes.
31296 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31298         Update copyright years.
31300 2018-01-03  Martin Liska  <mliska@suse.cz>
31302         PR ipa/83594
31303         * ipa-visibility.c (function_and_variable_visibility): Skip
31304         functions with noipa attribure.
31306 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31308         * gcc.c (process_command): Update copyright notice dates.
31309         * gcov-dump.c (print_version): Ditto.
31310         * gcov.c (print_version): Ditto.
31311         * gcov-tool.c (print_version): Ditto.
31312         * gengtype.c (create_file): Ditto.
31313         * doc/cpp.texi: Bump @copying's copyright year.
31314         * doc/cppinternals.texi: Ditto.
31315         * doc/gcc.texi: Ditto.
31316         * doc/gccint.texi: Ditto.
31317         * doc/gcov.texi: Ditto.
31318         * doc/install.texi: Ditto.
31319         * doc/invoke.texi: Ditto.
31321 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31323         * vector-builder.h (vector_builder::m_full_nelts): Change from
31324         unsigned int to poly_uint64.
31325         (vector_builder::full_nelts): Update prototype accordingly.
31326         (vector_builder::new_vector): Likewise.
31327         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31328         (vector_builder::operator ==): Likewise.
31329         (vector_builder::finalize): Likewise.
31330         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31331         Take the number of elements as a poly_uint64 rather than an
31332         unsigned int.
31333         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31334         from unsigned int to poly_uint64.
31335         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31336         (vec_perm_indices::new_vector): Likewise.
31337         (vec_perm_indices::length): Likewise.
31338         (vec_perm_indices::nelts_per_input): Likewise.
31339         (vec_perm_indices::input_nelts): Likewise.
31340         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31341         number of elements per input as a poly_uint64 rather than an
31342         unsigned int.  Use the original encoding for variable-length
31343         vectors, rather than clamping each individual element.
31344         For the second and subsequent elements in each pattern,
31345         clamp the step and base before clamping their sum.
31346         (vec_perm_indices::series_p): Handle polynomial element counts.
31347         (vec_perm_indices::all_in_range_p): Likewise.
31348         (vec_perm_indices_to_tree): Likewise.
31349         (vec_perm_indices_to_rtx): Likewise.
31350         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31351         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31352         (tree_vector_builder::new_binary_operation): Handle polynomial
31353         element counts.  Return false if we need to know the number
31354         of elements at compile time.
31355         * fold-const.c (fold_vec_perm): Punt if the number of elements
31356         isn't known at compile time.
31358 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31360         * vec-perm-indices.h (vec_perm_builder): Change element type
31361         from HOST_WIDE_INT to poly_int64.
31362         (vec_perm_indices::element_type): Update accordingly.
31363         (vec_perm_indices::clamp): Handle polynomial element_types.
31364         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31365         (vec_perm_indices::all_in_range_p): Likewise.
31366         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31367         than shwi trees.
31368         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31369         polynomial vec_perm_indices element types.
31370         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31371         * fold-const.c (fold_vec_perm): Likewise.
31372         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31373         * tree-vect-generic.c (lower_vec_perm): Likewise.
31374         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31375         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31376         element type to HOST_WIDE_INT.
31378 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31379             Alan Hayward  <alan.hayward@arm.com>
31380             David Sherwood  <david.sherwood@arm.com>
31382         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31383         rather than an int.  Use plus_constant.
31384         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31385         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31392         a HOST_WIDE_INT to a poly_int64.
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         * calls.c (load_register_parameters): Cope with polynomial
31399         mode sizes.  Require a constant size for BLKmode parameters
31400         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31401         forces a parameter to be padded at the lsb end in order to
31402         fill a complete number of words, require the parameter size
31403         to be ordered wrt UNITS_PER_WORD.
31405 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31406             Alan Hayward  <alan.hayward@arm.com>
31407             David Sherwood  <david.sherwood@arm.com>
31409         * reload1.c (spill_stack_slot_width): Change element type
31410         from unsigned int to poly_uint64_pod.
31411         (alter_reg): Treat mode sizes as polynomial.
31413 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31414             Alan Hayward  <alan.hayward@arm.com>
31415             David Sherwood  <david.sherwood@arm.com>
31417         * reload.c (complex_word_subreg_p): New function.
31418         (reload_inner_reg_of_subreg, push_reload): Use it.
31420 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31421             Alan Hayward  <alan.hayward@arm.com>
31422             David Sherwood  <david.sherwood@arm.com>
31424         * lra-constraints.c (process_alt_operands): Reject matched
31425         operands whose sizes aren't ordered.
31426         (match_reload): Refer to this check here.
31428 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31429             Alan Hayward  <alan.hayward@arm.com>
31430             David Sherwood  <david.sherwood@arm.com>
31432         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31433         that the mode size is in the set {1, 2, 4, 8, 16}.
31435 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31436             Alan Hayward  <alan.hayward@arm.com>
31437             David Sherwood  <david.sherwood@arm.com>
31439         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31440         Use plus_constant instead of gen_rtx_PLUS.
31442 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31443             Alan Hayward  <alan.hayward@arm.com>
31444             David Sherwood  <david.sherwood@arm.com>
31446         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31447         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31448         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31449         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31450         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31451         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31452         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31453         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31454         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31455         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31456         a poly_int64.
31457         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31458         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31459         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31460         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31461         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31462         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31463         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31464         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31465         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31466         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31467         function.
31468         * expr.c (emit_move_resolve_push): Treat the input and result
31469         of PUSH_ROUNDING as a poly_int64.
31470         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31471         (emit_push_insn): Likewise.
31472         * lra-eliminations.c (mark_not_eliminable): Likewise.
31473         * recog.c (push_operand): Likewise.
31474         * reload1.c (elimination_effects): Likewise.
31475         * rtlanal.c (nonzero_bits1): Likewise.
31476         * calls.c (store_one_arg): Likewise.  Require the padding to be
31477         known at compile time.
31479 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31480             Alan Hayward  <alan.hayward@arm.com>
31481             David Sherwood  <david.sherwood@arm.com>
31483         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31484         Use plus_constant instead of gen_rtx_PLUS.
31486 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31487             Alan Hayward  <alan.hayward@arm.com>
31488             David Sherwood  <david.sherwood@arm.com>
31490         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31491         rather than an int.
31493 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31494             Alan Hayward  <alan.hayward@arm.com>
31495             David Sherwood  <david.sherwood@arm.com>
31497         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31498         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31499         via stack temporaries.  Treat the mode size as polynomial too.
31501 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31502             Alan Hayward  <alan.hayward@arm.com>
31503             David Sherwood  <david.sherwood@arm.com>
31505         * expr.c (expand_expr_real_2): When handling conversions involving
31506         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31507         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31508         as a poly_uint64 too.
31510 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31511             Alan Hayward  <alan.hayward@arm.com>
31512             David Sherwood  <david.sherwood@arm.com>
31514         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31516 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31517             Alan Hayward  <alan.hayward@arm.com>
31518             David Sherwood  <david.sherwood@arm.com>
31520         * combine.c (can_change_dest_mode): Handle polynomial
31521         REGMODE_NATURAL_SIZE.
31522         * expmed.c (store_bit_field_1): Likewise.
31523         * expr.c (store_constructor): Likewise.
31524         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31525         and polynomial REGMODE_NATURAL_SIZE.
31526         (gen_lowpart_common): Likewise.
31527         * reginfo.c (record_subregs_of_mode): Likewise.
31528         * rtlanal.c (read_modify_subreg_p): Likewise.
31530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31531             Alan Hayward  <alan.hayward@arm.com>
31532             David Sherwood  <david.sherwood@arm.com>
31534         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31535         numbers of elements.
31537 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31538             Alan Hayward  <alan.hayward@arm.com>
31539             David Sherwood  <david.sherwood@arm.com>
31541         * match.pd: Cope with polynomial numbers of vector elements.
31543 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31544             Alan Hayward  <alan.hayward@arm.com>
31545             David Sherwood  <david.sherwood@arm.com>
31547         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31548         in a POINTER_PLUS_EXPR.
31550 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31551             Alan Hayward  <alan.hayward@arm.com>
31552             David Sherwood  <david.sherwood@arm.com>
31554         * omp-simd-clone.c (simd_clone_subparts): New function.
31555         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31556         (ipa_simd_modify_function_body): Likewise.
31558 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31559             Alan Hayward  <alan.hayward@arm.com>
31560             David Sherwood  <david.sherwood@arm.com>
31562         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31563         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31564         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31565         (expand_vector_condition, vector_element): Likewise.
31566         (subparts_gt): New function.
31567         (get_compute_type): Use subparts_gt.
31568         (count_type_subparts): Delete.
31569         (expand_vector_operations_1): Use subparts_gt instead of
31570         count_type_subparts.
31572 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31573             Alan Hayward  <alan.hayward@arm.com>
31574             David Sherwood  <david.sherwood@arm.com>
31576         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31577         (vect_compile_time_alias): ...this new function.  Do the calculation
31578         on poly_ints rather than trees.
31579         (vect_prune_runtime_alias_test_list): Update call accordingly.
31581 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31582             Alan Hayward  <alan.hayward@arm.com>
31583             David Sherwood  <david.sherwood@arm.com>
31585         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31586         numbers of units.
31587         (vect_schedule_slp_instance): Likewise.
31589 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31590             Alan Hayward  <alan.hayward@arm.com>
31591             David Sherwood  <david.sherwood@arm.com>
31593         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31594         constant and extern definitions for variable-length vectors.
31595         (vect_get_constant_vectors): Note that the number of units
31596         is known to be constant.
31598 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31599             Alan Hayward  <alan.hayward@arm.com>
31600             David Sherwood  <david.sherwood@arm.com>
31602         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31603         of units as polynomial.  Choose between WIDE and NARROW based
31604         on multiple_p.
31606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31607             Alan Hayward  <alan.hayward@arm.com>
31608             David Sherwood  <david.sherwood@arm.com>
31610         * tree-vect-stmts.c (simd_clone_subparts): New function.
31611         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31613 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31614             Alan Hayward  <alan.hayward@arm.com>
31615             David Sherwood  <david.sherwood@arm.com>
31617         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31618         vectors as polynomial.  Use build_index_vector for
31619         IFN_GOMP_SIMD_LANE.
31621 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31622             Alan Hayward  <alan.hayward@arm.com>
31623             David Sherwood  <david.sherwood@arm.com>
31625         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31626         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31627         for variable-length vectors.
31628         (vectorizable_mask_load_store): Treat the number of units as
31629         polynomial, asserting that it is constant if the condition has
31630         already been enforced.
31631         (vectorizable_store, vectorizable_load): Likewise.
31633 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31634             Alan Hayward  <alan.hayward@arm.com>
31635             David Sherwood  <david.sherwood@arm.com>
31637         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31638         of units as polynomial.  Punt if we can't tell at compile time
31639         which vector contains the final result.
31641 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31642             Alan Hayward  <alan.hayward@arm.com>
31643             David Sherwood  <david.sherwood@arm.com>
31645         * tree-vect-loop.c (vectorizable_induction): Treat the number
31646         of units as polynomial.  Punt on SLP inductions.  Use an integer
31647         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31648         cast of such a series for variable-length floating-point
31649         reductions.
31651 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31652             Alan Hayward  <alan.hayward@arm.com>
31653             David Sherwood  <david.sherwood@arm.com>
31655         * tree.h (build_index_vector): Declare.
31656         * tree.c (build_index_vector): New function.
31657         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31658         of units as polynomial, forcibly converting it to a constant if
31659         vectorizable_reduction has already enforced the condition.
31660         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31661         to create a {1,2,3,...} vector.
31662         (vectorizable_reduction): Treat the number of units as polynomial.
31663         Choose vectype_in based on the largest scalar element size rather
31664         than the smallest number of units.  Enforce the restrictions
31665         relied on above.
31667 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31668             Alan Hayward  <alan.hayward@arm.com>
31669             David Sherwood  <david.sherwood@arm.com>
31671         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31672         number of units as polynomial.
31674 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31675             Alan Hayward  <alan.hayward@arm.com>
31676             David Sherwood  <david.sherwood@arm.com>
31678         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31679         * target.def (autovectorize_vector_sizes): Return the vector sizes
31680         by pointer, using vector_sizes rather than a bitmask.
31681         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31682         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31683         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31684         Likewise.
31685         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31686         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31687         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31688         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31689         * omp-general.c (omp_max_vf): Likewise.
31690         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31691         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31692         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31693         * tree-vect-slp.c (vect_slp_bb): Likewise.
31694         * doc/tm.texi: Regenerate.
31695         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31696         to a poly_uint64.
31697         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31698         the vector size as a poly_uint64 rather than an unsigned int.
31699         (current_vector_size): Change from an unsigned int to a poly_uint64.
31700         (get_vectype_for_scalar_type): Update accordingly.
31701         * tree.h (build_truth_vector_type): Take the size and number of
31702         units as a poly_uint64 rather than an unsigned int.
31703         (build_vector_type): Add a temporary overload that takes
31704         the number of units as a poly_uint64 rather than an unsigned int.
31705         * tree.c (make_vector_type): Likewise.
31706         (build_truth_vector_type): Take the number of units as a poly_uint64
31707         rather than an unsigned int.
31709 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31710             Alan Hayward  <alan.hayward@arm.com>
31711             David Sherwood  <david.sherwood@arm.com>
31713         * target.def (get_mask_mode): Take the number of units and length
31714         as poly_uint64s rather than unsigned ints.
31715         * targhooks.h (default_get_mask_mode): Update accordingly.
31716         * targhooks.c (default_get_mask_mode): Likewise.
31717         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31718         * doc/tm.texi: Regenerate.
31720 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31721             Alan Hayward  <alan.hayward@arm.com>
31722             David Sherwood  <david.sherwood@arm.com>
31724         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
31725         * omp-general.c (omp_max_vf): Likewise.
31726         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
31727         (expand_omp_simd): Handle polynomial safelen.
31728         * omp-low.c (omplow_simd_context): Add a default constructor.
31729         (omplow_simd_context::max_vf): Change from int to poly_uint64.
31730         (lower_rec_simd_input_clauses): Update accordingly.
31731         (lower_rec_input_clauses): Likewise.
31733 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31734             Alan Hayward  <alan.hayward@arm.com>
31735             David Sherwood  <david.sherwood@arm.com>
31737         * tree-vectorizer.h (vect_nunits_for_cost): New function.
31738         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
31739         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
31740         (vect_analyze_slp_cost): Likewise.
31741         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
31742         (vect_model_load_cost): Likewise.
31744 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31745             Alan Hayward  <alan.hayward@arm.com>
31746             David Sherwood  <david.sherwood@arm.com>
31748         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
31749         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
31750         from an unsigned int * to a poly_uint64_pod *.
31751         (calculate_unrolling_factor): New function.
31752         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
31754 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31755             Alan Hayward  <alan.hayward@arm.com>
31756             David Sherwood  <david.sherwood@arm.com>
31758         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
31759         from an unsigned int to a poly_uint64.
31760         (_loop_vec_info::slp_unrolling_factor): Likewise.
31761         (_loop_vec_info::vectorization_factor): Change from an int
31762         to a poly_uint64.
31763         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
31764         (vect_get_num_vectors): New function.
31765         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
31766         (vect_get_num_copies): Use vect_get_num_vectors.
31767         (vect_analyze_data_ref_dependences): Change max_vf from an int *
31768         to an unsigned int *.
31769         (vect_analyze_data_refs): Change min_vf from an int * to a
31770         poly_uint64 *.
31771         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31772         than an unsigned HOST_WIDE_INT.
31773         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
31774         (vect_analyze_data_ref_dependence): Change max_vf from an int *
31775         to an unsigned int *.
31776         (vect_analyze_data_ref_dependences): Likewise.
31777         (vect_compute_data_ref_alignment): Handle polynomial vf.
31778         (vect_enhance_data_refs_alignment): Likewise.
31779         (vect_prune_runtime_alias_test_list): Likewise.
31780         (vect_shift_permute_load_chain): Likewise.
31781         (vect_supportable_dr_alignment): Likewise.
31782         (dependence_distance_ge_vf): Take the vectorization factor as a
31783         poly_uint64 rather than an unsigned HOST_WIDE_INT.
31784         (vect_analyze_data_refs): Change min_vf from an int * to a
31785         poly_uint64 *.
31786         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
31787         vfm1 as a poly_uint64 rather than an int.  Make the same change
31788         for the returned bound_scalar.
31789         (vect_gen_vector_loop_niters): Handle polynomial vf.
31790         (vect_do_peeling): Likewise.  Update call to
31791         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
31792         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
31793         be constant.
31794         * tree-vect-loop.c (vect_determine_vectorization_factor)
31795         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
31796         (vect_get_known_peeling_cost): Likewise.
31797         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
31798         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
31799         (vect_transform_loop): Likewise.  Use the lowest possible VF when
31800         updating the upper bounds of the loop.
31801         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
31802         rather than an int.
31803         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
31804         polynomial unroll factors.
31805         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
31806         (vect_make_slp_decision): Likewise.
31807         (vect_supported_load_permutation_p): Likewise, and polynomial
31808         vf too.
31809         (vect_analyze_slp_cost): Handle polynomial vf.
31810         (vect_slp_analyze_node_operations): Likewise.
31811         (vect_slp_analyze_bb_1): Likewise.
31812         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31813         than an unsigned HOST_WIDE_INT.
31814         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
31815         (vectorizable_load): Handle polynomial vf.
31816         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
31817         a poly_uint64.
31818         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
31820 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31821             Alan Hayward  <alan.hayward@arm.com>
31822             David Sherwood  <david.sherwood@arm.com>
31824         * match.pd: Handle bit operations involving three constants
31825         and try to fold one pair.
31827 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31829         * tree-vect-loop-manip.c: Include gimple-fold.h.
31830         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
31831         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
31832         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
31833         Add a path that uses a step of VF instead of 1, but disable it
31834         for now.
31835         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
31836         and niters_no_overflow parameters.  Update calls to
31837         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
31838         Create a new SSA name if the latter choses to use a ste other
31839         than zero, and return it via niters_vector_mult_vf_var.
31840         * tree-vect-loop.c (vect_transform_loop): Update calls to
31841         vect_do_peeling, vect_gen_vector_loop_niters and
31842         slpeel_make_loop_iterate_ntimes.
31843         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
31844         (vect_gen_vector_loop_niters): Update declarations after above changes.
31846 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
31848         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
31849         128-bit round to integer instructions.
31850         (ceil<mode>2): Likewise.
31851         (btrunc<mode>2): Likewise.
31852         (round<mode>2): Likewise.
31854 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31856         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
31857         unaligned VSX load/store on P8/P9.
31858         (expand_block_clear): Allow the use of unaligned VSX
31859         load/store on P8/P9.
31861 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31863         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
31864         New function.
31865         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
31866         swap associated with both a load and a store.
31868 2018-01-02  Andrew Waterman  <andrew@sifive.com>
31870         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
31871         * config/riscv/riscv.md (clear_cache): Use it.
31873 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
31875         * web.c: Remove out-of-date comment.
31877 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31879         * expr.c (fixup_args_size_notes): Check that any existing
31880         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
31881         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
31882         (emit_single_push_insn): ...here.
31884 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31886         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
31887         (const_vector_encoded_nelts): New function.
31888         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
31889         (const_vector_int_elt, const_vector_elt): Declare.
31890         * emit-rtl.c (const_vector_int_elt_1): New function.
31891         (const_vector_elt): Likewise.
31892         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
31893         of CONST_VECTOR_ELT.
31895 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31897         * expr.c: Include rtx-vector-builder.h.
31898         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
31899         directly on the tree encoding.
31900         (const_vector_from_tree): Likewise.
31901         * optabs.c: Include rtx-vector-builder.h.
31902         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
31903         sequence of "u" values.
31904         * vec-perm-indices.c: Include rtx-vector-builder.h.
31905         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
31906         directly on the vec_perm_indices encoding.
31908 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31910         * doc/rtl.texi (const_vector): Describe new encoding scheme.
31911         * Makefile.in (OBJS): Add rtx-vector-builder.o.
31912         * rtx-vector-builder.h: New file.
31913         * rtx-vector-builder.c: Likewise.
31914         * rtl.h (rtx_def::u2): Add a const_vector field.
31915         (CONST_VECTOR_NPATTERNS): New macro.
31916         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
31917         (CONST_VECTOR_DUPLICATE_P): Likewise.
31918         (CONST_VECTOR_STEPPED_P): Likewise.
31919         (CONST_VECTOR_ENCODED_ELT): Likewise.
31920         (const_vec_duplicate_p): Check for a duplicated vector encoding.
31921         (unwrap_const_vec_duplicate): Likewise.
31922         (const_vec_series_p): Check for a non-duplicated vector encoding.
31923         Say that the function only returns true for integer vectors.
31924         * emit-rtl.c: Include rtx-vector-builder.h.
31925         (gen_const_vec_duplicate_1): Delete.
31926         (gen_const_vector): Call gen_const_vec_duplicate instead of
31927         gen_const_vec_duplicate_1.
31928         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
31929         (gen_const_vec_duplicate): Use rtx_vector_builder.
31930         (gen_const_vec_series): Likewise.
31931         (gen_rtx_CONST_VECTOR): Likewise.
31932         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
31933         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31934         Build a new vector rather than modifying a CONST_VECTOR in-place.
31935         (handle_special_swappables): Update call accordingly.
31936         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
31937         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31938         Build a new vector rather than modifying a CONST_VECTOR in-place.
31939         (handle_special_swappables): Update call accordingly.
31941 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31943         * simplify-rtx.c (simplify_const_binary_operation): Use
31944         CONST_VECTOR_ELT instead of XVECEXP.
31946 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31948         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
31949         the selector elements to be different from the data elements
31950         if the selector is a VECTOR_CST.
31951         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
31952         ssizetype for the selector.
31954 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31956         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
31957         before testing each element individually.
31958         * tree-vect-generic.c (lower_vec_perm): Likewise.
31960 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31962         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
31963         * selftest-run-tests.c (selftest::run_tests): Call it.
31964         * vector-builder.h (vector_builder::operator ==): New function.
31965         (vector_builder::operator !=): Likewise.
31966         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
31967         (vec_perm_indices::all_from_input_p): New function.
31968         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31969         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
31970         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
31971         instead of reading the VECTOR_CST directly.  Detect whether both
31972         vector inputs are the same before constructing the vec_perm_indices,
31973         and update the number of inputs argument accordingly.  Use the
31974         utility functions added above.  Only construct sel2 if we need to.
31976 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31978         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
31979         the broadcast of the low byte.
31980         (expand_mult_highpart): Use an explicit encoding for the permutes.
31981         * optabs-query.c (can_mult_highpart_p): Likewise.
31982         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
31983         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31984         (vectorizable_bswap): Likewise.
31985         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
31986         explicit encoding for the power-of-2 permutes.
31987         (vect_permute_store_chain): Likewise.
31988         (vect_grouped_load_supported): Likewise.
31989         (vect_permute_load_chain): Likewise.
31991 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31993         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
31994         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
31995         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
31996         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31997         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
31998         (vect_gen_perm_mask_any): Likewise.
32000 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32002         * int-vector-builder.h: New file.
32003         * vec-perm-indices.h: Include int-vector-builder.h.
32004         (vec_perm_indices): Redefine as an int_vector_builder.
32005         (auto_vec_perm_indices): Delete.
32006         (vec_perm_builder): Redefine as a stand-alone class.
32007         (vec_perm_indices::vec_perm_indices): New function.
32008         (vec_perm_indices::clamp): Likewise.
32009         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32010         (vec_perm_indices::new_vector): New function.
32011         (vec_perm_indices::new_expanded_vector): Update for new
32012         vec_perm_indices class.
32013         (vec_perm_indices::rotate_inputs): New function.
32014         (vec_perm_indices::all_in_range_p): Operate directly on the
32015         encoded form, without computing elided elements.
32016         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32017         encoding.  Update for new vec_perm_indices class.
32018         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32019         the given vec_perm_builder.
32020         (expand_vec_perm_var): Update vec_perm_builder constructor.
32021         (expand_mult_highpart): Use vec_perm_builder instead of
32022         auto_vec_perm_indices.
32023         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32024         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
32025         or double series encoding as appropriate.
32026         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32027         vec_perm_indices instead of auto_vec_perm_indices.
32028         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32029         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32030         (vect_permute_store_chain): Likewise.
32031         (vect_grouped_load_supported): Likewise.
32032         (vect_permute_load_chain): Likewise.
32033         (vect_shift_permute_load_chain): Likewise.
32034         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32035         (vect_transform_slp_perm_load): Likewise.
32036         (vect_schedule_slp_instance): Likewise.
32037         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32038         (vectorizable_mask_load_store): Likewise.
32039         (vectorizable_bswap): Likewise.
32040         (vectorizable_store): Likewise.
32041         (vectorizable_load): Likewise.
32042         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32043         vec_perm_indices instead of auto_vec_perm_indices.  Use
32044         tree_to_vec_perm_builder to read the vector from a tree.
32045         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32046         vec_perm_builder instead of a vec_perm_indices.
32047         (have_whole_vector_shift): Use vec_perm_builder and
32048         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
32049         truncation to calc_vec_perm_mask_for_shift.
32050         (vect_create_epilog_for_reduction): Likewise.
32051         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32052         from auto_vec_perm_indices to vec_perm_indices.
32053         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32054         instead of changing individual elements.
32055         (aarch64_vectorize_vec_perm_const): Use new_vector to install
32056         the vector in d.perm.
32057         * config/arm/arm.c (expand_vec_perm_d::perm): Change
32058         from auto_vec_perm_indices to vec_perm_indices.
32059         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32060         instead of changing individual elements.
32061         (arm_vectorize_vec_perm_const): Use new_vector to install
32062         the vector in d.perm.
32063         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32064         Update vec_perm_builder constructor.
32065         (rs6000_expand_interleave): Likewise.
32066         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32067         (rs6000_expand_interleave): Likewise.
32069 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32071         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
32072         to qimode could truncate the indices.
32073         * optabs.c (expand_vec_perm_var): Likewise.
32075 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32077         * Makefile.in (OBJS): Add vec-perm-indices.o.
32078         * vec-perm-indices.h: New file.
32079         * vec-perm-indices.c: Likewise.
32080         * target.h (vec_perm_indices): Replace with a forward class
32081         declaration.
32082         (auto_vec_perm_indices): Move to vec-perm-indices.h.
32083         * optabs.h: Include vec-perm-indices.h.
32084         (expand_vec_perm): Delete.
32085         (selector_fits_mode_p, expand_vec_perm_var): Declare.
32086         (expand_vec_perm_const): Declare.
32087         * target.def (vec_perm_const_ok): Replace with...
32088         (vec_perm_const): ...this new hook.
32089         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
32090         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
32091         * doc/tm.texi: Regenerate.
32092         * optabs.def (vec_perm_const): Delete.
32093         * doc/md.texi (vec_perm_const): Likewise.
32094         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
32095         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
32096         expand_vec_perm for constant permutation vectors.  Assert that
32097         the mode of variable permutation vectors is the integer equivalent
32098         of the mode that is being permuted.
32099         * optabs-query.h (selector_fits_mode_p): Declare.
32100         * optabs-query.c: Include vec-perm-indices.h.
32101         (selector_fits_mode_p): New function.
32102         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
32103         is defined, instead of checking whether the vec_perm_const_optab
32104         exists.  Use targetm.vectorize.vec_perm_const instead of
32105         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
32106         fit in the vector mode before using a variable permute.
32107         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
32108         vec_perm_indices instead of an rtx.
32109         (expand_vec_perm): Replace with...
32110         (expand_vec_perm_const): ...this new function.  Take the selector
32111         as a vec_perm_indices rather than an rtx.  Also take the mode of
32112         the selector.  Update call to shift_amt_for_vec_perm_mask.
32113         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
32114         Use vec_perm_indices::new_expanded_vector to expand the original
32115         selector into bytes.  Check whether the indices fit in the vector
32116         mode before using a variable permute.
32117         (expand_vec_perm_var): Make global.
32118         (expand_mult_highpart): Use expand_vec_perm_const.
32119         * fold-const.c: Includes vec-perm-indices.h.
32120         * tree-ssa-forwprop.c: Likewise.
32121         * tree-vect-data-refs.c: Likewise.
32122         * tree-vect-generic.c: Likewise.
32123         * tree-vect-loop.c: Likewise.
32124         * tree-vect-slp.c: Likewise.
32125         * tree-vect-stmts.c: Likewise.
32126         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
32127         Delete.
32128         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
32129         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
32130         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
32131         (aarch64_vectorize_vec_perm_const): ...this new function.
32132         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32133         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32134         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32135         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32136         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32137         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32138         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32139         into...
32140         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
32141         check for NEON modes.
32142         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32143         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32144         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32145         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32146         into...
32147         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
32148         the old VEC_PERM_CONST conditions.
32149         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32150         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32151         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32152         (ia64_vectorize_vec_perm_const_ok): Merge into...
32153         (ia64_vectorize_vec_perm_const): ...this new function.
32154         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32155         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32156         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32157         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32158         * config/mips/mips.c (mips_expand_vec_perm_const)
32159         (mips_vectorize_vec_perm_const_ok): Merge into...
32160         (mips_vectorize_vec_perm_const): ...this new function.
32161         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32162         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32163         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32164         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32165         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32166         (rs6000_expand_vec_perm_const): Delete.
32167         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32168         Delete.
32169         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32170         (altivec_expand_vec_perm_const_le): Take each operand individually.
32171         Operate on constant selectors rather than rtxes.
32172         (altivec_expand_vec_perm_const): Likewise.  Update call to
32173         altivec_expand_vec_perm_const_le.
32174         (rs6000_expand_vec_perm_const): Delete.
32175         (rs6000_vectorize_vec_perm_const_ok): Delete.
32176         (rs6000_vectorize_vec_perm_const): New function.
32177         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32178         an element count and rtx array.
32179         (rs6000_expand_extract_even): Update call accordingly.
32180         (rs6000_expand_interleave): Likewise.
32181         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32182         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32183         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32184         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32185         (rs6000_expand_vec_perm_const): Delete.
32186         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32187         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32188         (altivec_expand_vec_perm_const_le): Take each operand individually.
32189         Operate on constant selectors rather than rtxes.
32190         (altivec_expand_vec_perm_const): Likewise.  Update call to
32191         altivec_expand_vec_perm_const_le.
32192         (rs6000_expand_vec_perm_const): Delete.
32193         (rs6000_vectorize_vec_perm_const_ok): Delete.
32194         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
32195         reference to the SPE evmerge intructions.
32196         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32197         an element count and rtx array.
32198         (rs6000_expand_extract_even): Update call accordingly.
32199         (rs6000_expand_interleave): Likewise.
32200         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32201         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32202         new function.
32203         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32205 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32207         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32208         vector mode and that that mode matches the mode of the data
32209         being permuted.
32210         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32211         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
32212         directly using expand_vec_perm_1 when forcing selectors into
32213         registers.
32214         (expand_vec_perm_var): New function, split out from expand_vec_perm.
32216 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32218         * optabs-query.h (can_vec_perm_p): Delete.
32219         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32220         * optabs-query.c (can_vec_perm_p): Split into...
32221         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32222         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32223         particular selector is valid.
32224         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32225         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32226         (vect_grouped_load_supported): Likewise.
32227         (vect_shift_permute_load_chain): Likewise.
32228         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32229         (vect_transform_slp_perm_load): Likewise.
32230         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32231         (vectorizable_bswap): Likewise.
32232         (vect_gen_perm_mask_checked): Likewise.
32233         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32234         implementations of variable permutation vectors into account
32235         when deciding which selector to use.
32236         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32237         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32238         with a false third argument.
32239         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32240         to test whether the constant selector is valid and can_vec_perm_var_p
32241         to test whether a variable selector is valid.
32243 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32245         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32246         * optabs-query.c (can_vec_perm_p): Likewise.
32247         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32248         instead of vec_perm_indices.
32249         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32250         (vect_gen_perm_mask_checked): Likewise,
32251         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32252         (vect_gen_perm_mask_checked): Likewise,
32254 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32256         * optabs-query.h (qimode_for_vec_perm): Declare.
32257         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32258         (qimode_for_vec_perm): ...this new function.
32259         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32261 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32263         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32264         does not have a conditional at the top.
32266 2018-01-02  Richard Biener  <rguenther@suse.de>
32268         * ipa-inline.c (big_speedup_p): Fix expression.
32270 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32272         PR target/81616
32273         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32274         for generic 4->6.
32276 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32278         PR target/81616
32279         Generic tuning.
32280         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32281         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32282         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32283         cond_taken_branch_cost 3->4.
32285 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32287         PR tree-optimization/83581
32288         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32289         TODO_cleanup_cfg if any changes have been made.
32291         PR middle-end/83608
32292         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32293         convert_modes if target mode has the right side, but different mode
32294         class.
32296         PR middle-end/83609
32297         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32298         last argument when extracting from CONCAT.  If either from_real or
32299         from_imag is NULL, use expansion through memory.  If result is not
32300         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32301         the parts directly to inner mode, if even that fails, use expansion
32302         through memory.
32304         PR middle-end/83623
32305         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32306         check for bswap in mode rather than HImode and use that in expand_unop
32307         too.
32309 Copyright (C) 2018 Free Software Foundation, Inc.
32311 Copying and distribution of this file, with or without modification,
32312 are permitted in any medium without royalty provided the copyright
32313 notice and this notice are preserved.